Вы находитесь на странице: 1из 512

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016

Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
EVALUATION OF RELATIVE
DENSITY AND ITS ROLE IN
GEOTECHNICAL PROJECTS
INVOLVING COHESIONLESS
SOILS

A symposium
presented at the
Seventy-fifth Annual Meeting
AMERICAN SOCIETY FOR
TESTING AND MATERIALS
Los Angeles, Calif., 25-30 June 1972

ASTM SPECIAL TECHNICAL PUBLICATION 523


E. T. Selig and R. S. Ladd, editors

List price $30.75


04-523000-38

AMERICAN SOCIETY FOR TESTING AND MATERIALS


1916 Race Street, Philadelphia, Pa. 19103.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agree
(~) BY AMERICAN SOCIETY FOR TESTING AND MATERIALS 1973
Library of Congress Catalog Card Number: 72-90704

NOTE
The Society is not responsible, as a body,
for the statements and opinions
advanced in this publication.

Printed in Baltimore, Md.


July 1973

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
Foreword
A two-session symposium on evaluation of relative density and its role
in geotechnical projects involving cohesionless soils was held 25-30 June
1972 at the Seventy-fifth Annual Meeting of the American Society for
Testing and Materials in'Los Angeles , Calif. The sponsor of the symposium
was Committee D-18 on Soil and Rock for Engineering Purposes under the
chairmanship of E. B. Hall. The format for each session consisted of a
keynote address, followed by presentation of selected papers and then a
panel discussion on the session topic. Session I concerned the factors
affecting relative density including the measurement of maximum, mini-
mum, and in situ density. Session II concerned the correlation between
relative density and properties or performance of soils and gives examples
of the use of relative density.
The keynote address for the first session was given by W. G. Holtz,
Consulting Civil Engineer, Wheat Ridge, Colorado; while for the second
session, Yves Lacroix, Director, Woodward-Clyde Consultants, New York,
N. Y., presented the keynote address. The symposium chairman and also
moderator of Session I was E. T. Selig, Department of Civil Engineering,
State University of New York at Buffalo; R. S. Ladd Woodward-Moor-
house & Associates, Inc., Clifton, New Jersey, served as cochairman and
moderator of Session II.
These published proceedings contain all of the accepted papers dealing
with the symposium topic, most of which were not presented orally, and the
two keynote addresses. A concluding paper summarizes the program
discussion and provides recommendations.
The chairman and co-chairman would like to express their appreciation
to the staff members of ASTM who assisted in the presentation of this
symposium and Special Technical Publication, and especially to Miss
Jane Wheeler.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Ag
Related
ASTM Publications

Underwater Soil Sampling, Testing, and Construction


Control, STP 501 (1972), $15.50

Special Procedures for Testing Soil and Rock for En-


gineering Purposes, STP 479 (1970), $15.75

Laboratory Shear Testing of Soils, STP 361, (1965),


$24.50

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement
Contents

Introduction 1

Determination of Relative Density Considering the


Measurement of Maximum, Minimum, and In Si t u Density
The Relative Density Approach--Uses, Testing Requirements, Reliability,
and Shortcomings--w. o. HOLTZ 5
Accuracy of Relative Density Measurements: Results of a Comparative Test
Program--F. h. TAVENAS, R. S. LADD, AND P. LA ROCHELLE 18

Variability of Laboratory Test Results--D. A. TIEDEMANN 61


Statistical Significance of the Relative Density--YOSnlXKI YOSHIMI A~D
IKUO TOHNO 74
Effect of Variations in Minimum Density on Relative Density--R. c. GUVrA
AND J. D. MCKEOWN 85

Factors Controlling Maximum and Minimum Densities of Sands--T. L. YOUD 98


Influence of Grain Shape and Size upon the Limiting Porosities of S a n d s - -
E, A. D I C K I N 113
Some Observations on the Control of Density by Vibration--E. w. BRAND 121

Laboratory Studies of Maximum and Minimum Dry Densities of Cohesionless


Soils--M. M. JOHNSTON 133
Maximum Density Determination of Subbase Materials--o. CUMBERLEDGE
AND R. J'. COMINSKY 141
Compaction of Sand on a Vertically Vibrating Table---RICARDO DOBRY AND
2. V. WHITMAN 156

Vibratory Compaction in the Laboratory of Granular Materials in Long


Columns--A. I. JOHNSON AND D. A. MORRIS 171
Uniformity of Saturated Sand Specimens--J. J. EMERY, W. D. LIAMFINN, ANn
K. W. LEE 182
Errors of In-Place Density Measurements in Cohesionless Soils--
D. F. GRIFFIN 195

Some Testing Experiences and Characteristics of Boulder-Gravel Fill in


Earth Dams--R. J. FROST 207
Relative Density Tests on Rock Fill at Carters Dam--R. z. STEPHENSON 234

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
vi CONTENTS

Correlation Between Relative Density and Measured


Performance or Properties of Granular Soils
Direct Determination and Indirect Evaluation of Relative Density and Its
Use on Earthwork Construction Projects--YVES LACROIX AND
H. M. HORN 251
Prediction of Drained Strength of Sands from Relative Density Measure-
ments--D. H. CORNFORTH 281
Effect of Particle Shape on the Engineering Properties of Granular Soils--
I. ttOLUBEC AND E. D~APPOLONIA 304

Effect of Relative Density on the Liquefaction Susceptibility of a Fine Sand


under Controlled-Stress Loading--G. N. DURHAMAND F. C. TOWNSEND 319
Influence of Relative Density on the Strength and Deformation of Sand under
Plane Strain Conditions--~. M. A L - H U S S A I N I 332
Comparisons of Vibrated Density and Standard Compaction Tests on Sands
with Varying Amounts of Fines--F. c. TOWNSEND 348
Determination of Relative Density of Sand Below Groundwater T a b l e - -
J . O. OSTERBERG AND SERGE VARAKSIN 364
Discussion 376
Use of Relative Density in Geotechnical Projects
Experience with Relative Density as a Construction Control Criterion--
D. J . LEARY AND R. J . WOODWARD~ I I I 381
Density Measurements in a Hydraulic F i l l - - s . J. POULOS AND A. HED 402
Field and Laboratory Determination of Maximum Density in Coarse Sands
and Gravels for Mica D a m - - w . i. LOW AND C. SENER 425
Correlation Between Gradational Parameters and Limiting Densities for
Cohesionless Materials Placed Hydraulically--m M. REITZ 444
Comparison of Relative Densities Estimated Using Different Approaches--
R. A. BELL AND J. P. SINGH 455
Relative Density--Three Examples of Its Use in Research and P r a c t i c e - -
K . - J . MELZER 463
Difficulties in the Use of Relative Density as a Soil Parameter--F. A. TAV~NAS 478
Summary
Evaluation of Relative Density Measurements and Applicati0ns--E. T. SELm
AND 1~. S. LADD 487

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agree
STP523-EB/Jul. 1973

Introduction

Recognizing the importance of relative density in engineering problems


dealing with cohesionless soils and the number of American Society for
Testing and Materials standards related to this topic, ASTM Committee
D-18 on Soil and Rock for Engineering Purposes considered it desirable to
organize and sPonsor this symposium. Emphasis was placed on the following
four items concerning the evaluation of relative density and its role in
geotechnical projects involving cohesionless soils: (1) determination of
relative density considering the measurement of maximum, minimum, and
in situ or sample density as well as the reliability of relative density and the
factors influencing it, (2) correlation between relative density and measured
performance or properties of granular soils, (3) applications of relative
density to geotechnical pro~ects involving cohesionless soils and the
usefulness of this concept, and (4) evaluation of the existing ASTM
standards concerning relative density with recommendations for improve-
ments in the future.
The symposium was divided into two sessions. The morning session
primarily covered the measurement of relative density. The session began
with an address by W. G. Holtz which defined relative density and
applicable soils, explained the history of the test standards, and discussed
the reliability of the concept. One of the highlights of this session was the
presentation of F. A. Tavenas of the results of the cooperative testing
program carried out by forty-one (41) soil laboratories in the United States
and Canada to evaluate the accuracy and precision of relative density
measurement. Fifteen papers on the topic of Session I were accepted by the
symposium committee. Several of these were selected for presentation to
provide further background prior to the panel discussion. Panel members in
Session I were: E. T. Selig (moderator), State University of New York at
Buffalo; D. F. Griffin, Naval Civil Engineering Laboratory; W. G. Holtz,
Consulting Engineer; R. S. Ladd, Woodward-Moorhouse & Associates;
R. J. Stephenson, South Atlantic Division Laboratory, U.S. Army Corps of
Engineers; F. A. Tavenas, Laval University; and D. A. Tiedemann,
U.S. Bureau of Reclamation.
The afternoon session dealt with applications of relative density,
including correlation with properties and performance and an assessment
of usefulness. The session began with illustrations of applications of relative

Copyright by ASTM Int'l


Copyright9 by(all
ASTMrightsInternational
reserved); Fri Mar 11 16:13:06 EST 2016
www.astm.org
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
2 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

density by Yves Lacroix. Twelve papers were accepted on the topic of


Session II. Following the presentation of several of these papers, an active
discussion was held by the panel as well as persons attending the symposium
concerning the reliability of relative density and its usefulness. The panel
members for this session were: R. S. Ladd (moderator), Woodward-
Moorhouse & Associates; W. G. Holtz, Consulting Engineer; I. Holubec,
E. D'Appolonia Consulting Engineers; Yves Lacroix, Director, Woodward-
Clyde Consultants; S. J. Poulos, Geotechnical Engineers; E. T. Selig, State
University of New York at Buffalo; R. J. Stephenson, South Atlantic
Division Laboratory , U.S. Army Corps of Engineers.
The papers have been grouped in these proceedings by session topic
following the session keynote address. The final paper evaluates the entire
written and oral content of the symposium and, on the basis of this
information, summarizes the results and recommends future action by
ASTM. It is the hope of the authors and sponsors of this symposium that
this ASTM Special Technical Publication will provide a comprehensive
enough evaluation of relative density and its role in geotechnical proj ects to
enable practicing engineers to function more effectively on projects
involving cohesionless soils. It is also hoped that the accumulated knowledge
of test techniques, apparatus, and applications will be used as a basis for
improving the many test standards associated with relative density.

E. T. Selig
Associate Professor
Department of Civil Engineering
State University of New York at Buffalo
Buffalo, N. Y.

R. S. L~dd
Laboratory Director
Woodward-Moorhouse& Associates, Inc.
Clifton, N. J.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreeme
Determination of Relative Density
Considering the Measurement of Maximum,
Minimum, and In Situ Density

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further repr
W. G. Holtz ~

The Relative Density Approach---Uses,


Testing Requirements, Reliability, and
Shortcomings

REFERENCE: Holtz, W. G., "The Relative Density Approach--Uses,


Testing Requirements, Reliability, and Shortcomings," Evaluation of
Relative Density and Its Role in Geotechnical Projects Involving Cohesionless
Soils, A S T M S T P 523, American Society for Testing and Materials, 1973,
pp. 5-17.
ABSTRACT: To evaluate the density of free draining cohesionless soils in
terms of percent relative density, there are three parameters that must be
determined. These are (1) the minimum density that the soil can be structured,
(2) the maximum density that the soil can be structured, and (3) the density
of the soil being evaluated. Errors or variations can be obtained in determining
each of these three values. The reasons for such variations include (1) lack of
specimen similarity, (2) dissimilar test procedures used, (3) testing equipment
not similar or not properly maintained, and variations in the training and
expertise of operator. The results of a large number of tests by numerous
laboratories indicate that the variations associated with the minimum and
maximum density tests are about the same as those associated with the impact
compaction tests. The requirement for determination of three parameters to
determine the relative density value can lead to the compounding of errors and
variations in the worst cases. The use of proper test methods, closely followed;
specified equipment, well maintained; well trained test personnel, and care in
sampling should provide reproducible relative density values within 10 percent
relative density.
KEY WORDS: cohesionless soils, density (mass/volume), tests, soils, repro-
ducibility, soil compacting, reliability, soil mechanics

T h e s u b j e c t of t h e r e l a t i v e d e n s i t y a p p r o a c h to c e r t a i n g e o t e c h n i c a l
p r o b l e m s is e x t r e m e l y t i m e l y a n d of g r e a t i m p o r t a n c e to soils engineers.
I w a n t t o b e g i n t h i s p a p e r b y d e s c r i b i n g w h a t we a r e t a l k i n g a b o u t w h e n we
s p e a k of r e l a t i v e d e n s i t y , w h y r e l a t i v e d e n s i t y c r i t e r i a a r e sensible a p -
p r o a c h e s t o c e r t a i n soil p r o b l e m s , t h e origin of t h e p r e s e n t A S T M T e s t for

i Consulting civil engineer, and former assistant chief, Division of Research, U. S.


Bureau of Reclamation, retired; Wheat Ridge, Colo. 80033.

Copyright by ASTM Int'l


Copyright9 by(all
ASTMrightsInternational
reserved); Fri Mar 11 16:13:06 EST 2016
www.astm.org
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
6 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

Relative Density of Cohesionless Soils, (D 2049-69), and what we would


like to accomplish.
I am sure that most everyone is familiar with the relative density concept,
but before I begin, I wish to define it. According to ASTM Definitions of
Terms and Symbols Relating to Soil and Rock Mechanics (D 653-67),
relative density is defined as the ratio of (1) the difference between the void
ratio of a cohesionless soil in the loosest state (e~x) and any given void
ratio (e), to (2) the difference between its void ratios in the loosest (emax)
and in the densest (emln) states. In equation form this becomes
emax m e
Dd ---- • 100, in terms of void ratio
eraax- emin

or

max (~ - ~/min)
Dd = X 100, in terms of density,
y (~, max - ~ min)

which is the form most of us are concerned with in engineering project


applications.
Thus, there are three parameters which must be determined: (1) emax or
Ymln, which describes the most loose state that a particular soil can be
structured; (2) emln or y . . . . which describes the most dense state that the
soil can be structured; and (3) e or -y which is the in-place density of a
natural deposit or fill, or perhaps the density of a laboratory sample or
research test specimen, which we desire to describe in terms of that soils
densest and loosest states. Therefore, the most loose state is zero percent
relative density, and the most dense state is 100 percent relative density,
and, theoretically, no lower or higher density states, respectively, should
exist without changes in gradation.
We also need to consider what soils are applicable to the relative density
approach. While the title of this symposium refers only to cohesionless
soils, it is certain that cohesionless silts or sandy or gravelly soils containing
excessive amounts of cohesionless silt are not applicable to the relative
density criteria. A more descriptive term might include the words pervious,
free-draining, or perhaps near free-draining.
Some years ago when we were working on vibrated sand backfill for the
San Diego Aqueduct and other large conduits, I proposed some limits of
soil types for compaction-by-vibration purposes [1]2. These limits which
were based on tests of several different sand and sand-gravel soils with
different amounts of - 2 0 0 fines [2, 3] were as follows:
1. GW, GP, SW, and SP soils are suitable. (The fines in these soils are
limited to 5 percent by definition.)

The italic numbers in brackets refer to the list of references appended to this paper.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to
HOLTZ ON THE RELATIVE DENSITY APPROACH 7

2. Borderline GW-GM, GW-GC, GP-GM, and GP-GC soils containing


less than 8 percent fines are usually suitable.
3. Borderline SW-SM, SP-SM, and SP-SC soils are suitable. (Fines in
these soils are limited to 12 percent by definition.)
4. SM and SC soils require special consideration and suitability depends
upon gradation of the sand and the plasticity of the fines. Some SM soils
with fines as high as 16 percent have been found to be suitable.
In this case suitability was defined as limiting amount of fines where the
specified minimum percent relative density became less than the specified
minimum percent compaction based on ASTM Tests for Moisture-Density
Relations of Soils, Using 5.5-1b Rammer and 12-in. Drop, (D 698-70).
Based upon literature recently studied, I can find no reason to modify the
preceeding criteria. We should remember that when we consider compaction
by vibration, we are really talking about soils that are adequately pervious
to drain excess water during the vibration period. We learned long ago, for
instance, that loose, silty, loessial, foundation soils could not be compacted
by vibroflotation methods which incorporated a large vibroflot (sand = 44
percent, Silt = 47 percent, -0.005 m m = 9 percent) [4]. The research
reported by F. C. Townsend in this symposium provides some excellent
additional information on this subject.

Why Relative Density


There are several ways by which the denseness of sand and gravel soils
can be expressed. These may be simple density, void-ratio, or degree of
compaction based upon some impact energy or vibration condition. Many
years ago it was suggested that relative density would be an appropriate
means to define the looseness and denseness of sand or sand-gravel soils in
a meaningful way, because important properties such as shear and con-
solidation could be correlated quite well by this means [5]. Relative density
has been found useful in liquifaction studies and other seismic studies for
sand and gravel soil foundations and embankments subject to earthquakes
or other vibrational conditions. Relative density is useful to control
laboratory test specimens, to use as a specification density requirement for
fill and foundation construction, and as a means to assess the competence
of natural sand-gravel soil deposits for foundation uses.

Definitions and Standards


As soon as engineers started to use relative density as a soil parameter we
began to confuse ourselves, because we did not have a common definition or
set of standards to work from. William T. Cavanaugh, Managing Director
of ASTM, wrote in the April 1972 issue of Materials Research and Standards:
"The evolution of language in the slow victory over Babel was man's
greatest foreward move in his eternal struggle to bring order out of chaos.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License
8 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

TABLE 1--Various terms used to describe the state of denseness.

Relative Density, percent 0 10 20 30 40 50 60 70 80 90 100


I I I I I I I I I
USBR [6] very very
Lambe and Whitman [7] loose loose medium dense dense

very
Burmister [5] loose medium ~compact compac

very very
Meyerhoff [8] loose loose compact dense dense

very
Hough [9] loose firm compact compact

Tschebotarioff [10]
Plummer and Dore [11] loose medium dense

Standardization, of course, is nothing more than the forced-draft creation


of language. Language itself is a set of standard symbols for things and
concepts. The standards that ASTM produces comprise a language for
commerce, a language for research, a language for regulation, and a
language, if you will, for accommodating the fruits of science and technology
to our culture."
We are still going through a certain amount of Babel with respect to
describing the looseness and denseness of sand-gravel soils. Throughout the
soils engineering literature one can find various words used to describe the
state of denseness, some of which are listed in Table 1. The differences
which can be construed from qualitative word groupings can readily be
seen. For instance, "compact" has an entirely different meaning to
Burmister than to Meyerhoff. If these word descriptions were used on drill
hole logs, the user of the logs could not be sure of the true meaning that was
to be conveyed. Some effort should be made b y Subcommittee D 18.92 on
Nomenclature for Soil and Rock Mechanics to provide uniform definitions.
There also was no agreement as to how the relative density of a soil was
to be determined quantitatively. If one is to analyse a foundation or
structure situation, particularly when seismic effects are involved, on the
basis of soil properties, rather than test indices such as penetration N
values, it is necessary to assign quantitative relative density values to the
soil. Thus, the need for adequate test procedures is readily apparent.
The determination of the relative density of any soil specimen or soil
mass requires three density determinations; this is one of the shortcomings
of the relative density approach. Errors, or different results obtained

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant t
HOLTZ ON THE RELATIVEDENSITY APPROACH 9

through differences in test procedures, involved in the three determinations


can significantly magnify the errors or differences determined for the
"percent relative density."
Recognizing that some kind of standards would be necessary, if we were
all to work on the same basis and be able.to interpret the data of others,
Section D, Subcommittee 3, of Committee D-18 was established in June
1954 under the Chairmanship of Earl Felt to work on standards for
determining the minimum and maximum densities of sand soils and gravel
soils. The work of the Subcommittee resulted in todays ASTM D 2049-69
which was approved by Committee D-18 in 1964. Probably more coopera-
tive research was conducted on this standard test method, prior to adoption,
than most any other adopted by Committee D-18 at that time.
Earl Felt first organized a group of subcommittee members from eight
organizations who were willing to participate in a cooperative testing and
research study. Samples of six different soils, varying from fine sand to open
graded crushed rock were sent to each of the members. Some twelve differ-
ent methods, including vibration, impact hammer, dropping, tapping (dry
and saturated), etc., were experimented with to determine the maximum
density. Some six different methods for minimum density determinations
were also conducted. The results of this work were reported at the 1958
Annual Meeting of the Society [12]. The results of similar tests by individ-
uals and different laboratories were compared. Tentative selections for the
most promising and consistent methods were made, and a second program
of testing, involving a limited number of laboratories, was undertaken. On
the basis of the cooperative programs, the Subcommittee wrote the test
procedures essentially as they are now written in ASTM D 2049-69.
It took several years until 1964 to get this standard through
Subcommittee 3 and Committee D-18. It was recognized that there were
some shortcomings. For instance, there may be a few situations where field
compaction of certain soils, that are applicable to the relative density
approach, may produce densities in excess of the laboratory maximum
values obtained by both dry and saturated procedures. However, these
appeared to be few, and reproducibility appeared to be as good as those
being obtained by other compaction methods such as ASTM D 698-70 and
ASTM Tests for Moisture-Density Relations of Soils, Using 10-1b Rammer
and 18-in. Drop, (D 1557-70). It was the opinion of the majority of
Subcommittee members that the benefits of a reasonably good standard
method, by which we all spoke the same language, far outweighed any
disadvantages of the particular test methods then proposed. In addition,
it was felt that by getting a tentative standard method published, many
persons and organizations would work with it, evaluate its suitability, and
suggest beneficial modifications and improvements. In other words, it gave
us something to shoot at. It is worth noting that in the eight years since the
adoption of ASTM D 2049-69 no significant, practical changes have been

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licen
]0 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

suggested to the Subcommittee. This has been disappointing. As usual,


there have been some criticisms, but no one has really brought forward
better methods that can be used for day-to-day engineering project work.
It is hoped that this symposium will stimulate further study and improve-
ments in the present standards, if it is determined that this is advisable.

Reliability of Relative Density Values


General
If we are to use the relative density approach, we must be able to
measure the three required parameters--sample or in-place density,
absolute minimum density, and absolute maximum density--with sufficient
accuracy to develop a meaningful relative density reliability. This con-
stitutes the theme for this first session of the symposium, I will try to bring
some of the symposium papers into focus.
Before we try to analyse the reliability and meaningfullness of the
relative density parameters and the relative density value, I would like to
review the reliability of some other laboratory soil tests and reasons why
duplication of test values become difficult or impossible.
First, we have three conditions that we must evaluate for comparisons:
(1) how close can one individual duplicate his test values; (2) how close can
several individuals in one laboratory or organization duplicate each other's
test values, and (3) how close can several organizations, utilizing oae or
more individuals, duplicate each other's test values. In an evaluation of
this kind, we must presume (a) that each test is conducted on similar
specimens and materials, (b) that the identical test procedures are used,
(c) that the same equipment is used and that the equipment is properly
maintained and installed in the same manner, and (d) that each operator
has adequate and similar expertise in conducting the tests. All of these
criteria often are not met, and they are difficult to assess and determine.

Specimen Similarity
During the conduct of the American Council of Independent Labora-
tories (ACIL) comparative tests for consistency limits, specific gravity,
gradation, and compaction, one of the most difficult problems was to assure
the likeness of specimens tested by each laboratory. In the papers by
Tiedeman, and Tavenas, Ladd and LaRochelle for this symposium,
variations in the gradation data for supposedly similar specimens of "fine
sand" and "medium sand" are shown. For example, with his wider graded
medium sand, Tiedeman reports varations as high as 23 percent on the basis
of cumulative percent passing the No. 16 screen or 13 percent on this
individual screen size. It can be seen, however, that the variations in
duplicate tests were less than for duplicate specimens, indicating that there
was considerable variation in the specimens. The true effect of such

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Li
HOLTZ ON THE RELATIVE DENSITY APPROACH 11

variations on the maximum and minimum values is not readily determin-


able. In his paper in this symposium, Youd concluded that the effect of
particle size (which would include minor particle size variations) was
negligable with the minimum and maximum densities being controlled
primarily by particle shape particle size range and variances in gradational-
curve shape. Dicken also reported that the effect of particle size on limiting
porosities was small compared with variations caused by particle shape
(paper is included in this symposium).
Generally speaking, it is usually more difficult to produce uniform
samples and uniform specimens of the coarse sand and gravel soils than fine
silty and clayey soils. Also, the difficulty usually becomes greater when the
soils are cohesionless and as the range of particle sizes become greater. This
is primarily caused by segregation. Segregation even can effect the uni-
formity of sampling when closely controlled quartering and splitting
methods are used. The scooping and placing of cohesionless, coarse-grained
soils in molds for the standard minimum and maximum density tests can
cause segregation which can produce nonhomogeniety of gradation within
a test specimen and gradation variations between so-called duplicate
specimens. This is a problem we are faced with in conducting the maximum
and minimum density tests. Emery, Finn and Lee have discussed their
studies and problems in developing uniform sand specimens for shake-table
and laboratory shear tests (included in this symposium).
Identical Test Procedures
Of course, a principal reason for the Committee D-18 interest in this
symposium is to re-evaluate ASTM D 2049-69. The standard was developed
so that more uniform and more meaningful results could be obtained by
individuals making duplicate tests, by individuals in one organization being
able to secure similar results, and for different organizations to secure
reasonably similar results. This is our objective. To achieve this it is very
important that the ASTM procedure be followed very closely until some-
thing better is developed. It will be noted that other procedures were used
in some of the studies reported in this symposium. Such practices make
evaluation and comparison of their work difficult.
Equipment
If we are to secure reasonable uniformity of results between different
laboratories, it is very important that the equipment be the same and that
it be installed and be maintained adequately. It is often surprising to find
that the vibrating table may not be adequately anchored to the floor, that
the cylinders, weights, and measuring devices are not standard, that the
power supply is not adequate, or that vibrating amplitudes are not correct.
Rusted or roughened cylinders may effect the results. The pouring devices
may not be as specified.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agre
12 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

Expertise of Operators
One could hardly expect a novice with no training to produce the same
reliability of results as a trained laboratory engineer or technician who has
had considerable experience. Yet there are instances where practically
untrained personnel are performing these tests. One of our missions should
be to upgrade the quality of soils laboratory work in general. Most of our
laboratory tests can be performed by nonprofessional technicians if they
are properly trained by professional or highly experienced technical
personnel. However, professional guidance always should be available, and
the test results should be reviewed by professional engineers at all times
along with their analysis of the data for the problem at hand. Some of the
test results obtained during the ACIL original program [13], which involved
some 99 laboratories, were so extreme as to be almost ludicrous, if the
problem was not so serious. Here, there was not entirely a question of
specimen dissimilarity, but almost certainly involved were some great
variances of procedures or poorly trained operators who did not have the
slightest comprehension of what they were doing.

Reliability of Compaction Tests


Before we look at our ability to duplicate relative density tests, let us
review the results of the ACIL supplementary program on the standard
compaction tests ASTM D 698-70 and D 1557-70. I summarized the values
which were obtained on the basis of Table 2 of the report on supplemental
testing for the ACIL program.

TABLE 2--Summary of results of the A CIL supplementary program.

Maximum Dry Density, l b / f t 3


D-698 D-1557

All Umpire All Umpire


Labs ~ Labs b Labs Labs

Low plasticity L soil,


avg 105.9 106.0 112.5 111.8
range 15.6 2.2 13.8 1.4
range, -4-% avg 5.1 1.0 6.1 0.6
Medium plasticity M soil,
avg 109.7 109.5 115.8 117.7
range 18.8 1.8 26.6 1.6
range, • avg 8.6 0.8 11.5 0.7
High plasticity H soil,
avg 99.6 98.6 113.3 114.0
range 14.8 2.7 8.15 3.4
range, • v7o avg 7.5 1.4 8.2 1.5

a All laboratories.
b Three well known umpire laboratories.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to
HOLTZ ON THE RELATIVE DENSITYAPPROACH 13

T A B L E 3--Variations from mean, percent.

Soil Standard Avg of Avg,


Deviation Extremes 10 to 90%a
Max 3" Min 3" Max 3" Min 3' Max -y Min ~/

Johnson SW A ...... 4-1.7 4-2.5 4-1.5 4-2.4


(9 USED labs) SP (B) ...... 4-1.4 4-2.6 4-1.4 4-2.1
Tiedeman Fine Sand 1.3 1.4 4-2.1 -4-2.9 4-1.2 4-2.0
(14 USBR Labs) Medium Sand 1.8 1.8 4-3.1 4-3.2 4-2.3 -4-1.9
Tavenas et al. Fine Sand 2.7 1.7 4-6.3 4-3.7 4-2.1 4-2.4
(41 general labs) Gravelly Sand 4.5 2.5 4-8.3 4-5.2 4-3.7 4-2.8

a Average dry density (3') variations from mean for 10 and 90 percent cumulated fre-
quencies.

Briefly, from this data, one can conclude that when the tests are carefully
performed, as by the umpire laboratories, a maximum density 5= 1.5 percent
(or less than 5=1.5 percent) can be achieved. On the other hand, when
considering all participating laboratories the variations were extremely
large, up to as high as 5= 11.5 percent. This would lead one to believe that,
between laboratories, there must be great differences in equipment,
maintenance of equipment, and the type and training of personnel perform-
ing the tests. No consistent differences in these respects were apparent as
regards using ASTM D 698-70 or D 1557-70 methods.
Some later supplemental studies by two different laboratories, each using
three operators and duplicating tests, also showed some wide variations;
however, their ASTM D 698-70 average values were so different from the
other average ACIL values that there is considerable question as to their
techniques or other factors. Their ASTM D 1557-70 average density and
range values were close to those of the umpire laboratories.
In. the paper b y Tavenas, Ladd, and LaRochelle, standard compaction
tests (ASTM D 689-70 and D 1557-70) were made on dry fine sand and dry
gravelly sand by the 41 participating laboratories. These results were
reported in their Table XII. For about 95 percent of the tests, a maximum
density 5=2 percent would cover all types of tests and materials. This is
slightly higher variation than that shown for the umpire tests, but is
exceedingly better than shown for all laboratories participating in the
ACIL tests. I t appears that the 5=2 percent variation is a reasonable
variation to expect with reasonably good testing practices. Others have
expressed the same order of variation.

Reliability of Minimum and Maximum Density Tests


The papers b y Tiedeman, Tavenas, Ladd, and LaRochelle, and M. M.
Johnston (included in this symposium) report on a large number of tests

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to L
14 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

conducted by several laboratories for maximum and minimum densities of


sand and gravel soils. In some cases several operators performed the tests in
one laboratory or duplicate tests or both were made. These data provide an
excellent background on the range of data that can be expected for both the
maximum and minimum tests.
From the data given in these papers the variations obtained in maximum
and minimum densities by ASTM D 2049-69 for dry soils were summarized.
From this data it appears that, within organizations such as the U.S. Corps
of Engineers (USCE) and the U.S. Bureau of Reclamation (USBR), their
field laboratories can produce more uniform results than when a large group
of laboratories are taken at random such as those reported by Tavenas
et al. The results should be in this order, because of more uniform training
of the organizational operators; however, we also must keep in mind the
number of laboratories involved in each of the comparisons, the random
group having a much greater number.
Tiedeman and Tavenas et al also report on the results of duplicate tests.
As would be expected, the variations were much less, which points up the
importance of consistent techniques.
Based on the random results of single tests by a large number of labora-
tories, the results of the cooperative relative density tests indicate less
variability and better uniformity than those obtained during the original
ACIL study of standard compaction tests. However, the standard compac-
tion tests performed by the umpire laboratories showed less variability and
greater uniformity than the relative density tests performed by the several
field laboratories of the USCE and USBR.
Tiedeman concluded in his paper that the variations associated with
minimum and maximum density tests are about the same or less than those
associated with the impact type compaction test. Tavenas et al concluded
that the quality of the results would seem satisfactory (coefficients of
variation in the order of 2.5 percent and coefficients of reproducibility in the
order of 0.8 percent), but that the standard and modified impact compac-
tion tests exhibit a better quality than the maximum density of the relative
density test.
Other Considerations
The two papers by Tiedeman and by Tavenas, Ladd, and LaRochelle
point out how errors or variations in the maximum and minimum density
tests are greatly magnified in the calculation of relative density. Both papers
have illustrations showing how widespread the variations in calculated
relative densities could be. To illustrate this point, the data of Tavenas et al
for the maximum and minimum densities of the fine sand, which fell within
the limits of 10 and 90 percent cumulated frequencies (assumed to be
representative of reasonably good testing), was used to develop Fig. 1. In
1958, Merriman [14] determined that under good conditions and using good

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licens
HOLTZ ON THE RELATIVE DENSITY APPROACH 15

RELATIVE DENSITY - % t
i.-0 tO 20 30 40 , 50 60 ,70 80 99 I00
bU . _ ....... 150

'~~ :, i :, ii i i ! ~ ! ::::::::::::::::::::::::::::::
~3^111111111[ l;i'iiiiiii~II!11iiiiiiiiiiiiilii3 o
u . . . . . . . . .. .. . . . . . ,. ., . ,. ., . . . . , . :. . . . . . . . . . . ~,:,, ,, ,, ~. .~ . ' . ' . ,. .', . I . ',;',
. . ', ', ', ; ~R,
,~

| it i IE; LLI ,i t t t t I rl ~l|l ~ ii


20 .......................... ~20
b I I I I I I I I I I ] I I 1 ] I [

~Assumed in-place density ~, | ! : I I i i I 1 I ll~PT~,, Ir ',J~....~'~-..L.~. ^


10|i i [ ~i , i ~ ~ [ ~ ~ ~ - , , ]~" [., . I I I I I L , , ~ , ~ ,.,..-~-~ = , L-,."~ . ~ L I dtu
'if [ i~,~P'~ I I IJLpI I I ] ~k,,..P~l [ I t i L
l L~k~r i i [ I,Ii~'l ~ r~b~r- I [ F ] I I I I
[ I I I I 1 ~ I I i I % I~,l ff"l I : I~I~'T I I LI ,I I
II111,[,, l I I I l I I I I I L ~
J.~,~"], L . . . . . . . . . . .I .
,i,~l q I I l ~ , b ~ [ I I I
I I I I I I i i i i i I I I~.,~PT i i I I..I T ] I I ] ~ I I ] I
I I I i I I J.,~,~r [ ] I L ~ F I i I IJ.,'f'-I I 1 ~I
] L..ap.ff~l I I L i ~#P"T 1 i

~v~' ~ I I I J. 't~l I I IJ,'~W-I I I I l'VV r "

r'r"[ I I I l.b.l"T I I I L..,I~'VI I I I I I I


^~%l [ LI~I I I I L~I I I l I I l ] >.
9. x / ~ I II I -

1>
< 901 I II
II
II,
I
lI
I
I
If
II
LI
I
III

,,I
I
1190 ~

I II I I
I
I
I
I
III
IFI
II
I
L I'] I l I
I
I [I i I I II
I
I
zI
]1
I I,
I
I
L
i*iI I L
[
H
II 118o
I I
J It I I I II
[I I ] t It
I tl ] ]
I ][ I
I I1 I 'I
I II I
I II I L
I II I I I
T~ II i II I J L
, ,/MINIMUM .AXi.U~"
DENSIT Y p -J
i
~--~ D E N S I T Y
;,SGALE SGALEj
LABORATORY SAMPLE NO.
RELATIVE DENSITY VS DRY DENSITY
HOLE NO DEPTH (Scaled to plot as a straight line)

FIG. 1--Variations in relative density as related to variations in minimum, maximum,


and in-place densities.

sand-pouring or sand-cone procedures, in-place soils densities could be


duplicated within about =t=2 percent. Of course, duplicated data in this
order cannot be approached when poor field testing practices are allowed to
be, used; for example, stomping around a test hole in wet sand. Referring to
Fig. 1, if we assume the average in-place density is 108 lb/ft 3 and the
accuracy =t=2 percent, and the worst conditions of relative density testing

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to L
16 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

and field testing all exist at the same time---lowest maximum and minimum
values and highest in-place values by one operator, as compared with the
highest maximum and minimum values and lowest in-place values by
another operator--the range of relative densities obtained by the two
operators could be as great as from 46 to 91 percent. This represents a
range from a medium dense to very dense condition, or, perhaps, from an
unsuitable density condition to a very satisfactory condition.
The magnitude of variations that can be obtained in relative density
values due to variations in the related parameters is not new. As an
example, this was discussed by Elio D'Appolpnia [15] in 1953 on the basis
of research he had conducted at that time. He decided that the relative
density values could be kept to about 10 percent, if good practices and
suggested criteria were to be followed in setting specification limitations.
We do not anticipate obtaining all of the most unfavorable test situations
at the same time, and the chances are good that this would not hal~pen.
However, in an effort to improve the soil mechanics profession, we should
look towards the best possible means to assure competent and reliable end
results. This is a reason for taking a hard look at our present relative
density practices, as well as our relative density standard, ASTM D 2049-69.
I am sure that you will be confronted with ideas for other approaches to
the problem. These could include the use of a certain percent of t h e
maximum density value as obtained by ASTM D 2049-69 in lieu of the
relative density approach. Other test procedures will undoubtedly be
proposed. We may find that the relative density approach is not the best
from the reliability standpoint. However, we must not reach this decision
hastily. We have much correlative information as to relative density versus
other soil properties, relative density versus competence of sand deposits
and fills, relative density versus blow counts from sounding devices, etc.
Perhaps the changeover would be easy, perhaps not.
One final point to remember is that the studies reported deal with the
variations of individual tests. Decisions of design or construction control
are not made--or should not be made--on the basis of one or two tests.
They are normally made on the basis of an adequate number of tests. For
instance, we do not require a contractor to remove 10 000 yd ~ of compacted
sand fill on the basis of one test. It is usually done on the basis of a dozen
tests or more. If the standard test methods were followed to reasonable
degrees by an operator with a reasonable amount of correct training, the
test results should be reasonably close, and the odd-ball determinations
would be readily apparent. A wide scattering would indicate poor variable
techniques or poor equipment.
I think we could do a great deal to answer some of our reliability and
reproducibility problems by making greater efforts to improve soil testing
practices, and to more closely follow the present ASTM methods.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lic
HOLTZ ON THE RELATIVEDENSITY APPROACH 17

References
[I] Holtz, W. G., in Papers on Soils, A S T M S T P 206, American Society for Testing and
Materials, 1957, pp. 50-66.
[2] Merriman, J., "Research Tests to Investigate Criteria for Selection Between Vibra~
tory or Impact Compaction Methods," Laboratory Report No. EM-441, U. S.
Bureau of Reclamation, 1955.
[3] 1972 Annual Book of A S T M Standards, Part 11, pp. 779-784.
[~] Meissner, V. S., "Vibroflotation Experiments at Enders Dam," Laboratory Report
No. 178, U.S. Bureau of Relcamation, 1948.
[5] Burmister, D. M., Proceedings, American Society for Testing and Materials, Vol. 48,
1948, p. 1249.
[6] Gibbs, H. J. and Merriman, J., "Second Progress Report of Research on Determin-
ing the Relative Density of Sands by Spoon Penetration Testing," Laboratory Re-
port No. EM-356, U. S. Bureau of Reclamation, 1953.
[7] Lamle, T. W. and Whitman, R. V., Soil Mechanics, Wiley, New York, 1969, p. 31.
[8] Meyerhoff, G. G., Journal of the Soil Mechanics and Foundations Division, American
Society of Civil Engineers, Jan. 1956, p. 17.
[9] Hough, B. K., Basic Soils Engineering, Ronald Press Co., New York, 1957, p. 357.
[10] Tschebotarioff, G. P., Soil Mechanics Foundations and Earth Structures, McGraw-
Hill Book Publishing Co., New York, 1951, p. 57.
[11] Plummer, F. L. and Dore, S. M., Soil Mechanics and Foundations, Pitman Publish-
ing Corp., New York, 1940, p. 33.
[12] Felt, E. J. in Symposium on Application of Soil Testing in Highway Design and Con-
struetion, A S T M S T P 239, American Society for Testing and Materials, 1958, pp.
89-110.
[13] Johnson, A. W. and Guinnee, J. W., "Report on a Supplemental Testing Program
for ACIL Standard Reference Soil Samples," for the Research Steering Committee
of ASTM Committee D-18.
[14] Merriman, J., "Laboratory Evaluation of Volumeter No. 770 for Field Density Test
Determinations," Laboratory Report REM-2, U. S. Bureau of Reclamation, June
1958.
[15] D'Appolonia, Elio in Symposium on Dynamic Te~ting of Soils, A S T M S T P 156,
American Society for Testing and Materials, 1953, pp. 138-154.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
F. A. Tavenas, 1 R. S. L a d d / and P. La l~ocheUe~

Accuracy of Relative Density Measurements:


Results of a Comparative Test Program

R E F E R E N C E : Tavenas, F. A., Ladd, R. S., and La Rochelle, P., " A c c u r a c y


of Relative Density Measurements: Results of a Comparative Test
P r o g r a m , " Evaluation of Relative Density and Its Role in Geotechnical Projects
Involving Cohesionless Soils, ASTM ST P 5~3, American Society for Testing and
Materials, 1973, pp. 18-60.

ABSTRACT: A cooperative testing program was conducted, under the sponsor-


ship of the American Society for Testing and Materials (ASTM), by 41 soil
laboratories in the United States and Canada to determine variations associated
with gradation tests, minimum and maximum density tests used to define the
relative density, and Proctor compaction tests on identical~specimens of a fine
sand and a gravelly sand. D a t a were analyzed statistically on the basis of
variations among laboratories and between duplicate tests. Results indicated
that: (1) all tests were affected by a large variability and a low reproducibility,
(2) variations among laboratories were two to three times greater than vari-
ations between duplicate tests, (3) resulting variations on the relative density
were such that this parameter cannot be used efficiently in practice, and
(4) relative compaction, even though not very reliable, is a better parameter
for expressing the density of a soil.
KEY WORDS: soil tests, density (mass/volume), analysis of variance,
cohesionless soils, soil compacting, tests, statistical analysis, variability

In the last twenty years the concept of relative density has been used
more and more often in the investigation of the properties of cohesionless
soils. As a matter of fact, the relative density has become one of the basic
parameters of these materials. It is almost systematically used as a reference
parameter in laboratory investigations on the mechanical behavior of sands,
and compaction specifications are increasingly given in terms of a minimum
relative density required. Also, the evaluation of the liquefaction potential
of sand deposits submitted to earthquakes is based on the relative density.

1 Associate professor and professor, respectively, Civil Engineering Department, Laval


University, Quebec, Canada.
Woodward Moorhouse & Associates Inc., Clifton, N. J. 07012.

18

Copyright by ASTM Int'l


Copyright9 by(all
ASTMrightsInternational
reserved); Fri Mar 11 16:13:06 EST 2016
www.astm.org
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
TAVENAS ET AL ON RESULTS OF A COMPARATIVE TEST PROGRAM 19

Finally, the relative density is the soils parameter most commonly measured
in situ, and, therefore governs the great majority of designs involving
cohesionless soils.
Along with this increasing use of the relative density, the requirements
for its accurate determination have necessarily become more stringent. For
example, in the analysis of the liquefaction potential of a natural, or
recompacted deposit, the measured value of the relative density has to be
compared to a critical value to give a "yes or no" type of answer. Such
procedure bears, naturally, on the premise that the in situ relative density
is determined with an accuracy sufficient to ensure a satisfactory reliability
of the answer.
The accuracy of relative density measurements has, for the most part,
been taken for granted without thorough analysis. Indeed the methods of
measuring the maximum and minimum densities have been investigated by
different authors (Kolbuszewski [1],3 Felt [2], Pettibone and Hardin [3]),
but the purpose of these investigations was to compare the relative merits
of different methods and to make a selection which would finally lead to the
proposal of a standard such as ASTM Test for Relative Density of
Cohesionless Soils (D 2049-69). Even though these investigations showed
the great sensitivity of the maximum and minimum densities not only to
the testing method but also to the operator, it was not until very recent
years that the question of the accuracy of relative density measurements
was seriously raised.
The relative density, Dr, is usually defined as:

Dr = Ydma____~X "~d--'Y~mi.
*~d '~d m a x -- "~d r a i n

Due to the relative magnitude of the maximum (Td max), the minimum
(~,d mi,), and the actual dry unit weight (~,d), the relative density is
computed from the ratio of small differences between large numbers. This
implies that small variations of the large numbers will be magnified to
produce a great variability in the computed result. The simple application
of the theory of errors led Tavenas and La Rochelle [4i]to conclude that any
laboratory determination of Dr would be affected by a large variability
(hDr mi. = 6 percent), even if the ASTM D 2049-69 standard method
were used. These values of hDr would be increased by about 10 percent in
the case of the in situ measurement of the relative density. With such
characteristics the determination of any satisfactory value of Dr alJpeared
problematic. The results of a comparative test program reported by
Tiedemann [5] confirmed these conclusions. In this program, 15 U.S.
Bureau of Reclamation (USBR) laboratories performed the ASTM
D 2049-69 standard tests on identical specimens of two sandy materials. All

a The italic numbers in brackets refer to the list of references appended to this paper.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lice
20 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

laboratories worked with the same type of equipment and uniformly trained
operators. The variability of the results was approximately the same as
found for impact-type compaction tests: the standard deviations were of the
order of ~1.6 lb/ft 3 for the minimum density and • lb/ft 3 for the
maximum density. However, the resulting width of the 95 percent interval
for the relative density was 37 percent if no variation of the actual dry unit
weight was considered. In terms of reproducibility, the standard deviations
were respectively lower at =t=1 lb/ft 3 leading to a width of the 95 percent
interval for Dr of 21 percent.
While these two investigations indicated clearly that the accuracy of
relative density measurement is limited and may have a determining
influence on the applicability of this soil parameter, it appeared necessary
to check the validity of these conclusions on a much broader basis. For this
purpose it was suggested by the senior author and accepted by the American
Society for Testing and Materials (ASTM) Committee D-18 on Soil and
Rock for Engineering Purposes, Subcommittee D18.09 on Dynamic
Properties of Soils, and the Organizing Committee of the 1972 ASTM
Symposium, to perform a large comparative test program. The present
paper describes the scope and organization of this test program and presents
the results obtained and their analysis.

Scope of the Investigation


In order to ensure the broadest possible basis for the comparative test
program, 87 United States and Canadian laboratories, equally distributed
in government agencies, industry, and universities were invited to partici-
pate. The proposed plan was to send to all participants identical specimens
of two cohesionless soils, a uniform fine sand, and a well graded gravelly
sand, and to have the following tests performed: gradation test, minimum
and maximum density tests according to ASTM D 2049-69 or the partici-
pants own method or both, and two one-point compaction tests with
standard and modified Proctor compaction procedures. Thus, with the
results obtained on perfectly identical specimens by a representative
selection of all laboratories, a valuable analysis of the variability and
reproducibility of these testing methods would be possible.
Fourty three laboratories accepted the invitation and were sent the
specimens in October 1971. Fourty-one laboratories, as listed in Appendix I,
completed the program.
Materials Tested
The materials were selected so as to complete and enlarge the results of
the investigation reported by Tiedemann [5].
Fine Sand--The fine sand used was the material designated as 24E-11 in
the investigation by Tiedemann [5]. The material was prepared at the
USBR Laboratory in Denver. All specimens were individually composited

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licens
TAVENAS IET AL ON RESULTS OF A COMPARATIVE TEST PROGRAM 21

pe+iC~ cliam {~) I sieve n o i s,eve climens,on in inch I

200 I00 60 40 20 10 ~4 I 3
/
9(2

8C

70
/
6O
/
/
~o
il ! ~ i /
4O

3(3
/2
IC

0
I TIIIl .......i,,,,I, ,Ithl,,h
006 o+ 02 06 ~ 2 6 I 2 I0 20 60 I00

[ s,i [ ~o~ g..... I

FIG. 1--Grains size disiributions of the samples.

from screened fractions of a local stream deposit. The screening was done in
an aggregate processing plant which separates sands into Nos. 4, 8, 16, 30,
50, and 100 sizes. The material passing the 100 sieve was washed on a 200
sieve. The individual sizes were then combined to match the mean grain
size distribution of the soil 24E-11 as reported by Tiedemann [5]. Figure 1
shows the actual grain size distribution of the fine sand specimen.
The total weight of the specimen was 50 lb; the weighing of the individual
sizes was made with a fan-type balance with a 30 lb capacity, reading
directly to 0.01 lb. For shipment the 30 and 50 sizes were individually
bagged as sacks 1 and 2, respectively, with total weights of 8.60 and
12.62 lb. These bags and the remaining of the specimens were double-sacked
and shipped.
Gravelly Sand--The material was prepared at Lava] University, Quebec,
Canada. All samples were individually composited from screened fractions
of a local glacial deposit. After a first mechanical screening, the final
processing was done by hand to separate into 1, 3/~, and 3/~ in., and Nos. 4,
10, 20, 40, 60, 100, and 200 sizes. The material passing the 100 sieve was
washed on a 200 sieve to remove all fines. The individual sizes were then
combined to match the selected grain size distribution shown on Fig. 1. The
weighing was done on a 20-kg solution balance reading directly to i g. The
50olb specimens were single-sacked and put in a wooden box for shipment.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to L
22 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

Test Procedure
The test sequence and procedures were specified so as to have untested
material used whenever possible and to ensure uniformity between the
participants. The instructions given to the participants are presented in
Appendix II.
Data Analysis
After all the data had been received on 29 February 1972, the results
were rechecked and each participant was arbitrarily assigned an identifica-
tion number.
The results were analyzed statistically with respect to the variations
between laboratories and, whenever possible, to the variations between
duplicate tests within laboratories.
Variations between laboratories--The following statistical characteristics
were computed:
1. The mean, 4, was calculated as ~ = ~ x / n where ~-'~xis the summa-
tion of n individual values of x.
2. The range, R, is the difference between the highest and lowest values
o f X.
3. The cumulative distribution function gives the distribution of the
observations in selected intervals within the range.
4. The standard deviation, S, was calculated as:
ix- 2) 2
n--1
5. The coefficient of variation is defined as the ratio of the standard
deviation to the mean. It is also referred to as the coefficient of variability,
as opposed to the coefficient of reproducibility.
Variations Between Duplicate Tests--The minimum, maximum, and
Proctor densities were measured in two duplicate tests by each participant.
The differences between duplicate tests were compared by computing the
following parameters:
1. The average R was calculated as:

R=Eldl
k
where ~ [ d I is the summation of the absolute differences between dupli-
cate tests and k is the number of pairs of duplicate tests.
2. The combined standard deviation, S', was calculated as:
S' = , / - 2 d 2
2K

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursu
TAVENAS ET AL ON RESULTS OF A COMPARATIVE TEST PROGRAM 23

3. The coefficient of reproducibility is defined as the ratio of the com-


bined standard deviation to the mean as determined above.
Composite Results--From the test results on minimum, maximum, and
Proctor densities, the relative density, the relative compaction based on the
maximum density, and the relative compaction based on the standaYd and
modified Proctor densities were computed assuming an actual dry unit
weight of 108 lb/ft 3 for the fine sand and 122 lb/ft 3 for the gravelly sand.
The composite results were analysed only with respect to the variations
between laboratories.

Scale Accuracy
The two small sacks of the fine sand were included to obtain an indication
of the accuracy of the scales used by the participants. The following results
were obtained:

Sack Number Number of Mean Weight, Standard Deviation,


Observations lb (g) lb (g)

1 37 8.59 0.05
(3896) (24)
2 37 12.61 0.05
(5720) (24)

These variations are considered to be small and any effects they might
have on the density determinations would be minimal.

Gradation Tests
Fine Sand
Participants were asked to perform gradation tests on 100-g specimens
taken by quartering or splitting from: (a) the unused material (Specimen
4-S) (b) the material after relative density tests (Specimen 1-S), and
(c) the material after Proctor tests (Specimen 3-SB). The material had to
be sieved for 15 min on U.S. Standard sieves 10, 20, 40, 100, and 200. If
possible a powered sieve shaker was to be used.
The results of the gradation tests are presented in Table 1. Figure 2
shows the mean gradation curves and the limits of plus and minus two
standard deviations from the mean which contain about 95 percent of the
observed values.
Tests on Specimen AS--The tests results on the unused material allow for
an evaluation of the accuracy of the gradation test. The mean gradation
curve on Fig. 2 is in good concordance with the actual grain size distribu-
tion. However, it appears that this is not due to the good quality of the test
results but essentially to their quantity. As a matter of fact, the variability

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to
24 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

T A B L E 1--Statistical analysis of the gradation test on the fine sand specimen.

Statistics Percentage Passing Sieve Size


10 20 40 60 100 200

Specimen 4S
Mean 94.7 79.9 56.4 34.8 12.4 2.5
Standard deviation 0.9 5.2 5.0 4.1 3.2 1.4
Coefficient of variation 1.0 6.5 8.9 11.7 25.5 54.5
Maximum 96.7 84.4 66.1 43.8 20.4 8.7
Minimum 93.0 63.2 41.7 22.6 4.5 0.2
Range 3.7 23.2 24.4 21.2 15.9 8.5
Specimen 1S
Mean 94.7 79.7 57.7 35.8 12.4 2.5
Standard deviation 1.0 4.4 4.9 4.6 2.6 O. 9
Coefficient of variation 1.1 5.5 8.5 12.7 21.3 35.7
Maximum 96.8 86.5 66.6 46.6 17.7 4.7
Minimum 92.0 60.7 44.3 25.0 5.4 0.4
Range 4.8 25.8 22.3 21.6 12.3 4.3
Specimen 3SB
Mean 94.5 79.9 57.7 37.2 14.7 3.8
Standard deviation 1.1 3.4 5.3 4.4 3.2 1.5
Coefficient of variation 1.2 4.3 9.2 11.8 21.7 40.0
Maximum 96.1 84.8 65.4 44.7 20.4 8.7
Minimum 90.5 68.6 41.0 21.5 5.9 0.3
Range 6.6 16.2 24.4 23.2 14.5 8.4

L p.,,c, diom (ram) { sieve no 1 ~le'~. c i l m . ~ ' , in inch 1

3,~ 3

s
i i

; L l i l 1 /,/ZI// , !
,o Ji! ~ i ~4~ ' ' ~ --

~o !~"I i i i,i
I,?/-

,o'~-~ ~-~-,'/ 1 ~_ F
I
, ',,i ,i~ . . . . . . . . . i ~ , , l , , , , J , I,i,IJ,hl ......... I .... I , I,l,l,hhhl ............. I , I,l~l~l,hl,t ............ I ,Jl,l,hl,lhh
006 Ol 02 06 I 6 I 2 6 I0 20 60 I00

FIG. 2--Results of gradation tests, fine sand sample 4S.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agree
TAVENAS ET AL ON RESULTSOF A COMPARATIVE TEST PROGRAM 25

of the reported results is large: the standard deviations are greater than
4 percent, and the ranges greater than 20 percent on the percentage passing
the sieves 20, 40, and 60 which retain about 60 percent of the material. As
noted by Tiedemann [5] the magnitude of the variations is related to the
percentage of material retained on the sieve rather than to its size.
Effect of the Compaction Tests--As shown on Table 1 none of the statistical
parameters computed on Specimen 1S after relative density tests or on
Specimen 3SB after Proctor compaction tests, are significantly different
from those observed on the unused Specimen 4S. Under such circumstances
and taking into account the great variability observed on the results on
Specimen 1S, it appears logical to neglect the small variations on the mean
gradation curve and to conclude t h a t the compaction tests had no influence
on the grain size distribution of the material.

Gravelly Sand
Gradation tests were performed on: (a) the unused material (Specimen
4G), (b) the material after compaction tests (Specimen 1G) (Due to the
testing sequence adopted by most of the participants, it was impossible to
consider the influence of the relative density tests on the grain size distribu-

TABLE 2--Statistical analysis of the gradation test on the gravelly sand specimen.

Statistics Percentage Passing Sieve Size


in. ~ in. 4 10 20 40 60 100 200

Specimen 4G
Mean 90.0 70.8 55.7 40.7 28.1 17.0 11.3 6.2 0.9
Standard deviation 4.2 5.8 5.7 6.0 5.5 4.1 3.7 2.0 0.5
Coefficient of variation 4.6 8.2 10.2 14.7 19.7 24.3 32.4 32.8 53.0
Maximum 98.3 83.5 70.0 52.9 40.5 27.5 24.1 11.3 2.1
Minimum 78.9 53.7 37.5 24.6 15.2 8.7 5.0 2.6 0.1
Range 19.4 29.8 32.5 28.3 25.3 18.8 19.1 8.7 2.0
Specimen 1G
Mean 90.5 72.0 57.8 42.9 29.7 19.0 12.4 6.7 1.3
Standard deviation 3.2 3.6 3.6 4.1 4.4 3.6 2.4 1.4 0.8
Coefficient of variation 3.6 5.0 6.2 9.5 14.7 19.0 19.1 20.0 58.0
Maximum 94.4 78.6 65.6 50.2 38.9 27.9 17.4 9.6 3.0
Minimum 73.5 57.9 47.5 33.6 19.1 11.4 7.3 3.8 0.0
Range 20.9 20.7 18.1 16.6 19.8 16.5 10.1 5.8 3.0
Specimen 3G
Mean 91.8 73.4 60.4 46.2 32.5 20.7 13.9 7.8 2.0
Standard deviation 1.9 4.6 4.4 4.8 5.1 4.1 3.0 2.2 1.1
Coefficient of variation 2.1 6.2 7.3 10.4 15.7 19.8 21.7 27.9 55.9
Maximum 95.6 82.5 69.3 55.9 42.7 28.9 19.9 12.2 4.3
Minimum 88.1 64.3 51.5 36.6 22.3 12.5 7.9 3.5 0.0
Range 7.5 18.2 17.8 19.3 20.4 16.4 12.0 8.7 4.3

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Li
26 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

tion of the specimens separately), and (c) the material after standard and
modified Proctor compaction tests only (Specimen 3G) (in 30 cases it was
possible to isolate the effect of the Proctor compaction tests).
The entire specimen, weighing 1500 g for Specimen 4G and about 7000 g
for Specimens 1G and 3G, was first sieved on the sieves i in., 3/~ in., a/~ in.,
No. 4, and pan. One hundred grams of the material passing No. 4 was then
sieved for 15 rain on U.S. Standard sieves 4, 10, 20, 40, 60, 100, and 200.
The results of the gradation tests are presented in Table 2. Figures 3 and 4
show the mean gradation curves and the limits of plus and minus two
standard deviations.
Tests on Specimen 4G--The agreement between the observed mean
gradation curve and the actual grain size distribution is not as good as for
the fine sand, and the deviations between the two curves are observed on
sieves 10, 20, 40, and 60, amount to 5 percent. This is possibly due to the
testing technique and particularly to the selection of 100 g in the fraction
passing the No. 4 sieve after the first sieving.
The variability of the results is large with standard deviations greater
than 5.5 percent on the sieves a/~ in., 4, 10, and 20 on which 60 percent of the
material is retained and ranges up to 32.5 percent on sieve No. 4.
Effect of Compaction Tests--As in the case of the fine sand it is difficult to
draw any conclusion as to the influence of the relative density tests on the
gradation of the material since the variations of the mean curves are very
l pacticle diarn (ram) l sieve n o i s~e~ d i ~ o n in inch I

% 3/4 t~

/////~//1 I

// / /

,ol !,1 ,/

x m~n ~urve
~ l , i ': I ) X p l u s ond minus two stondocd
devlotions from meon voiues
/ __

2o f,', r 'I ,,/-" f//


t - ~//

..... J. . . . . . . . . . . . . . . . h,.I,l,l,hl,hl ......... i,,,,I , 1,1,],],],1,1 ........ hi ~Ll'l ]' ' ' , ~ n ~ , l l]l l l , I , I [,I,I,
006 oI 02 06 I 2 6 I 6 I0 20 60 IOO

coo~se fine r flr~ ~ mc'{h u m i coo,se


FIG. 3--Results of gradation tests, gravelly sand sample 4G.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
TAVENAS ET AL ON RESULTSOF A COMPARATIVETEST PROGRAM 27

! poeficle ~iom. (ram) I sie,R no. I ,~eve dimensK)n in ,nch ]

I00

90

80
//
70

6,O

4C

2~

IG

0
006 OI 02 06 I 2 .6 2 6 I0 20 60 I00

medium coorse fine echum medium

FIG. &--Results of gradation tests, gravelly sand sample 1G.

small as compared to the standard deviations. Any comparison is made


even more difficult by the fact that the size of Specimen 1G was four times
larger than Specimen 4G. The quantity of material tested had an important
influence on the variability of the results. For the larger Specimen 1G the
standard deviation averaged 4 percent and the range 18.5 percent, as
compared to 5.7 percent and 30 percent for the smaller Specimen 4G.
More definite conclusions can be drawn as to the effect of the Proctor
compaction tests. The analysis of the results on those Specimens 3G which
were submitted to Proctor compaction tests only, shows that a certain
amount of particle breakage has occurred, leading to an increase of 2 to 5
percent of the mean percentage passing sieves 4 to 60. Since no such
increase was noted on Specimen 1G after relative density and one standard
Proctor compaction test, it can be concluded that this particle breakage
occurred essentially in the modified Proctor test performed on these
specimens. The degradation during one test is however limited.
Disc~on
Against the well accepted idea that the gradation test on cohesionless
soils with zero percent passing sieve 200, is a simple and, therefore, accurate
test, the results reported above show that the variability of such test
results is very important. This variability can be related to various factors.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License A
28 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

1. The size of the sample. The results on Specimens 4G and 1G show the
definite reduction of the variability of the results with an increasing weight
of the specimen. This fact would advocate for the largest possible speci-
mens, and it has been taken somewhat into account in ASTM Particle-Size
Analysis of Soils (D 422-63). However, the results on Specimen 1G have a
large variability even though the weight of tested specimen was much
larger than required in ASTM D 422-63.
2. The weight of material retained on each sieve. As previously men-
tioned the standard deviations and ranges increased with an increasing
weight of soil retained on the sieve. To eliminate this source of error the
size of the specimen, particularly in the case of poorly graded materials,
should be kept to a minimum. Since this requirement is in complete
opposition to the preceding one, it appears that the variability of any
gradation tests will be very large, and of the same order as reported herein.
3. The method of selection of the specimen. Quartering or splitting
procedures are difficult and have an increased influence on the results as the
weight of the specimen decreases and the maximum grain size increases.

M i n i m u m and M a x i m u m Density Tests


The main purpose of this part of the comparative test program was to
investigate the variability and reproducibility of the ASTM D 2049-69
standard method for determining the minimum and maximum density of
cohesionless soils. In addition, the participants were also allowed to use
their own testing procedures, thus, the variability of the limiting densities
with the testing technique could also be analyzed. In each case the
maximum and minimum densities were to be measured on two specimens
of each material so that the reproducibility of the results could be evaluated.
All results are summarized in Figs. 5 and 6.

Variability of A S T M D 2049-69 Test Results


The ASTM standard procedure was used by 75 percent of the partici-
pants. The number of reported test results is about 60, tl~at is, large enough
to be statistically analysed.
Minimum Density--The minimum density was determined by 62 tests
for the fine sand and 63 tests for the gravelly sand. The statistics are shown
on Table 3. The computed standard deviations are 1.7 ]b/ft 3 (0.03 tonne
per cubic meter (t/m 3) ) for the fine sand and 2.5 lb/ft ~ (0.04 t / m a) for the
gravelly sand. These values are larger than those reported by Tiedemann [5],
but would normally be considered as acceptable since the coefficients of
variation are of the order of 2 percent, namely, much smaller as those
usually observed on other soil mechanics tests [6].
The ranges are 7.1 lb/ft 3 (0.11 t / m 8) and 11.8 lb/ft 3 (0.19 t/m3), for the
fine sand and gravelly sand, respectively. These values are slightly larger
than four times the standard deviations. The cumulated frequency

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licen
}24 I I I I I [

o i i ;
m [
120 I mean + S
~w ! o I
m i t
$ 8 o , o I
o i
e!
116 I r
9 o 8
9 i
| '
' mean Yd max
i <
9 i
i : ! E
I*
9 i
)

7 --- i " mean- S


b i _ o
~ ~08
) ! ,. Z
W=wef method J@
~w i 9 i
I i m
r
:~ 104 o Xmax part -I
cf~
9 7rnax ASTM
v ? o )~m~n port
I i i 9 Xmm ASTM
O0
T
o
-e :1~ ~i mean + S
)6
: : ~ 1 7-6 - , . ~I "; :: 9. I 9 9
! ~ , !I mean 7 d ~ i n
i 8,~ iL 9
I 9 I
----~-- -- -- --'~ -- f:--o-- 1- --'~ mean - S -I
92 |
B I ; a I : m
-,I
I t
i o i
8
B8 I L I I i I I o
4 20 24 28 32 36 40 44 48 52 ~
identification no of parficipont

FIG. 5---Minimum and maximum densities of fine sand. ,o

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
144 , i 9 : e i ~ i i t l I[ I l l t I I I r I r I ' I I r r I I i ~ E I ~ I r
0

i
0(3
~40

9 ~ mean § S
0
t 9 i
n56 | 8
o
z
(~ --4
a
9
9 i mean Yd max
152
9 <
9 I 9 $'9 O
0 P
<
9
D_ 128 +
9 I 0
--O ~meon ~ S
---I
I W : wet method
) O
-r

~ 124 9 o )'max port


m
! 9 )'max ASTM O
u )'mln oart z
A I
: ~ 9 )'rain ASTM F.
120
9 !
I
l o i o_
D i
H6

112
" ' ~'-; . . . . 4~oo~s ! q
' . "~ I F I
108 ~ i L I i I I L ~ I I I I I I I. I I I I I I I I I i i I
12 16 20 24 28 32 56 40 44 48 52 56
qdentificohon no of peHictponi

FIG. 6--Minimum and maximum densities of gravelly sand.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
TAVENAS ET AL ON RESULTS OF A COMPARATIVE TEST PROGRAM 31

T A B L E 3--Variability of A S T M D 20~9-69 test results.

Statistic M i n i m u m Density, M a x i m u m Density,


lb/ft a t/m 3 lb/ft 3 t/m 3

Fine Sand
N u m b e r of tests 62 58

Mean 95.3 1,53 114.9 1.84


Standard deviation 1.7 0.03 2.7 0.04
Maximum 98.7 1.58 120.8 1.93
Minimum 91.6 1.47 106.4 1.70
Range 7.1 0.11 14.4 0.23

Gravelly Sand
N u m b e r of tests 63 59

Mean 114.2 1,83 133.8 2.14


S t a n d a r d deviation 2.5 0.04 4.5 0.07
Maximum 120.6 1.93 140.8 2.26
Minimum 108.8 1.74 118,5 1.90
Range 11.8 0.19 22.3 0.36

I00

90

80

70

~. 6O

50

40

30

20

0
88 90 92 94 96 98 100 102
Jry densiCy, Ib/cuf~

FIG. 7--Frequency distribution curves, minimum density of fine sand (ASTM D ~049-69).

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreeme
32 RELATIVE DENSITY I N V O L V I N G COHESIONLESS SOILS

I00
/
r.i
number of
ob~lrvofiO~$= 65
/
90 meon= 1142
S=25

80

~80L t
i s
$
40 ~ [ I
/
it m~/-meon
20 I
I.
[
I
0
I
I08 I10 112 ( 14 116 II~ 120 122
dry density,lb/cuf t
FIG. 8--Frequency distribution curves, minimum density of gravelly sand (ASTM D
~0~9-e9).

distribution curves, Figs. 7 and 8, show that the observations are normally
distributed within the range and are nearly symmetrical about the mean.
Maximum Density--The dry method maximum density was determined
by 58 tests for the fine sand and 59 tests for the gravelly sand. As shown in
Table 3 the standard deviations are 2.7 lb/ft 8 (0.04 t / m 3) and 4.5 lb/ft s
(0.07 t/m3), respectively, corresponding to coefficients of variation of 2.3
percent and 3.4 percent, that is, 60 percent larger than those observed on
the minimum density. Within ranges of 14.4 lb/ft 3 (0.23 t / m 3) and 22.3
lb/ft s (0.36 t/mS), the observations, as shown by the frequency distribution
curves in Figs. 9 and 10, are not as normally distributed as in the case of the
minimum densities.
Reproducibility of A S T M D ~0~9-69 Tests Results
As mentioned before, each participant performed duplicate tests on each
material. The analysis of the differences between duplicate tests gives an
indication of the reproducibility of the considered tests. The results of this
analysis are given in Table 4.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
TAVENAS ET AL ON RESULTS OF A COMPARATIVE TEST PROGRAM 33

T A B L E ~--Reproducibility of A S T M D 20~9-69 test results.

Statistic M i n i m u m Density, M a x i m u m Density,


lb/ft 8 t/m a lb/ft 3 t/m 3

Fine Sand
N u m b e r of tests 31 29

Average 0.56 0. 009 0.59 0.009


Range
Combined 0.54 0. 009 0.67 0.011
standard deviation
Gravelly Sand
N u m b e r of tests 32 31

Average 1.10 0.018 1.45 0. 023


Range
Combined 1.05 0. 017 1.37 0. 022
standard deviation

I00

90

80

70

~60

i
so

30

20

0
I00 104 ~08 112 116 120 124 128
dry density, i%
u L tL

FIG. 9--Frequency distribution curves, maximum density of fine sand (ASTM D ~0~9--69)

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further
34 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

I00
~,, ~ I II -/
90 meOn=1338 /

80 S= 4,5 t

70 I -
~ 6O
' /
I
o2 -S S ~!
40 /-
/ ~.mean
I
!
ZO
/

0 ,-~-- I I
dry density, Ib/cuft

FIG. lO--Frequency distribution curves, maximum density of gravelly 8and (ASTM D


eo49-69).

Minimum Density--The combined standard deviations are 0.54 lb/ft 3


(0.009 t / m s) and 1.05 lb/ft 8 (0.017 t / m s) for the fine sand and gravelly
sand, respectively. They are significantly larger than those reported by
Tiedemann [5]. This can be explained by the fact that the group of operators
participating in the USBR investigation was more homogeneous than in
the present case. Thus, the present data are not indicative of the repro~
ducibility of the tests itself, but are also influenced in some way by a
variability between laboratories; the reproducibility observed here is an
average reproducibility. The variations in the minimum densities between
duplicate tests are about one third of the variations between laboratories.
Maximum Density--As for the variations between laboratories, the
combined standard deviation for the maximum densities are larger than for
the minimum density, with values of 0.67 lb/ft a (0.011 t / m s) for the fine
sand and 1.37 lb/ft 3 (0.022 t / m s) for the gravelly sand. They represent only
30 percent of the standard deviations of the results obtained by different
laboratories.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License
TAVENAS El" AL ON RESULTS OF A COMPARATIVE TEST PROGRAM 35

Variability Between Different Test Methods


Seventeen participants used test methods different from the A S T M
D 2049-69 standard. A complete description of these 17 methods cannot be
given here. The main differences between these methods and the standard
are the following: for the minimum density test, the use of a different
funnel (4) 4, a s c o o p (5), a different mold (6), and the application of
Kolbuzeswski [I] method (1); for the maximum density, the use of a
different mold (4) and a different vibrating table (2), an additional
vibrator attached to the mold (1), vibration b y layers procedure (1),
Proctor type compaction (5), or compaction in layers with a vibrating
hammer (4). The results of the statistical analysis of the observations
reported b y the different participants are presented in Table 5.
Minimum Density--The statistics are very close to those computed for
the A S T M standard tests. T h e mean is slightly smaller for the fine sand,
at 94.3 l b / f t a (1.51 t / m a) instead of 95.3 l b / f t 3 (1.53 t / m a) b u t is exactly the
same at 114.2 l b / f t a (1.83 t / m 3) for the gravelly sand. The standard
deviations and ranges are also very similar.
This similarity is related to the minor differences in the testing techniques.
T h e use of a different funnel or of a scoop for placing the material, or of a
larger or smaller mold did not seem to have a real influence on the result.

TABLE 5---Variability of minimum and maximum densities measured by differentmethods.

Statistic Minimum Density, Maximum Density,


lb/ft ~ t/m 3 lb/ft 3 t/m 3

Fine Sand
Number of tests 32 38

Mean 94.3 1.510 114.3 1.83


Standard deviation 2.1 0.034 7.9 0.13
Maximum 97.2 1.560 138.8 2.22
Minimum 88.2 1.410 102.2 1.64
Range 9.0 0.154 36.6 0.58

Gravelly Sand
Number of tests 25 27

Mean 114.2 1. 830 130.3 2.09


Standard deviation 2.6 0.042 6.9 0.11
Maximum 118.1 1.890 140.5 2.25
Minimum 109.1 1.740 114.2 1.83
Range 9.0 0.150 26.3 0.42

Numbers in parenthesis refer to the number of participants using a given method.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
36 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

Maximum Density--On the contrary, the maximum density is seriously


influenced by the testing technique. However, this influence is not as
evident in the mean values, which happen to be less than in the ASTM
tests, as in the standard deviations and ranges which are about twice as
large as in the ASTM tests. Such important differences are due to the large
differences between the different test procedures reported. Incidently, it
appears that the maximum density for the fine sand as obtained by
compaction with a vibratory hammer is higher than that obtained by the
ASTM standard method, 116.4 lb/ft 3 (1.86 t / m 3) as compared to 114.9
lb/ft 3 (1.84 t/m3). However, for the gravelly sand, no difference between
the two methods is observed.

Discussion
Before considering the use of the minimum and maximum densities to
compute the relative density, some remarks concerning the measurements
of these parameters ought to be made.
Testing Method--The most common reason why some laboratories use
their own method instead of the standard ASTM procedure is that they
think they can get a lower minimum and a higher maximum density. The
results reported above show that this is not the case, at least on the average.
Furthermore, due to the much larger variability of the results obtained
from nonstandard methods, a comparison between tests results from
different laboratories will be very difficult, if at all possible. Therefore, the
generalized use of the ASTM D 2049-69 standard procedure can only be
encouraged, at least in this respect.
Influence of the Tested Material--It is evident from the test results that
the material has a major influence on the quality of the result in terms of
reproducibility as well as of variability. The trend already mentioned by
Tiedemann [5] is confirmed here: the coarser the material tested, the larger
the variability. The standard deviations (Table 3) for the gravelly sand are
60 percent larger than for the fine sand, and the combined standard
deviation (Table 4) are 100 percent larger. This fact should be taken into
account, either by increasing the number of individual tests as the maxi-
mum particle size of the material increases or by simply being more careful
when dealing with the relative density of gravelly materials.
Quality of the Test Results--The standard ASTM method for determining
the maximum and minimum density of cohesionless materials yields results
characterized by an apparently acceptable variability (coefficient of
variation of the order of 2.5 percent), reproducibility (coefficient of variation
of the order of 0.8 percent), and by a normal distribution of the observations
within the corresponding range. As compared to other soil mechanics tests,
the ASTM D 2049-69 standard could be considered as a rather good and
reliable test.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licens
TAVENAS El" AL ON RESULTS OF A COMPARATIVE TEST PROGRAM 37

Relative Density
The minimum and maximum densities of cohesionless soils are not used
directly as usual soils parameters. They were defined and are measured
only to form the basis of the relative density determination, by means of the
well-known formula:
Dr -- ~'d max X "~d ~ 'Yd rain

Td Td max ~ T d rain

Based on the results of the present comparative test program, the quality
of the minimum and maximum density measurements was analyzed in the

relative densdy,%
0 t I0 20 30 40 =50 t 60 70 80 t 90 tOO
150 _ i i I [ I I I I i i I I I I I ] I I I I I I I I I I I I I.
Z
145

i40 _ ~"

/
_ ~ 135

v~ 130

r / ' / 125

/ /
~ r -~ /
/ / / ~ / t.. Izo

/ minimum densilies

~,,o
g
~
.~I~IIus ond rninu~ tWO slondord deviofions /
~
~
.o~
__,, ,ore ,, . . . . . . . . . . ~ m,.~ . . . . . o"/ / / -
| ::l~o~ ..I / ~.- ~" .

,co~ / / / /. ..-"~/ / ,oo

/ 95

80 i i i I i i t t i i i i ~ i i i i i L I I I iiii t I I I I I
0 I0 20 30 .~0 50 60 70 80 90 IO0

FIG. ll--Variation in relative density between laboratories AHTM tests.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No fu
38 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

previous section of this paper. The next step is to investigate how this
quality reflects on the computed value of the relative density. This can be
done, either by considering the reproducibility and variability of the
limiting densities, or by analyzing the relative densities as computed from
each participant's results.
Influence of the Variability of .yd m in and "yd max on the Relative D e n s i t y - - T o
analyze the influence of the variability of the limiting densities on the
relative density, the use of the relative density graph proposed by Bur-
mister [7] is best suited. Only the results obtained from the ASTM
D 2049-69 standard will be considered. The mean and the mean plus and
minus two standard deviations of the maximum and minimum densities

relOfive density ,%
0 I0 20 130F aO 50 rSOl 70 BO 1901 I00
I~ .i i] i i t I i i i [ i i l [ i F I I [ I I I I I I [ 50

145

140

"- .~ ~

~ " Mean of ,*he


ond minimum
moximum
densit*es
~
~ i ~
~ ~

-Plus and m~nus two sfondord deviations from

~ I05
1he mo=imurn and minimum mean densities
J~J =~

95~ 95

90

F
i I i I i i i i i i ] I ] i [ i ] i i ~ [ i i i i i i I i i t t I
0 I0 20 50 4O 50 60 70 80 9O I008~

relohve densdy, %

FIG. 12--Variation in relative density between duplicate A S T M tests.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No fu
TAVENAS ET AL ON RESULTS OF A COMPARATIVE TEST PROGRAM 39

TABLE 6--Variability of the relative density.

Statistics Indirectly Determined Directly Determined


Prom Figs. 14, 15, 16, Prom Each Participant's
and 17 Results
Variability Reproduci- Variability Reproduci-
bility bility

Fine Sand
Mean 69.0% 69.0% 70.0% 70.0%
Standard deviation 11.0% 3.5% 11.8% 2.5%
Coefficient of variation 15.9% 5.1% 16.9% 3.6%
Gravelly Sand
Mean 43.5% 43.5% 45.9% 45.9%
Standard de;ciation 18.0% 6.5% 21.2% 5.7%
Coefficient of variation 41.4% 14.9% 46.2% 12.4%

were reported on Figs. 11 and 12, respectively, for the variations of the
results between laboratories (variability) and for the variations within the
different laboratories (reproducibility). The solid line joining the means
represents the average relative density as a function of the dry density.
Since the distribution of the observations was normal, the dashed lines
drawn at plus and minus two standard deviations from the means are the
limits within which approximately 95 percent of the observations will be
confined. The probable standard deviation of the relative density can be
computed by dividing the width of this interval by four. Table 6 gives the
probable standard deviations of Dr as computed in this way for the two
tested materials along with the corresponding coefficients of variations.
These parameters were determined for an assumed dry density of 108 lb/ft 3
(1.73 t / m 3) for the fine sand and 122 lb/ft 3 (1.95 t / m 3) for the gravelly sand.
The magnification of the inaccuracies of the maximum and minimum
densities in the resulting inaccuracy of the computed relative density, as
suggested by Tavenas and La Rochelle [4], is perfectly evident here. With
standard deviations and coefficients of variations of the limiting densities
which could be considered as very satisfactory, the resulting minimum
coefficients of variation (corresponding to Dr -- 100 percent) are equal to
11 and 18 percent for the fine sand and the gravelly sand, respectively, and
amount to one third of these values if the reproducibility of the results is
concerned. Since the width of the 95 percent interval is approximately
constant for the full range of relative densities, the usual coefficient of
variations will be much larger, at orders of magnitude of 20 percent for the
fine sand and 40 percent for the gravelly sand.
Variability of the Relative Density as Measured by Each Participant--By
assuming dry densities of 108 lb/ft 3 (1.73 t / m 3) and 122 lb/ft 3 (1.95 t / m 3)
for the fine sand and the gravelly sand, respectively, relative density values

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licen
40 RELATIVE DENSITY INVOLVING COHESIONLESS SOfLS

were computed from each pair of minimum and maximum densities reported
by the participants. The results of the statistical analysis are given in
Table 6, while the cumulated frequency distribution curves for the fine sand
and the gravelly sand are shown on Figs. 13 and 14, respectively.
The statistics are in good agreement with those established indirectly in
the previous section. This is logical considering the normal distributions
observed on the minimum and maximum density values. The means and
standard deviations are only 1 to 3 percent higher, indicating a variability
slightly larger than expected from the preceding section. The combined
standard deviations are I percent smaller, showing a better reproducibility.
This can possibly be explained by the fact that the so-called reproducibility
of ~d mln and ~d . . . . determined previously, actually includes the true
reproducibility plus variations of these between laboratories, both parts of
the total reproducibility being magnified in Fig. 12 while only the first part
is magnified here.
Accuracy of Relative Density Meazurements---Some very important if not
dramatic conclusions can be drawn from the preceding findings.
1. Even though the ASTM D 2049-69 standard tests for determining the
minimum and maximum density of cohesionless materials can be considered
as "normally accurate" soil mechanics tests with observed coefficient of

~00

9O
" [* n I

80 §

70

i ~176
~
o
5C)
/
I" I
.r

I
I
.
...... _M
i

E
8 4o -
I/ 1
I
50 I
I I i

20
t f r e e sond )'d = 108 PCF
~-meon P2.2~I
(3 relohve densdqes
io - I - - - - ~ ~'~ST~-2~DO4S
__ relohve ~nslhe$
o11 volues
o
o IO 20 50 40 50 6Q 70 ~,O 90 I00 I10 120 bS0
relative density,%

FIG. 13--Frequency distribution curve, relative density of fine sand.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No fu
TAVENAS El" AL ON RESULTSOF A COMPARATIVE TEST PROGRAM 41

] ! -s@l~ *sr I ,~ _ ,%

! r~i ,; I i ~ J
I I ' i / ' !

i
i[ I I'fr I i J

i II V,," i F
--- -r -,/TY .... ~ t
~o . . . . . . f_J __~,'_ i __• , , :

0 I0 20 30 40 50 60 70 80 90 I00 ltO 120 130


relative d en sit y,%

FIG. 14---Frequency distribution curve, relative density of gravelly sand.

variability of the order of 2.5 percent and coefficient of reproducibility of


the order of 0.8 percent, the use of these parameters in the relative density
formula leads to a result of poor quality, since it is characterized by
coefficients of variability of the order of 15 to 40 percent and by coefficients
of reproducibility of the order of 3 to 15 percent in most of the usual cases.
Thus, and simply due to the formulation of the relative density, the
variability is multiplied by a factor of 10.
2. As for the limiting densities, the variability of the relative density
increases as the maximum grain sizes of the tested material increases. In the
present case the standard deviations were found 60 to 100 percent larger for
the gravelly sand than for the fine sand. Since the fine sand tested is close
to the ideal material with a small maximum grain size and a coefficient of
uniformity of the order of 3, and no particles passing sieve 200, the vari-
ability and reproducibility observed on this material are the best possible
with the existing testing technique. Thus, it cannot be expected to deter-
mine any relative density with a width of the 95 percent internal, less than
10 percent if the results obtained by one technician only are considered, and
less than 40 percent if the results obtained by different laboratories are
analyzed. As shown by Tiedemann [5], the results obtained by different
operators in the same laboratory fall in between.
3. A third basic parameter influencing the determination of the relative

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licens
42 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

density is the actual dry density. This parameter also is affected by a


certain error. In the most important case of the measurement of the in situ
density, the existing methods (undisturbed piston sampling, gamma-ray
method, sand cone method, Washington method) are such that any value
of ~ cannot be defined with an error less than 4-2 lb/ft ~ (4-0.003 t/m*).
This accuracy was evidenced, for example, by Waterways Experiment
Station [8] and Meigh and Skipp [9].
The error on the in situ dry density has to be combined to the previously
discussed variability to give the final variability of the relative density.
This was done in Fig. 15. On this figure the dashed zone represents the 95
percent interval for the correlation between the in situ dry density and the

relative dewily,%
0 (0 1201 1301 40 SO 1601 ?0 80 i~)1 ICO
. I I I I I I I I I I I I I I I I I I I I I I I I I i_ I'~
- -42%; -~ *52% ~ :
145
I
[, J ,40

I
i / L35

/ J
~J ~ 130
gfoveLiy s~nd .I ~

125

i
fine sond / / / ~t

.!
J

90 90

85 85

-5I% - ~ +54%

i i I i i i i i i i I i i i i i i i i i i L i t i i i i i i i t i IIIIi co- R~
o I0 20 30 ~0 50 60 "/0 80 90
reJotive (:lensit y, */o

FIG. 15--Error on the in situ relative density.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agree
TAVENAS ET AL ON RESULTS OF A COMPARATIVE TEST PROGRAM 43

relative density. The width of the 95 percent interval of the relative density
is shown to be 65 percent for the fine sand and 94 percent for the gravelly
sand. In this second case it is evident that the 95 percent interval is close to
the full range of possible values for the relative density. Under such
circumstances the probability of evaluating the correct relative density by
a wild guess is at least equal to that of measuring it by the standard
method !
4. Due to the very large variability of the relative density between
l~boratories, the comparison of relative densities measured by different
laboratories will be totally nonsignificant. There are important practical
implications of this fact: all established correlations between the relative
density and various properties of cohesionless soils such as the standard
penetration index, the point resistance in a static penetration test, the
friction angle, the modulus of compressibility, the shear wave velocity, etc.,
are useless to anyone but the operator who has established them, since he is
the only one who can reproduce the relative density of the considered soil
with sufficient accuracy. Similarly, the existing liquefaction criteria can
only be properly used by those who have proposed them. Finally, the
control tests for a compaction job will be only valuable when the same
operator who has performed the reference tests also carries out the control
o
test. In this respect the meaning of a check test by an independent labora-
tory to solve a controversy between a compaction contractor and the
controlling laboratory may be put in serious doubt.
5. It appears, therefore, that due not so much to the variability of the
minimum and maximum densities but essentially to the formulation of the
relative density, the resulting accuracy of this parameter is so poor that its
use will be related to major uncertainties (the best case is of ideal material
such as the present fine sand, and will be practically meaningless in most
of the other cases).
Standard and Modified Proctor Tests
The participants were asked to perform two tests each on the fine sand
and the gravelly sand following the standard and modified Proctor
compaction procedures. It was specified to perform these tests on oven-dried
material, at zero percent water content. In this way the influence of the
optimum water content on the variability of the optimum dry density was
eliminated. At the same time it was expected to get densities close to the
usual optimum since dry cohesionless materials can be easily compacted.
Variability of the Proctor Test Results
Seventy-seven tests were performed on each material with each of the
standard and modified Proctor procedure. The statistics concerning the
variations of the Proctor densities between laboratories are given in
Table 7.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
44 RELATIVE DENSITY iNVOLViNG COHESIONLESS SOILS

As expected the densities obtained b y the modified Proctor procedure are


higher than those produced b y the standard procedure, but the differences
are not as large as normally observed, particularly for the gravelly sand
where ~ increases only from 128.3 l b / f t 3 (2.06 t / m 3) to 129.8 l b / f t 3
(2.08 t/m3). The computed standard deviations vary from 2.0 l b / f t ~
(0.032 t / m 3) to 2.7 l b / f t 3 (0.043 t / m 3) and represent about 19 percent of the
observed ranges. Here again, the variability increases with an increasing
maximum grain size, but this increase is significant only for the results of
modified Proctor tests. The variability observed here is very similar to that
reported b y Liu and Thompson [6] for tests performed in complete accord-
ance with the ASTM Tests for Moisture-Density Relations of Soils, Using
5.5-Lb R a m m e r and 12-In. Drop, Method D (D 698-70) with a variable
water content thus indicating that this additionnal variable has no signifi-
cant influence on the variability of the optimum density.

Reproducibility of the Proctor Test Results


Thirty-eight pairs of duplicate tests can be analyzed to define the
reproducibility of the Proctor tests. The corresponding statistics are given
in Table 8.
The best reproducibility is obtained for the modified Proctor compaction
on the fine sand with a combined standard deviation of 0.76 lb/ft ~ (0.012
t/mS); the worst is observed for the standard compaction one, the same
material with a standard deviation twice as large. Results for the gravelly

TABLE 7--Variability of Proctor test results.

Statistic Standard Proctor, Modified Proctor,


lb/ft s t/m 3 lb/ft 3 t/m 3

Fine Sand
Number of tests 77 77

Mean 110.5 1.77 114.3 1.83


Standard deviation 2.3 0. 037 2.0 0.032
Maximum 117.6 1.88 120.6 1.93
Minimum 105.5 1.69 110.2 1.77
Range 12.1 0.19 10.4 0.16

Gravelly Sand
Number of tests 77 78

Mean 128.3 2.06 129.8 2.08


Standard deviation 2.4 0.038 2.7 0.043
Maximum 134.3 2.15 135.5 0.043
Minimum 120.1 1.92 121.3 1.94
Range 14.2 0.23 14.2 0.23

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agre
TAVENAS ET AL ON RESULTSOF A COMPARATIVETEST PROGRAM 45

TABLE 8--Reproducibility of Proctortest results.


Statistic Standard Proctor, ModifiedProctor,
lb/ft8 t/m 3 lb/ft3 t/m s

Fine Sand
Number of pairs of tests 38 38

Average range 1.85 0.030 0.83 0.013


Combined standard deviation 1.53 0.025 0.76 0.012
Gravelly Sand
Number of pairs of tests 38 38

Average range 1.35 0.022 1.48 0.024


Combined standard deviation 1.25 0.020 1.37 0.022

sand fall between these limits with a reversed tendency. Such values of the
combined standard deviation are larger than those reported by Johnson and
Guinnee [10] for series of tests performed by a single operator. As for the
maximum and minimum density tests, it is probable that the reproduci-
bility observed here is the real reproducibility plus a variability of this
reproducibility between laboratories.

Discussion
It should be first emphasized that the Proctor compaction tests were
performed on oven-dried material, and no mention was made by the
participants of any problem having occurred during the tests.
Comparison Between the Standard and Modified Procedures--From the
reported results no clear-cut difference between the two methods can be
made. However, even if the material tested has an influence, it seems that
the modified procedure is a better test in terms of a better reproducibility.
The coefficients of reproducibility are, 1.38 and 0.97 percent for the fine
sand and the gravelly sand, respectively, in the case of the standard Proctor
test as compared to 0.66 and 1.06 percent in the case of the modified Proctor
test. Such a difference is certainly not negligible. On the other hand the
coefficients of variability are nearly equal for both tests, at least on the
average, and of the order of 2 percent.
Comparison Between the Proctor and Maximum Density Tests---The main
reasons for introducing the use of a vibratory table in performing the
maximum density test procedure were, that it would give maximum dry
densities higher than the modified Proctor test, that it would reduce or
eliminate the problem of particle breakage often associated with the impact
compaction method, and that it would be of higher reproducibility. The
present investigation shows that these assumptions are partly correct but
to an extent smaller than anticipated.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to L
46 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

TABLE 9---Compar/son of A S T M D ~0~9-69 and A S T M D 1557-70 test results.

Statistics Modified Proctor Maximum Density


ASTM D 1557-70, ASTM D 2049-69,
lb/ft s t/m 3 lb/ft 3 t/m a

Fine Sand
Mean 114.3 1.830 114.9 1.840
Standard deviation (vari- 2.0 O.032 2.7 O.043
ability)
Standard deviation (repro- O. 76 O.012 O.67 O.011
ducibility)
Gravelly Sand
Mean 129.8 2.08 133.8 2.14
Standard deviation (vari- 2.7 O.043 4.5 O. 07
ability)
Standard deviation (repro- 1.37 O.022 1.37 O.022
ducibility)

As discussed in the section on gradation tests, only the modified Proctor


compaction tests produced some particle breakage on the gravelly sand
specimen. This effect cannot be neglected even though it is very limited,
since the modification is the percentages passing sieve 4 to 60 was of the
same order of magnitude as the standard deviation on these percentages.
The analysis of the mean of the maximum density obtained b y the
A S T M D 2049-69 and b y the ASTM Tests for Moisture-Density Relations
of Soils, Using 10-Lb R a m m e r and 18-In. Drop (D 1557-70) procedures
also shows that the modified Proctor compaction gives lower maximum
densities. ' However, as shown in Table 9, the differences are not very
important particularly for the fine sand.
T h e third assumption of a better quality of the ASTM D 2049-69 tests
results is proven wrong, since the corresponding standard deviations
applicable to the variability are 30 to 60 percent higher than for the
modified Proctor tests results, while the reproducibility is the same for
both tests.

Relative Compaction and Relative Density


In a preceding section the relative density was shown to be affected b y
a high variability and low reproducibility such that its use would be most
difficult, if not practically impossible. This does not necessarily mean t h a t
the concept of expressing the density of a soil relatively to particular
densities of that soil is not applicable, but that the application of this
concept in the form of the relative density is affected b y important
weaknesses.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licens
TAVENAS ET AL ON RESULTSOF A COMPARATIVETEST PROGRAM 47

Other expressions of the relative state of compactness of a soil are either


possible or in use (such as the relative compaction). The quality of this
soil's parameter will be evaluated here.

Relative Compaction
The concept of relative compaction is normally used in conjunction with
the Proctor test for the control of the compaction of fills. The relative
compaction RC is defined as:

RC - ~/d
'Yd max

where Yd is the in situ density and -~ max is the maximum density obtained
by the standard or modified Proctor test; however, one could also think of
using the maximum density as a reference.
The relative compactions were computed from each participant's data on
the basis of the standard Proctor, the modified Proctor, and the maximum
density (ASTM D 2049-69 test only) with an assumed unit weight of 108
lb/ft 3 (1.73 t / m 3) and 122 lb/ft 3 (1.95 t / m 3) for the fine sand and the
gravelly sand, respectively. The corresponding statistics are shown in
Table 10, and the cumulated frequency distribution curves are given on
Fig. 16 and 17.
As noted before, the relative compaction based on the ASTM D 2049-69
maximum density is the lowest, but at the same time the most variable. On
the other hand, RC based on the modified Proctor has intermediate values,
which are close to the preceding for the fine sand but are significantly less
variable. Thus, it appears that the modified Proctor (one point compaction

T A B L E 10--Comparative anatysis of the relative compactions.

Statistics Relative Compaction in Percent Based on:


Standard Proctor Modified Proctor A S T M D 2049-69
Maximum Density

Fine Sand
Mean 97.82 94.54 94.08
Standard deviation 1.99 1.66 2.29
Coefficient of variation 2.03 1.75 2.44
Maximum 102.37 98.00 101.50
Minimum 91.84 89.55 89.40
Range 10.53 8.45 12.10
Gravelly Sand
Mean 95.10 94.05 91.26
Standard deviation 1.77 2.02 3.22
Coefficient of variation 1.86 2.14 3.53
Maximum 101.58 100.58 102.95
Minimum 90.84 90.04 86.65
Range 10.74 10.54 16.30

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lice
48 RELATIVE DENSITY INVOLVING COHESIONLESS ,SOILS

I00

1- I,p, "' . /.._/


/
90 ~' , / / /'
,~o~ / i r
II :"
80

70
! ,, 1 ' i
I i
! !
6O

50
,;
;'
t
I
/
r
E i ,: k
8 40 I/ !
, . . . .
30

,Y~
20
/) /, ~ ~'~;"_~,r; '~
[ (~) modified proctor density

s ; (~) sfondaedprOCtOr ~',~Sity

88 90 92 94 96 913 I00 102 104


relotive compoctlon,~

FIG. 16--Compar/.sonof the relative compactione based on Proctor and maximum density
te~, fine sand.

test on oven-dried soil) leads to relative compactions of a better overall


quality.

Relative Compac$ion versus Rela$%e Density


At a first glance, when comparing the statistics concerning the relative
compaction (Table 10) and the relative density (Table 6), it would seem
that the variability of the relative compaction is much better, since the
coefficients of variations are one order of magnitude lower. However, as
stressed by Lee and Singh [11], the possible ranges of the relative compac-
tion are muchqower than that of the relative density, because no soil can be
found in a state of zero unit weight which would correspond to zero percent
relative compaction, and the minimum value of RC corresponds to the
minimum possible unit weight of the soil considered, that is, of the order of
80 percent. Therefore, a proper evaluation of the variability of both
parameters should be based on the ratio of the standard deviation to the

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to
TAVENAS ET AL O N RESULTS OF A COMPARATIVE TEST PROGRAM 49

possible range, rather than on the coefficient of variation. This ratio is of the
order of 11 to 21 percent for the relative density, as compared to 8 to 10
percent for the relative compaction based on the modified Proctor density.
Thus, it can be concluded that the concept of relative compaction is
significantly better than the relative density in terms of accuracy and
practicability of the result.
It is often considered that a major advantage of the relative density over
the relative compaction is that the relative density magnifies small varia-
tions of the in situ density, thus allowing a better control of such variations,
particularly on compaction works. From the present findings, this supposed
major advantage appears to be a major disadvantage, since the relative
density also magnifies the errors on the unit weights to such an extent that
the computed result is barely better than that obtained by a pure guess.
Therefore, the use of the relative compaction can only be encouraged, not
only in compaction specifications but also in the analysis of natural
deposits.
I00

9C

80

70

60

a,
50
o

40

30

20

0
86 88 90 92 94 96 98 IO0 t02
relative compoctlon~%

FIG. 17--Comparison of the relative compactions based on Proctor and maximum density
te~ts, gravelly sand.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
50 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

Conclusions
Forty-one United States and Canadian Laboratories representing
governments, universities, and industry have participated in a comparative
test program, the purpose of which was to evaluate the variability and
reproducibility of the most usual tests performed on eohesioniess soil
samples, namely, the sieve analysis, the measurement of the minimum and
maximum densities and of the resulting relative density, and the measure-
ment of the maximum density by the standard and modified Proctor
compaction tests.
The first and most general conclusion of this investigation is that none of
the considered tests are really reliable (a high variability and a low
reproducibility are typical of all these soil mechanics tests). This finding
confirms those yielded by similar investigations applied to other geo-
technical tests such as the identification tests for cohesive soils. This general
characteristic of soil mechanics tests has been largely ignored up to now,
but should be considered most seriously in the future, since it throws some
doubts on the validity of our general approach to the evaluation of soils
properties. More specifically, the results of the present investigation
certainly emphasize the necessity of making use of some of the basic
principles of the probability and decision theories.
As to the quality of the various testing methods considered in this
program, the conclusions may be summarized as follows:
1. The sieve analysis is affected by a large variability with average
coefficients of variations of the order of 20 percent. For the material tested
the range observed was nearly as wide as the ranges specified for materials
to be used as selected fills in runways, roads, or filters. The variability of
such important parameters as the dl0, de0, or d~ is such that serious questions
may be raised as to the validity of the usual criteria based on them.
2. The minimum and maximum densities can be best evaluated with the
ASTM D 2049-69 standard method. This method yields results Which are,
on the average, as good as those obtained from other methods, but which
can be more easily compared when measured by different laboratories
since their variability is much less. The quality of these results would seem
satisfactory with coefficients of variation of the order of 2.5 percent and
coefficient of reproducibility of the order of 0.8 percent. The quality of the
result is better for the minimum density test and tests performed on fine
materials.
3. Due to its formulation the relative density is affected by a variability
and a reproducibility 10 times worse than those of the limiting densities.
Due to this high variability, which can be best emphasized by the width of
the 95 percent interval of from 40 to 100 percent, it appears practically
meaningless to try to measure a relative density, since a pure guess is more
likely to give a correct answer.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License A
TAVENAS ET AL ON RESULTS OF A COMPARATIVE TEST PROGRAM 51

4. Standard and modified Proctor tests exhibit a better quality than the
maximum density test. Thus, because of its more satisfactory formulation,
the relative compaction based on the modified Proctor density appears to
be a slightly better tool for evaluating the state of compactness of a
cohesionless soil deposit.
Acknowledgments
The authors are indebted to C. B. Crawford for his assistance, to
E. B. Hall, Chairman of the ASTM Committee D-18 who accepted to
provide the moral support of ASTM to this comparative test program, and
to E. T. Selig, Chairman of this Symposium who helped in providing a
frame for this investigation.
The participation of the 41 United States and Canadian laboratories,
without which this investigation would have been impossible, is gratefully
acknowledged.
The very important contribution of D. A. Tiedemann, U.S. Bureau of
Reclamation is acknowledged. His participation in the preparation of the
test program, the processing of the fine sand specimens, and the discussion
of the present paper was of major importance.
The preparation and shipping of the specimens was supported by the
U.S. Bureau of Reclamation, the National Research Council of Canada
Grant A-7724, and Laval University, Quebec. The preparation of this paper
was supported by the National Research Council of Canada Grant A-7724.

APPENDIX I
List of Participants

United States
Universities
State University of New York at Buffalo, Buffalo, N.Y.
University of New Mexico, Albuquerque, N.M.
Federal
Soil Conservation Service, Lincoln, Neb.
U.S. Bureau of Reclamation, Denver, Colo.
Missouri River Division Laboratory, Omaha, Neb.
U.S. Army Engineers, Waterways Experiment Station, Vicksburg, Miss.
Tennessee Valley Authority, Knoxville, Tenn.
Consulting Engineers
Joseph S. Ward & Associates, Caldwell, N.J.
Woodward-Moorhouse & Associates Inc., Clifton, N.J.
Dames & Moore, San Francisco, Calif.
E. D'Appolonia Consulting Engineers Inc., Pittsburg, Pa.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreeme
52 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

Law Engineering Testing Co., Jacksonville, Fla.


McClelland Engineers, Houston, Tex.
Shannon and Wilson, Seattle, Wash.
Haley & Aldrich, Inc., Cambridge, Mass.
Commercial Testing Laboratories
The H. C. Nutting Company, Cincinnati, Ohio
Geo-Testing Inc., San Rafael, Calif.
Smith-Emery Company, Los Angeles, Calif.
Canada
Universities
Universit~ Laval, Quebec, P.Q.
Ecole Polytechnique, Montreal, P.Q.
Universit~ de Sherbrooke, Sherbrooke, P.Q.
Queen's University, Kingston, Ont.
University of Manitoba, Winnipeg, Man.
University of Alberta, Edmonton, Alta.
University of British Columbia, Vancouver, B.C.
Government's Organizations
National Research Council of Canada, Ottawa, Ont.
Minist~re de la Voirie, Quebec, P.Q.
Hydro-Quebec, Montreal, P.Q.
Department of Highways, Ontario, Downsview, Ont.
Ontario-Hydro, Toronto, Ont.
Canada Department of Regional Economic Expansion, Saskatoon, Sask.
Department of Highways, Edmonton, Alta.
Canadian National Railways, Montreal, P.Q.
Manitoba-Hydro, Winnipeg, Man.
Consulting Engineers
Geocon Limited, Dorval, P.Q.
Terratech Limit~e, Montreal, P.Q.
Montreal Engineering Co., Montreal, P.Q.
H. Q. Golder and Associates, Mississauga, Ont.
R. M. Hardy and Associates, Edmonton, Alta.
Warnock Hersey, International Limited, Darmouth, N.S.
Dames & Moore, Don Mills, Ont.

APPENDIX II
Sequence and Procedures for Testing Specimens

General Remarks
It is our intention to have each participant use untested material whenever
possible in performing the maximum-minimum density tests and the compaction
tests. However, we realize that we have not shipped enough soil to accomplish

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
TABLE ll--Testing sequence for fine sand specimen when maximum-minimum density testing will be done by either A S T M procedure or the
participant's standard procedure.

Weigh two sacks together and individually


Mix sample
By quartering
Specimen 4S
(excess material), O
z
sieve
Specimen 1S Specimen 2S Specimen 3S
(6.5 to 7 kg) (6.5 to 7 kg) (6.5 to 7 kg)

Oven dry Oven-dry By quartering o

1
Minimum by ASTM Minimum by ASTM . Specimen 3S-A Specimen 3S-B
or participant's or participant's (3-~ kg) (3-[- kg)
Procedure procedure
Oven-dry Oven-dry
<.
Maximum by ASTM Maximum by ASTM Standard compaction Modified compaction
or participant's or participant's
procedure procedure Modified compaction Standard compaction

Sieve Sieve O

O1
to

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
t~
4~

TABLE 12--Testing sequence for fine sand specimen when maximum-minimum density by participant will be done by both the A S T M procedure
and the participant's standard procedure using a large mold.
<,
Weigh two sacks together and individually
1 z
Mix sample -4
-r
By quartering
1
Specimen 4S <
(excess material), z
o
1 sieve-minimum
Specimen 1S Specimen 2S Specimen 3S weight = 100 g t~
o-r
(6.5 to 7 kg) (6.5 to 7 kg) (6.5 to 7 kg)
O
Oven dry Oven dry By Quartering Z
F~
r

Minimum by ASTM Minimum by participant's Specimen 3S-A Specimen 3S-B


procedure (3q- kg) ( 3 + kg)
Minimum by participant's
procedure Minimum by ASTM Oven dry Oven dry

Maximum by ASTM J, Standard compaction Modified compaction


Maximum by participant's
Maximum by participant's procedure Modified compaction Standard compaction
procedure
Maximum by ASTM Sieve
Sieve

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
TABLE 13--Testinq sequencefor fine sand sample when maximum-minimum density te,~tinq by participant will be done by both the, A S T M
procedure and the participant's standard procedure using a small mold.

Weigh two sacks together and individually


Mix sample
By quartering
- - Specimen 4S
(exc~ material),
t Sieve-m{nimum
Specimen 1S Specimen 2S Specimen 3S weight = 100 g
(6.5 to 7 kg) (6.5 to 7 kg) (6.5 ot 7 kg) O
z
Oven dry Oven dry By quartering
4
-O~
1
Minimum by ASTM Minimum by ASTM Specimen 3S-A Specimen 3S-B o
(3+ kg) (3+ kg)
1
Maximum by ASTM Maximum by ASTM Oven dry Oven dry O
1
Standard compaction Modified compaction
i Minimum by participant's Minimum by participant's -4
Sieve procedure procedure .<
Maximum by participant's Maximum by participant's
procedure procedure "u

1
Standard compaction Modified compaction O

Sieve

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
56 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

this. To establish a standard sequence for using tested material we would like each
participant to follow a specified testing sequence.
Fine Sand Specimen
Testing Sequence--Three separate testing sequences have been prepared; (see
Tables 11 through 13. One of these testing sequences should be followed in per-
forming the required testing. The sequence which the participant selects should
be based on how the participant plans to perform the maximum-minimum density
test.
In Table 11 the testing sequence outlined is for a participant to follow if the
participant plans to perform the maximum-nimimum density testing using either
the ASTM procedure or their standard procedure, but not both.
In Table 12 the testing sequence outlined is for a participant to follow if the
participant plans to perform the maximum-minimum density testing using both
the ASTM procedure and a procedure which is frequently used by the participant
and in which a large mold is used.
In Table 13 the testing sequence outlined is for a participant to follow, if the
participant plans to perform the maximum-minimum density tests using both
the ASTM procedure and a procedure which is frequently used by the participant
and in which a small mold is used.
Sampling Preparation--In opening the sack of fine sand you will find two small
sacks. The sand in these sacks is part of the sample. Without opening them, weigh
them together and individually using the scale that will be used in conducting the
relative density tests. Report the weights obtained on Form ]3-1 in Appendix II.
After weighing add the contents of these smaller sacks to the contents of the large
sack from which they were taken and thoroughly mix the sample.
Next, from this thoroughly mixed sample, obtain three specimens weighing
between 6.5 and 7.0 kg by using a sample splitter, riffle sample, or hand quartering
procedure. In addition, one of these specimens should be split in half using a sample
splitter if possible and label one of these specimens 3S-A and the other 3S-B. The
other two specimens should be labeled 1S and 2S, respectively.
Then place these four specimens in separate containers and oven dry (110 C or
230 F) to a constant weight. If a large force draft oven is available, drying over
night should be adequate. The excess material should also be placed in a con-
tainer and oven dried to a constant weight. This material should be labeled Speci-
men 4S.
Test Pracedures
M a x i m u m and M i n i m u m Density Tests
Maximum-minimum density tests are to be performed in accordance with (1)
ASTM D 2049-69 using the dry method only and a 0.1 ft a mold; (2) the test pro-
cedure usually used by the participant; or (3) both methods 1 and 2. In addition,
the sequence in which the tests are to be performed on the test specimens should
be done in accordance with one of the three testing sequences outlined in Tables 11
through 13. It should be noted that in all cases the tests should be performed on
oven dried material. Record the data and make the necessary calculations that
are required to complete Form B-2 in Appendix II, and attach the data sheets used
in performing these tests to this form.
One-Point Compaction Tests
One-Point Compaction tests are to be performed using oven dried material and
compacting the material in a 4-in. (1/30 ft ~) mold. The sequence to be followed is

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
TABLE 14--Test sequence for gravelly sand specimen when maximum-minimum density testing by participant will be done by either the A S T M
procedure or the participant's standard procedure.

Mix sample
$
By quartering
Specimen 4G
(excess material),
$ O
Specimen 1G Specimen 2G Specimen 3G sieve Z
(6.5 to 7 kg) (6.5 to 7 kg) (6.5 to 7 kg)
$
Oven dry Oven dry Oven dry
l o
Minimum by ASTM Minimum by ASTM Standard compaction
or participant's or participant's
procedure procedure Modified compaction O
Maximum by ASTM Maximum by ASTM Sieve
or participant's or participant's
procedure procedure

Standard compaction Modified compaction

Sieve O
O

r
"4

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
co

TABLE 15--Testing sequence for gravelly sand specimen when maximum-minimum density testing by participant will be done by both the -4
A S T M procedure and the participant's standard procedure.

Mix sample
-4

By quartering E,
Specimen 4G
(excess material), o
<
z
Specimen 1G Specimen 2G Specimen 3G Sieve O
(6.5 to 7 kg) (6.5 to 7 kg) (6.5 to 7 kg)
O
Oven dry Oven dry Oven dry
O
z
Minimum by ASTM Minimum by ASTM Minimum by participant's
[ [ procedure
Maximum by ASTM Maximum by ASTM Maximum by participant's _o
J, ~ procedure
Modified compaction Standard compaction
Sieve Standard Compaction
Modified compaction
Sieve

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
TAVENAS El' AL ON RESULTS OF A COMPARATIVE TEST PROGRAM 59

that given in the same figure used for performing the maximum-minimum density
tests previously described. Record the data and make the necessary calculations
that are required to complete Form B-3 in Appendix II, and attach to this form
the data sheets used in performing these tests.
Both the "standard" and "modified" compaction tests should be performed
using the efforts described in ASTM D 698-70 (Method A) and D 1557-70 (Method
A), respectively.

ParticLe-Size Analyses
Particle-size analyses are to be performed on Specimens 1S, 3S-B, and 4S in
accordance with the testing sequence which the participant followed in performing
the maximum-minimum density tests, that is, one of the sequences outlined in
Tables 11, 12, or 13.
Approximately 100 g should be taken from these specimens using a splitter or
similar procedure, weighed after oven drying, and thoroughly washed on a 200
sieve. Transfer this washed material to a suitable container, oven-dry over night,
and perform a sieve analysis using U.S. Standard sieves 10, 20, 40, 60, 100,
200, and pan. If a powered sieve shaker is used, sieve for 15 min. Record the data
and make the necessary calculations that are required to complete Form B-3
presented in Appendix II, and attach the data sheets used in performing the par-
ticle size analyses to this form.

Gravelly Sand Specimen


Testing sequence--Two testing sequences have been prepared. They are pre-
sented in Tables 14 and 15. One of these testing sequences should be followed in
performing the required testing. The sequence which the participant selects should
be based on how the participant plans to perform the maximum-minimum density
testing.
In Table 14, the testing sequence outlined is to be followed by the participant if
the participant plans to perform the maximum-minimum density testing using
either the ASTM procedure or their standard procedure, but not both procedures.
In Table 15, the testing sequence outlined is to be followed by the participant if
the participant plans to perform the maximum-minimum density tests using both
the ASTM procedure and a procedure which is frequently used by the participant.
Sample Preparation--Before proceeding with any testing, (1) thoroughly mix
the sample; (2) prepare three specimens weighing between 6.5 and 7.0 kg using a
sample splitter, riffle sampler, or hand quartering procedure; (3) place these three
specimens in separate containers; and (4) oven dry (110 C or 230 F) to a constant
weight. All of the excess material (weighing approximately 1.5 kg) should also be
placed in a container and oven dried to a constant weight. This excess material
should be labeled Specimen 4G, while the other three specimens should be labeled
1G, 2G, and 3G, respectively.

Test Procedures
Maximum-Minimum Density Tests
The instructions given in the section under fine sand can be followed here except
for the following: (1) testing sequence outlined in Tables 14 and 15 should be fol-
lowed and (2) Form B-5 used instead of Form B-2. Note, the test specimen should
not be scalped on the ~-in. sieve, that is, the -t-3~-in. material should be included
in the test specimen.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
60 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

One-Point Compaction Test


The instructions given in the section under fine sand specimen can be followed
here except for the following: (1) a 6-inch mold should be used; (2) the compaction
efforts described in Methods D, of A S T M D 698-70 and D 1557-70 should be used
instead of Methods A; (3) testing sequence outlined in Tables 14 and 15 should be
followed; and (4) Form 6-B used instead of Form B-3. Note, the test specimen
should not be scalped on the ~ - i n . sieve; that is, the T ~ - i n . material should be
included in the test specimen.
Particle-Size Analyses
Particle size analyses are to be performed on Specimens 1G, 3G, and 4G in ac-
cordance with the testing sequence which the participant followed in Tables 14
and 15.
In performing the particle-size analyses on these specimens, after over drying
and recording the weight of the specimen, all of the material in each specimen
should be sieved on the following U.S. Standard sieves: 1 in. ~ in., No. 4, and
pan. Next, from the material retained in the pan, select a specimen weighing ap-
proximately 100 g and record the weight. Then thoroughly wash this specimen on
the 200 sieve and transfer the wash material to a suitable container, oven dry to a
constant weight, and perform a sieve analysis using U.S. Standard sieves 4, 10,
20, 40, 60, 100, 200, and pan. If a powered sieve shaker is used, sieve for 15 min.
Record the data and make the necessary calculations that are required to com-
plete Form B-6 in Appendix II, and attach to this form, the data sheets used in
performing the particle-size analyses.

References
[1] Kolbuszewski, J. J., Proceedings, 2nd International Conference on Soil Mechanics
and Foundation Engineering, Vol. 1, Rotterdam, 1948, pp. 158-165.
[P] Felt, E. J. in Symposium on Application of Soil Testing in Highway Design and Con-
struction, A S T M S T P 239, American Society for Testing and Materials, 1958 pp.
89-110.
[3] Pettibone, H. C. and Hardim, J., "Research on Vibratory Maximum Density Tests
for Cohesionless Soils," paper presented at the 67th Annual Meeting, American
Society for Testing and Materials, 1964.
[~i] Tavenas, F. and La Rochelle, P., "Problems Related to the Use of the Relative
Density," Report S-21, Laval University, Quebec, Canada, 1970.
[5] Tieder~nn, D. A., "Variability of Laboratory Relative Density and Gradation
Tests," Report REC-ERC-71-17, Bureau of Reclamation, Denver, 1971.
[6] Liu, T. K. and Thompson, M. R. in Proceedings, National Conference on 'Statistical
Quality Control Methodology in Highway and Airfield Construction, University of
Virginia, Charlottesville, Va., May 1966, pp. 375-395.
[7] Burmister, D. M., Proceedings, American Society for Testing and Materials, Vol. 48,
1948, pp. 1249-1268.
[8] "Density Changes of Sand Caused by Sampling and Testing," Potamology Investi-
gations Report No. 12-1, Waterways Experiment Station, Vicksburg, Miss., 1952.
[9] Meigh, A. C. and Skipp, B. O., Geotechnique, Vol. 10, No. 3, 1960, pp. 110-126.
[10] Johnson, A. W. and Guinnee, J. W., "A Report on the Consideration of the Test
Results from the ACIL Standard Soil Sample Program Along with the Supple-
mental Testing Conducted in Cooperation with ASTM D-18," Internal Report,
Committee D-18, American Society for Testing and Materials, 1966.
[11] Lee, K. L. and Singh, A., Journal of the Soil Mechanics and Foundation Division,
American Society of Civil Engineers, Vol. 97, SM7, July 1971.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
D. A . Tiedemann 1

Variability of Laboratory Relative Density


Test Results

REFERENCE: Tiedemann, D. A., "Variability of Laboratory Relative


Density Test Results," Evaluation of Relative Density and Its Role in
Geotechnical Projects Involving Cohesionless Soils, ASTM S T P 5~3, American
Society for Testing and Materials, 1973, pp. 61-73.
ABSTRACT: Fourteen Bureau of Reclamation soils laboratories conducted
duplicate minimum, and wet and dry method maximum density tests on a fine
and a medium sand. Data were analyzed on the basis of variations between
laboratories and between duplicate tests. Results indicated that: (1) variations
between laboratories were two to three times greater than variations between
duplicate tests; (2) variations between laboratories for minimum and maxi-
mum densities were the same, or less than, variations reported for impact-type
compaction tests; and (3) expressing the degree of compaction in terms of rela-
tive density, as compared to percent of maximum density, requires using
different standards for both the level of compaction required and the limits
within which compaction would be considered acceptable. Gradation tests on
the two sands were also conducted by 16 laboratories and analyzed in the
same manner.
KEY WORDS: soil tests, density (mass/volume), sieve analysis, eohesionless
soils, sands, statistical quality control, soil compacting, density measurement,
soil mechanics, reproducibility.

The concept of relative density [1]2 is widely used for expressing the
state of compactness of cohesionless granular soils. It involves comparing
the natural, or compacted, density of a soil to the minimum and maximum
densities of which it can be placed in the laboratory. Various tests have
been devised to determine the two limiting densities. While most of the
test methods have been in use for more than ten years, very few studies
have been conducted to determine the variations associated with their use.
Knowledge of these variations is needed: (1) for developing statistical

1 Research civil engineer, Engineering and Research Center, Bureau of Reclamation,


Denver, Colo. 80225.
The italic numbers in brackets refer to the list of references appended to this paper.

61

Copyright by ASTM Int'l


Copyright9 (all rights
by ASTM reserved); Fri www.astm.org
International Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreemen
62 RELATIVEDENSITY INVOLV/NG COHESIONLESS SOILS

quality control methods, and (2) for determining the reliability of relation-
ships between a soil's relative density and its physical behavior, such as
load-settlement characteristics, permeability, penetration resistance, and
static and dynamic strengths. This information becomes of increased im-
portance when data developed by one laboratory is used by another.

Purpose and Scope of Investigation


In 1960, the Bureau of Reclamation (Bureau) adopted the vibratory
table method as a standard test procedure for determining the maximum
density of cohesionless soils. This method is essentially the same as the
ASTM Test for Relative Density of Cohesionless Soils (D 2049-69). The
minimum density test procedure of pouring dry soil has been in use for a
longer period of time. And, of the two limiting tests, it has the lesser
variation between laboratories. Relative density investigations conducted
by the Bureau [~-5] and others [6-16] have been concerned mainly with
comparing the results obtained using different maximum density test pro-
cedures or with studying the effects of variables such as the magnitude,
time, and direction of vibration have on the maximum density obtained.
In only two studies [17, 18] has the variability in results obtained by any
one given method been investigated.
To obtain this information, a Bureau cooperative testing program was
conducted [19]. Fourteen Bureau soils laboratories conducted minimum,
and wet and dry method maximum density tests, in duplicate, on a fine
and a medium sand. In conjunction with these tests, 16 laboratories con-
ducted gradation tests on the sands. The results obtained from this pro-
gram are presented in this paper.
Materials Tested
Two gradations of clean, poorly graded sand (SP) were prepared from
screened fractions of a local stream (Clear Creek) deposit material. One
sample, designated as 24E-11, was predominantly a fine sand and the
other, 24E-12, was predominantly a medium sand. These soils contained
less than 5 percent of particles finer than the No. 200 sieve or coarser than
the No. 4 sieve.
Fifty-pound samples of each sand were sent to the participating labora-
tories. The samples were prepared by shoveling the soil at random from
thoroughly mixed stockpiles. To prevent the loss of fine particles, plastic-
lined sacks were used for shipment.

Data Presentation and Analyses


After all the data had been received each laboratory was assigned an
alphabetical designation. When more than one set of duplicate tests was
conducted by the same laboratory, the operator or operators were given a
numerical designation.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Ag
TIEDEMANN ON VARIABILITYOF LABORATORYTEST RESULTS 63

The results were analyzed in two ways: (1) the reproducibility between
laboratories, and (2) the repeatability between duplicate tests. In analyzing
the results, the following statistics were calculated [20].
Reproducibility Between Laboratories
Range--The range (R) is the difference between the highest and lowest
values.
Average--The average (:~) or arithmetic mean was calculated as
= r.x/n

where Zx is the summation of the individual values, and n is the number


of these values.
Standard Deviation--The standard deviation (s) which expresses the
degree of variation with respect to the mean was calculated as
s = %/Y.(x - X)2/(n - 1)
For normally distributed data, the interval between two standard devi-
ations on both sides of the mean will contain approximately 95 percent of
the values, and if the results used in determining the standard deviation
are representative, 95 percent of any future tests would be expected to
fall within these limits. For purposes of discussion, the results from these
tests were assumed to meet these requirements.
Repeatability Between Duplicate Tests
Average Range--The average range (/~) was calculated as
[~ = Z I d l/k
where Z I d I is the summation of the absolute differences between duplicate
tests, and k is the number of duplicate tests.
Combined Standard Deviation--The combined standard deviation, s', was
calculated as
s'= v -5/ek
While it was recognized that the variability between laboratories is
influenced by the repeatability within each laboratory [~I], no attempt
was made to separate the two when analyzing the reproducibility between
laboratories.

Gradation Tests
Procedure
The gradation tests were conducted primarily as a check on the uni-
formity of the samples. Results were obtained from 16 laboratories, about
one half of which performed duplicate tests. The tests were conducted by
dry sieving the samples as received for 10 to 15 rain.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
64 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

TABLE 1--~ummary of gradation test uariations between/aborator/es. =

Percentage passing, based on dry weight of total sample. (Percentage retained on


individual sieves, based on dry weight of total sample.)
Statistical feature Sieve Size

No. 4 No. 8 No. 16 No. 30 No. 50 No. 100 No. 200

Fine Sand
Range 100-98 97-92 90-84 75--66 50-42 14--9 5-2
(0-2) (3-6) (7-10) (15-20) (21-30) (29-38) (7-10)
Average 100 96 87 70 45 11 3
(0.2) (4.2) (8.5") (17.1) (25.0) (33.8) (7.8)
Standard devia- 0.6 1.0 1.4 2.4 2.4 1.6 0.9
tion (0.6) (0.8) (1.0) (1.3) (1.8) (1.9) (0.9)
Medium Sand
Range 100-98 85-75 56-33 33-14 13-4 7-2 4-0
(0-2) (15-25) (32-47) (17-30) (9-16) (2-6) (1-4)
Average 100 81 43 22 9 4 2
(0.3) (19.0) (37.9) (29.3) (12.7) (4.8) (2.4)
Standard devia- 0.6 3.1 5.6 4.6 2.2 1.2 0.7
tion (0.6) (3.2) (3.7) (2.6) (2.0) (1.1) (0.9)

= Data based on results of 25 gradation tests.

U.S. STANDARD SIEVE S E R I E S


=200 =100 #50 ~r #16 #8 #4

i , 'i

~-~-MEDIUM SAND
(24E-12)

i
I
.037 .074 .149 297 .590 1.19 2.38 4.76 9.52
DIAMETER OF PARTICLE IN MILLIMETERS

I FINE ISAND
MEDIUM ICOARSE]

FIG. 1--Average gradation curves for test samples.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No fu
TIEDEMANN ON VARIABILITY OF LABORATORY TEST RESULTS 65

T A B L E 2---Summary of gradation test variations between duplicate te~ts.

Percentage passing, based on dry weight of total sample


Sieve Size

No. 4 No. 8 No. 16 No. 30 No. 50 No. 100 No. 200

Fine Sand a
Maximum 0 3 5 6 5 1 1
Minimum 0 0 0 0 0 0 0
Average range 0 1.1 2.1 2.6 1.7 0.6 0.4
Standard devi- 0 1.2 1.9 2.3 1.7 0.5 0.5
ation
Medium Sand b
Maximum 1 5 8 5 2 2 2
Minimum 0 0 0 0 0 0 0
Average range 0.2 2.2 4.0 2.4 1.2 0.6 0.4
Standard devi- 0.4 2.1 3.5 2.0 1.0 0.7 0.6
ation

a Based on the results from seven pairs of duplicate tests.


b Based on the results from eight pairs of duplicate tests.

Test Results
The variations between laboratories were analyzed in two ways: on the
basis of the percentage passing each sieve and the percentage retained on
the individual sieves. The summary of these results is presented in Table 1,
and the average gradation curves are drawn in Fig. 1. The variations
between duplicate tests were analyzed only on the basis of the percentage
passing each sieve. These results are presented in Table 2.

Discussion
When the results are analyzed on the basis of the percentage passing
each sieve, the standard deviations are as high as 5.6 percent and, as found
in similar investigations [8, 22, 28], do not appear to have any definite
relationship to the average values. However, when analyzed on the basis
of the percentage retained on the individual sieve, the standard deviations
are decreased and, as shown in Fig. 2, are related to the average percentages
retained.
The duplicate test and between-laboratory standard deviations are
similar for the fine sand, but differ for the medium sand. These differences
are attributed to the variability in medium sand samples furnished the
participating laboratories. The medium sand, unlike the fine sand, tended
to segregate when mixed. Difficulty in obtaining identical samples has been
encountered in other cooperative testing programs [8, 22].

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No furthe
66 RELATIVE DENSITY I N V O L V I N G COHESIONLESS SOILS

Symbol 9 o o 9 0 t x + ~ A
Reference (8) (8} (8) (8) (8) {8) (19) (19) (23) (23)
Soil designation i 2 3 4 5 8 ~1 *z s G

..c_
z2~6

q[ O~

i ouo~' O 4
A 0
+
0 O'
+ o ~ t
9 o t + 8 , A.
O , 0
2 I(
0 ' i x
{DATA OBTAINED FROM REFS. {8, 19, 2:5){
I I
Oo ,5 IO 15 20 25 30 35 40 45
AVERAGE PERCENTAGE RETAINED ON INDIVIDUAL S I E V E S

FIG. 2---Variation in average percentageof sample retained on individual sieves.

Minimum and Maximum Density Tests


Procedure
All of the laboratories that received samples were requested to conduct
duplicate minimum, and duplicate wet and dry method maximum density
tests following the procedures outlined in Designation E-12, Part B, of
the Earth Manual [24]. This test designation is essentially the same as
ASTM (D 2049-69) and utilizes a vibratory table for determining the
maximum density. The 0.1 ft 3 mold was used in all tests.
Results
The results from 14 laboratories are presented in Fig. 3. Each set of
duplicate tests is represented by a hatched bar; the ends of the bars are
drawn at the individual test values. These results were analyzed on the
basis of the reproducibility between laboratories and the repeatability
between duplicate tests.
Reproducibility Between Laboratories---In determining the variability be-
tween laboratories, the following values were selected from the data
reported by each laboratory: the lowest minimum density, the highest
dry method and the highest wet method maximum densities, and the
highest maximum density of the two methods. When a laboratory reported
more than one set of duplicate tests by the same test method for either
sample, the first set of tests was used. The results of the statistical analyses
are summarized in Table 3. Standard deviations for the minimum and
maximum (of either method) densities are 1.3 and 1.1 lb/ft 3, respectively,

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License
TIEDEMANN O N VARIABILITY OF LABORATORY TEST RESULTS 67

LABORATORY A B C D E F G H I J K L M 0
I 7 ~-~' ~ ~-ql r'-'~ ~--'q [-'-q 1"-7 ~ ~ r---I [ - - 1 C - 7 I - - 1
OPERATOR I 2
~ ~-~ ~-~ r J 7 r-J~ ~ 1, - ~ ] ,
125

I'~-
DRY METHOD
WET M E T H O D
(24E -II s,,,,. 1
)

120

F],,"
115

t-,
I
I10
I-
$
z r-
(3 105
>-
E

1
I00

~n ==
11 ~ in

95
J.

AVERAGE MINIMUM DENSITY AVERAGE MAXIMUM DENSITY


90

LABORATORY A B C D E F G H 1" J K L M 0
r 7 I--'--] r - - 7 [-'--'-~ ~ 17 I~ I--7 ~ ~'--] ~ i i r'~] i i
OPERATOR I 2 5 I I t I
r-'--I i---'7 r-'~ ~1--7 ~ [----7

125 ~ MEDIUM sAND I


124E -J2) j

115

(.;i
cL

~-
I
IlO
J
I,--
(/)
Z
~.J
c~ 105

e:
r
|
IOC ;
a= []!
95

AVERAGE MINIMUM DENSITY AVERAGE MAXIMUM DENSITY

9r

FIG. 3--Summary of relative density tests.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
68 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

TABLE 3--Between-laboratory variations in minimum and maximum density tests.

Minimum Maximum MaximumDry Density


Dry Dry
Density, Density, Dry Method, Wet Method,
lb/ft~ lb/fta lb/ft~ lb/fta

Fine Sand
Number of tests 14 14 14 14
Range 92.9-98.5 113.9-118.8 113.6-118.8 110.0-117.8
Average 96.2 115.8 115.4 114.2
Standard deviation 1.4 1.3 1.3 2.1

Medium Sand
Number of tests 14 14 14 14
Range 95.2-101.5 117,4-124.9 117.4-122.7 113.5-124.9
Average 97.8 120.4 119.7 118.6
Standard deviation 1.8 2.2 1.8 3.2

for the fine sand, and 1.6 and 1.8 lb/ft 3, respectively, for the medium
sand. In comparison, standard deviations ranging from 1.9 to 2.9 lb/ft ~
have been reported [22] for impact-type compaction tests on fine-grained
soils.
Of the two maximum density procedures, the dry method had the lower
standard deviations. This would be expected since the wet method is
subject to the same sources of errors as the dry method plus those associ-
ated with the addition of water. Also, in only 3 of the 14 laboratories was
the maximum density obtained by the wet method. Thus, it would appear
that the dry method was not only more consistent but that it also generally
produced the maximum density. However, this may not be entirely true.
In the wet method, the density obtained is related to the degree of saturation
[10, 24], as shown in Fig. 4 where the wet method dry densities are plotted
as a function of the water content. Also plotted in this figure are the de-
grees of saturation for a specific gravity of 2.65. The majority of specimens
had degrees of saturation varying from 80 to 90 percent which would
explain, in part, the lower densities and higher standard deviations associ-
ated with the wet method tests. It is interesting to note that the lower
densities were obtained for the higher water contents, indicating that
sufficient water was present to saturate the samples but that the entrapped
air and excess water could not be vibrated free. This behavior would not
normally be expected for the types of material tested because of the small
amount of fines contained.
Repeatability Between Duplicate Tests---By analyzing the results on the
basis of the differences between duplicate tests, the variations in samples,
errors in equipment calibrations, and differences in equipment performance
are minimized and an indication of operator repeatability is obtained.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lic
TIEDEMANN ON VARIABILITY OF LABORATORY TEST RESULTS 69

I I II I II I I II I III I I I II I Itl III I I I I I III I I II llll fill

.\,,!",,
:\\\ \

\ x\ , k.~\' \
o
\ \ \\ x \\ \\,IX\\\~ "\ _
I
ra-

w
\~' \\ \ \\ \ l_\/} " \~ -
~,\, \,,l~:
a

,\ - \, }\ < ~ , \
a

\\ ,,l~XX/$. I" \ 'b x~' \:

ii I I [ * I I t t i / i ~ i I i t ~ ~ i ; z i r i i i I I I I i ~ I I i I i i 1111 7
~ ~ /'~
12 13 14 15 16 17 II 12 13 14 15 16 I

WATER CONTENT- PERCENT

FIG. 4---Variation in maximum density as determined by wet method.

The results of the statistical analyses of the duplicate minimum, and wet
and dry method maximum tests are presented in Table 4.
The variations between duplicate tests are one half, or less, of the vari-
ations between laboratories and are in agreement with results reported
by others for the same test procedures [18] and for impact-type compaction

TABLE 4---Variations between duplicate minimum and maximum density tests.

Minimum Maximum Dry Density


Dry Density,
lb/ft ~ Dry Method, Wet Method,
lb/ft 3 lb/ft 8

Fine Sand
Number of tests 18 18 18
Average range O. 3 O. 5 1.3
Combined standard deviation O. 2 O. 5 1.1

Medium Sand
Number of tests 18 18 17
Average range 0.6 1.0 1.2
Combined standard deviation 0.6 0.9 I. 1

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
70 RELATIVE D E N S I T Y I N V O L V I N G COHESIONLESS SOILS

tests [17, 26]. The minimum density variations are also a b o u t the same as
those obtained by a single operator for similar materials [3]. The minimum
and dry method maximum test variations were smaller for the fine sand
than for the medium sand; however, the wet method maximum test vari-
ations were about the same for both sands. This is attributed to the greater
sensitivity of the density of the fine sand to changes in the degree of
saturation [25] which offset the differences between the medium sand
samples.

Relative Density
In the preceding section, the variations associated with the minimum
and maximum density test procedures were discussed. Since these tests are
the basis for determining the relative density, the concern in the test
variations is in their influence on the calculated relative density. In Fig.. 5,
the average minimum and maximum densities determined for the medium
sand are plotted as 0 and 100 percent relative density, respectively. From
the solid line joining these two limiting densities, the relative density of
any intermediate density can be readily determined. The dashed lines
connect the plus and minus two standard deviation intervals about the
minimum and maximum densities as determined for the medium sand.
The outer set of lines represents the between-laboratory variations and
the inner set the duplicate test variations. Approximately 95 percent of
the test results would be expected to fall within these intervals. For the
extreme variations in both the minimum and maximum density a dry
density of 110 lb/ft 3 could be reported as a relative density varying from
40 to 76 percent on a between-laboratory basis, or from 52 to 66 percent
if the limiting densities were determined by a single operator.
Compaction control testing is usually conducted by one or more oper-

130,

---- OUPLaCATE TEST VARIATION-'---~ ~ ~'~


120
i r j r r f ~ / ~ f ~
o. --~ - 7- :T. . . . T - ~- / 11 ~ :
~ I10
P

kU
Q
~ IO0' -- } ] l o 0
er

9C
MINIMUM MAXIMUM
DENSITY RELATIVE DENSITY, % DENSITY
SCALE SCALE

FIG. 5---Variation in minimum and maximum densities for medium sand samples.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
TIEDEMANN ON VARIABILITY OF LABORATORY TEST RESULTS 71

ators in the same laboratory. The variations that might occur can not be
determined directly from this study since no one laboratory conducted a
sufficient number of tests. However, these values would be expected to fall
between the two cases studied and at best would be like those for the
duplicate tests. For a particular engineering feature, the minimum allow-
able relative density would be that which experience or laboratory testing
has shown to provide a satisfactory condition for the application involved.
Because of the test variations, the average relative density would have to
be greater than the minimum allowable value. On the basis of the duplicate
test variations, an average relative density of approximately 92 percent
would be needed to meet a compaction requirement of not less than 85
percent relative density. Included in the test results would be a number of
relative densities of 100 percent or greater.
Since most compaction control thinking is oriented towards the concept
of percent maximum density, or percent Proctor, this may appear to be a
large variation. However, in terms of percent maximum density, the
minimum allowable value would be 97 percent and the average 98 percent
which is only a variation of 1 percent. Therefore, in using the relative
density concept for compaction control instead of the percent maximum
density, it is not only necessary to change the degree of compaction re-
quired, but also the limits of acceptance.
While these examples might be considered unrealistic since the extreme
conditions were assumed, they do indicate the variations that could occur.
It should be noted that the tests were performed in laboratories using the
same standard methods and most of the operators, or their supervisors,
had received formal training in the test procedures. Had the tests been
performed by laboratories using different methods or by inexperienced
operators the variations would have probably been considerably greater.
In an ASTM cooperative relative density testing program reported in
1958 [8], five methods were used to determine the minimum density, and
six methods were used to determine the maximum density. The data ob-
tained were analyzed on the basis of between-laboratory variations and
the results are presented in Table 5. The standard deviations for the
minimums are about the same as found for the Bureau laboratories; how-
ever, the standard deviations for the maximums are considerably higher
varying from about 3 to 6 lb/ft 3. These values would cause very large
variations in the relative density.
Another factor not considered in this program was the sensitivity of the
relative density to variations in the in-place dry density. For the sands
tested, a variation in dry density of 1 lb/ft 3 corresponds to a variation in
relative density of about 3 ~ percent. Other studies [27] have indicated
that the variation in relative density for a 1 lb/ft 8 change can be as high
as 5 percent. Thus, small changes in the in-place dry density can greatly
affect the calculated relative density.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License
7"2 RELATIVEDENSITY INVOLVING COHESIONLESSSOILS

TABLE 5---Between-laboratory variations in minimum and maximum density tests."

Minimum Density Maximum Density


8oil
Average, lb/ft s Standard Average,lb/ft a Standard
Deviation, Deviation,
lb/ft a lb/ft 3

1 90.8 1.6 112.0 4.3


2 100.4 2.1 121.7 2.9
3 98.2 1.5 124.0 6.3

" Data obtained from l~f. 8.

Summary
Based on the results obtained from the Bureau cooperative test program
reported in this paper and other studies, it appears that the variations
associated with the minimum and maximum density tests investigated are
about the same as, or less than, those associated with the impact-type
compaction test. However, when the results are used to compute the rela-
tive density, large variations can occur. The magnitude of these variations
must be considered when using relative density for controlling the com-
paction or relating the physical behavior of cohesionless soils.
Acknowledgments
This study was conducted under the supervision of C. W. Jones, Head,
Special Investigations and Research Section, Earth Sciences Branch. H. J.
Gibbs is Chief of the Earth Sciences Branch. The cooperation of the
personnel of the foUowing Bureau laboratories who participated in this
study is appreciated: Earth Sciences Branch, Division of General Research,
Denver, Colo.; Grand Coulee Third Powerplant Construction Office, Grand
Coulee, Wash.; Columbia Basin Project, Othello, Wash.; Chief Joseph
Dam, Oroville, Wash.; Fresno CVP Construction Office, Fresno, Calif.;
San Luis Unit, CVP, Coalinga, Calif.; San Luis Unit, CVP, Los Banos,
Calif. ; Lahontan Basin Project, Carson City, Nev.; Sacramento River
Division, CVP, Willows, Calif. ; Mead Construction Office, Boulder City,
Nev.; Soldier Creek Dam, Central Utah Project, Duchesne, Utah; Silver
Jack Dam, Curecanti Unit, Montrose, Colo.; San Juan-Chama Project,
Chama, N. M.; Glen Elder Construction Division, Kansas River Project,
Beloit, Kan.; and Fryingpan-Arkansas Project, Salida, Colo.

References
[I] Burmister, D. M., Proceedings, American Society for Testing and Materials, Vol. 48,
1948, pp. 1249-1268.
[$] McNeel, O. F., "Mechanical Dynamic Compaction Experiments," Earth Materials
Laboratory Report No. EM-208, Bureau of Reclamation, Denver, Colo., May 1949.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reprod
TIEDEMANN ON VARIABILITY OF LABORATORY TEST RESULTS 73

[3] Pettibone, H. C., "Development of a Maximum Density Test for Cohesionless Soil
by a Vibratory Method," Earth Laboratory Report No. EM-557, Bureau of Recla-
mation, Denver, Colo., May 1961.
[~] Pettibone, H. C. and Hardin, J. in Compaction of Soils, ASTM, STP 377, June 1964,
pp. 3-19.
[5] Hardin, J., "Laboratory Tests to Refine the Maximum Density Procedure for Co-
hesionless Soils Using a Vibratory Table," Soils Engineering Branch Report No.
EM-697, Bureau of Reclamation, Denver, Colo., March 1965.
[6] Kolbuszewski, J. J., Proceedings, 2nd International Conference Soil Mechanics and
Foundation Engineering, Vol. 1, June 1948, pp. 158-165.
[7] Maxwell, A. A. and Burns, C. D., "Miscellaneous Laboratory Tests," Soil Compac-
tion Investigation, Report No. 5, U. S. Army Corps of Engineers, Waterways Ex-
periment Station, Vicksburg, Miss., June 1950.
[8] Felt, E. I. in Symposium on Application of Soil Testing in Highway Design and Con-
struction, A S T M STP 239, 1958, pp. 89-110.
[9] Selig, E. T. in Proceedings, 2nd Pan American Conference Soil Mechanics and Foun-
dation Engineering, Vol. 1, July 1963, pp. 129-144.
[10] Forssblad, L., "Investigations of Soil Compaction by Vibration," Civil Engineering
and Building Construction Series No. 34, Acta Polytechnica Scandinavica, Stock-
holm, 1965.
[11] D'Appolonia, D. J. and D'Appolonia, E., Proceedings, 3rd Asian Regional Conference
Soil Mechanics and Foundation Engineering, Vol. 1, Sept. 1967, pp. 266-268.
[12] Woodward-Clyde-Sherard and Associates, "Results of Laboratory Tests on Soil
Samples," in Detailed Investigation of Bolsa Island Site, Appendix H, Bechtel,
Oct. 1967.
[13] Forssblad, L., Highway Research Record, No. 177, 1967, pp. 219-225.
[1~] Dobry, R. and Whitman, R. V., "Densification of Sand by Vertical Vibrations in
'Standard' Molds," Progress Report No. 7, Soils Publication No. 251, Massachu-
setts Institute of Technology, Dee. 1969.
[15] Dove, R. P., Williamson, T. G., and Walsh, H. R. I., Highway Research Record, No.
284, 1969, pp. 37-50.
[16] Travenas, F., Capelle, J. F., and La Rochelle, P., Canadian Geotechnieal Journal.
Vol. 7, No. 37, Feb. 1970, pp. 37-53.
[17] "Summary Report of Soil Studies," Potamology Investigation, Report No. 12-2,
U. S. Army Engineers, Waterways Experiment Station, Vicksburg, Miss., 1952.
[18] Tavenas, F. and La Rochelle, P., "Problems Related to the Use of the Relative
Density," Report S-21, Dept. of Civil Engineering, Laval University, Quebec, 1969.
[19] Tiedemann, D. A., "Variability of Laboratory Relative Density and Gradation
Tests," Soils Engineering Branch Report REC-ERC-71-17, Bureau of Reclamation,
Denver, Colo., Feb. 1971.
[20] Bennett, C. A. and Franklin, N. L., "Statistical Analysis in Chemistry and the
Chemical Industry," Wiley, New York, 1954.
[21] Mandel, J., Materials Research and Standards, Vol. 11, No. 8, Aug. 1971, pp. 8-15
and 52.
[22] Liu, T. K. and Thompson, M. R. in Proceedings, National Conference on Statistical
Quality Control in Highway and Airfield Construction, University of Virginia,
Charlottesville, Va., May 1966, pp. 375-395.
[23] Tavenas, F. A., Ladd, R. S., and LaRochelle, P., "The Accuracy of Relative
Density Measurements: Results of a Comparative Test Program," included in this
symposium.
[2~] Earth Manual, Bureau of Reclamation, First Edition, Revised i963, Denver, Colo.
[25] Broms, B. B. and Forssblad, L. in Proceedings, Specialty Session 2, 7th International
Conference Soil Mechanics and Foundation Engineering, Aug. 1969, pp. 101-118.
[26] Sherman, G. B., Watkins, R. O., and Prysock, R., Highway Research Record, No.
177, 1967, pp. 157-185.
[27] Lee, K. L. and Singh, A., Proceedings, American Society of Civil Engineers, Vol.
97, No. SMT, July 1971, pp. 1049-1052.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
Yoshiaki Yoshimi 1 and Ikuo Tohno ~

Statistical Significance of the Relative


Density

R E F E R E N C E : Yoshimi, Yoshiaki and Tohno, Ikuo, " S t a t i s t i c a l S i g n i f i -


c a n c e o f t h e R e l a t i v e D e n s i t y , " Evaluation of Relative Density and Its Role
in Geotechnical Projects Involving Cohesionless Soils, ASTM S T P 523, American
Society for Testing and Materials, 1973, pp. 74-84.
A B S T R A C T : On the basis of a series of laboratory tests and a literature survey,
the maximum and minimum densities for the determination of the relative
density of sands involve random errors from a fraction of one percent to one
percent, and systematic errors reaching several percent.
Concerning random errors in density measurements, the coefficient of vari-
ation of the relative density is expressed as a function of those of the specimen
density and the limiting densities. For given dispersion in the densities, the
coefficient of variation of the relative density increase rapidly as the relative
density decreases.
A numerical analysis of the influence of systematic errors in the limiting
densities on the reliability of the relative density shows that the relative devi-
ation in the relative density may reach tens of percent if arbitrary methods
are used.
In view of the sensitivity of the relative density to variations in the density
measurements, the need for rigidly standardized test methods for the limiting
densities is emphasized, and criteria for the standard methods are suggested.
K E Y W O R D S : eohesionless soils, density (mass/volume), coefficient of vari-
ability, error analysis, sands, tests

I t h a s b e e n r e c o g n i z e d t h a t s m a l l c h a n g e s in t h e m i n i m u m d e n s i t y or in
t h e s p e c i m e n d e n s i t y can cause c o n s i d e r a b l e v a r i a t i o n s in t h e r e l a t i v e
d e n s i t y . T h i s f a c t p l u s t h e l a c k of r i g i d l y s t a n d a r d i z e d p r o c e d u r e s for
d e t e r m i n i n g t h e m i n i m u m a n d m a x i m u m densities t e n d s to d e t r a c t f r o m
t h e r e l i a b i l i t y of t h e r e l a t i v e d e n s i t y . I n t h i s p a p e r t h e a u t h o r s i n t e n d to
s h o w h o w t h e r e l a t i v e d e n s i t y is influenced b y r a n d o m a n d s y s t e m a t i c
errors in t h e l i m i t i n g densities a n d t h e s p e c i m e n d e n s i t y .

1 Professor and assistant, respectively, Tokyo Institute of Technology, O-okayama,


Meguro-ku, Tokyo, Japan.

74

Copyright by ASTM Int'l


Copyright9 by(all
ASTMrightsInternational
reserved); Fri Mar 11 16:13:06 EST 2016
www.astm.org
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
YOSHIMI AND TOHNO ON STATISTICAL SIGNIFICANCE 75

Factors Affecting the Reliability of Relative Density


The relative density of a given soil, D~, is computed from the maximum
dry density, y . . . . the minimum dry density, Ymin, and the dry density
of the specimen, y(the usual subscript, d, being dropped for the sake of
simplicity) as follows:

D. = 1/~m,. - 1/~ = ~m~x(~ - ~ i ~ ) (I)


1/~ml. - 1/~x ~ ( ~ m . ~ -- ~i~)

In this paper Eq 1 is preferred to the alternative form in terms of the


void ratio, D~ = (em~ - e)/(em~z - emi~), because the dry density is re-
lated more directly to measured quantities than the void ratio.
The dry density is computed from two measured quantities, that is, the
dry weight and the volume of the soil, both of which involve random errors
inherent in any measurement. In addition to the random errors, there may
be systematic errors due to lack of rigidly standardized methods for the
determination of the maximum and minimum densities.
Because the relative density of a given soil is proportional to (~ - ~in)/~',
even a small variation in Ymi~ or in 7 may cause a considerable variation
in the relative density when "r - ~ , is small, that is, when the relative
density is low. For example, when ~mi~ = 1.350 g/cm 3, y ~ = 1.637 g/cm 3,
and ~/= 1.425 g/cm 3, the relative density is 30.02 percent. If ~/~i~ is
increased by 1 percent to 1.3635 g/cm s, the relative density becomes
25.83 percent which is 14 percent lower than the initial value. Thus, the
relative deviation in the relative density is 14 times that in the minimum
density, that is,
AD~/D~ = - 14.0Ay~in/5'ml~ (2)
The fact that the relative density is sensitive to variations in the density
measurements has prompted the authors to examine the reliability of the
relative density.

Random Errors in Dry Densities


Specimen Density--The precision of measurements of the volume and
dry weight for the determination of the specimen density may be different
for the laboratory and field conditions. For a careful research in the labora-

TABLE l--Physical propertiesof soils.

Soil Grain Shape SpecificGravity EffectiveSize, Uniformity


of Solids, G, D10 (mm) Coefficient,U~

Toyoura s a n d rounded 2.65 0.147 1.39


Tokyo s a n d subangular 2.73 0.178 1.68

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreemen
76 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

TABLE 2-----Coe~cientof variation of maximum dry density determined by vertical vibration


with no surcharge.

Soil No. of Tests Maximum Dry Density


Mean, g/cm 3 Standard Coefficient Range/Mean,
Deviation, of Variation, %
g/cm* %

Toyoura sand 16 1.538 0.0035 0.23 0.8


Tokyo sand 16 1.423 0.0075 0.53 1.9
Quartz sand~ 12b 1.615 2.4

~ Specific gravity -- 2.64; mesh 10-60; subangular.


b A c c e l e r a t i o n -- 1.31 g; frequency = 32 Hz.

t o r y using compacted specimens, both the volume of the container and the
dry weight can be measured with sufficient accuracy.
On the other hand, field measurements of undisturbed specimens or the
in situ measurements of density m a y involve considerably greater errors.
Laboratory tests were conducted to evaluate the variations in the limiting
dry densities of T o y o u r a sand and T o k y o sand whose physical properties
are given in Table 1.
M a x i m u m Dens/ty---For the maximum density, oven-dry sand was
placed in a compaction mold (100 mm in diameter and 127 mm deep),
and subjected to vertical vibration without surcharge at an acceleration
of 2.0 g and a frequency of 42.5 Hz for 10 rain.
According to previous studies [1-3] 2, the procedure just described was
expected to give a dense packing without crushing the soil particles.
As shown in Table 2, the coefficient of variation of the maximum density
for 16 tests each on Toyoura and T o k y o sands was in the order of a fraction
of one percent. Also given in the table is the result of 12 tests on a quartz
sand b y D o b r y and Whitman [1], showing a similar degree of dispersion.
M i n i m u m Density--For the minimum density, the tilting test and the
spoon test [4, 5] were conducted. In the tilting test, a fixed quantity
(weight) of oven-dried sand was placed in a 1000-ml graduate, which was
slowly tilted several times holding one hand over the open end, and the
volume was read directly. In the spoon test, oven-dried sand was poured
gently in a container (59.93 mm in diameter and 39.65 m m deep) from a
negligible height. After the container had been filled above the rim, excess
sand was removed b y quickly sliding a straightedge, taking extreme care
not to jar the container. Then the sand in the container was weighed.
The results of the minimum density tests are summarized in Table 3.
All the data b y the tilting test were obtained using one measuring cylinder.

The italic numbers in brackets refer to the list of references appended to this paper.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
YOSHIMI AND TOHNO ON STATISTICALSIGNIFICANCE 77

Therefore, there m a y be additional variation due to errors in volume


graduations if different cylinders are involved. Such variation m a y be
roughly estimated on the basis of the tolerance of the graduate. Besides,
the precision in reading the volume of dry sand in the cylinder is rather
poor, and there is marked segregation of soil grains. Thus, the overaU
reliability of the tilting method is probably not as good as Table 3 m a y
indicate.
For T o y o u r a sand, the spoon test yielded smaller scatter t h a n the
tilting method, although the method of test had little influence on the
mean value. For T o k y o sand, however, the spoon test yielded 2 to 3 percent
smaller mean values t h a n the tilting test, although the method of test
had little influence on the coefficient of variation.
As a whole, Table 3 shows t h a t the coefficient of variation of the m i n i m u m
density m a y be considered to be from a fraction of one percent to one
percent.

Systematic Errors in Dry Densities


A systematic error in the specimen density m a y arise from lack of
proper calibration of measuring instruments or from improper handling

TABLE 3--Coe~wient of variation of minimum dry density.

Soil Method Tech- No. of Minimum Dry Density Dry Weight


nician a Tests of Specimen,
Mean, Standard Coefficient g
g/cm 3 Deviation, of Varia-
g/cm a tion, %

Toyoura sand tilting B 16 1.38 0.0146 1.06 353.1


B 16 1.37 0.0061 0.45 406.5
B 16 1.36 0.0072 0.53 484.3
B 16 1.36 0.0045 0.37 555.5
spoon A 17 1.349 0.0031 0.23
B 15 1. 356 9.0038 0.28
C 16 1.353 0.0C,52 0.46
Tokyo sand tilting A 16 1.23 0.0060 0.49 320.5
A 16 1.24 0.0067 0.54 386.1
A 16 1.24 0.0072 0.58 438.4
A 16 1.23 0.0077 0.63 481.0
B 16 1.23 0.0068 0.56 437.1
spoon A 19 1.202 0.0099 0.83
B 15 1.212 0.0114 0.94
C 18 1.210 0.0078 0.65

A -- skilled geologist; B = graduate student in engineering; C -- young woman with


no technical background.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
78 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

TABLE 4--Systsmatic errors in density measurements.

Cause of Errors A~//~, AT=axl h-~i=/ Ref Remarks


% ~.... % ~=~., %

Shock during handling + +2


Lack of experience -4- -t-1
Surcharge on soil --5 [2]
Insufficient acceleration below -- [1, 3]-
optimum
Excess acceleration above op- -5 [1, 3]
timum
Insufficient frequency below -- [3]
optimum vertical vibration
Excess frequency above opti- -- [3] with zero sur-
mum charge
Saturation (acceleration = -8 [1]
1.3 g)
Saturation (acceleration = -4-4 [1]
3.5 g)
Moisture -- [1]
Crushing of soil grains -t-
Segregation of soil grains

of soil specimens prior to measurements. Although the former, of course,


should be avoided, the latter m a y be unavoidable when undisturbed
specimens of loose cohesionless soils are involved.
As long as different methods are used for the determination of the
limiting densities, there m a y be systematic errors of considerable magni-
tude. Probable causes of the systematic errors are listed in Table 4, with
rough estimates of the relative deviations where data are available. The
systematic errors m a y be defined as deviations from the density values
determined with an "ideal" method which is yet to be established.

Influence of Random Errors in Dry Densities on the Relative


Density
According to statistics the standard deviation of the relative density,
SDr, can be expressed as a function of the standard deviations of the dry
densities, S ~ , S~min, and S~, as follows [6]:

(3)
\ 0~ /

F r o m Eqs 1 and 3 the following expression is derived:

v2D. = C 2~ v ~~ = + C ~~ v 2~ , ~ + C~v~ 2 (4)

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
YOSHIMI AND TOHNO ON STATISTICAL SIGNIFICANCE 79

in which
SDr S~ x S~min S~
v~.=--, v~,~,.= , v~=-- (5a)
Dr ' Tmax Train T

C .1,ma x ~ "Ymin
"~max -- "groin

,,,,,~,,x(1 -- D,) 1 -- D,
C,y,=,,, = (y .... -- ",/m,~,)D, = (C~,=,,,, -t- 1) D----~ (5b)

7"m~ C~,+ 1
C'z = (')'max - - ~ / m i n ) D r - - 1 = D, -- 1

The terms v's give the coefficient of variation, and C's may be called the
"error propagation" factors.
Equations 5b show that the error propagation factors for the minimum
density and specimen density, C~mi, and C~, increase as D, decreases until
they become infinite when Dr equals zero, whereas C ~ is independent
of D~. The error propagation factors are plotted in Fig. 1 against DT for
C ~ x = 4.7 which is selected as a representative value for clean sands
having low uniformity coefficients [4, 7].

50

tfl
25
n,,"
0
0
20
Z
0
V--
15
0
n,.
n

0 IO

% 20 40 60 80 IOO

RELATIVE DENSITY, D r , %

FIG. 1--Error propagation factors for relative density.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further
80 RELATIVE DENSITY INVOLVING COHESIONLES5 SOILS

In a special case involving equal coefficients of variation, that is, v~.= =


v~.i. = v~, Eq 4 yields

v~, = v~, = v~__z~= 0 = ~ / C ~ , . + C2~ml. + C~' (6)


~Tmsx VTmin V7
The value of C is computed for the case where C , . = = 4.7, and the result
is plotted in Fig. 2 against the relative density.

Required Number of Measurements for the Dry Densities


It can be seen in Fig. 2 that the coefficients of variation of the dry
densities must be limited to a small value if one wishes to attach quanti-
tative significance to the relative density of loose sands, as in the lique-
faction problem where the cyclic shear stress required to cause liquefaction
in saturated sand in a given number of stress cycles for a given initial
confining stress is reported to be proportional to the relative density [8, 10].
In order to reduce the coefficient of variation of the dry densities, one
must first eliminate any systematic errors by establishing rigidly standard-
ized procedures. That having been achieved, one can reduce the standard
deviation by increasing the number of measurements for each of the dry
densities according to the following relationship:
~. = ~/~/m (7)
in which a is the standard deviation of a population, m is the number of
measurements (or "specimen size"), and Sm denotes the standard deviation
of the mean of m measurements.
If the coefficient of variation of the relative density must be limited to

I00
80
60
~'mln ,- 4 7 --
40 - ~ rmox-r,.,, " -

20 i\ f i
|0
8
6
4
0 20 40 60 80 I00

RELATIVE DENSITY, Dr , %

FIG. 2--Rat/o of coeffwient of variation of relative density to that of dry density.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
YOSHIMI AND TOHNO ON STATISTICAL SIGNIRCANCE 81

40
v, = v,,=; v,.,,.
~'mln - 4 7
20 \\\\ - ~'m.x ~',.i. " -

,o \ \ \ \'\~
2"-<"'-4

m- 4 0 ~

20 40 60 80 IOO
RELATIVE DENSITY. Dr, %
FIG. 3--Number of measurements required for a given coe.ffwient of variation of relative
density.

v'D,, each of the coefficients of variation of the dry densities must be


limited to Vj),/C, provided that V~m~x= V~mln= V~. On the basis of Eqs 6
and 7, the number of measurements for each of the dry densities is given by
m = (r ~ (8)

The relationship of Eq 8 is evaluated for ~'mln/(Tm~ -- 7mi,) -----4.7 and


plotted in Fig. 3. For example, if D, = 50 percent, v'j), = 5 percent, and
v~ = 1 percent, seven tests each must be made for the maximum, minimum,
and specimen densities, and the mean value must be used for the determi-
nation of the relative density from Eq 1. For the same combination of
v'D, and v~, the required number of measurements will be two if D , = 85
percent, and 40 if Dr = 22 percent.
Influence of Systematic Errors in Dry Densities on Relative Density
There has been a considerable accumulation of information on the
relationship between mechanical properties of granular soils and the relative
density [9]. But difficulty arises when test data by different researchers
are compared, because different methods for the maximum and minimum
densities are often involved. If two different methods give different values
of the limiting densities for the same soil, the difference constitutes a
systematic error which cannot be treated statistically.
The relative deviation in the relative density, A D , / D , , due to a deviation
in any one of the dry densities, Aym~x, &Y~n, A% may be derived from

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to
82 RELATIVEDENSITYINVOLVINGCOHESIONLESSSOILS

Eq 1, and expressed as follows:


7min A7max
(9a)
Dr A7min_AT~ 0 7~,x - - 7rain Jr ATmax 7max

: _ ,,.,:,.

~,~---,,~o D,(Tm.x -- 7rain -- ATtain) 7rain

AD__rZ] = 7")'mi. A7 (9c)


D , J ~T~--~Tmi.--O (7 -- 7m~.)(7 + A7) 7
The preceding equations are reduced to the following form if A')'m,= and
hTmln are negligible compared t o 7ma~- 7rain, and if ~7 is negligible
compared to 7:
AD,] - - CT.,. ATm_,___~__ (10a)
A Tmin~ AT==0,ATmax((Tn~x.-- Train 7max

A'Tmin
(10b)
/)rJ A Tma~A 7~0 ,A Train(( Tmar.. Train 7min

~D._~__.rl "~ C 7 A....~.~ (10c)


/ ) , J ATn~x==ATmin~0 ,AT(<,}, 7
in which C's are the same as the error propagation factors of Eq 5b.

5o w-rT---
Dr " 50%
Dr = 3 0 %
I//
/

jail/ X;'o lt//


30 - - --/-Jl //~r'~
7 1 r )%

f! /// 0 s I00% /
20 --
t/I/,9
Dr =

/ i~-min- o
0
0 2 0 2 4 0 2 4

I Atilt ' % I A~ ) - m x " %

FIG. 4.--Relative deviation of relative density for "rmin/('Y~x -- ~min) = 4.7.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreemen
YOSHIMI AND TOHNO ON STATISTICAL SIGNIFICANCE 83

' ~ Dr .f~ 5~

5 h~. 0

-IO : \
-I0 -5 0 5 I0
A~'ma~max (%)

FIG. 5--Influence of systematic errors in limiti~l densities on relative density.

The relationship of Eq 9 is shown in Fig. 4 in full lines, and that of Eq


10 in dashed lines. It is evident in the figure that for small deviations in
the dry densities, say less than 1 percent, the full lines agree closely with
the dashed lines. Although the full lines diverge from the dashed lines at
greater relative deviations, Eq 10 can still give a correct order of magnitude
of 5Dr/D~. The numerical example of Eq 2 is checked by substituting
7mi, = 1.350 g/cm 3, 7m,x = 1.637 g/cm ~, D~ = 30.02 percent, and AT~i, =
0.0135 g/cm a into Eq 9b.
For the case in which both maximum and minimum densities are subject
to deviations while the deviation in the specimen density is ignored, the
relationship among hD,/D~, hT~x/Tm,~ , and ATmin/'Ymin is evaluated from
Eq 1 for D~ = 50 percent and C,m~. = 4.7, and the result is plotted in
Fig. 5, as a family of constant ~,D,/Dr. If ~7m=/Y~a~ = AYmi-/7~,
AD,/D~ is approximately 10 times the relative deviations of the limiting
densities. On the other hand, if Av~,~/vm~ = -~7~i,/7~i=, AD~/Dr is
small due to fortuitous cancellation of errors.
In view of relatively large values of ~ 7 ~ o r a~/min as indicated in Table
4, it is conceivable that ~ / D ~ may reach tens of percent. If, for example,
AD~/D~ = :i:20 percent, a relative density of 50 percent reported by two
researchers using different methods for 7m~ and 7~= may really represent
40 and 60 percent. Quantitative comparison of such relative densities
should be done with great caution.
Criteria of Limiting Density Tests
In view of the fact that the relative density is sensitive to small changes
in the limiting densities, it is essential that a rigidly standardized test
method be established for each of maximum density and the minimum

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licen
84 RELATIVEDENSITY INVOLVING COHESIONLES5 5OILS

density. I n selecting the standard methods, emphasis should be placed


more on minimizing dispersion t h a n on obtaining the absolute m a x i m u m
and m i n i m u m densities. The standard test should satisfy the following
conditions:
1. T h e test should be free from crushing of the soil particles, so t h a t the
test can be repeated on the identical specimen.
2. T h e density should be as insensitive as possible to variations in the
test conditions.
3. T h e test should be easily reproducible b y a technician of ordinary
skill.
T o ensure t h a t the technician has the required skill, it would be desirable
to establish a control test in which a standard value of limiting densities
are to be obtained on an identical specimen supplied from the same batch.

Conclusions
T o improve reliability of the relative density of granular soils, it is
essential t h a t rigidly standardized tests be established for the m a x i m u m
and m i n i m u m densities. For low relative densities, it is also necessary to
make sufficient n u m b e r of measurements for the limiting densities and the
specimen density, the required n u m b e r being determined on the basis of
statistics.

References
[I] Dobry, R. and Whitman, R. V., "Densification of Sand by Vertical Vibrations in
'Standard' Molds," Research Report R70-05, Soils Publication No. 251, Massa-
chusetts Institute of Technology, Cambridge, 1969.
[2] D'Appolonia, D. J. and D'Appolonia, E., "Determination of the Maximum Density
of Cohesionless Soils," Proceedings, 3rd Asian Regional Conference on Soil Mechan-
ics and Foundation Engineering, VoL 1, 1967, pp. 266-268.
[3] Selig, E. T., "Effect of Vibration on Density of Sand," Proceedings, 2nd Panameri-
can Conference on Soil Mechanics and Foundation Engineering, Vol. 1, 1969, pp.
129-144.
[4i] Hutchinson, B. and Townsend, D., "Some Grading-Density Relationship for Sands,"
Proceedings, 5th International Conference on Soil Mechanics and Foundation Engi-
neering, Vol. 1, 1961, pp. 159-163.
[5] Kolbuszewski, J. J., "An Experimental Study of the Maximum and Minimum
Porosities of Sands," Proceedings, 2nd International Conference on SOil Mechanics
and Foundation Engineering, Vol. 1, 1948, pp. 158-165.
[6] Worthing, A. G. and Geffner, J., Treatment of Experimental Data, Wiley, New York,
1943, pp. 205-214.
[7] Whitman, R. V., "Hydraulic Fills to Support Structural Loads," Journal, Soil
Mechanics and Foundation Division, American Society of Civil Engineers, Vol. 96,
No. SM 1, 1970, pp. 23-47.
[8] Lee, K. L. and Seed, H. B., "Cyclic Stress Conditions Causing Liquefaction of
Sand," Journal, Soil Mechanics and Foundation Division, American Society of
Civil Engineers, Vol. 93, No. 1, 1967, pp. 47-70.
[9] Lambe, T. W. and Whitman, R. V., Soil Mechanics, Wiley New York, 1969.
[10] Lee, K. L. and Singh, Awtar, "Relative Density and Relative Compaction,"
Journal, Soil Mechanics and Foundation Division, American Society of Civil Engi-
neers, VoL 97, No. SM7, 1971, pp. 1049-1052.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
R. C. Gupta, and J. D. McKeown ~

Effect of Variations in Minimum Density on


Relative Density

R E F E R E N C E : Gupta, R. C. and McKeown, J. D., "Effect o f V a r i a t i o n s i n


M i n i m u m D e n s i t y o n R e l a t i v e D e n s i t y , " Evaluation of Relative Density and
Its Role in Geotechnical Projects Involving Cohesionless Soils, A,~TM S T P 5~3,
American Society for Testing and Materials, 1973, pp. 85-97.
ABSTRACT: This paper describes the results of an investigation carried out
to study the effect of variations in minimum density on relative density test
results. The material selected for the investigation was obtained during con-
struction of Kettle Generating Station. The test results were statistically
analysed. For the materials tested, the influence of variations in minimum
density on relative density results is startling. This creates a dilemma for ef-
fective quality control in the field in terms of enforcing the requirements of
design as spelled out in a contract specification based on percentage relative
density.
K E Y W O R D S : cohesionless soils, density (mass/volume), tests, earth fills,
soil compacting, statistical analysis, construction control

Relative density is used quite extensively for control of compaction of


cohesionless soils in earth fill structures. To utilize the method, maximum
and minimum densities are obtained as standards, and relative density is
calculated as follows:
~d msx('Yd -- "Yd rain)
Rd = X 100
'~d('~d max -- "Yd rain)
where
R d = relative density, percent,
~/d -----in-place density, l b / f t s,
~d =~ --- maximum density, lb/ft 3, and
3'd rain = minimum density, lb/ftL
Geotechnical engineer, Construction Division, Manitoba Hydro, Winnipeg, Man.,
Canada.
2 Soils and instrumentation supervisor, Kettle Generating Station, Manitoba Hydro,
Gillam, Man., Canada.

85

Copyright by ASTM Int'l


Copyright9 (all rights
by ASTM reserved); Fri www.astm.org
International Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreemen
86 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

At the Kettle Generating Station being built in Northern Manitoba


and nearing completion, the contract specifications required that the semi-
pervious materials should be compacted to obtain acceptable densities as
determined according to ASTM Test for Relative Density of Cohesionless
Soils (D 2049-69).
During the field control of compaction it was possible to correlate the
field density test results when expressed as a percentage of maximum
vibrated density to the movement of the compaction equipment. However,
it was virtually impossible to visually detect any direct relationship be-
tween changes in relative density to changes in field compactive effort. In
such cases, considerable time was spent on the interpretation of relative
density data, particularly where these data were slightly below or border-
lining the specification limits. In accordance with the design requirements
for quality control, values of relative density were computed and compiled
statistically. However, in actual practice the day to day construction
control reverted to the percentage compaction method, whereby the field
density test value was compared to and expressed as a percent of the
laboratory maximum vibrated density for the material.
Because of the irregularities in the relative density test results, a program
was commenced to investigate the possible sources of error, and the effect
it had on the field application of this test method. Variations affecting this
standard method could be attributed to one or more of the following:

1. the possible error in the minimum density test procedure,


2. the possible error in the maximum density test procedure,
3. the possible error in the field density procedure, the apparatus and
the technique for using them in the field,
4. the effect of variation in the grain size distributions, moisture content
and climatic conditions on the density of the in-place materials, and
5. the human factor.

It was decided to investigate initially the minimum density test pro-


cedure. This paper describes the influence of variations in minimum density
on the relative density values. The results of the maximum density program
are in a study stage and may be reported in the future.

Materials

The semipervious materials used in the earth fill construction were


obtained from two kame-eskers, and as such consisted predominantly of
rounded to subrounded sorted sands and gravel [1]3. The classification of
the materials according to the Unified Soils Classification System was
(SP-SW). The quality of these materials were considered excellent for

s The italic numbers in brackets refer to the list of references appended to this paper.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agree
GUPTA AND MCKEOWN ON VARIATIONS IN MINIMUM DENSITY 87

I SIEVE ANALYSIS .... ]


s~No I oR~v~L I
Fint I M~lil/m ~ Coorse I Fin 9 'i Cgoqe I
_200 iO0 50 40 50 16 IO s 4 3/8 5/4 i q2 3

~ I/!l .... 1
60--

,o
:t " '39 " -- 40 Z
"I -

- ! ,2 ..... I .... ~o ,-i


' 4C
/
L;iW/ Cl I C l
LiMiTsl
I
A/ : V / I I
Z
/ ' SAMPLs I
I -1ii1 1
"
m
i,M
o
........ iU

z o # l / !~rl l~j, ft , _ / 139 / 7.0 I I.I I


-! : Z f l Vt" - 1-~ r ,,., I o , I '~
/ ~-.:r I / i 3~/7. 1 Io 1

: i , II l ,oo/o
90 . . . . . . . -i- . 4,,,,r .l ~ + - .---~l-.,,,IG[l~----l-~ I0
I/t" I i ,,T~,,, //~'7 I /'1"1
80 . . . . . . @ ~
, 1 t 1i ----
2o
i I i I 9925- -
I ~2~
To - - T ao

~ 6 o ~ - - 4o z

o: 40 . . . . . . . . . I z
~' f ', I i/ix ~ ,,/// ,---, c c: '~ -~
~o---/-~ - h - ? i"/V/--- - - ' >'---= . . . . . . . . . . . . ,o
-/i .o,
708
.I oI.O o.
1.3
.oT1 ,~. ;.Uy .., ... ,.. ,o

o~ / " 1 I ,oo
074 i49 297* 59 119 236 476 952 191 38.1 762
D t A M E T E R OF P A R T I C L E S IN M I L L I M E T E R S

F I G . 1--Grain size curves of samples tested.

earthworks construction. The petrographic analysis carried out on ma-


terials obtained from the esker revealed the percent predominant rock and
mineral types to be as follows: limestone 45, granite 21, quart~ 15, basic
igneous 12, and feldspars 5.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
T A B L E 1--Minimum density, statistical summary.

Laboratory Technician Number Computed Limits for s (95 Standard Coefficient Range (90% Limits), Number of
Sample Nos. of Tests N Averages, 2, Percent Confidence Deviation, ~, of Variation lb/ft3 Tests for 1%
lb/ft3 Limits), s 4- l b / f t 3 lb/fts v, % Variation

139 1 10 114.49 114.494-0.82 1.081 0.94 112.99-116.21 8 -4


-<
2 10 113.60 113.60-I-0.64 0.853 0.75 111.83-114.86 5
3 10 115.26 115.264-0.50 0.666 0.58 114.40-116.57 3
4 10 113.20 113.204-1.07 1.426 1.26 111.03-114.53 14 r
All 40 114.14 114.144-0.42 1.290 1.13 (111.50-115.80) ~ 11 <
z
352 1 10 111.55 111.554-0.35 0.463 0.41 110.91-112.29 1 o
2 10 110.97 110.974-0.35 0.468 0.42 110.08-111.83 1
3 10 109.66 109.664-0.42 0.555 0.51 108.96-110.65 2
8
"I"

4 10 110.55 110.55=1=0.66 0.872 0.79 109.20-111.72 5


All 40 110.69 110.694-0.29 0.913 0.82 (109.10-112.00) ~ 6 o
z
483 1 10 120.84 120.844-0.31 0.409 0.34 119.87-121.28 1 o~
2 10 118.77 118.774-0.62 0.823 0.69 116.94-119.72 4
3 10 120.39 120.394-0.56 0.741 0.62 119.24-121.21 3 o_
4 10 120.48 120.484-0.56 0.744 0.62 119.17-121.50 3
All 40 120.12 120.124-0.34 1.048 0.87 (118.00-121.30)~ 6
502 1 10 117.31 117.314-1.26 1.678 1.43 115.01-120.20 18
2 10 115.61 115.614-1.13 1.503 1.30 112.60-117.75 15
3 10 117.81 117.814-0.94 1.243 1.05 115.29-119.57 10
4 10 115.93 115.934-0.89 1.178 1.02 114.13-117.86 9
All 40 116.66 116.664-0.53 1.648 1.41 (114.30-119.50) ~ 17
9751 1 10 110.50 110.504-2.52 3.347 3.03 105.54-114.46 82
2 10 111.45 111.454-1.49 1.977 1.77 107.56-113.26 28
3 10 110.78 110.784-2.02 2.682 2.42 106.11-113.02 52
4 10 108.91 108.914-3.25 4.313 3.96 102.45-113.15 141
All 40 110.41 110.41 4 - 1 . 0 4 3.214 2.91 (103.30-I 13.80) 9 76

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
9873 1 10 110.13 110.132:2.77 3.675 3.34 104.00-113.69 100
2 10 111.26 111.26::i:1.43 1.897 1.70 107.49-113.15 26
3 10 111.23 111.232:2.75 3.651 3.28 104.07-116.04 96
4 10 109.22 109.22::1::4.00 5.305 4.86 101.95-115.45 212
All 40 110.46 110.462:1.22 3.776 3.42 (102.70-115.00) 9 105
9925 1 10 115.94 115.942:1.39 1.841 1.59 112.99-118.56 22
2 10 115.33 115.332:1.63 2.161 1.87 110.96-118.08 31
3 10 115.74 115.742:1.13 1.501 1.30 112.47-117.57 15
4 10 113.75 113.752:1.71 2.265 1.99 109.86-116.76 35 O
All 40 115.19 115.192:0.67 2.079 1.80 (110.50-118.00)~ 29
129 1 10 114.64 114.642:4--1.90 2.526 2.20 110.39-116.98 43
z
2 10 115.49 115.492:4-1.25 1.662 1.44 112.58-117.64 18
3 10 115.28 115.282:1.85 2.459 2.13 111.28-117.35 40 3:
4 10 112.36 112.362:4-4.34 5.759 5.13 103.70-117,86 236
All 40 114.44 114.442:1.15 3.570 3.12 (104.40-117.60)~ 87
345 1 10 116.80 116.802:0.56 0.748 0.64 115.95-118.38 3 z
2 10 115.94 115.942:+-0.58 0.770 0.66 114.97-117.31 3 0
3 10 116.37 116.372:0.54 0.719 0.62 115.45-117.49 3 z
4 10 116.61 116.612:0.37 0.486 0.42 115.78-117.31 1
All 40 116.43 116.432:::0.24 0.738 0.63 (115.25-117.50) ~ 3
708 1 10 117.17 117.172:0.68 0.901 0.77 115.91-118.84 5 o
2 10 115.79 115.792:0.75 0.995 0.86 114.79-117.97 6 z
3 10 118.79 118.792:4-0.95 1.264 1.06 115.80-120.29 10
4 10 118.16 118.162:0.84 1.109 0.94 116.32-119.46 7
All 40 117.48 117.482:0.50 1.544 1.31 (114.90-120.00)" 15
C
3:
Represents the middle 90 percent spread of test d a t a from frequency distribution curves, Fig. 2.

',O

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
90 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

LeD Sample No. 545 Lob. Sample No. 708

, F-H
w o 8 L--H
7 , II
?'e II
~ Ns #11
f. 4
. i
.1~ tl 3 II
o ii 2
I I

IOQ

m
9 o
o
o

z ~.

o
I-

114 I|ll ItS I17 lie il4 Ill lie I~O


DRY DENS|TY (P.C-F) DRY D F N S I T ~ 1 6 2( P . C , F . )

Lob. Sample No. 9873


IIII I I. I. I. I. I.
III I I
"I'll I iiiii i
14 [ I
II
~'il I !i!ii i
i, illl : ..... Frequency Histogram
lit, ~ i~i 1 ~(Columns erected on
iollll IIEI ~
coil boundaries )
. s l'',l.l ''~
iIi i
o[Ill II ,I . I.I , , I I

~
LI J I
"|;=';,,:,,
' ' ,~I=' ,=,
,,~,,, 9
N=NumPer o f t e s t s with Test r e s u l t s in range
alga 9 I 9 , - ~ inn
Ill 9 " Ill too corresponding to square width
Graph below histogram i$ O c u m u l a t i v e
frequency curve of the histogram
SO

SO SO
NOTE:
The test results shown are the minimum
,o ~ densities obtained in the study.

9 "~
iOa 104 lOs 112 lie
DRY D[NSITY (PC F.)

FIG. 2--Typical histogram and cumulative frequency curves.


M i n i m u m Density Investigation
Material for the minimum density investigation was selected from
quality control storage specimens during earthworks construction. Com-
parative testing on ten specimens taken at random indicated the material

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lice
GUPTA AND MCKEOWN ON VARIATIONS IN MINIMUM DENSITY 91

to be fairly representative of the fill placed during construction. Figure 1


presents the grain size distribution and the specification limits for semi-
pervious materials. Also shown on the figure are the correspondirig coeffi-
cients for uniformity (Cu) and coefficients of curvature (Cc).
The minimum density tests were performed in accordance with ASTM
D 2049-69. To provide for possible human error, four technicians carried
out ten tests on each specimen. In this way a universe of 40 test results
was created for each specimen.
The minimum density test results were statistically analysed [2, 3] using
a computer program for each of the ten tests and the universe of 40 tests
to obtain the computed averages (~), limits for 2 for 95 percent confidence
level, standard deviation (z), coefficient of variation (v), range, and number
of tests required for =t=0.5percent variation. The results of the analysis are
shown in Table 1. The frequency histograms and accumulated frequency
distribution curves for each specimen for a universe of 40 test results were
plotted. Figure 2 shows such plots for laboratory specimens 345, 708, and
9873. Subsequent analysis will indicate that these specimens, in the above
order, represent the best, average, and worst in-test variations in terms of
the spread in the minimum density test results.

Analysis of Data
For the purposes of this analysis the data from four technicians was
pooled to create a universe of 40 test results. In some cases the data did not
meet the statistical tests for homogeneity at a specified level. An inference
about human error could be made in such cases. However, in practice
only one minimum density test is performed to establish the relative
density. The one result, as such, could fall anywhere in the range. In the
analysis of the test results only those within the middle 90 percent spread
were used.
The relative degree of accuracy of each technician is presented in Table 2.
The comparison was obtained by selecting the largest and smallest standard
deviation values recorded within each specimen group.
The statistical analysis of the data consisting of 40 tests is based on
certain assumptions and qualifications.
To make practical use of this investigation, it was necessary to know the

TABLE 2--Minimum density, technician accuracy comparison.

Technician Most Accurate Least Accurate PercentInfluence

1 3 ... 15
2 3 15
3 3 '4" 35
4 1 6 35

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreeme
9'~ RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

required number of tests it would take to obtain a desired degree of ac-


curacy. The number of tests was calculated in accordance with the recom-
mended practice detailed in ASTM Practice for Choice of Sample Size to
Estimate the Average Quality of a Lot or Process (E 122-58) [3]. This
value can be considered as an amplification of the standard deviation for
any group of data and is an ideal parameter by which other variabilities
can be correlated.
It is generally believed that some variation in relative density test
result data can be attributed to the grain size distribution characteristics
that exist between different materials. A comparison of the values of
coefficient of uniformity and coefficient of curvature for the materials
tested and the number of tests required to obtain a desired degree of
accuracy of one percent, however, did not show any such definite relation-
ship. Furthermore, it has also been suggested elsewhere [4] that in moisture-
density relationships, maximum vibrated density tests, or unit weight
measurements, there is a direct relationship between the resulting density
value and the percentage greater than the No. 4 screen size. It is also of
interest to note that such a relationship was not clearly evident in the
case of minimum density.
Without having made an examination of the possible error in each of
the procedures previously mentioned, a reasonable estimate was believed
to be • percent (1 percent range). When equated however, the •
percent resulted in relative density value outside the desirable limits. For
example, assuming that +0.5 percent error was added to a field density
that corresponds to 50 percent relative density and compared to a - 0 . 5
percent error in maximum and minimum density, the relative density
would be equal to 55 percent. The inverse of this situation would equal
45 percent. Thus, a 50 percent relative density could range from 45 to 55,
or • 10 percent error. The specified requirement for compaction at Kettle
Generating Station was 75 percent relative density. The range for this
value resulting from the previous assumptions would be from 70 to 79,
or • percent error. Both examples are calculated with typical maximum
and minimum density values obtained from this investigation.
Three specimens were selected to illustrate the best, average, and worst
in-test variations obtained from this program and are shown in Fig. 3.
The relative density plots show the influence that the 90 percent limits of
minimum density test values have on relative density for a universe of 40
tests. This is for the condition that maximum vibrated density and field
density remain constant. The values given were calculated using the actual
maximum density for the particular material.
Figure 3(a) shows the results for the best case for Specimen 345. A 50
percent relative density value could, in fact, vary from 47 to 52, or •
percent error. At 75 percent relative density this variation would be from
74 to 76, or • percent. The results for an average case, Specimen 708,

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
GUPTA AND MCKEOWN ON VARIATIONS IN MINIMUM DENSITY 93
Lab Sample NOI 3 4 5
I~0 1gO

ii

~O>0 ~0

z~

l 47 ~o 5z *

o [ I J
TEC~ I"[CH.Z TECH.~* TECH 4 aLL TECH. Z5 50 7~ IOO
%Rd
(o) (Actual)

Lob Sample NO 708


~" ,2o ioo

E
./
:E
I

~10 ~ 5o
r
o
o
i
Z5

~ ,
TECH. i T [ c ~ 12 TECH 3 TECH4 ALL TECH O Z5 sO ?~ I0o
% Rd
(b) (Actuot)

Lab Sample NO. 9 8 7 3

I
I I I ! ' ,oo ] o: /jj~
//

i
15
;~
i I
TECH. I TECH.2 TECH.3 TECH 4 ALL TECH.
%Rd
LEGEND (C) (Actual}

The rsghf hand graphs show influence treat tt~e m e a s u r e d


Lorgelt Tett U 90%Lrmil$
error ~n m i n i m u m density test has upon relative density

where maximum density and field density remain constant.


Comp~lld ~ O,vio~o.s~
A,,toCe, ~ ~6 ~No of Tes,*
Rlq'a 1%~
Smole~ *it Acc.racT
Colculoled from Upper ofld Lower 90 P I t { i n t Ll~itl
TECH, I , 2 ' e l r ALL Ts

FIG. 3--Effects of variations in minimum density on relative density.

are shown in Fig. 3(b). A 50 percent relative density value could, in fact,
vary from 41 to 57, or i 1 6 percent. At 75 percent relative density this
variation would be from 71 to 79, or =l=5.7 percent error. The worst case,
Specimen 9873, is presented in Fig. 3(c). A 50 percent relative density
valne varies from 38 to 64, or • percent. At 75 percent relative density

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Ag
94 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

this variation is from 69 to 80, or • percent. I t must be pointed out


t h a t for Specimen 9873, the relative density border limits are not equi-
distant from the center. This difference is attributed to the skewness in
the normal distribution of the test result data where the lower 90 percent
limit is further away from the median t h a n the upper limit.
The preceding comments have been made on the basis of the analysis
carried out on the test results ~4thin the middle 90 percent spread in a
universe of 40 results. If consideration was given to the range, the largest
and smallest test data recorded b y any one technician in a set of ten tests,
the error in relative density would be greatly increased.
Furthermore, it was interesting to note t h a t the lowest value of the
m i n i m u m density was obtained in the first test trial for 20 out of 40 speci-
men groups. This will definitely influence the field control based on only
one test. The authors, however, have no reasonable explanation for this
phenomenon at the present time.

Maximum Density
M a x i m u m density tests were performed on the materials used in the
minimum density program according to A S T M D 2049-69. The resulting
values as well as the corresponding coefficients of uniformity and coeffi-
cients of curvature are given in Table 3.
I t is interesting to note t h a t Specimen 345 achieved the greatest maxi-
m u m density value, and in the minimum density investigations, the same
material resulted in the lowest percentage error. I t can be seen from the
Table 3 t h a t b y and large the values of coefficient of uniformity show an
increasing trend with increasing m a x i m u m densities. Such a relationship
was not apparent in the case of minimum density program. I n the case of
coefficient of curvature values, no definite relationship was apparent for
increasing m a x i m u m density.

TABLE 3--Maximum density and values of Cu and Cc.

Laboratory Sample No. Maximum Density, Cu Cc


lb/ft a

352 128.0 7.1 1.1


139 131.5 7.0 1.0
708 136.0 24.4 O. 8
483 138.2 11.0 1.3
9873 138.2 20.0 0.6
502 140.0 14.4 1.1
9925 140.0 21.8 1.3
9751 141.0 19.5 1.2
129 141.3 21.8 1.5
345 145.0 22.6 1.3

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
GUPTA AND MCKEOWN ON VARIATIONS IN MINIMUM DENSITY 95

M0xl mum Oensity (p~c.f)


. oo o- o-?. o_o o o o~ j
o - m t ~ --

t
!

i :-
...1 o - -
o o o o
. . _o ~" :~._,~o o
= "

;o o
=

1
Moximum Density (~cJ) ~'~
.~ o o_ or#_ ~ : Oo - ~ ;

, ..... -.

I II 1@~741Vf
IIIl- .lllllllllll,
IIl[]llllll,
o., ~ ~ ." .3

[I IIIIIIl,ll~lllilll[llll, :
l llllli IIIIII,IFllllllllllll, .} |

I II lllJ'Li"l]
L!LSll~,ilt~: !i~;
Moxlmum Density (p.c.f,) | |174174

[ ]l I t],lll]l],L[ti~1t: 0

II I L[ I I IIIIIIIIILr
il I
I II 11 IIIlllllltt,I~llllill~
II 1,1 t ', l ', I','~' l~l ;111111~
~L

tlil~
I ~-~TL I II1 'll Ilii!
i I~'l'Ii I ill
I l {,t']/I [ I [ I i]ll IIIl]ii~
(l'~'d) ~ ! s u e 0 w n w ! u !

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreemen
96 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

It may be of interest to mention here that during the earlier stages of


the program, pilot tests were carried out for determining maximum density.
Because of the particle breakdown resulting in each progressive test,
progressively higher values of maximum density were obtained upon
reusing the material.
Field Density
The field densities at Kettle Generating Station were obtained by using
the Washington Densometer and the suggested ASTM procedure [5] for
this equipment.
It has been reported elsewhere [6] that there may be an error of close to
• percent Of the volume in the Washington Densometer procedure. It
must be pointed out that this error has been reported for tests carried
out under laboratory conditions. Figure 4 shows the influence or effect
that the variations in minimum density have on relative density combined
with a possible error in the field density measurement. The presentation is
self-explanatory. Figure 4(a) shows the results for the best case for Speci-
men 345. A 50 percent relative density value, in fact, could vary from 32
to 65, or -4-33 percent error. The result for the average case, Specimen 708,
and for the worst case, Specimen 9873, are shown in Figs. 4(b) and (c),
respectively.
It may be pointed out that the results from Specimen 9873 were termed
as the worst case because of the spread in minimum density test results.
If the combined error due to field density and minimum density results
taken together is compared with the results from Specimen 708, the total
percentage variation is higher in the case of Specimen 708. This is attributed
to the fact that the average minimum density in the case of Specimen 9873
is lower than that obtained for Specimen 708. The lower average value of
the minimum density, as obtained in the case of Specimen 9873, is con-
sidered closer to the actual value. Under these conditions the trend ob-
served in Fig. 4(c) is explainable.
Figure 4 furthermore does not take into account any error which, may
or may not exist in the maximum density determination. Any such error
is likely to amplify the spread in the values of relative density.

Conclusions
In terms of statistical significance, the effects of variations in minimum
density on relative density are startling. Although this variation will
decrease at increasing values of relative density, it still creates a dilemma
for effective quality control in the field in terms of enforcing the require-
ments of design as spelled out in a contract specification. A problem
definitely exists if the case in question is that of a contract specification
asking for lower relative densities, say in the range of 50 to 75 percent.
Seemingly, there is an urgent need for establishing an effective criterion

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lic
GUPTA AND MCKEOWN ON VARIATIONS IN MINIMUM DENSITY 97

in view of the extreme sensitivity of the minimum density test values.


I t is considered that the minimum density test results are difficult to
reproduce.
The tests described herein were conducted in an on-site laboratory, and,
therefore, were not performed under the strict control normally associated
with research work; however, conditions were Certainly typical of normal
field construction control--the one which is most likely to be encountered
in practice.
It may also be pointed out that the results are based on tests carried out
on a specific material. The statistical analysis was performed under certain
assumptions. In addition, such things as compaction equipment, field
moisture content, grain size distribution, climatic conditions, and test
procedure are some of the variables which may have a bearing on the
measured value of relative density. Consequently, caution must be exer-
cised in using the data.

Acknowledgments
The research program reported herein was carried out in an on-site
laboratory under the supervision of the second author. The authors acknowl-
edge the encouragement and cooperation received during the program
from W. H. Kasperski, resident manager, Kettle Generating Station and
A. Koropatnick, geotechnical manager, Construction Division. The authors
are thankful to K. J. Fallis, director of the Construction Division for the
permission to publish the results.

References
[1] Flint, R. F., Glacial and Pleistocene Geology, Wiley, New York, 1969.
[2] Neville, A. M. and Kennedy, J. B., Basic Statistical Methods for Engineers and Scien-
tists, International Textbook Company, New York, 1964.
[3] Manual on Quality Control of Materials, A S T M S T P 15-C, American Society for
Testing and Materials, 1951.
[4] Burmister, D. M. in Field Testing of Soils, A S T M S T P 322, American Society for
Testing and Materials, 1963, pp. 67-97.
[5] Humphres, H. W. in Special Procedures for Testing Soil and Rock for Engineering
Purposes, A S T M S T P ~79, American Society for Testing and Materials, 1964, pp.
451-457.
[6] "A Study of In-Place Density Determinations for Soils," TM No. 3-415, U. S. Army
Corps of Engineers, Waterways Epxeriment Station, Vicksburg, Miss., 1955.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No furt
7'. L. Youd~

Factors Controlling Maximum and Minimum


Densities of Sands*

REFERENCE: Youd, T. L., "Factors C o n t r o l l i n g M a x i m u m a n d M i n i -


m u m D e n s i t i e s o f S a n d s , " Evaluation of Relative Density and Its Role in Geo-
technical Projects Involving Cohesionless Soils, ASTM S T P 5~8, American
Society for Testing and Materials, 1973, pp. 98-112.
A B S T R A C T : Maximum and minimum density tests, conducted on a variety
of clean sands, show that the minimum and maximum void-ratio limits are
controlled primarily by particle shape, particle size range, and variances in
the gradational-curve shape, and that the effect of particle size is negligible.
Curves were developed for estimating minimum and maximum void ratios from
gradational and particle-shape parameters. Estimates for several natural and
commercially graded sands agree well with minimum and maximum void
ratios measured in the laboratory. Minimum densities (maximum void ratios)
Were determined by the standard American Society for Testing and Materials
(ASTM) minimum density test method (Test for Relative Density of Cohesion-
less Soils (D 2049-69)), except that smaller molds were used. Maximum den-
sities (minimum void ratios) were determined by repeated straining in simple
shear, a method which has been shown to give greater densities than standard
vibratory methods.
K E Y W O R D S : cohesionless soils, compacting, density (mass/volume), grain
structure, grain size, sands, soil mechanics.

T h e determination of relative density values for sands would be g r e a t l y


facilitated if reliable estimates of m a x i m u m and m i n i m u m densities could
be m a d e f r o m soil index properties r a t h e r t h a n f r o m elaborate l a b o r a t o r y
tests. Criteria presently available for estimating m a x i m u m a n d m i n i m u m
densities are i n a d e q u a t e because, (1) t h e y are too imprecise, (2) i m p o r t a n t
p a r a m e t e r s are n o t a d e q u a t e l y considered, and (3) the l a b o r a t o r y tests
t h a t were used to develop the criteria do n o t provide an a d e q u a t e measure
of m a x i m u m density.

* Publication authorized by the Director, U. S. Geological Survey.


1 Research civil engineer, Engineering Geology Branch, U. S. Geological Survey,
Menlo Park, Calif. 94025.

98

Copyright by ASTM Int'l


Copyright9 by(all
ASTMrightsInternational
reserved); Fri Mar 11 16:13:06 EST 2016
www.astm.org
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No f
YOUD ON FACTORS CONTROLLING DENSITIES OF SANDS 99

This report describes which soil properties control maximum and mini-
mum density limits and, on the basis of these properties, develops gener-
alized criteria for estimating maximum and minimum densities of clean
sands.

Previous Work
Burmister [1-3] 2 reported that the most important factor controlling
maximum and minimum densities is the range of particle sizes--the greater
the range of particle sizes, the greater the density. Other factors he con-
sidered important were the type of grading curve, particle size, and particle
shape. In Burmister's study, maximum density values were obtained either
by vibrating-tamper or impact procedures, and minimum density values
by pouring sand through a funnel into a mold [4]-
The application of curves developed by Burmister for estimating maxi-
mum and minimum densities is limited because, (1) the methods used to
obtain maximum density values do not provide an adequate measure of
maximum density [5] and, (2) the single-value density correction he
specifies for soils with more than 35 percent angular fragments or crushed
materials is too arbitrary and too imprecise to account adequately for the
influence of particle shape.
Kolbuszewski and Frederick [6] demonstrated that the density limits of
sands increase with increasing particle size and decrease with increasing
angularity; however, insufficient data were given in their report to formu-
late criteria for estimating density limits for sands in general. They ob-
tained minimum densities by the tipping method described by Kolbuszewski
[7] and maximum densities by depositing sand in a vacuum.
Maximum and minimum density values for a number of sands from
Ontario, Canada, were statistically correlated with Bagnold grading pa-
rameters [8] by Hutchison and Townsend [9]. A reasonably good correlation
was found for maximum density, but only a poor correlation was obtained
between minimum density and the Bagnold parameters. Maximum density
values were obtained by the vibrating-tamper method [7], and minimum
densities were obtained by both the tipping method [7] and the spooning
method described by Wu [10].
The application of Hutchison and Townsend's data is limited because
the compaction method used to obtain maximum density values does not
provide an adequate measure of maximum density [5], Bagnold grading
parameters are not definable for all sands, and the influence of grain shape
was not considered.
Kabai [11] showed that there was a good correlation between maximum
and minimum densities determined in the laboratory and the coefficient
of uniformity for Danube River sands. Minimum densities were determined

2 The italic numbers in brackets refer to the list of referencesappended to this paper.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
100 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

by pouring sand through a funnel into a mold, and maximum densities


were obtained by impact-compaction tests, using procedures similar to
those outlined in the ASTM Test for Moisture-Density Relations of Soils
Using 10-Lb. Rammer and 18-In. Drop (D 1557-70).
The value of Kabai's criteria is limited because the impact-compaction
test does not generally give maximum density values for sands that are
as high as those given by vibratory or shear-straining methods [5, 12, 18],
and the influence of particle shape was not considered.

Laboratory Investigation
In the first part of the laboratory investigation, maximum and minimum
density tests were conducted on sieved fractions and artificially propor-
tioned gradations of commercially available sands in order to confirm and
extend previous findings concerning the factors that control maximum and
minimum densities and to construct a generalized set of curves for esti-
mating these density limits.
In the second part of the laboratory study, the maximum and minimum
densities of a variety of natural and commercially graded sands were de-
termined in order to test the validity of the generalized prediction curves.

Procedure
Minimum densities were determined by the procedures outlined in the
ASTM Test for Relative Density of Cohesionless Soils (D 2049-69), except
that a 948-cm 3 (0.033-fP) mold was used rather than the standardized
2832-cm ~ (0.10-fP) mold. (Because particle sizes were very small compared
to the mold size and because the technique used gave values consistent
with minimum densities obtained by the standard method on two of the
sands tested, the influence of the smaller mold size is believed to be very
small.) Minimum density is sensitive to the procedures used to determine
it [12, 14]; however, because the ASTM standardized procedures give
consistent and generally lower values than other methods [5], they were
used in this study.
Maximum densities were determined by the repeated straining in simple
shear procedure described previously by Youd [13, 15]; this procedure
was shown to give greater densities than the procedures outlined in the
ASTM D 2049-69. The procedures outlined by Youd [13] were followed
for sands that were highly resistant to crushing and included the application
of 10 000 cycles of shear strain (2 to 5 percent strain) to specimens con-
tained in a simple shear device. The normal stress applied was 9.6 • 104
N / m 2 (2000 psf).
For specimens that were less resistant to crushing, the procedure was
modified to reduce crushing by reducing the normal stress to 4.8 X 104
N / m 2 (1000 psf) and the number of cycles of strain to 4000. The modified
procedure held crushing to what was considered to be an acceptable mini-

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licen
YOUD ON FACTORS CONTROLLING DENSITIESOF SANDS 101

mum. I n all specimens the increase of particles passing the No. 200 sieve,
which was found to be a good index of crushing, was held to less t h a n 1.5
percent of the total specimen weight, and the m a x i m u m increase of particles
passing any sieve was generally less t h a n 2 percent. Tests on crushing-
resistant specimens b y the modified procedure yielded a relative density
of at least 98 percent; thus, any error due to incomplete compaction was
small and would have been offset in p a r t b y the small increase in density
due to crushing.

Soil Properties

The parameters used to describe the physical properties of the sands


included the mean particle size (ds0), the coefficient of uniformity (Ca),
the type of grading curve, the particle roundness (R), and the specific
gravity (G,),
where
ds0 = the mesh size through which 50 percent of the particles, b y weight,
passes,
Cu = d6o/dlo, where d60 and dl0 are the mesh sizes through which 60
percent and 10 percent of the sand particles pass, respectively, and
R = the ratio of the average of the radii of the corners of a sand grain
image to the radius of the m a x i m u m circle t h a t can be inscribed
within the grain image [16].
The type of grain-size distribution curve was described qualitatively in

TABLE 1--Roundness criteria and values.

Roundness Description b Roundness Mean


Classa Intervals Roundness ~

Very angular Particles with unworn fractured surfaces and 0.12-0.17 0.14
multiple sharp corners and edges
Angular Particles with sharp corners and approxi- 0.17-0.25 0.21
mately prismoidal or tetrahedral shapes
Subangular Particles with distinct but blunted or slightly 0.25-0.35 0.30
rounded corners and edges
Subrounded Particles with distinct but well-rounded edges 0.35-0.49 0.41
and corners
Rounded Irregularly shaped rounded particles with no 0.49-0.70 0.59
distinct corners or edges
Well rounded Smooth nearly spherical or ellipsoidal par- 0.70-1.00 0.84
ticles
a After Powers [17].
b Descriptions represent the classification criteria used in this study and are not based
on a recalculation of Wadell roundness .values.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
102 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

terms of the statistical characteristics of the curve, for example, normally


distributed, skewed, bimodal, or gap graded, etc. These soil properties
were chosen as descriptors because of their common usage, and because
each is uniquely definable for any given sand specimen.
The R values were determined by the procedure suggested by Powers
[17] but modified to consider particles from each size fraction rather than
the specimen as a unit. Briefly, the procedures used for this study consisted
of examining under a microscope a representative number of sand particles
(at least 50) from each sieve fraction and visually assigning each particle
to a category listed in Table 1 (also modified from Powers). Figure 1
shows typical particles assigned to each category. The average R value
for the sieve fraction (R j), was then calculated from the equation R~. =
(ZRi)/n, where Ri is the roundness value from Table 1 assigned to particle
i, and n is the number of particles examined. The average value for the
sand specimen was then calculated from the relation R = (F,PjR~)/IO0,
where Pi is the percent (by weight) of particles retained in the applicable
sieve fraction.

FIG. 1--Particle shape classes.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
TABLE 2--Data on commercial sand fractions.

Code and Description Specific Size Range, Roundness, Density Limits


Sieve Range Gravity, G, mm R
emax emin

LL 5-10 Lapis Lustre sand: 45% quartz, 15% chert, 2.62 2.0-4.0 0.44 0.754 0.460
15% granite particles, 15% fine grained vol- C
canic rock particles, 10% quartzite particles
o
MS 10-18 Monterey sand: 65% quartz, 30% feldspar, 2.64 1.0-2.0 0.39 0.772 0.469 z
5% quartzite and chert
MS 18-35 Monterey sand: Minerals same as MS 10r18 2.64 0.5-1.0 0.34 0.799 0.458
OS 18-35 Ottawa sand: 100% quartz 2.65 0.5-1.0 o. 60 o. 7o4 o. 408
OS 35-60 Ottawa sand: 100% quartz 2.65 0.25-0.5 0.42 0.772 0.407
OS 60-120 Ottawa sand: 100% quartz 2.65 0. 125-0.25 0.38 0.830 0.460
DM 35-60 Del Monte white sand: 80% quartz, 15% 2.65 0.25-0.5 0.27 0.971 0.503
feldspar, 5% chert
DM6O-120 Del Monte white sand: Minerals same a s D M 2.65 0.125-0.25 0.23 1.082 0.550
35-60 Z
DM120-200 Del Monte white sand: Minerals same as DM 2.65 0.074-0.125 0.21 1.203 0.636 O
35-60
CB 5-10 crushed basalt: 100% basalt particles 2.85 2.04.0 0.20 1.19 0.700 ~
CB 10-18 crushed basalt: 100% basalt particles 2.85 1.0-2.0 0.20 1.26 0. 722 I
CB 18-35 crushed basalt: 100% basalt particles 2.85 0.5-1.0 0.19 1.26 0.705 N
CB 35-60 crushed basalt: 100% basalt particles 2.85 0.25-0.5 0.19 1.32 0.692 0
CB 60-120 crushed basalt: 100% basalt particles 2.85 0.125-0.25 0.18 1.35 0.747
CB 120-200 crushed basalt: 100% basalt particles 2.85 0.074-0.125 0.17 1.42 0.803 ~
Z
O!

f~

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions auth
10.4 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

-
'~176
/ ,,~. \ \
,,\ "
\ ,

'\ \ ,\ ~/--MIXl
'\ '\', \ ~,/~MIX 2
sop " \ , \ \ ' ~ MIX 3

,ix

='~
,-7

40 ~ ,\'\)

'\

I
I
! \ \~'~ "\ "\
ol ,
I0 I0 Ol
PARTICLE DIAMETER, mm

FIG. 2--Grain-size distribution curves for artificially proportioned sand mixtures.

T A B L E 3--Data on artificially proportioned sand mixes.

Code Mix No. G, C. R Density Limits

emax emin

MOL" 1 2.64 1.4 0.34 0. 799 0.458


MOL 2 2.64 2.5 0.35 0.688 0. 370
MOL 3 2.64 4.3 0.37 0. 577 0.300
MOL 4 2.64 8.0 0.37 0.491 0.271
MOL 5 2.64 4.3 0.37 0. 641 0.335
CB b 1 2.85 1.4 0.19 1. 257 0. 705
CB 2 2.85 2.5 0.19 1. 099 0.590
CB 3 2.85 4.3 0.19 0.993 0.480
CB 4 2.85 8.0 0.19 0. 800 0.439

- M O L mixes were proportioned from sieve fractions of Monterey, Ottawa, Lapis


Lustre and Del Monte sands, in t h a t order.
b CB mixes were proportioned from sieve fractions of crushed basalt.

f
I
\ " O S I09
\ 1\
\ yDM#
TR 2

k
\

,<
%\.
~0 OI IO 1.0 O.I 1.0 0.1
PARTICLE DIAMETER, mm

FIG. 3--Grain-size distribution curves for natural and commercially graded sands.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
YOUD ON FACTORS CONTROLLING DENSITIES OF SANDS 105

Sands
The commercially available sands that were fractioned and combined
into artifically proportioned mixes included Ottawa sand (Ottawa Silica
Co., Ottawa, Ill.), Monterey sand (Monterey Sand Co., Monterey, Calif.),
Del Monte white sand (Wedron Silica Co., Pacific Grove, Calif.), Lapis

TABLE 4--Data on natural and commercially graded sands.

Code Location and Approximate Specific Mean Coefficient Roundness,


Composition Gravity, Particle of Uniform- R
G~ Size, Ds0 mity, Cu

0190 Ottawa sand A S T M C190: 2.65 0.68 1.3 0.60


100% quartz
0109 Ottawa sand A S T M C109: 2. 65 O. 36 1.8 0.42
100% quartz
RC #2 Rodeo Cove, Calif., beach sand: 2.67 0.57 2.9 0.42
minerals same as RC # 1
RC #1 Rodeo Cove, Calif., beach sand: 2.67 0.79 3.9 0.37
60% chert, 20% quartz, 20%
"green stones" (generally ba-
saltic particles)
SGB San Gregorio, Calif., beach 2.65 0.31 1.9 0.30
sand: 80% quartz, 10% feld-
spar, 10% other
SGD San Gregorio, Calif., dune sand: 2.65 0.23 1.7 0.28
minerals same as SGB
SFB Ravenwood Point (San Fran- 2.65 0.76 4.5 0.28
ciseo Bay), Calif., 6 m below sur-
face: 50% shale particles, 50%
quartz and feldspar
AC #1 Alameda Creek, Calif., sand: 2.65 0.62 3.1 0.28
50% shale fragments, 50%
quartz, feldspar and chert
AC #2 Alameda Creek, Calif., sand: 2.65 0.55 2.0 0.25
minerals same as AC # 1
DM Del Monte white sand: 80% 2.65 0.37 2.4 0.25
quartz, 15% feldspar, 5%
chert
TR # 1 Trinity River, Calif., sand: 2.98 1.10 2.8 0.23
85% amphibolite particles, 5%
quartz, 5% epidote particles,
5% talc
TR # 2 Trinity River, Calif., sand: 2.98 0.64 3.4 0.21
minerals same as TR # 1
CB Washed crushed basalt, Napa, 2.85 1.02 5.6 0.19
Calif.: 100% bazsalt particles

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Li
106 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

9 CRUSHED BASALT 9 OTTAWA SAND


16 Cu : 14 0 DEL MONTE WHITE SAND
9 MONTEREY SAND
1.4 X LAPIS LUSTRE SAND

o E 12 0
(D 0
~ 1.0
0

::'< 1.8 9 9
9 X
0G
G 1.0
~ 0.8
Q::
~ ~ 0.6
~ E 0
9 0
9 X
O4
us,5 SIEVE RANGE
~ 02 2OO 120 60 35 18 I0 ,5
I ~ t----q
::~ 0.0 i
o.I i'.o
MEAN PARTICLE DIAMETER, D50, IN mm

F I G . 4--Density limits versus mean grain size for laboratory fractions with Cu = 1.4.

9 Ottawa sand
\ 0 Del Monte while sand
1.4 9 9 Monterey sand
k x kopis Lustre sand
9 Crushed bosoll
Cu:l.4

~ - Maximum void ratio, emax.

10
O o
I-'-.

e'-,0.8
~Jl~ O'X--'-~O-

\o
0.E ~ Mtntmumvoid ratio, emin
e ~
-nm~ x
0.4 Very
Angular Subangular Rounded
L :,~,~ J' Subrounded ~1.
o2 o'l olz o13 o',~ o's o'e
ROUNDNESS, R
F I G . 5--Dens#y limits as a function of grain shape for laboratory fractions w#h C. =
1.4.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreemen
YOUD ON FACTORS CONTROLLING DENSITIESOF SANDS 107

Lustre sand (Pacific Cement and Aggregates, San Francisco, Calif.), and
crushed basalt (Basalt Rock Quarry, Napa, Calif.).
Properties of the sand fractions are listed in Table 2. Gradation curves
and descriptive data for the artificial mixtures are presented in Fig. 2 and
Table 3. Properties of the natural and commercially graded sands are
given in Fig. 3 and Table 4.

Results
The influence of specific gravity on the test results was normalized by
converting maximum and minimum densities to minimum and maximum
void ratios (emia and emax), respectively. The emax and emi~ are tabulated
in Table 2 for the sand fractions and in Table 3 for the artificial mixes.
The data plotted in Fig. 4 show that no unique relation exists between
mean particle diameter and ema~ or emi,. However, the same void-ratio
data plotted against R on Fig. 5 form a well defined curve. These results
show that grain shape is an important factor controlling emaxand emi, and
that particle size per se has no significant influence.
In Fig. 6, e.... and emin for the artificial sand mixtures are plotted against

9 MOL mixes J,2,5,4


9 X MOL mix 5
!2 9 CB mixes ;,2,5,4

F-- LO R=02

c:::
3o=
X 0,6
~o~
~o~

J.-- ~'--~A R:02

>So4
Z
0.2 , ~ ~ . . . . . . .
4 6 I0
COEFFICIENT OF UNfFORMITY, Cu

FIG.6--Density limits as a function of gradation for artificially proportioned sand mixes.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreeme
108 RELATIVE DENSITY INVOLVING COHESiONLESS SOILS

Cu. For constant values of R and for gradational curves of the normally
distributed type, these data form well defined curves, indicating that range
of particle sizes is another important factor controlling the density limits.
Data from MOL Mix 5 (gap graded) are also included on Fig. 6 and
plot appreciably above the data from 1VIOL Mix 3 (normally graded),
which is characterized by an equivalent Cu. Thus, the type of gradational
curve is also a controlling factor.
A generalized set of curves (Fig. 7), relating em~x and emin to Cu with
particle roundness as a parameter, were constructed from the data in
Figs. 5 and 6. The validity of these curves for estimating emaxand emin for
normally or approximately normally distributed sands was examined by
comparing estimated void ratios for the sands listed in Table 4 with
maximum and minimum void ratios measured in the laboratory. Two
methods of estimation were used. First, emaxand emln were estimated from
the index properties C~ and R. Second, an R value was estimated from

:,< 14
E

O" 1.2
I.-..-

I.C

=E
X
i

06,

R O U N D E D ~

o"

~ 0.4

~:~ 02 i
I
7 BRo
2 3 4 6

COEFFICIENT OF UNIFORMITY, Cu
IO 15

FIG. 7--Generalized curves for estimating em~xand emlnfrom gradational and particle
shape characteristics. Curves are only valid for clean sands with normal to moderately skewed
grain-size distributions.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
YOUD ON FACTORS CONTROLLING DENSITIES OF SANDS 109

TABLE 5--Measured and estimated density limits for sands listed in Table 4.

Code Estimates
Measured
From Fig. 7 Burmister~ Hutchison Kabai ~
and
Townsend b

emax emin emax d emin d emia e emax emin emax emin emax emia

0190 0.70 0.42 0.71 0.42 0.42 0.82 0.66 ...... 0.76 0.64
0109 0.75 0.41 0.70 0.39 0.40 0.78 0.66 ...... 0.70 0.58
RC #2 0.55 0.31 0.60 0.33 0.31 0.74 0.50 0.81 0.50 0.59 0.48
RC #1 0.57 0.29 0.61 0.31 0.31 0.70 0.40 0.86 0.47 0.53 0.42
SGB 0.79 0.41 0.81 0.42 0.42 0.81 0.60 0.94 0.57 0.70 0.59
SGD 0.81 0.48 0.88 0.48 0.43 0.83 0.60 0.97 0.64 0.73 0.62
SFB 0.74 0.39 0.70 0.36 0.38 0.70 0.46 0.76 0.38 0.51 0.41
AC #1 0.73 0.39 0.75 0.39 0.37 0.73 0.49 0.88 0.50 0.58 0.47
AC #2 0.90 0.54 0.93 0.50 0.48 0.98] 0.73] 0.91 0.49 0.67 0.56
DM 0.91 0.51 0.88 0.48 0.50 1.00s 0.70] 0.93 0.57 0.64 0.52
TRI 0.87 0.48 0.91 0.48 0.46 0.91] 0.63/ ...... 0.58 0.48
TR2 0.91 0.51 0.92 0.50 0.48 0.93] 0.62] 0.87 0.50 0.56 0.46
CB 0.96 0.45 0.93 0.48 0.50 0.98] 0.61] 0.82 0.42 ......

Estimated from Figs. 3 and 4 of Burmister [3], assuming G, = 2.67.


b Estimated from Hutchison and Townsend's Eqs 1 and 3 [9], assuming G, = 2.67.
c Estimated from Fig. 10 of Kabai [11].
d Estimated from R and C~ values in Table 4 (Method 1).
e Estimated from Cu and the measured em~x (Method 2).
] Includes 44N (10-1b) correction for crushed particles.

Cu, a n d t h e em~x m e a s u r e d in t h e l a b o r a t o r y ; this R v a l u e a n d C~


were t h e n used to e s t i m a t e eml,. T h e s e e s t i m a t e s a n d t h e c o m p a r a t i v e
l a b o r a t o r y results a r e l i s t e d in T a b l e 5. F o r t h e 13 s a n d s p e c i m e n s used in
t h e c o m p a r i s o n , t h e difference b e t w e e n t h e e s t i m a t e d a n d m e a s u r e d e ~
b y t h e first m e t h o d was 0.011 • 0.024, a n d t h e difference b e t w e e n t h e
e s t i m a t e d a n d m e a s u r e d e~,in w a s - 0 . 0 0 4 -4- 0.014. B y t h e s e c o n d m e t h o d ,
t h e difference b e t w e e n t h e e s t i m a t e d a n d m e a s u r e d e ~ n was - 0 . 0 1 1 =i=
0.018. T h e s e d a t a show t h a t t h e curves in Fig. 7 a r e v a l i d for e s t i m a t i n g
e ~ a n d emin for clean s a n d s w i t h n o r m a l t o m o d e r a t e l y s k e w e d g r a i n - s i z e
d i s t r i b u t i o n curves. C l e a n s a n d s are defined as h a v i n g less t h a n 5 p e r c e n t
p a r t i c l e s b y w e i g h t p a s s i n g t h e No. 200 sieve. N o r m a l to m o d e r a t e l y
s k e w e d grain-size d i s t r i b u t i o n curves i n c l u d e c u r v e t y p e s " S " a n d " L "
a n d u p to t h e m o d e r a t e l y c o n c a v e or convex f o r m s of c u r v e t y p e s " C "
a n d " E " , r e s p e c t i v e l y , as defined b y B u r m i s t e r [2]. T h e g r a i n size d i s t r i -
b u t i o n curves in Fig. 3 fall w i t h i n t h e s e categories.

Comparison With Previous Work


E x c e p t for t h e l a c k of influence of p a r t i c l e size, this s t u d y c o n f i r m e d t h e
findings of p r e v i o u s i n v e s t i g a t o r s [1-3] t h a t p a r t i c l e shape, p a r t i c l e size

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Li
1 10 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

range, and type of grading curve are the primary factors controlling em~x
and emin. The apparent discrepancy as to the influence of particle size may
possibly be explained by the fact that, for most natural sands, there is a
correlation between particle size and particle shape. That is, the larger
the particle size, the more rounded it tends to become through natural
processes [18]. Thus, for any given sand, e.... and emin tend to decrease
with particle size because of the shape factor; the data plotted on Fig. 4
follow this trend for each sand.
The em~xand emin estimated according to the criteria of previous investi-
gators for the sands listed in Table 4 are shown in Table 5 where they can
be compared with estimated and measured values from this study.
Because Burmister [3] used a procedure very similar to the procedure
used in this study to measure em,x (sand poured through a funnel into a
mold), one would expect that estimates from his curves would agree well
with values from this study. This is true for R values between 0.28 and
0.30 (subangular), which include sands SGB, SGD, A C # I , and SFB.
With the suggested correction for crushed and broken particles applied,
em~xvalues from Burmister's curves are consistant with measured data for
sands with R values between 0.19 and 0.21, for example, sands T R # 2
and CB. The agreement between estimates of em~xfrom Burmister's curves
and the measured values for the other sands is not very good, chiefly
because the single-value correction for crushed and broken particles does
not adequately account for all of the density variation due to particle
shape.
For all of the sands, emin values estimated from Burmister's curves are
greater than those measured in this study. This result is consistant with
the fact that the methods used by Burmister to obtain maximum densities
(vibrating tamper or impact) do not give values as great as the simple
shear method used in this study.
The criteria for estimating maximum and minimum density established
by Hutchison and Townsend [9] did not include any variance for grain
shape. Thus, a general direct comparison between estimated en~x and emin
from their criteria and the measured values can not be made. It is noted
that the em~xestimated by their criteria is approximately equal in value to
the measured em~x for sands with an R of 0.25 (sands A C # 2 and DM).
However, this R is considerably different from the R of 0.50 to 0.55 de-
termined from several specimens of beach and dune sands used by Hutchi-
son and Townsend in their study. Several factors could contribute to this
discrepancy, including, (1) possible differences in the methods used to
determine R, (2) the beach and dune specimens used for this study may
have been more rounded than the beach and dune sands used by Hutchison
and Townsend, and (3) the tipping test procedure they used for determining
minimum densities may have yielded appreciably lower e.... values than
the funnel method used in this study.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant t
YOUD ON FACTORS CONTROLLING DENSITIES OF SANDS 11 1

The determination of the Bagnold grading parameters is another possible


source of error in estimating maximum and minimum densities by the
Hutchison and Townsend criteria. These parameters are uniquely definable
for many sands, such as those with S-shaped gradation curves. For others,
such as those with linear, concave, or convex gradation curves, their
evaluation is subject to some subjective interpretation or not definable at
all, hence, the omission of some data from Table 5.
All of the e~in estimated by the Hutchison and Townsend criteria are
greater than those measured in this study, indicating that maximum
densities obtained by the vibrating tamper-compaction method, which
they used in deriving their criteria, gives smaller densities than the simple
shear method.
The values of ema~ estimated from Kabai's [11] curves agree approxi-
mately with the em~xfrom this study for R values between 0.37 and 0.42
(sands 0109, RC # 1 and RC # 2). Kabai's em~ values are based on funnel
method minimum density tests, as are those in this study. The emi~ for
these same sands estimated from Kabai's curves are considerably greater
than those measured in the laboratory by the simple shear method, indi-
cating that the simple shear method gives considerably greater maximum
densities than the impact method used by Kabai.
Because Kabai did not consider the influence of particle shape on maxi-
mum and minimum densities, further comparisons with data estimated
from his curves are not warranted.

Conclusions
1. The results of this study confirm the findings of previous studies that
the primary factors controlling the maximum and minimum void-ratio
limits of clean sands are particle shape, particle size range, and the shape
of the gradational curve. Contrary to previous studies, it was found that
particle size per se has no significant influence on the density limits.
2. The curves in Fig. 7 are valid for estimating e~ax and emin for clean
sands with normal to moderately skewed grain-size distributions.

Acknowledgments
The assistance of Terry Craven, who performed the compaction tests,
and Julius Schlocker, who assisted with the mineral identifications, is
gratefully acknowledged.

References
[1] Burmister,D. M., Proceedings, AmericanSocietyfor Testingand Materials, Vol. 38,
1938, pp. 587-596.
[2] Burmister,D. M., Proceedings, AmericanSocietyfor Testingand Materials, Vol. 48,
1948, pp. 1249-1268.
[3] Burmister, D. M. in Field Testing of Soils, A S T M STP 322, AmericanSocietyfor
Testing and Materials, 1962, pp. 67-97.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No fu
112 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

[$] Burmister, D. M. in Procedures for Testing Soils, American Society for Testing and
Materials, Philadelphia, Pa., 1964, pp. 175-177.
[5] Felt, E. J. in Application of Soil Testing in Highway Design and Construction, A S T M
S T P 239, American Society for Testing and Materials, 1958, pp. 89-108.
[6] Kolbuszewski, J. J. and Frederick, M. R., Proceedings, European Conference on
Soil Mechanics and Foundation Engineering, Wiesbaden, 1963, Vol. I, pp. 253-263.
[7] Kolbuszewski, J. J., Proceedings, 2nd International Conference on Soil Mechanics
and Foundation Engineering, Rotterdam, 1948, Vol. 1, pp. 158-165.
[8] Bagnold, R. A., The Physics of Blown Sand and Desert Dunes, Methuen and Co. Ltd.,
London, 1941, pp. 113-116.
[9] Hutchison, Bruce and Townsend, David, Proceedings, 5th International Conference
on Soil Mechanics and Foundation Engineering, Paris, 1961, Vol. 1, pp. 159-163.
[10] Wu, T. H., Journal of the Soil Mechanics and Foundation Division, American Society
of Civil Engineers, 1957, VoL 83, No. SM1, pp. 1161-1 to 1161-23.
[11] Kabai, I. in Proceedings, 3rd Budapest Conference on Soil Mechanics and Founda-
tion Engineering, Budapest, 1968, pp. 115-126.
[12] Johnson, A. W. and Sallberg, J. R., "Factors Influencing Compaction Test Results,"
Highway Research Bulletin 319, 1962.
[13] Youd, T. L., "Maximum Density of Sand by Repeated Straining in Simple Shear,
Highway Research Record No. 374, 1971, pp. 1-6.
[14i] Kolbuszewski, J. J., Proceedings, 2nd International Conference on Soil Mechanics
and Foundation Engineering, Rotterdam, 1948, Vol. 7, pp. 47-49.
[15] Youd, T. L., Journal of the Soil Mechanics and Foundations Division, American
Society of Civil Engineers, 1972, Vol. 98, No. SM7, pp. 709-725.
[16] Wadell, Hakon, Journal of Geology, Vol. 43, 1935, pp. 250-280.
[17] Powers, M. C., Journal of Sedimentary Petrology, Vol. 23, No. 2, June 1953, pp. 117-
119.
[18] Twenhofel, W. H., Principles of sedimentation, McGraw-Hill Book Co. Inc., New
York, 1950, pp. 302-311.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreeme
E. A . D i c k i n I

Influence of Grain Shape and Size upon


the Limiting Porosities of Sands

REFERENCE: Dickin, E. A., "Influence of Grain Shape and Size u p o n


the L i m i t i n g Porosities of S a n d s , " Evaluation of Relative Density and Its
Role in Geotechnical Projects Involving Cohesionless Soils, A S T M S T P 5~3,
American Society for Testing and Materials, 1973, pp. 113-120.
ABSTRACT: The state of packing of a mass of sand grains is described by
its relative porosity n~ which defines the packing relative to maximum and
minimum porosities of the material. These limiting porosities depend in turn
upon the physical characteristics of the grains themselves. In the research
described herein, the influence of grain shape and size upon the limiting po-
rosities of quartz sands and glass ballotini was studied.
The maximum porosities were determined by deposition of the sample
through water as suggested by Kolbuszewski (1948) while minimum porosities
were obtained by vibration under water. Shape parameters for the sands were
determined from a correlation between the time of flow of a 0.5-kg specimen
and the sphericity measured by examination of individual grains.
Both maximum and minimum porosities decreased with increasing sphericity
while tests on glass ballotini indicated that the effect of grain si~e was negligible.
The porosity interval was approximately 12.5 percent for all the sands and
11 percent for spherical ballotini. Mixtures of sand and ballotini gave reasonable
agreement with the trend shown by the separate materials.
KEY WORDS- eohesionless soils, density (mass/volume), vibration, measure-
ment, classifications, sands

Nomenclature
d Average grain diameter (mm)
n Porosity
nrilax Maximum porosity
nmin Minimum porosity
n, Relative porosity
q F l o w t i m e (s)
A n g l e of i n t e r g r a n u l a r friction
~0 Sphericity

: Lecturer in Civil Engineering, The University of Liverpool, England.

113

Copyright by ASTM Int'l


Copyright9 by(all
ASTMrightsInternational
reserved); Fri Mar 11 16:13:06 EST 2016
www.astm.org
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
1 14 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

Porosity has been widely used to describe the state of packing or irregular
sand grains although its inadequacy has been demonstrated by both
Kolbuszewski [1]2 and Amirsoleymanii [2]. Kolbuszewski [3] considered
that relative porosity (nr) gave a better assessment of packing. This is
defined as
nmax ~ n
nr ~-
nmax -- n m i n

where n~.x and n~in are the limiting porosities of a sand and n its actual
porosity. He proposed standard tests to determine the limiting porosities
and concluded that intensity of deposition and material properties would
influence the packing.
The effect of grain shape and size on the limiting porosities has been
studied by Kolbuszewski and Frederick [4] who found that:
(a) the limiting porosities increased as angularity increased;
(b) the maximum porosity decreased with decreasing uniformity coeffi-
cient; and
(c) the porosity interval n~az - nmlnwas of the order of 10 to 12 percent
for all the sands considered, the interval decreasing slightly as roundness
increased.
They assumed a single shape value for each sand, although different
sieve fractions were used. Tests on glass ballotini indicated that the maxi-
mum porosity decreased with increasing grain size, although the minimum
porosity appeared to be independent of size.
Alyanak [5] concluded that the porosity interval was only a reflection
of the relative merits of the methods by which the limiting porosities were
obtained and was independent of particle shape and size distribution.
However, Smith [6] found a 3 percent decrease in nmaz - - n m i n as grains
increased in roundness from very angular (crushed basalt) to rounded
(Erith sand), which is consistent with the trend observed by E1-Sohby [7]
who reported a 5.7 percent difference between the values for crushed feld-
spar and glass ballotini. E1-Sohby indicated that the porosity interval
increased as the angle of intergranular friction increased, but he defined
his maximum porosities as those of his loosest cylindrical compression test
specimens, and it is unlikely that his n~,z values would be as high as could
be obtained by the deposition methods of Kolbuszewski. Smith alone took
account of the variation of particle shape with size although he finally
used weighted values to describe the shapes of his graded sands. In the
research described herein, the variation of limiting porosities with shape
of sieve fractions of five natural sands--Biddulph, Erith, Ham River,
Leighton Buzzard, and Stonecourt--was considered.

2 The italic numbers in brackets refer to the list of references appended to this paper.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
DICKIN ON INFLUENCE OF GRAIN SHAPE AND SIZE 115

Experimental Procedure

Determination of Maximum Porosity


The maximum porosity was determined from the volume occupied by
1 kg of the specimen after deposition through water as suggested by
Kolbuszewski [3]. Values of specific gravity used in porosity calculations
were obtained for each sand by the British Standard density bottle method
(BS 1377).
Determination of Minimum Porosity
The minimum porosity was obtained by vibration under water in prefer-
ence to more severe methods which cause particle degradation. The me-

FIG. 1--Photomicrographs of sands.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreeme
116 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

chanical vibrator used was developed by Smith [6] who found that the
densest packings were obtained for a frequency of 37.8 Hz and an amplitude
of 0.6 mm. Specimens were prepared by vibrating for 5 rain in a 102 mm
diameter by 102 mm high mould and were later subjected to axisymmetric
compression tests [8]. The porosity was calculated from the volume of
water displaced by the specimen in the compression cell as described by
Lee [9].

Determination of Grain Shape


Grain shapes were determined by examination of photomicrographs of
individual grains and were defined in terms of Riley's inscribed circle
sphericity [10]. Typical photomicrographs of the sands are illustrated in
Fig. 1.
Direct methods of estimating particle shape become lengthy if a truly
representative specimen of a sand is to be examined and are therefore
unsuitable for routine use. Thus, indirect methods have been developed
in which a bulk property of the material, which is strongly influenced by
grain shape, is measured in a standard test. A good correlation was found
between the time taken for 0.5 kg of the specimen to flow through a 6.35-
mm diameter orifice and sphericity. The flow technique has been reported
by King and Dickin [11] and was based on a test devised by Rex and
Peck [12]. Sphericity (~b0) was related to flow time (q s) and average grain
diameter (d mm) by
196d+144-- q
~b0--
146d + 112
This emperical equation yielded a more reliable value of sphericity than
would be obtained by direct measurement alone since a larger number of
grains were considered. Insertion of d and corresponding q values given
in Table 1 into the empirical equation yielded the variation of sphericity
with grain size shown in Fig. 2. Stonecourt sand excepted, a marked de-

T A B L E 1--Flow times of sieve fractions of sands.

Average Grain Flow time, s


Diameter, mm Biddulph Erith Ham River Leighton Stonecourt
Buzzard

0.510 90.8 77.5 85.6 82.4 88.0


0.388 84.0 74.3 77.5 74.6 81.2
0.325 81.8 70.9 77.0 72.6 77.3
0.272 78.0 68.0 72.0 70.3 72.4
0.230 74.8 66.7 69.8 ... 70.2
0.181 69.8 62.8 66.5 ... 68.0
0.121 . . . . . . 63.5 ... 66.2

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant
DICKIN ON INFLUENCE OF GRAIN SHAPE AND SIZE 117

, ! ,
_iI !I f Ii I, i t i
0.90
ERITH
FI~OM FLOWTIHE$ ]
- - - F20H FNOTOHtCRO6NApH$
/

0 'g5
I
m
]13

--I BIDDULPH
-(
0.80
o

0,75
8.s. S,EvG FRAcT,oNS
J 721~ol5a I,~ I 25
0,7~
0 o. o.z o.3 0,4 0.5 0.6
AVERAQE PARTICLE DIAMI~"rl}~ d ('t'nm)

FIG. 2--Variation of sphericity with grain size for sands.

crease in sphericity accompanied decrease in grain size. Upon closer exami-


nation, Stonecourt sand was found to contain significant quantities of
angular nonquartz material, mainly haemetite and calcite, in the larger
sieve sizes. This material not only gave an inaccurate estimation of spher-
icity from photomicrographs but yielded spurious flow times due to differ-
ences in surface texture. Values of sphericity for Stonecourt sand were
taken as those obtained by individual examination.

Results and Discussion

The Influence of Grain Shape on the Limiting Porosities of Sands


The limiting porosities for sieve fractions of the five sands are plotted
against sphericity in Fig. 3. A general decrease in both maximum and
minimum porosities with increasing sphericity is observed, although the
porosity interval nmax -- nmin does not appear to change significantly with
grain shape over this limited range of sphericities. The interval is approxi-
mately 12.5 percent which is of the same order as that reported by Kolbus-
zewski and Frederick [4].
The Influence of Grain Size upon the Limiting Porosities
Tests were carried out on sieve fractions of glass ballotini in order to
investigate the effect of variation in grain size alone upon the limiting
porosities. The results are shown in Fig. 4, and it is seen that no significant
variation in either maximum or "minimum porpsity occurs over this range

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licens
118 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

55

~ q ~ : ~ L . ~~|. _ '~ ,~ w | n MAX.

50 9 ~ - 0 9

-~0

o~ 9 EI~ITH
9 HAM RIVEI~
O LEIGHTON BUZZA~O
40 9 STONECOU~'"F

~5

0,~ o. ?G 0,78 o.%0 0.8~ 0.84 0.86 0,88 0.90


5PHE~ICITY "v,,"o

FIG. 3--Variation of limiting porosities of sands with grain shape.

O O. 4 02 O.a 0.,4 0'5 Og


AVERAnl~ [31A,Vm'1"iR ~.. m.rn.

FIG. 4--Limiting porosities of glass ballotini.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further
DICKIN ON INFLUENCE OF GRAIN SHAPE AND SIZE 1 19

55

50

45

40
:3

35

9- - - - ~...._...

E~O
9 QUARTZ SANDS,
9 BALLOTINI / LEIEIHTON ~,UZZARD t~lX,
o ~LAf~ BALLOTINI
22 I I
o.~z ,, o.'z~ o.~ o.e4 o.ss o.~ o.~6 ~.oo

SPHERICITY "f*o

FIG. 5---Limiting porosities of quartz sands and glass ballotini.

of sizes. This result does not confirm that of Kolbuszewski and Frederick
who found that the maximum porosity of ballotini decreased with increasing
grain size.
The porosity interval for ballotini was approximately 11 percent which
confirms previous reports [4, 7] that smaller n~ax - - n m i n values are obtained
for ballotini than for sands. It is not clear whether the trend is due to shape
differences or to the considerably lower angle of intergranular friction for
glass ballotini ( ~ ~ 15 deg) compared with that of quartz sands ( ~ usually
between 24 and 28 deg).
The variation of limiting porosities for 100 percent quartz sands and
glass ballotini is illustrated in Fig. 5. Intermediate values of sphericity
were achieved by mixing Leighton Buzzard sand with glass batlotini in
equal parts by volume. Values of maximum porosity for the mixtures were
consistent with the general trend, but minimum porosities were not as low
as might have been expected.
Conclusions
Both the maximum and minimum porosities decrease with increasing
sphericity for the five sands considered in this research.
The effect of variation in grain size has been shown to be negligible for
glass ballotini, and if this result is assumed to apply to sands, the variation
in maximum and minimum porosities in Figs. 3 and 5 is attributable to

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licen
120 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

changes in grain shape. L i m i t i n g porosities decrease with increasing grain


size for a n y one sand as a result of the inherent increase in sphericity
with grain size in n a t u r a l q u a r t z sands.
T h e porosity interval was a p p r o x i m a t e l y 12.5 percent for the sands
considered, while t h a t for glass ballotini was only 11 percent. I t was n o t
clear whether this was due to differences in shape or in intergranular fric-
tion, a l t h o u g h the effect of different angles of intergranular friction would
p r o b a b l y be small at the relatively low stress levels encountered in these
tests.
Values of m a x i m u m porosity for glass ballotini/sand mixtures followed
the trend of other results more closely t h a n did values of m i n i m u m porosity.

References
[1] Kolbuszewski, J. J. in Proceedings, Midlands Soil Mechanics and Foundations Engi-
neering Society, Birmingham, No. 4, 1961, pp. 9-18.
[2] Amirsoleymanii, "Packing of Granular Materials with Special Reference to Triaxial
Testing," Ph.D. thesis, University of Birmingham, 1964.
[3] Kolbuszewski, J. J., Proceedings, 2nd International Conference of Soil Mechanics,
Vol. 1, 1948, p. 158.
[~] Kolbuszewski, J. J. and Frederick, M. R., "The Significance of Particle Shape and
Size on the Mechanical Behaviour of Granular Materials," European Conference of
Soil Mechanics and Foundations Engineering, Yol. 1, 1963, pp. 253-263.
[5] Alyanak, I. in Proceedings, Midlands Soil Mechanics and Foundations Engineering
Society, Birmingham, No. 4, 1961, pp. 37-42.
[6] Smith, D., "The Influence of Particle Shape on the Limiting Porosities and Shear
Strength of Sands," M. E. thesis, University of Liverpool, 1965.
[7] Ei-Sohby, M. A., "The Behaviour of Particular Materials under Stress," Ph.D.
thesis, University of Manchester, 1964.
[8] Dickin, E. A., "The Influence of Grain Shape and Size on the Shear Strength Com-
ponents of Quartz Sands," Ph.D. thesis, University of Liverpool, 1971.
[9] Lee, I. K., Proceedings, American Society of Civil Engineers, Vol. 92, SM2, 1966,
pp. 79-103.
[10] Riley, N. A., Journal of Sedimentary Petrology, Vol. 11, No. 2, 1941, p. 94.
[11] King, G. J. W. and Dickin, E. A., Materials and Structures, Vol. 4, No. 2, 1972.
[I~] Rex, H. M. and Peck, R. A., Journal of Public Roads, Yol. 29, No. 5, 1956 pp. 118-
120.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No furth
E. W . B r a n d ~

Some Observations on the Control of


Density by Vibration

REFERENCE: Brand, E. W., " S o m e O b s e r v a t i o n s o n t h e Control of D e n -


sity b y V i b r a t i o n , " Evaluation of Relative Density and Its Role in Geotechnical
Projects Involving Cohesionless Soils, ASTM S T P 523, American Society for
Testing and Materials, 1973, pp. 121-132.

A B S T R A C T : Because vibration techniques axe commonly used for the prepa-


ration of laboratory test specimens, this study was carried out to examine some
of the factors which affect the densities obtained. A simple, one-degree-of-
freedom vibrating table device, equipped for control of both amplitude and
frequency, was used to give a sinusoidal wave form of vibration. Tests were
carried out with four size ranges of sand and with single size glass balls, and
observations were made on the effects of time of vibration, intensity of vi-
bration, and container size.
Maximum relative densities were achieved at a single value of the maximum
acceleration in each case, these densities increasing with time of vibration up
to about 30 rain. Only low states of compaction were obtained for some of the
vibrated specimens, a n d in no case was 100 percent relative density achieved.
By far the most significant experimental results related to the variation of
density throughout the vibrated soil specimens. This was investigated using a
split cylinder, with four equal parts, which could be dismantled after vibration
enabling separate density measurements to be made. Considerable density vari-
ations existed throughout the specimens at vibration intensities below the opti-
mum value. Only where the average relative density was high was an approxi-
mately homogeneous specimen obtained. This fact is important to those con-
cerned with the preparation of test sections in the laboratory.

K E Y W O R D S : vibration, compacting, tests, sands, density (mass/volume),


porosity, cohesionless soils

Vibration techniques are commonly employed in the preparation of


laboratory test specimens of granular materials. Insufficient regard is often
paid to the complex nature of the mechanics of the process of dynamic
compaction, however, and the presumption that homogeneous sand speci-
mens of a desired relative density can be easily and consistently achieved

1 Associate professor, Geotechnical Engineering, Asian Institute of Technology,


Bangkok, Thailand.

121

Copyright by ASTM Int'l


Copyright9 by(all
ASTMrightsInternational
reserved); Fri Mar 11 16:13:06 EST 2016
www.astm.org
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
122 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

by means of a vibrating table is erroneous. Published literature pertaining


to compaction by vibration is scant, the only contributions which are
directly relevant to the control of density in the laboratory being those by
Mogami and Kubo [1]~, Felt [2], Alyanak [3], Selig [4], Pettibone and
Hardin [5], Kolbuszewski and Alyanak [6], and D'Appolonia and D'Appo-
lonia [7].
The perfect technique for preparing sand specimens would be one which
permitted absolute control over the relative density and which resulted in
the formation of completely homogeneous specimens. It would also enable
the achievement of relative densities close to 100 percent, and would
render as insignificant the effects of the mold in which the material was
contained. In order to assess the possibilities that a vibration device might
be developed into this perfect instrument of sample manufacture, a brief
experimental study was carried out which yielded the results reported in
this paper. It is hoped that a few observations made during tests using a
simple one-degree-of-freedom vibrating table device will be of some interest
to those concerned with measurements on granular materials.

T h e Granular Materials

Description
Vibration tests were conducted with four dry quartz sands and one
specimen of glass balls. The sand specimens were designated as Fine Sand,
Medium Sand, Coarse Sand and Graded Sand, and the particle size ranges
were controlled by U.S. Sieves 50-100, 30-50, 16-30, and 16-100, re-
spectively. The Graded Sand was composed of equal parts of the other
three sands. The Fine Sand was from a residual deposit and consisted of
sharp angular particles, while the other two single-size specimens were
from river beds and were composed of less angular and more rounded grains.
The glass balls were almost perfectly spherical with sizes governed b y
Sieves 20 and 30.

Maximum and Minimum Densities


In order to assess the results of the vibration tests in terms of relative
density, maximum and minimum densities were measured for each material
using the methods devised by Kolbuszewski [8]. The minimum density
(maximum porosity) was determined by placing 1 kg of the material in a
measuring cylinder of 2 litres capacity, shaking the cylinder, then quickly
inverting it twice, and reading the volume of the specimen. The maximum
density (minimum porosity) was measured by compacting the material
with 15 min applications of an electric vibrating hammer to each of three
layers of the material placed under water in a standard Proctor compaction

2 The italic numbers in brackets refer to the list of references appended to this paper.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
BRAND ON THE CONTROL OF DENSITY BY VIBRATION 123

TABLE 1--Properties of the granular materials.

Material Particle, Maximum, Minimum Minimum Maximum Specific


Size, U.S. Density, Density, Porosity, Porosity, Gravity
Sieve No. gm/cm~ gm/cma % %

Fine sand 50-100 1.67 1.35 37.1 48.9 2.64


Medium s an d 30-50 1.77 1.47 32.0 43.45 2.60
Coarse sand 16-30 1.78 1.49 31.15 42.2 2.58
Graded s a n d 16-100 1.83 1.53 30.0 41.5 2.61
Glass balls 20-30 1.88 1.72 36.1 41.55 2.95

mold. Degradation of the particles by this process was found to be a


problem in the case of the glass balls, and this permitted only a small
force to be exerted on the hammer.
The results of the measurements of the limiting densities, and of specific
gravity determinations made with a pycnometer, are shown in Table 1.
It should be remembered that the limiting densities obtained by the
preceeding methods outlined might not be the extreme values, but they
are used herein as defining 0 and 100 percent relative density for the
specimens used in the vibration tests.

T h e Vibration Tests

Equipment Used
The main features of the small vibration table constructed for the
experiments are shown in Fig. 1. The 30-cm square steel table was allowed
to execute only vertical displacements. The displacement was governed by
the two ball bearings with their outer shells screwed to the table. These
were fitted with an eccentric shaft driven by a variable speed d-c motor.
This arrangement resulted in a sinusoidal wave form of vibration of the
table. Two sets of bearings were available for use, the eccentricities of these
(and, consequently, the available amplitudes of vibration) being 0.794 and
1.588 mm. The maximum frequency that could be maintained by the
motor was about 1500 cpm, which meant that a maximum acceleration of
about 4 g could be applied to the table.
The molds used to contain the specimens during vibration were 27-cm
long perspex cylinders with base plates which could be bolted to the
vibrating table. For most of the tests, a 9.40-cm diameter cylinder was used
but, in order to investigate size effects, other diameters were later employed.

Experimental Procedure
A specimen was prepared by pouring the sand through a funnel at a
fixed height into the cylinder to give a low relative density. The surface
of the sand was levelled, and the filled cylinder was weighed prior to vi-

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to L
124 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

(o) Side Elevation --2 HR variable speed


DC motor

o) .o_
- - -

Bearings screwed to top __ __ _

o:7 -st., h ....,s

I. 3o am 'I

(b) Plan (top plate r e m o v e d )

FIG. 1--Details of the vibration table.

bration. A small rubber pouch containing sand was attached to the top of
the cylinder so that sand was continuously supplied to the cylinder as
compaction occurred. Apart from the negligible effect of the small head of
sand in the pouch, no surcharge loading was used during vibration. After
a period of vibration at a fixed frequency and amplitude, the cylinder was
reweighed after the pouch had been removed and the sand surface levelled.

Experimental Results
Effect of Time of Vibration
The effect of time of vibration on the state of compaction of each of the
four sands and the glass balls for a maximum acceleration of 1.75 g is
shown in Figs. 2 and 3. The states of compaction in Fig. 2 are given in
terms of porosity, while they are shown as relative densities in Fig. 3.
The general shape of these curves is the same as observed by previous
investigators. The relative density increased continuously with time to a
constant value in each case. Densification occurred rapidly at first, but
the rate of compaction gradually decreased with time of vibration. These
relative density changes reflect the stability of the particle arrangement at
any time which, in turn, is related to the shear strength of the granular
medium. It might be mentioned that the specimens of glass balls generally

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Li
BRAND O N THE CONTROL O F DENSITY BY VIBRATION ] 25

50
[ i ,
' : ....a/ Graded Sand

j
52
i

54
Coarse Sand
I : i m
Medium Sand
~56

o 58

1
I !
i
Fine Sand

-- T i
(a = 0.794 ram.)
i
44
OJ
~
0.2 05
E 1
I
i
5
r I
I0 20 50 Loo
Time of V i b r a t i o n , min.

F I G . 2 - - V a r i a t i o n s in porosity with time of vibration (~ = 1.75 g).

reached a stable packing under a given intensity of vibration much more


rapidly than the sands. This is thought to be indicative of the relatively
low shearing resistance of the spherical particles enabling reorientar to
take place rapidly under the disturbing force.
Figure 4 shows the density-time curves for the Graded Sand vibrated
i~

C)

_o
a)
c(

0.1 02 0,5 I 2 5 I0 20 50 I00

Time of Vibration , min

F I G . 3 - - V a r i a t i o n s in relative d~asity with time of vibration (~ = 1.75 g).

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agree
126 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

ioo 50

52
80

54
6o Legend
o
J f , cpm o , turn
$ 680 1.588
~6 ~_
4o 0:82 g - -- 0 1400 1.588
[3
&
1200
860
0.794
1588 138
X 1400 0794
j: 9 i
o{ 0 2 05 2 5 I0 20 50

Time of Vibration , rain.

FIG. 4--Effect of intensity of vibration on variations in relative density with time for dif-
ferent intensities of vibration (Graded Sand).

at several different intensities. The maximum accelerations (/~) were calcu-


lated from the relationship:
5) = (2~f)~.a (1)
where f and a are the frequency and amplitude of vibration. It can be seen
that a constant relative density was reached in all cases after a period of
about 30 rain, and this was adopted as the standard time of vibration in
all subsequent tests.

Effect of Intensity of Vibration


The intensity of vibration, which is expressed herein by the maximum
applied acceleration, greatly affected the rate of compaction of the samples
and the final relative density achieved, as can be seen in Fig. 4 for the
Graded Sand. The influence of intensity of vibration on the state of com-
paction of each of the single-size materials is summarized in Fig. 5, where
the porosity achieved is plotted against the maximum acceleration. In all
cases, the minimum porosity was obtained where an acceleration of about
1.4 g was used. The porosity increased rapidly as the intensity of vibration
decreased below the optimum value and decreased gradually as the inten-
sity increased.
The sharp peaks in the porosity-acceleration curves illustrate how critical
the acceleration is to the achievement of the maximum possible state of
compaction. The acceleration imparted to a medium enables reorientation
of the particles to the extent that the applied force is sufficient to overcome
the shear stresses between the particles. These stresses are a function of
the interparticle contact area and contact pressure, as well as the shape
and surface texture of the particles, and both of these increase as the
porosity decreases under the applied force. This decrease in porosity con-
tinues with increase in particle acceleration, therefore, until the rate of

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licen
BRAND ON THE CONTROL OF DENSITY BY VIBRATION 127

34 /•....•~ 7 3 v~

)~ = 7 6 %
! ~ -3. -. .- ., .I Sand
36

Gloss balls
58

:,:_ 4O
o
o ~ = 6 5 %
D-
42
, ----.
Fine Sand

46

48
0 I 2 5 4

Maximum Acceleration , g

FIG. 5--Effect of intensity of vibration on ultimate state of compaction.

I00 50
..... Vr = 95 %

- 52
8 0 - -

34
~ 60

g~
3 6 'N

> 40 g_
r,," 38

20

I 4O
7 -
0
0 I 2 3

Maximum Acceleration , g

FIG. 6--Effect of intensity of vibration and initial density on ultimate state of compaction
(Graded Sand).

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
128 RELATIVE D E N S I T Y I N V O L V I N G COHESIONLESS SOILS

interparticle shear becomes such that the particles can no longer occupy
their most favorable positions and, consequently, the porosity increases
with further increase in acceleration. At high accelerations, an equilibrium
porosity is reached which is independent of the acceleration, the rate of
shear being such that the particles exist in a state akin to liquifaction.
Specimens of the Graded Sand were made up at different initial relative
densities to ascertain whether this was an important influence on the
ultimate state of compaction. It is clear from Fig. 6 that this was very
important at accelerations below the optimum value. At the optimum
value and above, however, the initial density appeared to be unimportant
to the ultimate state of compaction obtained.
The maximum relative densities achieved by vibration for the five
materials are marked on Figs. 5 and 6. In no case was a relative density
of 100 percent obtained. Values in excess of 90 percent, however, were
obtained at the optimum acceleration in the cases of the glass balls and the
Graded Sand. For the single-size sands, the peak relative densities were
quite low, that of the Fine Sand being only 65 percent. The maximum
relative density that can be achieved by vibration, therefore, is obviously
a function of particle size and particle shape.

Effect of Container Size


To investigate container size effects on the state of compaction after
different times of vibration, four cylindrical molds were used, the internal
diameters of which were 3.87, 5.34, 9.40, and 14.60 cm. The results shown
in Fig. 7 were obtained using the Medium Sand. The only observations
that can be made on the basis of these results are that, where a small
cylinder was employed, erratic density values were obtained, and no con-

8O
F
70 1

lZ1 , 36 "~o
,....-------I
~ 50
.~_ 57 n
l
i
__~ cylinder diam. = 5.87 cm.

|
~ 4O
t 0 ,, 554 58
. ! [3 , 9.40 ,,
9 14,60
30 ,
0.1 0.2 0.5 I~) 2'0 50
Time of Vibration , min.

F I G . 7--JF,ffect of mold diameter on variations in density with time of vibration (Medium


Sand: ~ = 1.75 g ) .

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lic
BRAND O N THE CONTROL OF DENSITY BY VIBRATION 129

Porosity , %
55 54 55 52
22.5 I ' ! I !

c
o

1 5 - -
0

E
o

o
a3
E 7 . 5 - -

c
/ 5 I0 m i n
(o = 0.794 mm )
I
50 60 70 80 90

Relative Density , %

FIG. 8--Variations in density throughout specimen with time of vibration (Graded Sand:
= 1.75 g ) .

stant density was achieved after a considerable period of vibration. The


cause of these difficulties was undoubtedly the significant side friction
effects of the small cylinders. The two larger cylinders both appeared to be
satisfactory containers for the sand even though small differences in relative
density were measured for a given period of vibration.

Density Variations within Specimens


The experimental results reported so far were based on the average
values of density measured for the vibrated specimens. The most significant
results obtained from the study, however, related to the variations in
density throughout the specimens. This was investigated by usiag a 22.5-cm
long perspex cylinder which was split into four equal parts and held to-
gether by jubilee clips. After vibration for a given period of time, the
cylinder could be dismantled piece by piece so that separate density meas-
urements could be made on the sand contained in each section.
The results of the tests to determine the density variations within
specimens of the Graded Sand are shown in Figs. 8, 9, and 10. It can be
seen that the density varied greatly throughout the specimens with the
time of vibration (Fig. 8) and with the intensity of vibration (Figs. 9 and
10). In addition, a comparison of the data in Fig. 9 with that in Fig. 10

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
130 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

Porosity , %
40 38 36 54 32 30
,r I I i
N o v,brohon I I '

g i,,

- t

._~
a a = . mm.)

0
0 20 40 60 80 I00
Relative Density , %

FIG. 9--Variations in density throughout specimen with intensity of vibration (Graded


Sand: a = 0.79~ ram).

Porosity , %
4O 38 36 54 32 50
22.5 i w , i 3.48 g

.E
~3
i 5 - -
0
~a

o
rn

E 7 ' . 5 - -
o

C~
I (o = 1.588 mm.)
0
0 20 40 60 80 I00
Relative Density , %

FIG. l ( N - V a r i a E o n s in density throughout specimen with intensity of vibration (Graded


Sand: a = 1.588 ram).

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
BRAND ON THE CONTROL OF DENSITY BY VIBRATION 131

indicates the influence of the amplitude of vibration (at constant maximum


acceleration) in the process of dynamic compaction. Considerably more
longitudinal variation in density existed where the larger amplitude was
employed.
The main observations to be made from the results shown in Figs. 9
and 10 are that the preparation of specimens by vibrating at intensities
appreciably below the optimum value resulted in large inhomogeneities in
the specimens. Only where the intensity of vibration was at or above the
optimum were specimens obtained t h a t were even approximately homo-
geneous. At the optimum intensity (1.75 g) fairly homogeneous specimens
were obtained.

Conclusions
There are many problems involved in the use of vibration techniques
for the preparation of test specimens of granular materials. The relative
density is a function of time and intensity of vibration and appears to be
affected by secondary factors such as amplitude and mold size. The highest
density for a particular material can be achieved only by vibrating at the
optimum intensity, but the relative density obtained is unlikely to reach
a value of 100 percent. Large density variations throughout vibrated
specimens will probably exist where intensities of vibration below the
optimum are employed.
The results reported herein generally agree well with those obtained
earlier by Alyanak [3], Selig [4], and Kolbuszewski and Alyanak [6].

Acknowledgments
The experimental results presented in this paper were obtained at the
University of Nottingham, England by M. P. Rogers to whom the author
~ishes to express his thanks.

References
[1] Mogami,T. and Kubo, K., "The Behaviourof Soil during Vibration," Proceedings,
3rd International Conferenceon Soil Mechanics, Zurich, 1953, Vol. 1, pp. 152-155.
[~] Felt, E. J., "Laboratory Methods of Compacting Granular Soils," Symposium on
Application of Soil Testing in Highway Design and Construction, A S T M STP 239,
American Societyfor Testing and Materials, 1959, pp. 89-110.
[3] Alyanak, I., "Vibration of Sands with Special Referenceto the MinimumPorosity
Test for Sands," Proceedings, Midland Soil Mechanicsand Foundation Engineering
Society, Birmingham,England, 1961, pp. 37-72.
[4] Selig,E. T., "Effect of Vibration on Density of Sand," Proceedings, 2nd Panameriean
Conferenceof Soil Mechanics,Rio de Janeiro, 1963, Vol. 1, pp. 129-144.
[5] Pettibone, H. C. and Hardin, J., "Researchon Vibratory Density Test for Cohesion-
less Soils", Compaction of Soils, ASTM STP 377, AmericanSocietyfor Testing and
Materials, 1964, pp. 3-19.
[6] Kolbuszewski,J. and Alyanak,I., "Effects of Vibrations on the Shear Strength and
Porosity of Sands," Surveyor and Municipal Engineer, Vol. 123, No. 3756, 1964, pp.
23-27, and No. 3757, pp. 31-34.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No fu
132 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

[7] D'Appolonia, D. J. and D'Appolonia, E., "Determination of the Maximum Density


of Cohesionless Soils," Proceedings, 3rd Asian Regional Conference on Soil Mechan-
ics, Hails, 1967, Vol. 1, pp. 266-268.
[8] Kolbuszewski, J., "An Experimental Study of the Maximum and Minimum Porosities
of Sands," Proceedings, 2nd International Conference on Soil Mechanics, Rotherdam,
1948, Vol. 1, pp. 158-165.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
M . M . Johnston 1

Laboratory Studies of Maximum and


Minimum Dry Densities of Cohesionless
Soils

REFERENCE: Johnston, M. M., "Laboratory Studies o f M a x i m u m and


M i n i m u m Dry Densities of Cohesionless Soils," Evaluation of Relative
Density and Its Role in Geotechnical Projects Involving Cohesionless Soils, A S T M
S T P 523, American Society for Testing and Materials, 1973, pp. 133-140.
ABSTRACT: Some of the differences in results of tests for the maximum and
minimum dry densities of cohesionless soils are examined. A comparative test
program to investigate reproducibility of results of maximum and minimum
densities for two types of sands is discussed, and an empirical correlation of the
uniformity coefficient versus maximum and minimum densities is presented.
A comparison is made of the Providence Vibrated Density method and the
vibratory table method.
It is shown that one of the important variables in determining the maximum
density of cohesionless soils using the vibratory table method is the amplitude
of the vibrating mold.
KEY WORDS: density (mass/volume), tests, cohesionless soils, vibratory
compacting, coefficient of uniformity, value analysis

W i t h the increased need for defensive design of earth dams to resist


earthquake damage, the density of the cohesionless zones of embankments
a n d their foundations becomes increasingly important. Unfortunately,
there has been no general agreement on methods for determining the maxi-
m u m and minimum dry densities of cohesionless soils.
The concept of using relative density as a construction control method
began as far back as 1942 when K. S. Lane [1]z, then of the Providence,
Rhode Island District, U. S. A r m y Corps of Engineers, began developing
a standard test for determining the maximum density of cohesionless soils.

1 Civil engineer, Soil Mechanics Branch, Engineering Div., Officeof the Chief of Engi-
neers, U. S. Army Corps of Engineers, Washington, D. C. 20314.
2 The italic numbers in brackets refer to the list of references appended to this paper.

133

Copyright by ASTM Int'l


Copyright9 by(all rights International
ASTM reserved); Fri Marwww.astm.org
11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No fur
134 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

F r o m Lane's studies evolved a test procedure that is termed the Providence


Vibrated Density (PVD) test.
Research b y the U. S. Bureau of Reclamation in 1961 [2] and 1965 [3]
resulted in the development of a test method for determining the maximum
density of a cohesionless soil using an electromagnetic table-type vibrator
to which is affixed a steel cylinder containing the material. In 1965, the
U. S. Army Corps of Engineers [5] adopted the table-type vibrator method
(with Syntron vibrator) as a standard test procedure, but still permitted
the Providence Vibrated Method, somewhat modified, as an alternative
for use in locations when a vibratory table is not readily available.

Comparison of the Modified Providence Vibrated Density (MPVD)


and the Vibratory Table Density Test Results
Because of the basic differences in the modified Providence Vibrated
Density (PVD) and the vibratory table test methods, the Corps of Engi-
neers in 1970 made a comparison of the values of maximum dry densities
obtained using each procedure. Table 1 lists the results of a comparative
test program using oven-dry specimens of cohesionless soils tested in ac-
cordance with the two methods outlined in Engineer Manual (EM) 1110-2-
1906, Laboratory Soils Testing [5, 6, 7]. For the materials tested, the two
procedures give comparable results, although there is a slight trend for the
vibratory table method to give lower values.

Comparison of Original PVD and Modified PVD Test Methods


The original PVD test method included the use of a coiled spring to
deliver a surcharge having an average pressure of 26 psi. Because it is

TABLE 1--Comparison of results of M P VD and vibratory table test methods.

Soil Classification Maximum Dry Density, lb/ft 3 Referencea


Modified Vibratory
Providence Table Method
Vibrated Method

Well-graded gravel (GW) 139.2 139.7 5


Silty sandy gravel (GW-GM) 137.5 135.0 7
Well-graded sand (SW) 133.2 131.9 6
Well-graded sand (SW) 132.7 128.5 6
Well-graded sand (SW) 130.7 131.3 7
Poorly-graded sand (SP) 125.2 124.7 6
Poorly-graded sand (SP) 124.6 123.7 6
Well-graded sand (SW) 120.7 118.6 5
Well-graded sand (SW) 118.2 118.7 5
Poorly-graded sand (SP) 117.8 115.5 5

a Numbers refer to the list of references appended to this paper.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
JOHNSTON ON MAXIMUM AND MINIMUM DRY DENSITIES 135

TABLE 2--Comparison of results of original P V D and modified P VD test methods.

Maximum Dry Density, Ib/ft~


Providence Modified
Vibrated Providence
Density Method Vibrated
Soil Classification Density Method Referencea

Sandy gravel (GP) 143.2 139.3 7


Gravelly sand (SW) 141.2 137.4 7
Silty sandy gravel (GW-GM) 141.0 137.5 7
Sandy gravel (GP) 139.5 136.2 7
Gravelly sand (SW) 139.4 135.8 7
Gravelly sand (SW) 132.2 130.7 7
Gravelly sand (SW) 120.3 118.7 7
Gravelly sand (SW) 118.4 116.6 7
Gravelly sand (SW) 117.2 115.5 7
Poorly-gradedsand (SP) 113.2 117.2 6
Sandy gravel (GW) 133.7 137.8 6

Numbers refer to the list of referencesappended to this paper.

awkward to maintain this constant surcharge by a spring as the specimen


densities during vibration, the test was modified in 1965 to incorporate a
dead weight as a surcharge. For convenience in testing, a weight was
selected that delivers I psi surcharge. A comparison of the two procedures,
using oven-dry specimens of cohesionless soils, was made by the U. S. Army
Waterways Experiment Station (WES) and the U. S. Army Engineer
Division, New England, and the results are listed in Table 2 [6, 7]. For
the materials tested, the maximum densities derived from the original
PVD method are generally higher than those obtained using the modified
PVD method. This possibily may be attributed to the fact that the heavier
surcharge used for the original PVD procedure reduces rebound of the
individual granular particles and reduces segregation during vibration as
compared to lighter surcharge effects. Further, the higher amplitude in-
duced by the lighter surcharge in the modified method tends to produce
greater particle segregation.

Correlation of Grain-Size Distribution and M a x i m u m - M i n i m u m


Density of Cohesionless Soils
Since the maximum and minimum dry densities of cohesionless soils
are functions of their grain-size distribution and specific gravity, an em-
pirical relationship was sought. The coefficient of uniformity, C~, is one
indicator of grain-size distribution. A plot of C, on a logarithmic scale
versus the minimum and maximum dry densities on an arithmetic scale
is shown in Fig. 1. These data were collected from several Division labora-
tories within the Corps. The correlation is based on the results of tests on

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lic
136 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

COEFFICIENT OF U N I F O R M I T Y - Cu

2 5 10 20
14(

C)
13G MAXIMUM

120
| Je ~

110
I

k~

~ 1011

Q
~7
9(1 \--MINIMUM

Gffi2.65

FIG. 1--Empirical relationship between maximum and minimum densities versus coef-
ficient of uniformity.

subangular to rounded granular soils having all material retained on the


U. S. Standard 200 sieve and specific gravities ranging from 2.65 to 2.89.
The values have been normalized for a specific gravity of 2.65. The relation-
ship can serve to estimate values of maximum densities, if the grain-size
distribution and the specific gravity of the material are known. Enter the
chart with the value of Cu, select the dry density, and correct by multiplying
by GA divided by 2.65, where GA is the actual specific gravity of the
material.

Comparative Test Program to Determine Reproducibility of Results


Using Uniform Test Methods
To determine the variations in the minimum and maximum density
values among Corps laboratories, samples of two types of sand were pre-
pared at WES and sent to each of the nine Division laboratories. One
sample, designated Sand A, was a well-graded, sub-rounded clean sand;
the other, Sand B, was a poorly-graded, sub-rounded clean sand. The
laboratories performed grain-size distributions, maximum density, and

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lice
JOHNSTON ON MAXIMUM AND MINIMUM DRY DENSITIES 137

TABLE 3---G~rains~e d~tribut~vn.

Laboratory No. Sand A Sand B


Percent Passing U. S. Sieve Size Percent Passing U. S. Sieve S~e
4 10 20 40 60 100 200 4 10 20 40 60 100 200

1 100 58 38 17 11 4 1 100 87 65 33 13 4 1
2 100 58 37 16 8 4 1 100 86 66 35 14 4 1
3 100 60 39 17 12 5 1 100 85 65 34 15 4 0
4 100 61 40 17 12 4 0 100 86 66 34 13 3 0
5 100 61 40 18 10 4 1 100 86 64 33 12 3 1
6 100 58 37 17 10 4 1 100 86 63 32 13 3 1
7 100 61 40 17 12 4 0 100 86 66 34 13 3 0
8 100 63 40 19 11 4 0 100 85 66 33 14 3 0
9 100 61 39 17 12 4 1 100 86 64 34 13 3 1

minimum density tests on each sample using the procedures given in the
Corps Engineer Manual, " L a b o r a t o r y Soils Testing," E M 1110-2-1906,
1965 edition [4]. All tests were performed using oven-dry material. The
results are shown in Tables 3 a n d 4. As indicated b y Table 3, the grain size
distributions of the nine specimens are virtually identical. Table 4 shows
that large variations in m a x i m u m and minimum density values were
obtained.

A study [8] made at the Corps Southwestern Division laboratory indi-


cates that the amplitude delivered b y a vibratory table to the compaction
mold exerts considerable influence on the value of the dry density. A plot
of maximum dry density versus mold amplitude (Fig. 2) suggests that there
is an optimum amplitude for each type of granular material. This optimum
amplitude varies with the rheostat setting. The amplitudes obtained de-

TABLE 4--Results of maximum and minimum density Ses~.

Laboratory No. Sand A Sand B


Maximum Dry Minimum Dry Maximum Dry Minimum Dry
Density, lb/ft 3 Density, lb/ft 3 Density, lb/ft s Density, lb/ft s

1 132.1 115.1 124.3 107.2


2 131.6 114.7 124.0 108.6
3 130.1 115.3 123.4 110.3
4 130.0 115.2 122.8 107.1
5 129.8 113.3 122.7 108.7
6 129.0 111.9 122.7 108.0
7 128.4 112.1 122.5 105.0
8 128.4 108.8 122.3 107.1
9 127.8 110.8 120.9 107.0

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
138 RELATIVEDENSITY INVOLVING COHESIONLE$S SOILS

SAND B-~ ~SAND A

:=
tJ

o,j
X

:S

2
o
:E

0
110 115 120 125 130 135

MAXtMUM DRY D E N S I T Y - LSS/CU FT

FIG. 2--Re~tionshiy of mold amplitude versus maximum dry density using a Syntron
VP-80 vibratory table.

pend on the time given for the springs of a vibratory table to return the
deck of the table to or above its equilibrium elevation. Thus, the optimum
amplitude required must be determined by trial for each type of table,
surcharge, and type of material being tested in order to obtain the maxi-
mum dry density. One of the primary reasons for the differences shown in
Table 4 is that all Division laboratories did not determine the optimum
aptitude to give the maximum density using their particular vibratory
table.
It is probable that the large differences in minimum dry density values
were caused by variations in methods of "striking off" or leveling the sand
surfaces with a straight edge and by slight differences in height of fall of
the sand particles, permitting varying degrees of segregation.
Discussion
Because the value of the relative density is quite sensitive to small
changes in the values of maximum, minimum, and in situ densities, it is
important that testing techniques and equipment be universally stand-

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lice
JOHNSTON ON MAXIMUM AND MINIMUM DRY DENSITIES 139

ardized. For example, an increase in the apparent maximum density of


only 2 lb/ft 3 will lower the value of the relative density from 82 to 76
percent under certain circumstances. This could be the difference between
acceptance or rejection of a pervious fill or foundation. It should be noted
that while the maximum dry density of a cohesionless soil may be obtained
by use of vibratory methods, there are other methods, such as use of a
modified compactive effort using a falling weight, which give equal or
higher values.

Conclusions
For the specific material discussed in this paper, it is concluded that:
1. The MPVD method and the electromagnetic vibratory table method
for determining maximum dry density produce essentially the same values
considering the deviations with normal testing.
2. The original PVD method gives higher values for maximum dry
density than those derived from the MPVD method. A general correlation
may be found between the uniformity coefficient, C~, and the maximum
and minimum densities of cohesionless sands if less than 5 percent of the
material passes the No. 200 sieve.
3. The value of the maximum density using the electromagnetic vibra-
tory table method is dependent on the amplitude of the mold, with the
optimum amplitude being approximately 0.01 in. for the table tested.
The large variations in the minimum dry density values reported from
nine Corps laboratories using nearly identical techniques and samples are
probably due to segregation and varying degrees of disturbance in leveling
off the excess sand after initial pouring.
The importance of the relative density test indicates a need for standardi-
zation of techniques and equipment to produce consistent maximum a n d
minimum density values for cohesionless soils. More research using a wide
range of soils is required to examine the variables such as amplitude,
frequency, surcharge, effect of saturation, segregation, degradation during
vibration, and mold size encountered during the performance of the maxi-
mum density test.

References
[1] Lane, K. S. in Proceedings, SecondInternational Conferenceon Soil Mechanics and
Foundation Engineering, Rotterdam, 1948, pp. 243-247.
[2] "Development of a Maximum Density Test for CohesionlessSoil by a Vibratory
Method," Earth LaboratoryReport No. EM-557, U.S. Bureau of Reclamation,1961.
[3] "Laboratory Tests to Refinethe MaximumDensity Procedure for CohesionlessSoils
Using a Vibratory Table," Earth Laboratory Report No. EM-697, U.S. Bureau of
Reclamation, 1965.
[4] "Laboratory Soils Testing," Engineer Manual, EM 1110-241906, Headquarters,
Dept. of the Army, Officeof the Chief of Engineers, 1970.
[5] Compton, J. R. and Strohm, W. E., Jr., "Compactionof Cohesionless Materials,"

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
140 RELATIVE DENSITY INVOLVING COHESIONLESS ,SOILS

Miscellaneous Paper S-68-15, U. S. Army Engineer Waterways Experiment Station,


Corps of Engineers, 1968.
[6] "Comparison of Modified-Providence and Vibratory Table Methods for Determin-
ing Maximum Density," unpublished report by U. S. Army Waterways Experiment
Station, 1971.
[7] "Comparison of a Limited Number of Maximum Density Test Results," unpublished
Report by U. S. Army Division, New England.
[8] "Minimum-Maximum Density Tests of Standard Samples," unpublished report by
U. S. Army Corps of Engineers, Southwestern Division Laboratory, 1969.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No furt
G. Cumberledge 1 and R. J. Cominsky 1

Maximum Density Determination of


Subbase Materials

REFERENCE: Cumberledge, G. and Cominsky, R. J., " M a x i m u m Density


Determination of Subbase Materials~ ~ Evaluation of Relative Density and
Its Role in Geotechnical Projects Involving Cohesionless Soils, A S T M S T P 5~3,
American Society for Testing and Materials, 1973, pp. 141-155.
ABSTRACT: The objective of this study is to establish a satisfactory method
of determining maximum density of cohesionless material, such as subbase,
that will provide compaction criteria for these materials and be compatible
with current methods of density determinations in the field. To establish a
procedure for determining the maximum density of a subbase material, three
different methods of tests are employed. These tests include the standard
AASHO impact test methods (Methods C and D), a vibrating table type of
compactive effort (ASTM Test for Relative Density of Cohesionless Soils
(D 2049-69) ), and a rectangular mold vibrated from the surface.
The effects of mold size, amplitude, surcharge pressure, and duration of
vibration are investigated and analyzed. In addition, three types of subbase
materials--gravel, limestone, and slag--are used in this study to examine the
effects of subbase type on maximum density.
It is concluded that the interaction effects of aggregate type, mold size,
amplitude, surcharge pressure, and duration of vibration are statistically sig-
nificant on maximum dry density. The impact tests consistently produce higher
densities with the investigated subbase materials than the vibratory methods.
There is a very high degree of correlation between the densities obtained from
the impact methods and the vibratory method where the subbase materials are
vibrated at 0.030-in. amplitude and a surcharge pressure of 1 psi is applied.
Estimating equations are developed for predicting maximum density of subbase
material from the vibratory test data.

KEY WORDS. cohesionless soils, density (mass/volume), vibratory com-


pacting, impact tests, tests, molds, pavement bases, vibration, subgrades

The Pennsylvania Department of Transportation (PennDOT) currently


utilizes standard AASHO compaction tests for the control of field com-

1Assistant engineer of tests and soils research engineer, respectively, Bureau of Ma-
terials, Testing and Research, Pennsylvania Department of Transportation, Harrisburg,
Pa. 17120.

141

Copyright by ASTM Int'l


Copyright9 (all rights
by ASTM reserved); Fri www.astm.org
International Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement
142 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

paction of subbase materials. The present test methods (AASH0 T-99,


Methods C and D), which involve impact compaction do not, in many
cases, produce a "true" maximum density since certain types of subbases
generally become more dense under repeated traffic loading.
A research project was initiated to determine the maximum densities of
subbase material utilizing various vibratory compaction methods. Standard
AASHO compaction tests were also conducted for direct comparisons of
maximum densities among the various test methods with hopes that a
new density criterion for subbase materials might be established.
The effect of moisture on maximum density of subbase materials is not
of particular concern in this study, since moisture content is not really the
predominant factor in controlling the field density. It has been established
that greater density can be achieved at a lower moisture content by in-
creasing the compactive effort [I, 2, 3, ~]~.
The optimum moisture contents associated with subbase materials are
such that placement of the materials in the field at these moisture contents
would be physically impossible, since the conditions would be entirely too
wet for equipment maneuverability. Therefore, contractors will generally
place the subbase materials at a lower moisture content and rely on heavier
equipment (higher compactive energy) to achieve maximum density.
Since PennDOT's current method of pavement design considers the
strength of subbase materials, emphasis is placed on developing a better
understanding of the behavior of various subbase materials upon com-
paction and correlating the laboratory test findings to field compaction
results to provide more meaningful pavement design criteria.
Test Methods
Three methods of tests were utilized to establish a procedure for de-
termining the maximum density of a subbase material. Method I was the
impact method as described in the standard AASHO Procedure T-99,
Methods C and D, or in accordance with ASTM Tests for Moisture-Density
of Soils, Using 5.5-1b Rammer and 12-in. Drop, (D 698-70). Cylindrical
mold sizes of 4 in. (10.2 cm) in diameter and 6 in. (15.2 cm) in diameter,
0.03 ft 8 (850 cm 8) and 0.10 ft 3 (2832 cm3), respectively, were employed.
Method II was a vibrating table type of compactive effort and was
conducted essentially according to ASTM Test for Relative Density of
Cohesionless Soils (D 2049-69), except that tests were conducted at various
amplitudes and surcharge pressures. Cylindrical mold sizes of 6 in. (15.2
cm) and 11 in. (27.9 cm) in diameter, 0.10 ft 3 (2832 cm 3) and 0.50 ft 3
(14,161 cm3), respectively, were utilized.
Method III was essentially a vibratory type compactive effort utilizing
a rectangular metal mold with the vibrator force applied to the top of the

2 T h e italic numbers in brackets refer to the list of references appended to this paper.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licens
CUMBERLEDGE AND COMINSKY ON SUBBASEMATERIALS 143

specimen. The equipment consisted of a metal rectangular mold 15 in.


(38.1 cm) by 17 in. (43.2 cm) by 12 in. (30.5 cm). A rotary electric vibrator
was attached to a 15-in. (38.1-cm) by 17-in. (43.2-cm) by 0.50-in. (1.3-cm)
plate, which was placed on top of the specimen. The vibrator was equipped
with a variable speed control and was capable of producing a maximum
3000-1b (1364-kg) force at 6000 rpm. The specimens were vibrated for
various durations with predetermined surcharge pressures.
With the use of these three methods of tests, the effects of gradation,
amplitude, surcharge pressure, time of vibration, and type of subbase
material on maximum dry density were investigated and analyzed. The
analysis of variance procedure utilizing the F-test was employed to de-
termine if the effects of the variables considered within a particular test
were statistically significant on the maximum dry density obtained. Also,
linear correlation and regression analyses were performed on selected test
methods.

Material Types
The subbase materials selected for this study were chosen because they
were readily accessible and widely used in Pennsylvania. Three types of
material--gravel, limestone, and slag--were employed in this study. The
source of the subbase type was varied and a total of eight sources was
utilized.

Subbase Gradation
The gradation was chosen to be within the limits of subbase materials
according to PennDOT specifications (Form 408). The grading is shown
in Table 1. Methods I and II were employed to test the effects of the
amount of coarse and fine material on density. The percentage of material
retained on the No. 4 sieve (+No. 4) was designated as coarse material,
while the percentage of material that passed through the No. 4 sieve ( - No.

TABLE 1--Selected gradations to evaluate gradation effect on maximum dry density.

Sieve Designation Percent Retained PercentPassing

11/~in. (38.1 ram) 0 100


1 in. (25.4 mm) 12 88
3/~in. (19.1 ram) 15 73
in. (9.5 mm) 22 51
No. 4 (4.75 mm) 14 37
No. 10 (1.65 mm) 13 24
No. 20 (830 ~m) 7 17
No. 40 (420 ~m) 5 12
No. 60 (250 ~m) 5 7
No. 100 (149 ~m) 3 4

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
144 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

4) was characterized as fine material. The 6-in. (15.2-cm) mold size was
chosen, and the Amity Hall gravel was selected as the test material.
Reproducibility of Tests
The reproducibilities of Methods I and II were evaluated. With Method
II, the amplitudes of vibration were 0.0186 in. (0.047 cm), 0.0237 in.
(0.060 cm), and 0.030 in. (0.076 cm). The surcharge pressures applied were
0, 1 psi (0.07 kg/cm2), 2 psi (0.14 kg/em~), and 3 psi (0.21 kg/cm2). The
Amity Hall gravel was utilized as the test material. The arithmetic mean
(,~) of the maximum densities for each test was determined, and the
reproducibility of each test was expressed as the standard deviation (~)
about the arithmetic mean.
Effect of Air-Dried and Saturated Material on Maximum Density
The literature indicates that the highest densities with granular ma-
terials are obtained when the specimens are vibrated in either an air-dried
or a completely saturated condition [3, 5]. To evaluate the previous con-
ditions, tests were conducted employing the 6-in. (15.2-cm) mold of
Method II and vibrating the specimens at 0.030-in. (0.076-cm) amplitude
while applying various surcharge pressures. All eight subbase sources were
tested, first in an air-dried condition and then in a saturated state.
Mold Size, Amplitude, Surcharge Pressure, and Duration of Vibration
Forsbladd [3] stated that "there has been no systematized effort to de-
termine the relative effects of diameter and depth of mold individually and
collectively on the resulting maximum unit weight and optimum moisture
content." In this study, all three test methods were utilized to evaluate
the influence of mold size on maximum dry density. In addition, the test
procedures included all eight subbase sources.
The effect of amplitude was investigated only with Method II by using
various amplitudes of vibration, which were 0.0186 in. (0.047 cm), 0.0237
in. (0.060 cm), and 0.030 in. (0.076 cm).
The surcharge pressures were varied in Methods II and III. With
Method II, surcharge pressures of 0, 1 psi (0.07 kg/cm2), 2 psi (0.14
kg/cm2), and 3 psi (0.21 kg/cm ~) were applied, while with Method III,
the surcharge pressures were 0.33 psi (0.023 kg/cm~), 0.66 psi (0.046
kg/em~), and 1.17 psi (0.082 kg/cm~).
The effects of the duration of vibration were only evaluated with Method
III, and the selected periods of vibration were 1, 2, 3, 4, and 5 rain.
Results a n d D i s c u s s i o n
Subbase Gradation--The effect of varying the proportion of material
passing a No. 4 sieve on the maximum dry densities is illustrated in Figs.
1 and 2. A definite peak value for maximum density is achieved at a

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
CUMBERLEDGE AND COMINSKY ON SUBBASEMATERIALS 145

147t46 , , , , , i , ,~ , , ' , " , , ,

145 ~

~ i44

143

i40

138

137

' ' ' ' ' t i i i -,i I I i i i i


0 5 I0 15 .20 25 30 35 40 45 50 55 60 65 70 75 80 85
Percent Plus No. 4 Moterial

FIG. 1--Maximum dry density obtained by Standard AASHO Method (6-in. mold) by
varying the percentages of +No. $ material of the Amity Hall gravel.

i i i i F i r-" i

13E

132

120

116

0 5, lO
, 15
, 20
, 25, 30
, ;5 40
, 45, ;o 55
, 60
,
Percent Plus No. 4 MoteriO~

FIG. 2--Maximum dry density obtained by vibrating 6-in. cylindrical mold at O.030-in.
amplitude with 1-psi surcharge pressure and varying the percentages of +No..$ material of
Amity Hall gravel.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further
146 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

certain percentage of ~-No. 4 and - N o . 4 material for both methods.


This indicates that there is a ratio of coarse to fine material which should
be utilized when obtaining the greatest dry density possible by either the
impact or vibratory methods.
With lower percentages of fines, the aggregate experiences grain-to-grain
contact. However, it is very difficult to compact such an admixture and,
consequently, the density results can be quite variable. The aggregate
with sufficient fines will encounter grain-to-grain contact with increased
resistance against deformation. This condition will result in an increase
in density. The aggregate mixed with a large amount of fines will have
grain-to-grain contact destroyed and the aggregate will be "floating" in
the fines, which can lead to a substantial decrease in density.
Reproducibility of T e s t s - - N o attempt was made to determine the repro-
ducibility of the maximum dry densities obtained by Method III. Table 2
shows the reproducibilities of the maximum dry densities produced by
Methods I and II. The best reproducibility is achieved with the vibratory
table method and is ~-0.49 lb/ft 3 (• kg/cm3). With the vibratory
test, the application of a surcharge pressure of 1 psi (0.07 kg/cm ~) gives
the best reproducibility and the reproducibility of the test becomes better
with increasing amplitude, that is, a decrease in the standard deviation
(• lb/ft 3) about the mean value.
The reproducibility of • lb/ft 8 (• kg/cm 3) is probably due to
the proper combination of surcharge pressure and amplitude which prevent
the soil particles from moving in an erratic motion. With this vibration of
the total mass, more uniformity in the system is established, thus resulting
in a better degree of reproducibility.

TABLE 2--Reproducibility of tests performed on Amity Hall gravel.

Test Methods Reproducibility,lb/ft3

Impact, 4-in. mold •


Impact, 6-in. mold -4-0.80
Vibratory, 6-in. cylindricalmold
0 surcharge, 0. 0186-in. amplitude •
0 surcharge, 0.0237-in. amplitude •
0 surcharge, 0.030-in. amplitude •
1 psi surcharge, 0.0186-in. amplitude •
1 psi surcharge, 0.0237-in. amplitude •
1 psi surcharge, 0.030-in. amplitude • a
2 psi surcharge, 0.0186-in. amplitude •
2 psi surcharge, 0.0237-in. amplitude •
2 psi surcharge, 0.030-in. amplitude •
3 psi surcharge, 0.0186-in. amplitude •
3 psi surcharge, 0. 0237-in. amplitude •
3 psi surcharge, 0.030-in. amplitude •

a Test method showingthe best reproducibility.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement.
CUMBERLEDGE AND COMINSKY ON SUBBASE MATERIALS | 47

Statistically, however, when comparing the reproducibilities of the test


results between Methods I and II, the differences are not significant.
Therefore, for engineering purposes, the reproducibilities of the impact
methods are as satisfactory as the vibratory methods.
Effect of Air-Dried and Saturated Material on Maximum Density--In
general, with subbase-type material, the result is that maximum density
is achieved when the specimen is vibrated in the air-dried state. In those
few cases where the density of the specimen is greater when vibrated in a
saturated condition, statistical analysis indicates that the difference in
maximum densities is not significant [6]. At intermediate moisture contents,
the resultant densities are significantly lower. Therefore, for the remainder
of the research study, all specimens were vibrated in an air-dried condition.
Mold Size, Amplitude, Surcharge Pressure, and Duration of Vibration--
The analysis of variance approach, utilizing the F-test, is employed to test
the variables for statistical significance. The error term for the statistical
tests consists of a second order interaction variance estimate. The larger
the F-value, the greater the probability that the variance of a particular
factor is greater than the experimental error variance. When the F-value
is so large that it would have a probability of occurrence by chance of
1 or 5 percent, this would justify rejecting the null hypothesis that there
is no factor effect on maximum dry density. It then can be concluded that
the variable effect is real and significant.
Table 3 shows the results of the statistical tests performed on Methods
I, II, and III. For Method I, the type of subbase material (specimen)
has a definite effect on maximum density and this effect is significant at
the 1 percent level. The F-test indicates that the interaction effect between
mold sizes and specimens is nonsignificant. The statistical significance of
the effect of specimens indicates that the maximum densities vary among
the eight subbase materials. This should be expected even though the
same gradation is employed for each specimen. The variation in densities
can be attributed, at least in part, to the variation in particle shape among
the subbase materials. Moreover, the slag materials contain large quantities
of void space within the solid matrix which will greatly reduce the maxi-
mum dry density.
Table 3 also shows that for Method 1, the maximum densities are sta-
tistically significant when different mold sizes are utilized. The 6-in.
(15.2-cm) mold in Method I produced slightly higher densities for all of
the subbase materials. These results are in contrast to the results found
from similar studies conducted by Burmister [1] and Townsend and Do-
hancy [7]. The same compactive effort (12 395 ft.lbs/ft a) was utilized for
the 4-in. (10.2-cm) and 6-in. (15.2-cm) molds in this study. Individual
specimens were used for the determination of each point on the moisture-
density curve established by utilizing a 4-in. (10.2-cm) mold, whereas the
same specimen was employed for each point using a 6-in. (15.2-cm) mold.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Ag
T A B L E 3--Tabulated statistics~ employing analysis of variance technique on test methods.

Source of Variation Test Method

I II III
-4
Degrees of Degrees of Degrees of
F r e e d o m Fcalculated F.01 Freedom Fcal~ul~ted F.01 Freedom Fcal~ul~t~d F ,01

<
Main Effects z
o
Amplitudes ... ... 2 68.43 5.15
D u r a t i o n of v i b r a t i o n . . . . . .. . . . . . . ~ 82..'38 5.'68 0
M o l d size "1" ' 2.2..39 i. 82 . . . . . . . . . . . . . . . . . . X

Replications 2 2.17 6.32


546.31 4.03 "i 616: 89 3.'i0 "'i 59id.'77 2..98 o
Specimen types 7 z
Surcharge pressures ... ... 3 52.35 4.29 2 639.52 5.01 N

Interaction Effects
Amplitudes X specimens ... 9 14 4.32 2.54 . . . . . . . . .
Amplitudes • surcharge ... 9 6 3.06 b 3.26
D u r a t i o n X specimens ... . . . . . . . . . . . . "2i 3. 00 2..67
D u r a t i o n X surcharge . . . . . . . . . 8 2.40 b 2.85
Mold Size X specimen "7 " "2".'22 3.50
Specimens X surcharge ... ... "2"1' 89 b 2".'35 'ii i().'79 2".'i3
a Error t e r m for statistical tests consists of a second order i n t e r a c t i o n variance.
b Significant a t 5 percent level b u t n o t a t 1 p e r c e n t level.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
CUMBERLEDGE AND COMINSKY ON SUBBASEMATERIALS 149

Sieve Designation
LS" i" 514" 318" No. 4 NO I0 NO. 20 No. 40 NO, 60,
I00 , , , i , , , i J

90

8C

70
I--
6C
c
sAmity Hall Gravel
LE 50
r 7 ~ la~
t Sr . ple~._.' t 2~ . ~ /--Sheridan Slog
4c

~0

2O

I0

i i I I I i .

FIG. 3--Gradation curves illustrating aggregate breakdown of a gravel and slag samples.

Consequently, there is particle degradation with the 6-in. (15.2-cm) mold,


which results in higher densities due to the generation of extra fines (Fig. 3).
Table 3 illustrates that for Method II, all the first-order interactions are
significant at the 1 percent level except for the amplitude-surcharge pres-
sure interaction, which is significant at the 5 percent level. When an inter-
action is statistically significant, the corresponding main effects (amplitude,
surcharge pressure, specimen type) cease to have much meaning. The
nature and magnitude of variation for possible interactions should be
known, to some degree, before and while the experiment is being performed.
Therefore, it is always necessary to investigate what the significance of the
interaction represents.
Referring to Table 3 for Method II, the significant interaction effect
involving specimen type is to be expected, since the maximum densities
will vary greatly from material to material due to the variation of specific
gravity and particle shape of the materials. Figure 4 illustrates the graphical
presentation of the maximum density interaction effect between amplitudes
and surcharge pressures. In the absence of an interaction, it could be
expected that the four levels of surcharge pressure should follow the same
trend as the test proceeds from one amplitude of vibration to another.
In other words, they should be nearly parallel or at least resemble one
another in graphical form.
The divergence of the lines indicates that an interaction is taking place.
An interaction effect between amplitudes and surcharge pressures was
assumed to be occurring during the tests since the loading was a "dead
weight" type. When such a loading is used and the amplitude of vibration

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lice
150 RELATIVE DENSITY I N V O L V I N G COHESIONLESS SOILS

increases, the "dead weight" will rise and fall from the surface of the
material [2]. Consequently, the surcharge pressure possibly transmits
stresses to the material and causes an increase in density. Hence, the
existence of the interaction implies that the effect of amplitude on the
resultant density is markedly dependent on the level of the applied sur-
charge pressure. When quoting the effect of amplitude, it now becomes
necessary to specify the magnitude of the applied surcharge pressure.
It is interesting to note from Fig. 4 that the maximum densities increase
up to and including the application of a 2-psi (0.14-kg/cm 2) surcharge
pressure. With the application of a 3-psi (0.21-kg/cm 2) surcharge pressure,
the maximum densities are lower .than those produced by applying a 2opsi
(0.14-kg/cm ~) pressure. Evidently, the 3-psi (0.21-kg/cm ~) surcharge pres-
sure produces just enough effective stress before vibration that it is more
difficult to reduce the intergranular stresses sufficiently to permit particl e
movement to a more stable condition such as produced with the 2-psi
(0.14-kg/cm ~) surcharge pressure. Another explanation for this oddity is
that the vibrator associated with this particular apparatus might not have
been able to handle the 3-psi (0.21-kg/cm ~) surcharge pressure. Conse-
quently, the amplitude of vibration could have decreased to a lower value
causing a reduction in density. Hence, the effect of decreasing amplitude,
plus the reduction for the test materials to experience volume change
needed to accomplish a more stable particle position, create a compound
effect. This agrees with similar findings of Hardin and Pettibone [8]. When
no surcharge pressure is applied, the granular particles have a greater

Symbol Surcharge Pressure


155 None
I Psi
..... 2 Psi
X X X X 5Psi
o? 13o

~ x X X X ~,x
X X x
'~ 125
X
a X

_E izo
x

115

0.0186 0.0257' 0.050


Amplitude (in.)

FIG. 4--1nteraction effect between amplitudes and surcharge pressures developed by


Method I I .

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuan
CUMBERLEDGE AND COMINSKY ON SUBBASE MATERIALS 151

Symbol Surcharge Pressure


155 0.33 Psi
0.66 Psi
I.JT Psi
J
~; 130

v
.....-
I /
/

125
J //
f

E
~= J20

115

Duration of Vibration (Min.)

FIG. 5--Interaction effect between durations of vibration and surcharge pressures devel-
oped by Method I I I .

tendency to "float", producing lower densities than when a specific sur-


charge pressure is applied.
Essentially the same pattern of statistical results is established for
Method III (Table 3). Of particular interest is the significance of the
interaction effect between the durations of vibration and surcharge pres-.
sures. Figure 5 illustrates the interaction effect graphically. As the sur-
charge pressure increases from 0.33 psi (0.023 kg/cm ~) to 1.17 psi (0.082
kg/cm~), greater densities are achieved. However, these densities are de-
pendent on the duration of vibration, which implies that the significant
interaction effect must be accepted. This concept coincides with that
proposed by D'Appolonia et al [9], that increasing the surcharge pressure
requires longer periods of vibration to achieve the terminal density.
Correlation and Regression Analysis--There was a very high degree of
correlation (R = 0.99) between densities produced by employing the two
molds in Method I [6]. In all cases during this study, Method I produced
higher densities than the vibratory methods. These findings are contrary
to similar works reported in the literature. This anomaly can be explained,
at least in part, to particle degradation when employing the impact meth-
ods. However, the primary cause can be attributed to improper acceleration
of vibration of the vibratory table. This variable was not considered in the
investigation. Moreover, the frequency of vibration was fixed at 3600 cpm
as specified in ASTM D 2049-69. Felt [5] demonstrated, however, that
maximum density occurred with different soils at different critical fre-
quencies. Consequently, the frequency of vibration employed in this study
may not have been the critical frequency for the type of subbase materials
evaluated.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licens
152 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

To investigate and determine if any relationship existed between the


impact and vibratory tests, linear correlation and regression analyses were
conducted on the density results of Methods I, II, and III. The maximum
densities resulting from the impact methods were assigned as the dependent
Y-values, and the densities obtained by the various vibratory methods
were assigned as the independent X-values.
The highest degree of correlation was established by employing the
density results from vibrating a 6-in. (15.2-cm) cylindrical mold at 0.030-in.
(0.0760-cm) amplitude and applying a 1-psi (0.07 kg/cm ~) surcharge pres-
sure. Figures 6 and 7 show the scatter diagrams and the associated statisti-
cal parameters. Perfect linear correlation is unity. Since the linear corre-
lation coefficient (R-value) in this particular case represents how well a
given straight line describes the relationship between two test methods, it
can be inferred that there is a high degree of positive linear correlation
between densities. It is important to note that this vibratory test also
provided the best reproducibility (Table 2, -4- 0.49 lb/fta). The remaining
test combinations in Method II produced lower R-values. The test combi-
nations in Method III showed the lowest degree of correlation (average
R = 0.57).
The linear coefficient of determination (R 2) is defined as the square of
the linear coefficient of correlation and represents the ratio of explained
variation of the test to the total variation. For example, squaring 0.97

14C
R2=0.94 0y ~
S2ylx : 2228 . / ~
130
Sy.x =4172 f

>,,r
v 12C

E~:= I10

a I00
r

9o ,
I i I i i
9O I00 I10 120 130 140
Moximunn Dry Density (Pcf)- Vibratory

FIG. 6--Scatter diagram and regression line for Method I (4-in. mold) and Method I I
(6-in. mold).

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
CUMBERLEDGE AND COMINSKY ON SUBBASE MATERIALS 153

~'x=126.88 + 0 90iX-118.59) 0 J
140 R=096 Q//X
R2=O 92 / ~ /
S~'x = 2l'63 / 0
130
Sy x =465 /

~o
I I10

~, ,oo

90

90 I00 I10 120 130 140


Maximum Dry Density(Pcf.)- Vibrotory

FIG. 7--Scatter diagram and re~ression line for Method I (6-in. mold) and Method I I
(a-in. mold).

(R-value from Fig. 6) yields a coefficient of determination of 0.94. Sta-


tistically, this means that 94 percent of the total variation can be explained
by the variables investigated. Six percent of the total variation remains
unexplained. This could possibly be due to random fluctuation or to an
additional variable, such as acceleration or frequency, which has not been
considered.
The standard error of the estimate (Sy.z) for the two diagrams are
4.72 lb/ft s (75.75 kg/m s) and 4.65 lb/ft 3 (74.63 kg/mS), respectively. The
difference in the standard errors is not statistically significant. Therefore,
either estimating equation can be employed with satisfactory results.
However, the estimating equations are only applicable for vibratory test
apparatus used in this study, and they are also restricted to the subbase
materials investigated. Moreover, the test conditions must be such that
the test specimen is vibrating at 0.030-in. (0.076-cm) amplitude, in a 6-in.
(15.2-cm) mold while applying a 1-psi (0.07 kg/cm ~) surcharge pressure.
Conclusions
From the results of the different types of laboratory tests on the maxi-
mum dry density of cohesionless subbase materials the following con-
clusions may be drawn.
(a) There is an "optimum" gradation ratio of coarse to fine material

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lice
154 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

which should be utilized when obtaining the greatest dry density possible
b y either the impact or vibratory methods.
(b) The reproducibilities of the vibratory methods are as satisfactory as
the impact methods for engineering purposes.
(c) The interaction effects of subbase type, mold size, amplitude of
vibration, surcharge pressure, and duration of vibration are highly sig-
nificant on maximum dry density. Since an interaction effect implies a
differential response of the variable, it becomes necessary to state explicitly
the test conditions when reporting maximum density values of granular
materials as determined by vibratory methods.
(d) The impact methods consistently produce greater densities for the
subbase materials investigated. Moreover, there is a very high degree of
correlation between the 4-in. (10.2-cm) and 6-in. (15.2-cm) cylindrical
molds employed in the impact tests.
(e) Even though the vibrating table test methods produce lower den-
sities than the impact methods, there is a high degree of correlation between
the methods, thus enabling the maximum density to be calculated from
the vibratory test data.
(f) The vibrating rectangular mold method is the least desirable test
method for determining maximum density.
Acknowledgments
The authors express their appreciation to the Pennsylvania Department
of Transportation and the Federal Highway Administration for the sponsor-
ship and financial assistance which made this investigation possible; L. D.
Sandvig, director, W. C. Koehler, engineer of tests, W. L. Gramling,
research engineer, and R. K. Shaffer, research coordinator, of the Bureau
of Materials, Testing and Research for allotting the time and personnel to
conduct this study; and, especially A. C. Bhajandas for his helpful sug-
gestions during the preparation of this paper.
References
[I] Burmister, D. M., "Environmental Factors in Soil Compaction," Symposium on
Compaction, Sixty-Seventh Annual Meeting of the American Society for Testing and
Materials, Philadelphia, Pa., June 1964.
[2] D'Appolonia, E., "Behavior of Compacted Fills," Fifteenth Annual Soil Mechanics
and Foundation Engineering Conference, University of Minnesota, Minneapolis,
March 1967.
[3] Forsbladd, L., "Investigation of Soil Compaction by Vibration," Acta Polytechnica
Scandinavica, Civil Engineering Construction Series 34, 1965.
[4] Johnson, A. W. and Sallberg, J. R., "Factors Influencing Compaction Test Results,"
Highway Research Board Bulletin 319, 1962.
[5] Felt, E. J., Symposium on Application of Soil Testing in Highway Design and Con-
struction, A S T M S T P ~39, American Society for Testing and Materials, 1958, pp.
83-110.
[6] Cumberledge, G., and Cominsky, R. J., "Maximum Density Determination of Sub-
base Materials," Pensylvania Department of Transportation Research Report,
Bureau of Materials, Testing and Research, Research Project 67-15, May 1970.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further repro
CUMBERLEDGEAND COMINSKY ON SUBBASEMATERIALS 155

[7] Townsend, D., and Dohaney, W., "Relative Density Tests on Some Ontario Sands,"
Ontario Joint Highway Research Program, Ontario Department of Highways, Re-
port 20, Aug. 1963.
[8] Pettibone, H. C., and Hardin, J., "Research on Vibratory Maximum Density Test for
Cohesionless Soils," Symposium on Compaction of Soils, Sixty-Seventh Annual Meet-
ing of the American Society for Testing and Materials, Chicago, Ill., 1964.
[9] D'Appolonia, E. and D'Appolonia, D. J., "Determination of the Maximum Density
of Cohesionless Soils," Third Asian Regional Conference on Soil Mechanics and
Foundation Engineering, Haifa, Israel, Sept. 1967.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
Ricardo Dobry ~ and R. V. W h i t m a n 2

Compaction of Sand on a Vertically


Vibrating Table

REFERENCE." Dobry, Ricardo and Whitman, R. V., " C o m p a c t i o n of S a n d


o n a Vertically Vibrating T a b l e , " Evaluation of Relative Density and Its
Role in Geotechnical Projects Involving Cohesionless Soils, A S T M S T P 5~8,
American Society for Testing and Materials, 1973, 156-170.
ABSTRACT: A detailed study has been made of the factors contributing to
densification of dry sand on a shaking table. The variables considered were the
frequency of vibrations, amplitude of vibrations, and size and shape of con-
tainer. No surcharge was used in any test. The different processes affecting the
achieved density were found to be: (1) repeated change in vertical stress
owing to inertia forces within the sand when the peak acceleration is less than
1 g; (2) rearrangement of particles during free fall, when the peak acceleration
just reaches 1 g; (3) impact action at the end of free fall, when the peak ac-
celeration exceeds 1 g; and (4) spalling of the surface layer by stress wave
reflections, when the peak accelerations reach several g's.
K E Y W O R D S : cohesionless softs, tests, soil compacting, vibratory compact-
ing, sands, density (mass/volume)

Vertical vibration on a shaking table is the most widely used method for
determining the maximum density of a sand. The test included in the
ASTM Standard for Relative Density of Cohesionless Soils (D 2049-69)
is of this type. However, until a few years ago, the results of this kind of
test as reported by different authors often were contradictory [1].s Only
recently has a clearer picture emerged of the mechanisms involved in the
test [2-6]. Such better comprehension is essential in order to select, on a
rational basis, the correct combination of acceleration and frequency of
vibration, type and size of mold, moisture content, use of surcharge, and
time of vibration.

1 Professor, Soil Mechanics, Instituto de Investigaciones y Ensayos de Materiales,


University of Chile, Santiago, Chile.
2 Professor, Civil Engineering, Massachusetts Institute of Technology, Cambridge,
Mass. 02139.
8 The italic numbers in brackets refer to the list of references appended to this paper.

156

Copyright by ASTM Int'l


Copyright9 by(all
ASTMrightsInternational
reserved); Fri Mar 11 16:13:06 EST 2016
www.astm.org
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
DOBRY AND WHITMAN ON A VERTICALLY VIBRATING TABLE | 57

1.72 LT"~
l (~v)p= j2,~ =,-/.-~=

/,~'.e /

2.0

9~ /.0
o 1.6.
~

o 1.60

I I I I I I I I I
4o 80 120 160 s
Freque,~c/ f~ c p s

FIG. 1--Selig's densificatlan results: dry sand (Ref. 8).

Most researchers have concluded that the final density is controlled


primarily by the acceleration. Barkan [7] thought it to be the sole con-
trolling factor, but experimental results obtained by Selig [8] indicated
that both acceleration and frequency must be considered (Fig. 1). A major
step was the discovery that an important transition occurs when the
acceleration reaches 1 g. This fact, often masked by imperfect experimental
techniques, was shown by data presented in 1967 by D'Appolonia et al [9].
Since then, an impressive amount of evidence has accumulated showing
that the densitication process starts at I g, with very little or no compaction
produced below this value. This is true for both dry and moist sands with
no static surcharge, and it suggests that impacts between the soil and the
mold base are the main cause of compaction [2, 5, 6].
This paper presents the results of a detailed study made of the com-
paction behavior of a dry sand on a vibrating table. Only one soil was
tested, but the influence of all factors thought to be important, with the
exception of surcharge, was verified. Although some tests were also per-
formed using moist and saturated samples, they are not discussed here.
Theoretical work aided the interpretation of experimental results obtained
both by the writers and by other authors. Additional details may be found
in the original report [5].

Experimental Procedures and Testing Program


The test equipment in operation during a test with moist sand is shown
in Fig. 2. The shaking table produces vertical oscillations of approximately
sinusoidal shape, double amplitude 2y~ up to 0.15 in. within a range of
frequencies, f = 10 to 60 Hz. The amplitude is fixed before each test,
but f could be changed during operation. Indeed, the machine was always

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Ag
1 .$8 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

started at 10 Hz, and the frequency was then rapidly increased, manually,
to the desired value.
The nominal peak acceleration ap is
ap = 0.0511(2y~)J ~ (1)
where ap is in g, 2y~ is in inches, and f is in hertz. Direct measurements
were made of a~ by placing an accelerometer on a pedestal attached to the
mold as shown in Fig. 2. Measured values agreed with Eq. 1 within a
10 percent error. Accelerations reported in this paper are the nominal peak
acceleration a~.
Three cylindrical molds were used: (a) a steel Proctor mold, 6 in. diameter
and 6 in. high; (b) a collar mold consisting of a maximum of 12 steel collars,
each 1 in. high and 1~ in. thick, linked together to shape a container of
6.25 in. diameter and of variable height; and (c) a lucite mold (shown in
Fig. 2) 5 in. diameter and 45/~ in. high. Since all molds had comparable
dimensions, the importance of mold material could be studied. The collar
mold was useful for studying the influence of sample height.
All recording instruments appear in Fig. 2: an accelerometer; a cathode
follower to amplify the acceleration signal; and a two-beam oscilloscope.
In a typical test with dry sand, the amplitude control dial of the shaking
table was placed at the test value, and the mold was fixed to the table
with the accelerometer attached. First, the empty mold was vibrated at

FIG. 2---Equipment in operation, including recording instruments.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
DOBRY AND WHITMAN ON A VERTICALLY VIBRATING TABLE 159

TABLE 1--Factors studied.

Factor Studied Values Cited

Peak accelerationap Oto3g


Double amplitude 2yp o. 025, o. 050, and 0.150 in.
Mold type Lucite, Proctor, and collar
Sample height 3, 6, and 10 in.

the test frequency, and a photograph was taken of the acceleration versus
time display on the oscilloscope. Then the shaking was stopped and the
mold was filled with ovendried sand using a large scoop; this procedure
placed the soil at an initial density of about 1.39 g/cm 3, slightly above the
minimum density. The vibrator was turned on again, and the frequency
was increased as rapidly as possible to the desired steady-state value.
During the test a second photograph of the display was taken (see Fig. 3).
The test was stopped by turning off the switch, usually after 10 rain of
vibration. The mold was then taken off the platform, the upper collar(s)
removed, the top surface of the sand levelled, and the mold weighed to
give data for computing the density.
About 200 tests were performed. In this way the influence of the factors
indicated in Table 1 was studied.

Properties of Sand
The soil selected was a quartz sand of subangular grains. The particle
sizes ranged between 0.25 and 2 ram, and the uniformity coefficient was
1.7. The specific gravity of particles was 2.64, and the minimum density,
obtained by carefully pouring sand from a spoon, was 1.388 g/cm 3.
A series of tests was performed with a Harvard miniature mold using
different compaction techniques, to obtain independent information on the
maximum density of the sand. The value estimated from these results was
1.64 g/cm 3.

Test Results
One of the first conclusions of the research was that the mold type was
relatively unimportant as compared with the influence of the amplitude
for a v > 1 g. Therefore, results obtained with any mold can be taken as
being representative for all molds. Figure 4 shows the three plots of final
density -y versus ap obtained with the lucite mold. The most distinctive
features of these graphs are:
(a) Below 0.9 g there is little densification, and most of the densification
is produced in the range 0.9 to 1.1 g.
(b) In all cases there is a well defined peak density ~v and a corresponding

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lice
160 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreeme
v
0

z
0
z

<,

FIG. 3--Acceleration records, with and without sand, between ap = 0.9 and ~.5 g.
i
Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
162 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

optimum acceleration (a~)opt ~> 1 g; (ap)opt ranges from 1.1 to 1.3 g. The
loosening of the sand after the peak is not large and the sand remains in
a dense condition.
(c) At some point between 1.3 and 2 g the loosening process stops, and
either the density stabilizes or increases again.
Other observations confirmed that the behavior of the sand is different
below and above 1 g. The first observation arose from the fact that, in most
tests, the desired ap was achieved by increasing the frequency after shaking
started. Observing the sand as the acceleration was thus increased, there
was no major change in the sand until av ~ 1 g; then, in a few seconds
the surface settled appreciably. The second observation was that, for
av /> 1 g, the noise of impacts was clearly heard, especially when with
one's ear near the mold. As will be discussed subsequently, these impacts
occurred because the sand jumped free of the mold and then fell back
against the mold. The decisive proof of the existence of impacts came
when using the lucite mold. Owing to the flexibility of this mold, the im-
pacts were picked up by the accelerometer. Typical oscilloscope displays
showing the impacts are presented in Fig. 3; the impacts appeared at pre-
cisely 1 g and they were not present when the mold was vibrated empty.
The amount of densification produced below 1 g varied widely from one
series of tests to the next. This densification depends mainly on the im-
portance of high frequency vibrations superimposed to the main sinusoidal
shaking. In Fig. 5 all densities measured for av = 0.93 g have been plotted,
and they range from 20 to 70 percent relative density. Conversely, the
relative density for av = 1.11 g was 83 • 3 percent, with the exception of
one point. This means that the density reached immediately after impacts
began was notably constant, being independent of mold type and height

9 2y~ =0.02~"
Zuc/~e /"Io/o' -- ~- 2 yp = 0.o~o"
~ry -~no' o Z yp = O. / ~ o "

X6~

7,-.

q
/35 i i i i i i
o / 2 3
~k Ac-z2e/~ro~/of) j 9 ~

FIG. 4 - - T y p i c a l results for density as function of acceleration.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agree
DOBRY AND WHITMAN ON A VERTICALLY VIBRATING TABLE 163

D - Ib

& ~'~'Op= ZII9


0
/
/
/
/
ap = 0.959 .4 .~0\ /
\ /
\ /

Cq%, 1
I

L~ i i I I I i i i

Luc/fe Hold P/-oc~ ~o/d Co/Lot" M o l d


H=6 ~ H=6 ~ 14 = 6" /4=/0"

FIG. 5--Equilibrium d e n s i t y f o r up = 1.1 g .

and of vibration amplitude. This constant density (~ = 1.593 g/cm9 is


called the equilibrium density.
Theoretical Models
The simplest model for the behavior of the sand is [8, $, 5] a rigid block
sitting on a rigid table, as shown in Fig. 6. The table represents the mold
base; it oscillates vertically with a sinusoidal movement. Two situations
are possible for the block:
(a) If a~ < 1 g, the block displaces together with the table during the
whole cycle of vibration.
(b) If a~ > 1 g, the block displaces with the table only during part of
the cycle, and there is a period of free fall followed by an impact.
In either case, the whole history of displacements, velocities, and acceler-
ations of the block is completely determined if a~ and ~ --- 2~rf of the table
are specified. A detailed study has been made using this model [5]; some
results will be presented here.
There are two instants T1 and T~ which correspond to the take-off and
to the end of the free fall of the block. If T1 and T~ are defined in radians
( T = ~t = 2vft, where t is in seconds) and measured from the beginning
of the cycle, they depend only on a~/g:

sin T1 = -g (2)
a~

a~ [sin T1 -- sin T~ + (T~ -- T1) cos T1] -- ~ ( T ~ -- T,) ~ = 0 (3)


g

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
164 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

////

//// l
FIG. 6--Rigid body model: sketch.

T1 and T~ have been plotted versus ap/g in Fig. 7. To check the model,
a series of special tests with dry and moist sand were performed and T~
was carefully measured in photographs such as those shown in Fig. 3. T~
was defined as the distance between the beginning of the cycle and the
first spike of the impact. The experimental results have been superimposed
in Fig. 7, and they show the same trend as the theoretical curve, with the
measured values being slightly smaller, especially for high accelerations.
Another useful parameter computed from the rigid block model was the
velocity of impact Av, that is, the algebraic difference in velocities between
the block and the table at the time T~.

AV = a~ (cos T~ -- cos T1) -~ -g (T~ - T1) (4)


O~ OJ

~e NIoLS~ 5 o ~ d
8

/ T2 (thtore~,cQO
2~
6' I f o
0 Q
e

Peok Ar ,, 9 ~

FIG. 7--Times T, and T~ versus ap: theoretical and measured.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lic
DOBRY AND WHITMAN ON A VERTICALLYVIBRATING TABLE 165

5. 0 -~0 20 /6 /2

2.5

~ 2.0

K5

go
0 20 ~+O 60 80 I00 120 Ig*O /60

FIG. 8--Rela~ion among acceleration, frequency, and impact velocity.

As both T1 and T~ depend only on a~, hv is a function of two variables:


frequency and acceleration. This function has been plotted in Fig. 8 as a
family of curves on a graph of acceleration versus frequency.
The rigid block model provides useful kinematic information, but does
not tell what happens inside the sand mass. An additional insight comes
from assuming that the block actually is deformable. Then, according to
elastic theory, the maximum impact stress at the bottom is proportional
to Av:
,~ = p C A v (5)
where a~ = impact stress, p = mass density of material, and C = longi-
tudinal wave velocity of material. This impact stress is transmitted up-
wards as a compressive stress wave, and reflects from the top surface as a
tensile stress wave of equal magnitude. There are subsequent reflections,
but the stresses become smaller and finally disappear because of the
damping properties of the sand.
Thus, the impacts at the end of free fall induce d y n a m i c tensile stresses
proportional to pChv. On the other hand, there is at any depth z of the
column a static compressive stress ~g = pgz due to the weight of the
material. Dry sand has no tensile strength, and whenever a net tensile
stress appears inside the mass,.the phenomenon of "spalling" is produced;
that is, the grains separate. It is reasonable to assume that the loosening

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Ag
166 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

TABLE 2--Optimum acceleration measured and predicted using the equation


(AV)p = 0.2MH.

Mold Type SampleHeight, in. 2yp, in. Optimum Acceleration, g


Predicted Measured

Lucite 6 0.025 1.24 1.31


Proctor 6 0.025 1.24 1.23-1.31
Collar 6 0. 025 1.24 1.23
Lucite 6 0.050 1.18 1.24
Lucite 6 0.150 1.13 1.10
Proctor 6 0.150 1.13 1.10
Collar 6 0.150 1.13 1.10
Collar 3 0.025 1.12 1.10
Collar 3 0.150 1.06 • 1.10
Collar 10 0. 025 1.37 Flat cur/'e ~

There is a small peak for ap ~ 1.40 g, with a value of the density slightly larger than
for the rest of the curve.

of the sand after the peak (ap > (ap)opt) is caused by spalling within a
substantial fraction of the total depth H. If H / 2 is taken as a representative
depth, impacts will begin to loosen the sand when the net stress is zero
at z = H / 2 :
~ -- o-, = l~pgH -- BpC(Av)p = 0 (6)
and

g H (7)
(,~v)~ - 2 B C

where B < 1 is a coefficient incorporating a multitude of uncertain factors


which tend to reduce a~ (imperfect rigidity of the base, damping of the
stress wave, etc.). (5v)~ should be the critical value of the impact velocity
associated with (a~)opt; that is, Av = (Av)~ when ap = (a~)opt. According
to this interpretation, (hv)~ is the limit between two situations: for 5v <
(5v)~ increasing impacts densify the sand and for Av > (hv)p increasing
impacts loosen the sand.
By combining Eqs 2, 3, 4, and 7, it would be possible to solve for the
optimum acceleration as a function of y~ and H, provided that the quantity
g / 2 B C is known. To test this theory, the following procedure was adopted.
First, using the (a~)opt observed for the first eight series of tests in Table 2,
(5v)p was computed from Eq 4 and then g / 2 B C was computed from Eq 7.
The resulting values of g / 2 B C were averaged, giving the value 0.241.
Thus, Eq 7 became
(Av)p = 0.241H (8)
where H is in centimeters and (Av)p is in centimeters per second. Then,

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to
DOE, RY AND WHITMAN ON A VERTICALLY VIBRATING TABLE 167

this average result was used to backcompute (a~)opt for each test, giving
the results listed in Table 2. It may be seen that the theory correctly
predicts the observed trends; that is, the theory correctly predicts the
change in (ap)opt caused by changing 2yp and H.
The data by Selig [8] appear to provide further confirmation of the
theory. The (ap)opt from one of Selig's density versus acceleration, curves
was used to determine (~v)~ = 12.3 cm/s (the corresponding value of
g/2BC - 0.44). Since H was the same (11 in.) in all of Selig's tests, (Av)~
should be the same for all tests. This vMue of (Av)~ has been plotted in
Fig. 1, and this "line of spalling" does indeed appear to define the combi-
nations of frequency and acceleration giving maximum density.

e)
o- = O COMPACTION

F]:/j

Send oriqinally loose


Little energy needed
Rapid process (i.e. occurs in relatively
smell number of cycles)

FINAL DENSITY INDEPENDENT ON ENERGY INPUT

b)
i~ IMPACT COMPACTION

Send originally dense


Much energy needed
Slow process (i.e. requires reletively
large number of cycles)

FINAL DENSITY DEPENDENT ON ENERGY INPUT

FIG. 9~"~ = 0" and impact compaction processes.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No furt
168 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

Interpretation of Densifieation Processes


There appear to be two processes which lead to densification of sand on
a vertically vibrating table.
First, when ap ~ 1 g there is rapid densification, and the resulting density
(the equilibrium densityuabout 80 percent relative density for the sand
tested in this program) is independent of initial density, frequency, and
sample height. This densification occurs, apparently, because the initial
intergranular stresses are released during a portion of each cycle of motion,
thus allowing the particles to rearrange themselves in a denser packing.
It seems reasonable to assume that the equilibrium density represents the
densest possible packing which the grains can attain simply by releasing
their potential energies. This process, whose end product is the equilibrium
density, has been called " ~ = 0 Compaction" by the authors, and it is
depicted in Fig. 9a.
The obvious continuation of this picture for a~ > 1.1 g is that further
densification occurs because of increasing intensity of the impacts, which
provide the stresses needed to overcome the friction resistance from sur-
rounding grains. This process will then continue either until maximum
density is reached or until spalling appears. The compaction is now slower,
with particles advancing a little bit in each cycle. This process has been
called "impact compaction" and is depicted in Fig. 9b.
During impact compaction, it would be expected that the density would
be determined by the impact stress, which is in turn related to hr. The
experimental curves for ~ = 1.69, 1.74, and 1.75 g/cm 3 in Fig. 1 appear
very similar to the theoretical curves of Fig. 8. By computing the values
of hv corresponding to the coordinates (ap, f) of each point along the

/78

A m
v i

"~ L72

~ /.ZO
i v

/.r~ I I l I I I i
I g 3 ~ 5 6 ,5'

Irn,oocf" Ve/oc/~y A v ~ crns/sec

FIG. lO---Curve yd versus Av: from Selig's results.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement
DOBRY AND WHITMAN ON A VERTICALLY VIBRATING TABLE 169

. :s I ~ L / ~ oFspo/I/ng fSne of
I ~ . . ,t..~ peoks)
9 I o/1 {oosenec~ L/ i Av=(4v)p- ~ H
I
/rnt:~at
C OIT1 -
portion

I lJW ~X {=o~gerI m p ~e~l~,

r,~o~f, coZ>. ~- /W / / / / / / / / . ;W_~". @ z , ~ . ".~. " / / / / / / z -l-


Dens/IX P r o m I I
dense/ del~e~/~ I l
on teB/-)n,o I I
Cond/f/bnS A o L
FrequenCy/ cp~

FIG. l lIGeneral compaction behavior.

experimental curve, it was found that they are indeed curves of constant
~v. The Av necessary to achieve different densities have been plotted in
Fig. 10.
Figure 11 thus describes the densification behavior of dry sand in a
mold subjected to vertical vibration and with no surcharge. The effect of
varying one parameter, such as frequency or amplitude of vibration, may
be predicted by superimposing on Fig. 11 a trajectory describing the test
conditions. For example, a series of tests with constant frequency will plot
as a vertical straight line and a series of tests with constant amplitude as a
parabola. The conditions giving the peak density will be indicated by the
intersection of the trajectory and the line of spalling. For medium-coarse,
uniform quartz sands, a value of g / 2 B C = 0.30 s-1 is recommended for
preliminary estimates of the line of spalling.
Conclusions
1. A theoretical model has been developed to explain the compaction
behavior of dry sand in a mold subjected to vertical vibrations and no
surcharge. This model is summarized by Fig. 11 and checks reasonably
well with test results. The main factor controlling the final density of the
sand is the intensity of impacts which occur when the acceleration is larger
than 1 g.
2. For a test series with dry sand, the peak density will occur at the line
of spalling, which may be computed with Eq 7. For medium quartz sand,
a value of g / 2 B C = 0.30 s-~ can be used for preliminary estimates.
3. Peak density may coincide with or may be lower than maximum
density. The theoretical picture suggests that saturating the sand and

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lic
170 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

a d d i n g a surcharge weight on top of it could improve the efficiency of the


test b y reducing spalling. This suggestion agrees with past experience
( A S T M D 2049-69).

Acknowledgments
This work was carried out u n d e r the M . I . T . I n t e r - A m e r i c a n P r o g r a m
in Civil Engineering, and the financial s u p p o r t of the F o r d F o u n d a t i o n
a n d of the University of Chile are gratefully acknowledged. Various stu-
dents c o n t r i b u t e d to this s t u d y t h r o u g h special projects and t e r m projects,
especially Jos~ Paniagua, D a v i d Driscoll, and R a l p h Mittelberger.

References
[1] Luscher, U., Ortigosa, P., Rocker, K., and Whitman, R. V., "Repeated Load and
Vibration Tests upon Sand, Progress Report No. 1," Research Report R67-29 of the
Dept. of Civil Engineering, M.I.T., 1967.
[2] Ortigosa De Pablo, P., "Densification of Sand by Vertical Vibrations with Almost
Constant Stresses," Master thesis, Dept. of Civil Engineering, M.I.T., 1968.
[3] Whitman, R. V. and Ortigosa, P., "Densification of Sand by Vertical Vibrations,"
Technical Paper T68-5, Soils Publication 222, Dept. of Civil Engineering, M.I.T.,
1968.
[4] Lambe, T. W. and Whitman, R. V. in Soil Mechanics, Wiley, 1969.
[5] Dobry, R. and Whitman, R. V., "Densification of Sand by Vertical Vibrations in
'Standard' Molds," Research Report R70-05, Soils Publication 251, Dept. of Civil
Engineering, M.I.T., 1969.
[6] Krisek, R. J. and Fernandez, J. I., Journal of the Soil Mechanics and Foundations,
American Society of Civil Engineers, Vol. 97, No. SM8, Aug. 1971, pp. 1069-1079.
[7] Barkan, D. D. in Dynamics of Bases and Foundations, McGraw-Hill, 1948.
[8] Selig, E. T., Proceedings, 2nd Panamerican Conference on Soil Mechanics and
Foundation Engineering, Vol. 1, 1963, pp. 129-144.
[9] D'Appolonia, 1). J. and D'Appolonia, E. in Proceedings, 3d Asian Regional Confer-
ence on Soil Mechanics and Foundation Engineering, 1967, pp. 266-268.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
A. I. Johnson~ and D. A . Morris 2

Vibratory Compaction in the Laboratory of


Granular Materials in Long Columns*

R E F E R E N C E : Johnson, A. I. and Morris, D. A., "Vibratory Compaction


in t h e L a b o r a t o r y of G r a n u l a r M a t e r i a l s i n L o n g C o l u m n s , " Evaluation
of Relative Density and Its Role in Geotechnical Projects Involving Cohesionless
Soils, ASTM STP 523, American Society for Testing and Materials, 1973,
pp. 171-181.

A B S T R A C T : For a laboratory study of the drainage of long columns of porous


media, a maximum density, uniformly distributed throughout the colmnn,
was required. Research resulted in the development of a mechanical tech-
nique for the packing of drainage columns, as much as 60 in. long, with glass
beads and natural sands of various particle sizes. A vibratory packer used to
pack these columns, which are 1 to 8 in. in diameter, provided good repro-
ducibility of dry unit weight and porosity between duplicate columns as well as
a verticaJ uniformity of these properties within the same or duplicate columns.
To develop the standard method of packing columns, a study was made of the
effects of time, amplitude, and surcharge weight on the uniformity and repro-
ducibility of results. The technique was standardized at a packing period of 10
s and a vibratory amplitude of 0.09 cm.

KEY WORDS: density (mass/volume), compacting, porosity, void ratio,


specific yield, porous materials, tests, drainage, cohesionless soils

As part of a cooperative research project between the California Depart-


ment of Water Resources and the U. S. Geological Survey, specific yield
(the volume of water drained by gravity from saturated porous media)
was studied in the laboratory and in the field by the authors [1]3. One
phase of this study was the laboratory drainage of columns of porous
materials. To permit accurate comparison of drainage data from these
columns, a maximum, but uniformly distributed, density was required
throughout the porous media. Although the research reported in this
* Publication authorized by the Director, U. S. Geological Survey.
1 Assistant chief, Office of Water Data Coordination, U. S. Geological Survey, Washing-
ton, D. C. 20242.
2 Assistant district chief, U. S. Geological Survey, Anchorage, Alaska.
3 The italic numbers in brackets refer to the list of references appended to this paper.

171

Copyright by ASTM Int'l


Copyright9 by(all
ASTMrightsInternational
reserved); Fri Mar 11 16:13:06 EST 2016
www.astm.org
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
172 RELATIVEDENSITY INVOLVING COHESIONLESSSOILS

FIG. 1--Cam-actuated packer used for packing disturbed specimens of granular ma-
terials.

paper was not involved with a study of relative density, it is believed that
the data obtained and techniques developed will be of interest to those
involved in that subjeet.
In the U. S. Geological Survey's laboratory at Denver, Colo., a cam-
activated packer, (Fig. 1) designed by the senior author, had been used
since 1949 as a standard method of packing small cylinders of disturbed
specimens of granular materials [2]. However, this equipment was not large
enough to handle drainage columns 60 in. in length and 1 to 8 in. in diam-
eter, when filled with porous media.
During the first stages of the research in the laboratory, manual packing
by tapping the side of the columns with a rubber mallet did not provide
uniform porosity (the percentage of the total volume of the media that is
occupied by voids) throughout columns of porous media or reproducible
porosity in different cylinders of the same porous media. Porosities within
a column varied by as much as 10 percent--too great a range for the
research. Thus, a search was made for a mechanical method of packing
that would provide both uniform and reproducible porosities in columns
of porous media.
A review of literature produced little information on mechanical packing

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agree
JOHNSON AND MORRIS ON COMPACTION OF GRANULAR MATERIALS 173

methods for columns of porous materials, but some information was found
on related techniques. Pauls and Goode [3] used a vibrating table to pack
oven-dry materials and concluded that 20 rain of vibration was adequate
to obtain maximum density in the materials. Bartell and Albaugh [$]
investigated the use of vibrational methods for packing small volumes of
powder to maximum density and found that the vibrational method pro-
vided uniform and reproducible packing within a reasonable time.
Cusens [5] studied the vibratory packing in the laboratory of 2600-g
specimens of dry-mix concrete and found that a frequency of 3000 counts
per min and an amplitude of 0.004 in. provided optimum packing. Felt [6]
presented the results of a cooperative study of packing methods for pro-
ducing maximum density in six different granular materials. One free-fall
packer used 30 free falls of 18 in. and a surcharge of 4.2 psi for packing.
Also used were four vibratory packers with amplitudes of 0.01 to 0.02 in.,
frequencies of 3500, 3600 (twice), and 7200 counts per rain, and vibration
times of 10 to 45 rain. They concluded that the vibratory method with a
surcharge of 2 to 3 psi provided optimum packing.
The Earth Materials Laboratory of the Bureau of Reclamation [7, 8]
studied the use of mechanical vibrators as applied to determination of the
maximum density of soils. This study indicated that a mechanical vibrator
mounted under the container produced a more symmetrical vibration than
that of several vibrators attached to the sides of the container being
vibrated. This study also indicated that better reproducibility was ob-
tained by completely filling the packing container rather than filling and
vibrating by increments, and that higher densities were obtained by using
oven-dry media and surcharge weights.
Two methods of test for maximum and minimum density of granular
materials were suggested by Burmister [9] and Jones [10] in American
Society for Testing and Materials (ASTM) publications. Burmister's
method used a vibrating or drop-weight tamper, and Jones' method used a
foundry-type vibrator.
Following the library search, the authors concluded that additional
laboratory research was needed to evaluate the possible application of
mechanical vibrating equipment to the packing of long columns of granular
materials. Throughout the study, porosity was used as the property to
indicate the quality of packing.

Description and Operational Characteristics of Packer


After a preliminary study of several commercial packers, including jolting
as well as oscillatory equipment, the Syntron VP-60 vibratory packer 4
(Fig. 2) was chosen as most suitable for this study. This packer has a rheo-

The use of brand names in this report does not imply endorsement by the U. S.
Geological Survey. Their equivalents may be used for the same processes.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further r
174 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

FIG. 2--Vibratory packer and rheostat, with column clamped on packer platform.

star control for amplitude adjustment and a fixed frequency of 60 vibrations


per second (or line frequency). It is mounted on rubber, has a low noise
level, and is built for loads up to approximately 300 lb. It consists of a deck
or platform rigidly attached to an electromagnet. As a-c current energizes
the electromagnet, the platform is alternately attracted and repelled by a
permanent magnet rigidly attached to the base.
The amount of movement of the unloaded vibrator platform was de-
termined at five different settings of the rheostat by use of the General
Electric vibrometer shown in Fig. 3. This instrument, also used to de-
termine all vibrational amplitudes reported in this study, has a mechanical
amplification in a linkage between the vibration sensing element and the
stylus for the recording chart. The amplification factor was 10.5 for the
instrument used in this study. All reported amplitudes have been corrected

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agree
JOHNSON AND MORRIS ON COMPACTION OF GRANULARMATERIALS | 75

for this amplification. Table 1 shows the variations in amplitude obtained


in this study.
The change in amplitude of the packer platform at an arbitrarily chosen
rheostat setting (B), with the change of weight on the vibrator platform,
was then studied. This was done throughout the capacity of the packer by
adding the porous medium in 4.5-kg increments to a large cylindrical tube
which was attached rigidly to the packing platform. A decrease of 0.055 cm
in amplitude occurred as the load was increased from 6 to 135 kg. Seventy-
five percent of the change in amplitude took place in the first 35 kg of
loading. Only 15 percent additional change took place over the 35 to 80-kg
interval, and the remaining 10 percent change took place beyond 80 kg.
Effects of T i m e o f Vibration and Packing Method
The effect of length of time of vibration on porosity was studied using
columns about 2.5 cm in diameter and about 120 cm in height. Two columns
of 0.120-mm glass beads and one column of 20-mesh Del Monte sand
(Fig. 4) were filled by letting the beads and sand drop freely from the top
of the column. Two other columns of glass beads loaded with a tremie tube
were used. The columns, rigidly attached to the deck of the vibrator
(Fig. 2), were then vibrated for 300 s.
Although the initial porosity of the columns filled with a tremie was
considerably higher than the porosity of those loaded by free fall, this
study indicated that after 10 s of vibration the porosities in all the columns
closely approximated their final porosity. Because the columns filled by
the free-fall method and the columns filled with the tremie reached their
minimum porosity at about the same time, the tremie technique was

FIG. 3--Vibrometer used to measure amplitude of vibrations of packer~

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
176 RELATIVEDENSITY INVOLVING COHESIONLESSSOILS

T A B L E 1--Amplitude of vibration for unloaded vibra~ry packer.

Rheostat Amplitude, em
Setting
1st R u n 2d R u n 3d R u n 4th R u n

Minimum 0.04 0.04 0.04 0.04


A 0.07 0.07 0.06 0.06
B 0.09 0.09 0.09 0.09
C 0.13 0.13 0.12 0.12
D 0.18 0.18 0.19 0.18

j
I00

90

8 0 - -

8
o 70
z

<z 60

z S0

.J
~_

~
40

3o
E

/
/
10

Lo
d d d
PARTICLE-SIZE DIAMETER. IN MILLIMETERS

Distribution, in percent, for indicated particle size, in


Silt- and millimeters
Test material clay-size,
<0.062
Very fine Fine Medium I Coarse I Very coarse
0.062q).1____~ 0.125-0.2fi I 0.25-0.5 0.5-1 1-2

Glass beads .......... 52. 2 ~7. s ___~_~ .... ~_~ ...... i-~
Fresno sand ......... 4.3 5.15 17. 5
Del Monte sand ...... " I 9.61 85.01 4.8

FIG. 4--Particle size distribution of porous media used in packing study.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
JOHNSON AND MORRIS ON COMPACTION OF GRANULAR MATERIALS 177

adopted for filling the columns used later in the research study. The use of
the tremie eliminates the sorting which may occur during free fall.

Effect of Surcharge and Amplitude


Using a volumetric cylinder 14.6 cm high and 5 cm in diameter and
filled with glass beads 0.120 mm in diameter, the effects of surcharge and
amplitude on porosity were investigated. Using a tremie, the glass beads
were put into the cylinder, they were vibrated on the vibrating platform,
and the porosity of the medium was then determined after packing.
The effect of surcharge was determined by obtaining porosities with and
without the use of a 200-g weight placed on the glass beads during vibration.
The effect was determined at a rheostat setting of B (0.09-cm amplitude)
for vibration times of 25, 50, 100, and 200 s. A slightly lower porosity was
obtained when using a surcharge. The maximum and minimum porosities
obtained with the surcharge were 39.7 and 39.2 percent, respectively. The
maximum and minimum porosities obtained without surcharge were 40.0
and 39.5 percent. Since there was no indication of greatly increased uni-
formity or reproducibility between columns when using the surcharge, the
use of the surcharge technique was discontinued.
The effect of amplitude on porosity was determined over a range of
amplitudes from 0.05 to 0.17 cm. The effect of amplitude on porosity in
this range was small. The porosity again ranged from 39.2 to 40.0 percent.

Effect of Time of Vibration


The prior discussion on packing research suggested that an amplitude
and a vibration period of more than 10 s could be arbitrarily selected for
small columns of glass beads. However, a shorter period of packing may be
desirable, because a lesser amount of sorting may result.. Using the volu-
metric cylinder previously described and a vibrational amplitude of 0.09
cm, the porosity was determined over a 1 to 10-s period of packing. The
results indicate that the porosities were somewhat erratic over the range
of 1 to 4 s of vibration, but seemed to stabilize and give reproducible
results at a porosity of about 39.3 percent for vibration periods of 5 s
and over. The porosity was similar to that obtained when the longer period
of vibration was used.
To further check the reproducibility of porosities packed at the 0.09-cm
amplitude and a 10-s time interval, ten identical packing runs using glass
beads showed a variation of about 0.5 percent porosity and agreed very
closely with the packing results previously discussed. From the results of
this study, a technique for small volumes of porous media was standardized
at 10 s of vibration with an amplitude of 0.09 cm on the vibratory packer.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License
178 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

Packing of Long s
The column drainage research necessitated the use of 150-cm segmented
columns of porous media. Therefore, the application of the short column
technique to packing of long columns was then studied.
Duplicate segmented columns (Fig. 5), 2.5 cm in diameter, were filled
with 0.120-mm glass beads, with 20-mesh Del Monte sand, and with
Fresno medium sand (Fig. 4), and were packed by the standardized tech-
nique. After vibration, the dry unit weight and porosity of each column

D
i I/Pre ssure plate
t

Aluminum
/ channel

Plastic cylinder
section

,Clamp
\=
,t

um~num ~
angle
/

nsiometer

Outflow

[
I

FIG. 5---Assembly far segmented column.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reprod
HEIGHT ABOVE BASE OF COLUMN, IN CENTIMETERS
l --I l ~ I - r I r l
9 9 e 9 9 9 9 9 9 o 9 oo
f aJd P
~'~. ~,~
:~c:z
~
~'~ ~o
~~ = ~ ~ ~ e9
. 9 9 9 9 9 9 9 q 9
9 9
. 9
o9 el 9 9 9 9 9 ~
~ ~
~l ~ ' I ~ --~ ~---J~__ r I- ~ ' ]~ '

-I
]-
HEIGHT ABOVE BASE OF COLUMN. IN INCHES
HEIGHT ABOVE BASE OF COLUMN.IN CENTIMETERS
~
o
- - ~ I I -- f
Nn~
o''J:
n n o o o
-- I I I I --~-~LT~--I~I-~ i ~I, I I --~I t -- i
o I
~-..
HEIGHT ABOVE 8ABE OF COLUMN. ~N ~NCHES
t q3
HEIGHT ABOVE BASE OF COLUMN. IN CENTIMETERS
r i r--I I 1 I 1 ~--T-~ I t I
:o<
z~_ z
m~ m
NoN e9 9 o
o o eo
o o o

0~ " 9 =:- i~ = "~ 0 . o ,, o . o ' , 4lOOm


;S o I t e --/ J I r J t-- ~1 --J I I ~ I
HEIGHT ABOVE BASE OF COLUMN, ~N iNCHES
6Z| $1vRI~IIVW ~lvlrlNvaO -IO NOIL::)VdWO3 NO SRRIOW aNY NOSNHOr
Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
180 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

segment were determined. The porosities for the 0.120-mm glass beads
ranged from 40.0 to 42.5 percent (averaging approximately 41 percent),
for the 20-mesh Del Monte sand ranged from 34.5 to 37.0 percent (averaging
approximately 35.5 percent), and for the Fresno medium sand ranged from
35.5 to 40.0 percent (averaging approximately 37 percent). Figure 6 shows
that the vertical distribution of porosity had a maximum range of only 2.5
percent for the relatively homogeneous glass beads and Del Monte sand,
and 4.5 percent for the more heterogeneous Fresno sand. For all materials
tested, the reproducibility between columns varied only about 1 percent
in porosity. The porosity tended to increase toward the bottom of the
columns filled with Fresno medium sand; otherwise, there was no consistent
change in porosity throughout the columns.
The packing technique was then applied to columns of greater diameter.
The vertical distribution of porosity for columns of 0.120-ram glass beads,
2.5 to 20 cm in diameter, was then compared. Here again, the vertical
reproducibility was very good, the larger columns having a slightly wider
range of porosities than that of the 2.5-cm column. The average porosity
of the 10-cm column was approximately 1 percent less than the average
porosity of the 2.5-cm columns. The packing technique, 10 s of vibration
at an amplitude of 0.09 cm, now was considered to be applicable to long as
well as short column packing and was used as a standard method for all
future packing.
Summary
:~A mechanical technique using a commercial vibratory packer has been
developed for standardizing the packing of columns of porous media such
as glass beads and natural sands. The technique has been standardized at
a packing period of 10 s and at a vibratory amplitude of 0.09 cm.
A packing period of 10 s was short enough to keep sorting and inter-
mixing to a minimum and yet permit close to maximum settling, or mini-
mum porosity, to develop. For 60-in. long columns of 1 to 8-in. diameter,
the research indicated good reproducibility of porosity between duplicate
columns and good uniformity of porosity throughout any single column.

Acknowledgments
The authors are grateful for the assistance provided by their colleagues,
W. K. Kulp and R. C. Prill, during this study. The research was carried
out in cooperation with the California Department of Water Resources.
References
[1] Johnson, A. I., "SpecificYield--Compilationof SpecificYieldsfor Various Materi-
als," Water Supply Papers, 1662-D, U. S. GeologicalSurvey, 1967.
[2] Morris,D. A. and Johnson, A. I., "Summaryof Hydrologicand PhysicalProperties
of Rock and Soil Materials, as Analyzedby the HydrologicLaboratoryof the U. S.
Geological Survey, 1948-1960." Water Supply Papers, 1839-D. U. S. Geological
Survey, 1967.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further repro
JOHNSON AND MORRIS ON COMPACTION OF GRANULAR MATERIALS 18 |

[3] Pauls, d. T. and Goode, J. F., Public Roads, Vol. 20, No. 3, 1939, pp. 55-63.
[4] Bartell, F. E. and Albaugh, F. W., Proceedings, American Petroleum Institute,
Vol. 27, 1946, pp. 81-86.
[5] Cusens, A. R., Concrete Research, London, Vol. 10, No. 29, 1958.
[6] Felt, E. J. in Symposium on Application of Soil Testing in Highway Design and Con-
struction, A S T M STP ~39, American Society for Testing and Materials, 1958, pp.
89-110.
[7] "Development of a Maximum Density Test for Cohesionless Soil by a Vibratory
Method," Earth Lab. Report EM-557, Denver, U. S. Bureau of Reclamation, 1961.
[8] Pettibone, H. C. and Hardin, d., "Research on Vibratory Maximum Density Test
for Cohesionless Soils," Compaction of Soils, A S T M STP 377, American Society
for Testing and Materials, 1965.
[9] Burmister, D. M. in Procedures for Testing Soils, American Society for Testing and
Materials, 1950, pp. 111-113.
[10] Jones, C. W. in Procedures for Testing Soils, American Society for Testing and Ma-
terials, 1958, pp. 160-164.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
J. J. Emery, 1 W. D. L i a m F i n n / a n d 1~. W. Lee s

Uniformity of Saturated Sand Specimens

REFERENCE: Emery, J. J., Finn, W. D. Liam, and Lee, K. W., "Uniformity


of Saturated Sand Specimens," Evaluation of Relative Density and Its Role
in Geotechnical Projects Involving Cohesionless Soils, ASTM S T P 5~3, Ameri-
can Society for Testing and Materials, 1973, pp. 182-194.
ABSTRACT: Methods for checking the uniformity of saturated sand speci-
mens during their preparation and at various stages of cyclic loading triaxial,
simple shear, and shake table liquefaction tests are presented. These methods
involve the use of a gelatin solution to solidify the specimen so that it ean be
removed from the test apparatus for determination of the void ratio distribution.
A soil strain apparatus was also used to monitor the deformations induced within
the large shake table specimens during testing. The preparation methods
adopted for the liquefaction tests resulted in the production of uniform test
specimens. However, changes in the specimen's void ratio distribution during a
cyclic loading liquefaction test can result in the formation of a nonuniform
specimen with a loose top layer. Such loose zones can lead to a significant
reduction in the resistance to liquefaction of saturated sands.
KEY WORDS: cohesionless soils, density (mass/volume), sands, gelatins,
solidification, liquefaction, cyclic loads, variability, strain measurement, soft
mechanics

D a m a g e during m a j o r earthquakes caused b y the liquefaction of f o u n d a -


tion soils [1-3] 4 has focused a t t e n t i o n on research to determine the signifi-
cant parameters affecting the liquefaction potential of s a t u r a t e d sands.
M u c h of this research has involved cyclic loading triaxial tests [~, 5], cyclic
loading simple shear tests [6, 7], and shake table studies [8-10]. Field
studies have been limited to the observation of liquefaction a n d correlation
with the t y p e and properties of the f o u n d a t i o n soil involved [11]. This

i Assistant professor, Department of Civil Engineering and Engineering Mechanics,


McMaster University, Hamilton; formerly research associate, Faculty of Applied Science,
University of British Columbia, Vancouver, Canada.
2 Professor of civil engineering and dean, Faculty of A~plied Science, University of
British Columbia, Vancouver, Canada.
8 Graduate student, Department of Civil Engineering, University of British Columbia,
Vancouver, Canada.
4 The italic numbers in brackets refer to the list of references appended to this paper.

182

Copyright by ASTM Int'l


Copyright9 by(all
ASTMrightsInternational
reserved); Fri Mar 11 16:13:06 EST 2016
www.astm.org
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No f
EMERY ET AL ON UNIFORMITY OF SATURATED SAND 183

research has established the important variables controlling the incidence


of liquefaction and led to the development of various design guidelines
[12, 13].
To obtain meaningful data on the behavior of sand under both static and
cyclic loading, it is necessary to test homogeneous (uniform) specimens
under uniform states of stress and strain. However, there is still some
question about the possible development of loose zones within test speci-
mens, both during preparation and testing, that could initiate early failure
under cyclic loading conditions [14, 15]. For example, it has been suggested
by Castro [14] that the liquefaction behavior observed in many cyclic
loading tests may be due to the production of nonuniform void ratios
within the specimen. Finn et al. [9, 15] found that the resistance of saturated
sand specimens to further liquefactions (termed reliquefaction) was highly
dependent on the strains the specimens were subjected to during the
initial liquefaction. The loss in resistance to liquefaction of the now denser
specimens was attributed to the creation of a uniform metastable structure
or the development of a nonuniform specimen during the initial liquefaction.
This potontiaI loss in resistance to further liquefaction has significance in
the field, since deposits of saturated sand that have liquefied a number of
times may still be prone to liquefaction, even though they are now, on the
average, much denser [16]. This can probably be attributed to the formation
of loose zones within the deposit during the liquefaction process.
In order to determine whether significant nonuniformities are produced
in saturated sand specimens during their preparation and testing, it is
necessary to develop convenient and reliable techniques for measuring the
void ratio (density) distribution within the specimens. Various techniques
and apparatus that have been developed for determining the uniformity of
saturated sand specimens in an experimental study involving triaxial,
simple shear, and shake table liquefaction tests are described in this paper.
These methods involve the use of gelatin solutions to solidify the specimen.
A method for monitoring the strains induced within the large shake table
specimens during testing that utilizes a soil strain measuring system is also
described. Typical results obtained using the methods described are
presented along with suggestions for possible extension of the methods.
Procedure
Specimen Preparation
Full details of the procedure used during the triaxial, simple shear, and
shake table liquefaction tests are given in Refs 7, 9, 10, and 15. However,
it is pertinent at this point to briefly outline the main features of the
procedure used to prepare a uniform saturated sand specimen. Discussion
will be restricted to the preparation of shake table specimens, although
their large size involved some particular problems, because the general
procedure adopted is the same as for triaxial and simple shear specimens.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Ag
184 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

FIG. 1--Shake table test specimen preparation.

To prepare the large specimens for the shake table tests, a 45.7-cm high
extension enclosure was bolted to the open 183 by 45.7 by 17.8-cm specimen
container, and filled with water as shown in Fig. 1. A movable sand
spreader box with a submerged wire mesh bottom travels back and forth
automatically on top of the extension enclosure. As sand is poured into the
spreader box, the agitation of water flowing through a perforated pipe
around the lower inside of the apparatus is just sufficient to allow the sand
to flow through the wire mesh. The sand then falls through the water in the
extension enclosure to settle in the specimen container. The water flow in
the sand spreader box also assists in achieving a saturated specimen by
removing any air bubbles on the sand grains. This process is continued until
the specimen container is filled with sand. The specimen is then carefully
leveled to the required height of 17.8-cm by syphoning off any excess sand
above the final elevation without disturbing sand grains below the surface.
Great care is required during this step, since a disturbed looser layer of sand
at the top of the specimen could significantly reduce the apparent strength
of the specimen.
The resulting specimen is saturated, very uniform as will be shown later,
and has an initial relative density of 30 percent that is reproducible from
test to test. It is possible to vacuum saturate the sand required for the much
smaller triaxial and simple shear specimens, thus reducing the complexity
of the specimen preparation apparatus and procedure.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agree
EMERY El" AI, ON UNIFOP.MffY OF SATURATED SAND ] 85

Solidification of Specimens
The void ratio distribution within the specimens at various stages of
testing is obtained by solidifying samples or the total test specimen using a
gelatin solution. These solidified samples are then sliced into sections and
the distribution of void ratio within the test specimen is determined by
methods to be outlined in the following discussion.
Properties of Gelatin Solution
The gelatin solution used is made up of 2 to 5 percent gelatin crystals by
weight, (Kind and Knox Type 1720 or any other suitable type) which will
be referred to as the gel concentration, and water. This gelatin solution has
the following properties that make it very suitable for use in the determina-
tion of void ratios in saturated sand specimens:
1. The gelatin solution produces an homogeneous gel upon setting which
has a specific gravity approximately equal to that of water, that is, unity.
Specific gravities of 21/~ and 5 percent gels at various temperatures are
shown in Fig. 2. The specific gravity of the 2 ~ percent gel is 1.006 at 20 C,
which is the approximate laboratory temperature at which the tests are
made.
2. For the purpose of solidifying sand samples, varying gel concentra~
tions can be used depending on the desired time until solidification occurs.
A typical solidification time and gel concentration curve is shown in Fig. 3.
The gelatin solutions were prepared at approximately 65 C, which is the
1.02C
-'"'l,"' '"'r""l"',=" ..... 'l"" '"'l'"' ""l'"'l'"'l'"'-
:- I ---_
LOIC I/ I

I.OOC

r
,'7
0.99C
I I
_--- 9 5'% GEL CONCENTRATION
I 9 ~ --
"- 9 2 I/2 %GEL CONCENTRATION
O.98C - 9 WATER 9

0 I0 20 .30 40 50 60 70
TEMPERATURE, deg. C

FIG. 2--Specific gravity of gelatin solutions at various temperatures.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
186 RELATIVEDENSITYINVOLVING COHESIONLESSSOILS

15 I I I I I I I I I
(/) ,,,iw,,,i,,,,i,,,,I,,,,i,,,,i,,,,i,,,,
(2: GELATINSOLUTIONPREPAREDAT 65 C
o- r MAINTAINEDAT 20-*1C
z I0 ~ t
o_
g
b-

-J
0(D -

0 l|l| Jill Jill


0 I 2 3 4 5
GEL CONCENTRATION,PERCENT

FIG. 3--Solidification time of gelatin solutions at 20 C.

water temperature at which the gelatin crystals are dissolved quite readily,
and then put in a water bath at 20 C which represents the laboratory
temperature.
3. Apart from the small expansion or contraction of the gelatin solution
as its temperature is changed, it can be seen from Fig. 2 for the three points
where the gelatin solution has solidified that there is no measurable volume
change during the actual solidification stage. This characteristic is extremely
important, since it indicates that there will be no change in the void ratio
of the sample during the solidification process.
4. The solidified sample is soft enough so that it can be readily cut into
sections with a wire saw. When a low gel concentration has been used to get
a long solidification time, the sample's stiffness can be increased by storing
it in a refrigerator (approximately 5 C) for a few hours before the cutting
process. This ensures that there is very little disturbance of individual sand
grains during the cutting of sections.
5. The solidified gelatin solution filling the sample voids can be washed
out very easily using warm water (approximately 65 C). This enables the
weight of the sand making up the slice to be determined and hence the
slice's void ratio.

Solidification Procedures
There are two basic procedures for using gelatin solutions to form
solidified samples: the displacement method and the pore fluid method. In
the first method, the pore water in the saturated specimen is displaced by
the gelatin solution and then the sample is allowed to solidify. Use of the
displacement method is possible when the testing apparatus provides
drainage at both ends of the specimen, as is the case for the triaxial
apparatus shown schematically in Fig. 4. A small pressure, Ap, is applied
to the gelatin solution reservoir so that the solution can displace the pore

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lice
EMERY ET AL ON UNIFORMITY OF SATURATED SAND 187

water when the one-way valve in the top loading cap is open. The sample is
then allowed to solidify before removal from the cell. It is assumed that the
small gradient within the sample caused by ~p will not change its structure.
This displacement method can also be used to solidify samples in large
masses of sand such as the shake table test specimens. It is possible to use an
open-ended thin-walled tube to isolate a small sample of the shake table
test specimen as indicated in Fig. 5. A small gradient is created to displace
the pore water in the sample by filling up the top of the tube with gelatin
solution. Care must be exercised in placing the tube and removing the
surrounding sand to avoid disturbing the sample.
In the second method of sample solidification, the gelatin solution is used
as the pore fluid at all stages of preparation and testing so that a double
drainage system is not required. To use the pore fluid method, care must be
taken to select a gel concentration (Fig. 3) such that straining or shearing
of the specimen is completed before solidification occurs. Also, it must be
ascertMned that the pore fluid method does not affect the test results.
Comparisons using the simple shear apparatus (which has single drainage)
indicate that the use of the gelatin solution as the pore fluid has no measur-
able effect on the test results.
Void Ratio
The void ratio (e) is determined for each slice from the solidified sample
using the expression:

e --
Wd
in which wt is the weight of the slice containing gelatin, wd is the dry weight

O BALL VALVE
~ IP
,~ ONE-WAY VALVE
LOADING
p+/kp
CAF I |OVERLOW
~ _ _ ~ RESERVOIR

SOLUTION
RESERVOIR ~IAXIAL SPEL'-:'~CIMEN
~ POROUS STONE

O] PORE PRESSURE
NOTE: (P*AP) < 0"3 TRANSDUCER

FIG.4--Solidification of specimen in triaxial apparatus.


Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
188 RELATIVE DENSITY iNVOLVING COHESIONLESS SOILS

of the slice after the gelatin is washed out, G, is the specific gravity of the
sand, and Gg is the specific gravity of the gelatin solution. Care must be
taken to insure that no drying of the slice occurs before wt is determined.
The ratio (G,/Gg) can be taken as G, within the experimental limits
involved. It is advisable to check the average of the void ratio values for all
the slices with the overall void ratio of the solidified sample. In the present
research, it has been found that this difference in e is not more than 0.02.

Soil Strain Apparatus


The Bison soil strain apparatus that was used during the shake table tests
to monitor dynamic displacements within the specimen has been described
in detail for static applications by Selig and Grangaard [17]. For applica-
tions involving cyclic loading, the output from the soil strain apparatus is
recorded on a storage oscilloscope or light beam oscillographic recorder. The
location of the 2.54-cm diameter strain sensor pairs, the pore pressure
probes, and the accelerometer used to monitor the shake table tests is
indicated in Fig. 6 along with a typical oscillographic recorder output.
A multiplexer was not used with the soil strain apparatus so that only a
single pair of sensors was monitored on the oscillographic recorder. The
strain sensor pairs can be used to monitor either horizontal or vertical
displacements within the specimen. With the large size of test specimen
involved, the presence of the various probes in the soil mass should have a
negligible effect on its behavior during shaking.

FIG. 5--Sample solidification tubes in place.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
EMERY ET AL ON UNIFORMITY OF SATURATED ,SAND 189

FIG. 6--Monitoring of shake table test.

Results

Uniformity of Shake Table Specimens


The uniformity of shake table specimens at various stages of testing was
determined using the displacement method to solidify the samples and then
slicing the recovered samples parallel to their bottom. Void ratio distribu-
tions for four stages of testing are indicated on Fig. 7. Specimens 55 and 58,
prepared with no surcharge pressure or shaking applied, were extremely
uniform and showed little difference in void ratio. However, Specimens 53
and 60, which had been liquefied and subjected to about 15 cycles of
shaking after liquefaction, were not nearly as uniform, and the specimens

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agree
190 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

~ ~ z , i n u u t [ n n u u I~ v u u J i ~ u ~

SHAKING: SINUSOIDALACCELERATION OF +-025g AT


2 HZ, SURCHARGE PRESSURE : 0.14 Kg/crn2

N~ }AS PREPARED,NOSURCHARGEOR SHAKING


i NNeo.530
jLLIQUEFIEDONCE,NOTDRAINEDAFTER TESTING
9 No.62 LIQUEREDONCE, DRAINED AFTER TESTING
= No.64 LIQUEFIED TWICE,DRAINEDAFTER TESTING
2C
TOP
9 eD
~.--7, ' ~
0O

,/
~u
O
tJ
~
m
,,,~.
,o eo
-/
Tm

!
0
l"- F_

~LI_ 5-

O m N
=--I I i I I I I t I I I i ~ i I i I I I I i
BOTTOM 45 0.50 0.55 0.60 Q65 07tO
VOID RATIO,e

FIG. 7--Variation in void ratio of shake table test specimens.

were looser at the top and bottom. Specimen 53 was particularly loose at
the top. There had been a net decrease in void ratio throughout the
specimens as the sand grains settled during shaking after liquefaction, and
a layer of water formed under the membrane at the top of the specimen.
When Specimen 62 was liquefied once and then drained, it appeared that
if the surcharge pressure was high enough, that any looser layer at the top
was eliminated. This would explain, in shake table tests, why a decrease in
the resistance to further liquefactions had only been noted for extremely
low surcharge pressures [9]. The effect of further liquefactions towards a
more uniform void ratio distribution was indicated by Specimen 64. The
trend for the shake table specimens moved from a uniform as-prepared
specimen to a denser nonuniform specimen after the first liquefaction, and
then to an extremely uniform specimen with further liquefactions.

Uniformity of Simple Shear Specimens


The uniformity of simple shear specimens after preparation and at
various stages of cyclic straining was investigated using the pore fluid
method. The specimens were not drained after cycling. Tests indicated that
the variation of void ratio within the specimen is most pronounced with
height, and not length or width. For this reason, slices parallel to the

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to L
EMERY ET AL ON UNIFORMITY OF SATURATED SAND 191

bottom of the specimen were used. There are four stages of testing, A, B, C,
and D, at which the uniformity was investigated:
A. Specimen as formed with the seating load procedure completed [7].
This procedure consists of normally loading the specimen to 4 kg/cm ~,
reducing the normal load to 1 kg/cm ~, and finally, increasing the normal
pressure to 2 kg/cm 2 which is that used during testing.
B. Specimen subjected to 10 percent quasi-static undrained strain and
then brought back to the zero strain position.
C. Specimen just liquefied with strains limited to 4-3 percent, where
liquefaction is considered to have occurred when the pore pressure equals
the normal pressure.
D. Specimen liquefied and then subjected to an additional 15 cycles of
straining with the strain limited to + 3 percent.
The void ratio distribution along the height of the simple shear specimens
for stages A, B, C, and D is given in Fig. 8. It should be noted that averages
for several specimens are shown for each stage, and the specimens did not
all have the same initial average void ratio before cyclic straining. The
resemblance of the distribution for stages A, B, and C and the dissimilarity
of that for D shows up clearly. There is a tendency, which increases in
going from A to C, for the lower portion of specimens at stages A, B, and C
to be somewhat looser than the top portion. However, while a definite trend
shows up, the difference in void ratio is quite small and, in fact, approaches
the accuracy limitations of the method~about • to 5=3 percent.
The void ratio distribution for stage D is quite different. Since a layer

t
i ' I ' i ' I ' I ' I
9 A AS P R E P A R E D ( A V E R A G E OF 5 }
9 B 1 0 % Q U A S I - STATIC U N D R A I N E D S T R A I N ( AVERAGE OF 3 )
9 C JUST LIQUEFIED (AVERAGE OFh)
o D LIQUEFIED PLUS t5 CYCLES OF STRAIN (AVERAGE OF 3 )
(SEE TEXT FOR COMPLETE DETAILS )
, [

TOP-- D C B A - -

.0 9 4 9

m
d
t
I O 0
BOTTOM 0 I i I I I I I
0.50 0.52 0.54 Q56 Q58 G60 062 064
VOID RATIO,e

FIG. 8--Variation in void ratio of simple shear test specimens.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
192 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

of gelatin solution forms on top of the specimen as the sand grains settle, the
void ratio is much lower throughout the specimen than for stages A, B, and
C. Moreover, the void ratio distribution is not nearly as uniform, and the
specimen tends to become denser towards the center. It would appear that
continued cyclic undrained shear after liquefaction creates a layer of looser
sand on the top of the specimen that can influence the resistance to further
liquefaction. A comparison with the uniformity tests on shake table
specimens indicates that many of the same trends hold.
Uniformity of Triaxial Specimens
Uniformity studies on triaxial specimens have been limited, at present,
to checking the specimen as-prepared, since an effective one-way valve for
incorporation in the top loading cap, as shown in Fig. 4, has only been
developed recently. It is quite important that this one-way valve be in the
top loading cap to reduce the compliance of the system. The studies on the
as-prepared triaxial specimens used the pore fluid method and were
intended to check the use of a small vibrator to obtain dense specimens.
Initially, the vibrator was applied to the triaxial base plate, but it was
found that this method resulted in a specimen that was much denser near
the base. However, by using the vibrator along the side of the specimen
former, a relatively uniform dense specimen was produced. This vibration
method was also adopted for the preparation of dense simple shear
specimens.
These studies also indicated that great care is required in placing the top
loading cap on the specimen to avoid creating a denser top layer of sand
under the cap.
Soil Strain Measurements
Most of the soil strain measurements have been made in a series of shake
table tests with a sinusoidal acceleration of amplitude 0.25 q at 2 Hz, and
a surcharge pressure of 0.14 kg/cm 2 which represents a soil layer near the
ground surface. Both horizontal and vertical strain measurements have
been made with the strain sensor pairs in the positions indicated by Fig. 6.
The horizontal strains after liquefaction were an extension of 5.1 percent
for the top sensor pair and a compression of 0.6 percent for the bottom
sensor pair. The corresponding vertical strains were a compression of 2.7
percent and a compression of 4.7 percent. These values show the trend that
would be expected from the void ratio distributions for Specimens 53 and 60
in Fig. 7. While there is an overall decrease in void ratio as indicated by the
vertical strains, the extensional horizontal strain near the top is reflected in
a looser surface layer.
Discussion
Based on the void ratio distributions that have been determined for
solidified shake table and simple shear test specimens, it would appear that

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pu
EMERY ET AL ON UNIFORMITY OF SATURATED SAND 193

the preparation methods that have been developed result in the production
of uniform test specimens. However, continued shaking or shearing after
liquefaction causes the specimen to densify at constant overall volume,
giving a layer of water, and usually a layer of looser sand, at the top of the
specimen. This layer of looser sand can lead to a significant reduction in the
resistance of saturated sand specimens to further liquefactions. These
trends have been confirmed b y the measurements of strain in the shake
table tests.
The displacement method of simple solidification has been used in the
field with some success. ~ In this approach, the drill hole is filled with a
gelatin solution to a level providing a small head. When solidification is
completed, it is possible to recover a sample with a solidified undisturbed
portion. It was possible in this particular case to observe the foreset bedding
in the undisturbed sand sample.

Conclusions
The displacement and pore fluid methods of solidification using gelatin
solutions have proved very convenient for checking the uniformity of
saturated sand specimens. Various results that have been discussed indicate
how critical the preparation method is in achieving a uniform specimen.
Changes in the specimen's void ratio during cyclic testing can result in the
formation of a nonuniform specimen with a loose top layer. Further work
with triaxial specimens will be conducted now that a suitable one-way
valve has been developed for use in the displacement solidification method.

Acknowledgments
Research on the liquefaction of sands has been supported b y the National
Research Council of Canada under Grant No. 1498 at the University of
British Columbia since 1966. This continuing support is gratefully acknowl-
edged. The writers also express their appreciation to Y. P. Gupta for his
assistance with the soil liquefaction studies using the shake table.

References
[1] Duke, C. M. and Leeds, D. J., Bulletin of the Seismological Society of America, Vol.
53, No. 2, Feb. 1963, pp. 309-357.
[2] Seed, H. B. and Idriss, I. M., Journal of the Soil Mechanics and Foundations Division,
American Society of Civil Engineers, Vol. 93, No. SM3, May 1967, pp. 83-108.
[31 Seed, H. B. and Wilson, S. D., Journal of the Soil Mechanics and Foundations Divi-
sion, American Society of Civil Engineers, Vol. 93, No. SM4, July 1967, pp. 325-
353.
[4] Seed, H. B. and Lee, K. L., Journal of the Soil Mechanics and Foundations Division,
American Society of Civil Engineers, Vol. 92, No. SM6, Nov. 1966, pp. 105-134.
[5] Lee, K. L. and Seed, H. B., Journal of the Soil Mechanics and Foundations Division,
American Society of Civil Engineers, Vol. 93, No. SM1, Jan. 1967, pp. 47-70.

5 Personal communication, D. J. Picketing, partner, Cook, Pickering and Doyle Con-


sulting Engineers, Vancouver, Canada.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
194 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

[6] Peacock, W. H. and Seed, H. B., Journal of the Soil Mechanics and Foundations
D/vision, American Society of Civil Engineers, Vol. 94, No. SM3, May 1968, pp.
689-708.
[7] Finn, W. D. Liam, Pickering, D. J., and Bransby, P. L., Journal of the Soil Me-
chanics and Foundations Division, American Society of Civil Engineers, Vol. 97, No.
SM4, April 1971, pp. 639-659.
[8] Yoshimi, Y., Soils and Foundations, (Japan), Vol. 7, No. 2, March 1967, pp. 20-32.
[9] Finn, W. D. Liam, Emery, J. J., and Gupta, Y. P. in Proceedings, 3rd European
Symposium on Earthquake Engineering, Sofia, Sept. 1970, pp. 253-262.
[10] Finn, W. D. Liam, Emery, J. J., and Gupta, Y. P., Closed Loop, MTS Systems Corp.,
Vol. 3, No. 2, Fall/Winter 1971, pp. 14-18.
[11] Ohsaki, Y., Soils and Foundations, (Japan), Vol. 10, No. 2, June 1970, pp. 112-128.
I12] Lee, K. L. and Fitton, J. A. in Vibration Effects of Earthquakes on Soils and Founda-
tions, A S T M , S T P 450, June 1968, pp. 71-95.
[13] Seed, H. B. and Idriss, I. M., Journal of the Soil Mechanics and Foundations Division,
American Society of Civil Engineers, Vol. 97, No. SM9, Sept. 1971, pp. 1249-1273.
[14] Castro, G., "Liquefaction of Sands," Harvard Soil Mechanics Series, No. 81, Har-
vard University, Cambridge, Mass., 1969.
[15] Finn, W. D. Liam, Bransby, P. L., and Pickering, D. J., Journal of the Soil Mechanics
and Foundations Division, American Society of Civil Engineers, Vol. 96, No. SM6,
Nov. 1970, pp. 1917-1934.
[16] Ambraseys, N. and Sarma, S., Bulletin of the Seismological Society of America, Vol.
59, No. 2, April 1969, pp. 651-664.
[171 Selig, E. T. and Grangaard, O. H., Jr., Materials Research and Standards, American
Society for Testing and Materials, Vol. 10, No. 11, Nov. 1970, pp. 19-22.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreeme
D. F. Gri~n ~

Errors of In-Place Density Measurements in


Cohesionless Soils

REFERENCE: Griffin, D. F., "Errors of I n - P l a c e D e n s i t y Measure-


m e n t s i n C o h e s i o n l e s s Soils," Evaluation of Relative Density and Its Role
in Geotechnical Projects Involving Cohesionless Soils, A S T M S T P 5~3, American
Society for Testing and Materials, 1973, pp. 195-206.
ABSTRACT: This paper summarizes published investigations concerning the
deviations from a standard of in-place density measurements of cohesionless
soils. The effect of moisture content of the soil on in-place density measure-
ments also was investigated. Discussions about methods of calibrating meas-
uring devices and numerical values for test results are omitted. Some results
are shown graphically. It is hoped the reader will be stimulated to consult
the primary references if interested in more detail.
When a soil is physically sampled during the process of conducting an in-
place density measurement, a shearing action of the soft is unavoidable. Co-
hesionless soils are sensitive to volume change during shear; dense sands tend
to expand and increase in volume; loose sands tend to contract and decrease
in volume. Thus, in general, measured in-place density of dense sand is found to
be relatively low and for loose sand, it is found to be relatively high compared
to control values; the opposite relationship was found for cohesionless coarse
grain base course material. Any moisture present in such soils provides some
cohesion and in turn alters the error involved in a density measurement.
A plot of measured values versus control values provides a means for ad-
justing the value measured in the field for a specific soil more closely toward
the true (standard) value. All density measuring methods investigated require
such correction plots. Moreover, specific plots for a given sand generally are
not applicable to other sands of even slightly different character or water
content. Each method requires a separate plot for each soil and each soil mois-
ture condition.
K E Y W O R D S : density (mass/volume), cohesionless soils, tests, porous
materials, sampling, sands, density measurement

Reliability of values for in-place density of soil becomes a matter of


concern when the corresponding values for water content and specific
gravity of the soil sample fix the degree of saturation in excess of 100
percent. It has been known for many years that the volume of cohesionless
Research civil engineer, Naval Civil Engineering Laboratory, Port Hueneme, Calif.
93041.

195

Copyright by ASTM Int'l


Copyright9 (all rights
by ASTM reserved); Fri www.astm.org
International Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement
196 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

soils is altered by a shearing action; therefore, encountering such an


impossible relationship as saturation apparently greater than 100 percent
should not come as a surprise to the soil technologist.
In determining in-place density and moisture content of soil, there is
more to be considered than simply digging a clean hole, measuring its
volume, and weighing and determining the water content of the earth
removed. It should be self-evident that two specific problems exist in
connection with in-place soil density measurements. First, there is the
question of the accuracy of the density measuring device per se. Usually the
basic measurement required of the device is the volume of the cavity
formerly occupied by the soil specimen. Devices used to measure volumes
of cavities in soils can be calibrated by using simulations of soil cavities
formed in materials that are rigid enough so as not to deform by any
significant amount while a volume is being measured.
Calibration curves showing true (standard) volumes versus volumes
measured by a given device can then be established. An associated problem
is the determination of the true volume of a simulated cavity. If the
material in which the simulated cavity is formed is nonporous and not
water absorbent, the most probably true volume can be determined by
measuring the weight of distilled water or other fluid to exactly fill the
cavity at known temperature and density.
Finally, there is the question of the actual cavity in real soil. Are the soil
particles in the boundary of the cavity in the same identical location after
the cavity is formed as they were before the cavity was formed? If the soil
is cohesionless or is a cohesive soil in a state of liquefaction, it is quite
evident that the soil particles in the immediate vicinity of the cavity will
not be in the same positions after cavity formation as before.
The characteristic of soils that enables expansion or contraction while
undergoing shear of the type required to physically sample soil, confounds
the problem of obtaining reliable in-place soil density values. The problem
is made even more difficult because of the fact that all soils do not respond
in an identical fashion under shear action.
A given soil may be compacted to different values of known density, and
calibration curves established to show the relationship between the known
or control density and the density determined by a particular measuring
device. Such curves permit one to better estimate the true in-place density
than can be estimated without benefit of such curves.
This paper discusses (1) methods for achieving a range of known or
control densities of eohesionless soils, and (2) the development of calibration
curves for a selected soil.
Control Densities
Ottawa 20-30 sand and graded Ottawa sand were used as the testing
media in the laboratory [1].3
2The italic numbers in brackets refer to the list of referencesappended to this paper.
Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement.
GRIFFIN ON ERRORSOF IN-PLACE DENSITY MEASUREMENTS 197

In order to perform multiple tests at a given density, a large heavily


reinforced marine plywood box was constructed, with interior dimensions
of 3 by 3 ft in plan by 1 ft deep. The volume of the box was determined by
weighing it full of water at known temperature.
After considerable experimentation, five different methods were devel-
oped to reproduce five different densities fairly consistently within the test
box and these were adopted as the control densities. A brief discussion of
these methods follows:
Density S--A U. S. No. 10 sieve (2.000-ram opening) was hand-held
about 19 in. above the top of the test box. Sand was poured through the
sieve as it was slowly moved over the top of the test box.
Density F--A 5-gal capacity liquid fuel funnel was suspended from the
ceiling over the center of the test box with the small orifice, 1.3125 in. in
diameter, about 8 in. above the top of the test box. The funnel was kept
filled with sand as it was swung back and forth across the box in parallel
paths.
Density V--The test box was first filled to overflowing by pouring sand
into it from a gallon can. The stem of a concrete vibrator was then inserted
vertically until its tip reached the bottom of the box, the insertions forming
a grid pattern with spacings about 8 in. on centers. During the vibrating
period additional sand was added in order to maintain the level of the sand
above the top of the test box.
Density R--Sand was poured loosely into the test box in depth increments
of 3 in. Each increment was rodded with a a/~6_in" diameter steel rod in a grid
pattern about 2 in. on centers.
Density P - - T h e test box was filled by pouring sand loosely into it from
a gallon can.
In each case the box was filled to overflowing and the excess sand was
struck off with a steel straightedge. The relative density levels obtained are
listed in order of increasing densities for the two sands used 4n the labora-
tory, and appear as follows: Ottawa 20--30 sandS--P, F, R, V, S, and
graded Ottawa sand4--P, F, S, R, V.
Weight Measurements---The 9-ft 3 test box and contents were weighed on
a 2000-1b capacity scale. Sand specimens obtained during the testing
procedures were weighed on a balance with a capacity of 20 kg and a
sensitivity of I g, or on a balance with a capacity of 1600 g and a sensitivity
of 0.1 g.

Sampling Methods
Several methods used to sample soil for determining in-place density
were investigated.
s Sand from ASTM Test for Tensile Strength of Hydraulic Cement Mortars (C 190-
49), 1952 Book of A S T M Standards, Part 3, p. 167.
4 Sand from ASTM Test for Compressive Strength of Hydraulic Cement Mortars
(C 109-52), 1952 Book of A S T M Standards, Part 3, p. 119.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
198 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

Plastic Injection Method


Included was a hypodermic-like syringe and needle to inject a known
weight of quick-setting resin into the sand. Volumes of resin-sand lumps
were then determined by weighing them in air and in water. Weight of
resin was deducted from weight of lump to obtain weight of sand.

Wedge Method
Another method included a wedge apparatus. This consisted of a
supporting plate with guides attached through which metal plates could be
inserted to meet and isolate a wedge-shaped volume of sand. Sand was
excavated from the wedge and weighed. The volume of the wedge was
calculated from dimensional measurements of the wedge.

Tube Method
A cylindrical metal tube with open ends was pushed into the sand.
A waxed paper disc and plaster of paris were used to cap the upper end of
the tube. A concave dish was inserted into the sand to a position just
beneath the lower end of the tube. Tube and dish were carefully inverted
and excess sand struck off the tube. The sand inside the tube was removed
and weighed; the volume inside the tube was determined from the weight
and density of water required to fill it.

Sand-Funnel Method
The sand-funnel method investigated was similar to, but not identical to,
that described in ASTM Test for Density of Soil in Place by the Sand-Cone
Method (D 1556-64), known as the sand-cone method. A large cone was
devised for this investigation. The lower cone had a base diameter of 12 in.
The upper cone was attached to a cylinder 8 in. in diameter and 6 in. long.
This apparatus was used with a supporting annular plate alone and with a
supporting annular plate attached to an open tube 11.875 in. in diameter by
4 in. long, bevelled on the lower end. The tube was pressed into the sand
until the lower surface of the annular plate came into contact with the sand.
Sand was then excavated to a depth near the lower end of the tube, and
weighed. The volume of the cavity was then determined as a function of the
calibrated density of sand from the sand funnel required to fill the cavity.
In addition, a small sand-cone with quart jar attached to the upper cone
was used together with a small supporting annular plate attached to a tube
with an inside diameter of 3 in. and a length of 1.652 in. Weight of sand and
volume of cavity were determined in the same manner as above.

Rubber Balloon--Tube Method


The rubber balloon device investigated was very similar to that described
in ASTM Test for Density of Soil in Place by the Rubber-Balloon Method

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant
GRIFFIN ON ERRORS OF IN-PLACE DENSITY MEASUREMENTS 199

(D 2167-66). It was used with an annular base plate attached to a tube.


Two different diameter tubes (2.875-in. inside diameter and 4.875-in. inside
diameter) each 6-in. in Jlength were used. Weight of sand removed from a
tube and volume of cavity were determined in a manner similar to that
described in ASTM D 2167-66.

Comparison of Laboratory Results of Test Methods


In Table 1 [I] are summarized the data relevant to a comparison of the
various density-testing methods investigated. Table 1 should be evaluated
in con]unction with the type of information presented in Fig. "1. All
density-testing methods involved a shearing action on the soil and had the
usual effect of causing loose sands to contract and tight sands to expand.
Without exception, the results for every method tested showed an algebraic
decrease in percent deviation versus increasing control densities. The
critical density of a soil is that density at which no volume change occurs
during shear. At the critical density there is no error in measurement;
however, for some density-testing methods no critical density was observed
within the range of densities tested.
Figure 1 is presented as an example of the kind of response all of the
density=testing methods showed. As indicated, every method tested showed
an algebraic decrease in percent error with an increase in control densities.
The position of the least squares trend lines varied for each method as did
the range of deviation from each trend line.
There appears to be no direct correlation between the ranges of error and
the volumes of samples obtained by the various methods. In the majority
of methods investigated, the ranges of error were less in the graded Ottawa

12
II . . J JIJIJJlllll
I0 - - 0 Average of 4 T~sts in Ottawa 2 0 - : 5 0 Sand
9 Average of 4 Tests in Graded Ottawa Sand
9 ~ Least Squares Line (Approx)
Sample: 2.875 in. diam by 5.5 in. Long
8 Tube: 5 0 deg Inside Bevel
c
~ 7 P
F.s I R I I v =l , Densit~t Symbols

ff ,

W
5

4
= 2
_0 0
3I 0
0
2

I
P F R V s
0
1.56 1.58 1.60 1.62 1.64 1.66 1.68 1,70 1.72 1.74 1.76 1,78
Control Density, g per cu cm

FIG. 1--Rubber baloon-tube method, te~t density error versus control density.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License A
0
0

.<

o)

T A B L E 1--Comparison of test methods. I--

_<
z
O
Results of Various Test Methods with Ottawa 20-30 Sand with Graded Ottawa Sand
O
"1-

Method Deviation Band Critical ~ Average Sample Deviation Band Critical 9 Average Sample O
Width, percent Density, g / c m 3 Volume, cm ~ Width, percent Density, g/cma Volume, cm a z
7.
{/)

Plastic injection 3.0 1.678 140 4.6 134 o


Wedge
45-deg inside bevel 2.1 ... 8380 . . ~

45-deg double bevel 1.2 8380


45-deg outside bevel 2.9 11633 8380
Tubes: 30-deg outside bevel b
2. 875-in. I D X 1.5-in. sample 5.3 1.658 155 3.3 1.610 150
2.875-in. I D X 4.1-in. sample 1.9 1.668 439 0.9 1.618 411
2.875-in. I D )< 6.0-in. sample 0.6 1.675 637 2.5 1.615 613
4. 875-in. I D X 5.9-in. sample 3.0 1.678 1811 1.6 1.635 1838
Tubes: 30-deg double bevel b
2.875-in. I D X 4.4-in. sample 1.3 1.675 472 . . ~

2. 875-in. I D X 6.3-in. sample 1.0 1.673 670 ~

4.875-in. I D )< 6.2-in. sample 2.5 1.674 1908 . . 9 . ~

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
Tubes-30-deg inside bevel b
2. 875-in. I D X 4.2-in. sample 1.1 1.657 450 . . .

2.875-in. I D X 5.9-in. sample 1.3 19 631


4.875-in. I D • 6.5-in. sample 1.7 I9 1992
Tubes: 45-deg inside bevel b
2.875-in. I D X 4.4-in. sample 0.6 19 475
2.875-in. I D X 5.6-in. sample 2.0 19 590 . . ~

4. 875-in. I D X 6.2-in. sample 1.0 19 1901 . . 9 ,9

Tubes: 45-deg outside bevel b


2.875-in. I D X 4.2-in. sampl e 2.1 19 449 2.1 1.622 422
2.875-in. I D X 5.9-in. sample 2.3 19 628 3.2 1.622 613 0
m
4. 875-in. I D X 5.9-in. sample 1.9 19 1818 0.8 1.642 1828
Tubes: semicircular edge b 2:
2.875-in. I D X 4.5-in. sample 1.4 1.675 479 0
Z
2. 875-in. I D X 69 1-in. sample 0.5 1.665 650
4.875-in. I D X 6.2-in. sample 1.5 1.660 1899
Large sand funnel 6.1 2126 o
Large sand funnel-tube 37 1:736 6748 i:d 1:606 di48
Small sand funnel-tube 5.9 ... 158 1.7 19 155 o
R u b b e r ballon-tube
2. 875-in. I D X 6-in. tube 2.2 ... 554 1.9 559
4.875-in. I D X 6-in. tube 2.9 ... 1613 3.4 1566

a Critical density as used here is t h a t density where no error is detectable in measurement of in-place d e n s i t y - - n o volume change occurs
with shear 9
b Sample lengths shown are averages 9
3:
o~
C

to
o

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
202 RELATIVE DENSITY INVOLVfNG COHESIONLESS SOILS

TABLE 2--Variations from assumed true densities for each method.

Device Variation, lb/ft s

Water balloon
Pyenometer-funnel
Quart Jar and sand funnel
7-in. sand density cylinder and funnel
10-in. sand density cylinder and funnel

sand than in the Ottawa 20-30 sand. In view of the fact that most natural
sand deposits are graded, results in the graded Ottawa sand may be the
more important. The results rather conclusively show that the two sands
are quite different with reference to the position and slope of the least
squares lines of percent error versus control density for a given test method.
The Waterways Experiment Station (WES) has observed that "The only
part of the density test subject to serious error (other than the personal
element) is the determination of the volume of the hole from which the
sample is taken."[2] I feel that the error-potential involves both the
apparatus used to measure volume, and the excavation-induced volume
changes in the (then virtual) cavity that the soil sample occupied before
removal. WES conducted a study to determine the amount of error
inherent in each piece of apparatus used in making in-place density meas-
urements. Based on plots of true volumes of simulated cavities versus
volumes measured by the methods investigated, the approximate variation
from true density (assumed to be 120 lb/ft a) in pounds for each method is
shown on Table 2.
The soil density of 120 lb/ft ~ apparently was assumed simply for the
above comparison. If a different density value had been assumed, the
variations would have undoubtedly resulted in a different magnitude of
accuracy; however, the relative positions of the methods would have
remained the same. WES concluded among other findings that "The
accuracy of the density determined by any in-place test may be influenced
considerably by disturbances to the walls of the hole during digging."
Keeton [3] investigated the effects of moisture present in sand on the
accuracy with which in-place densities can be measured. For this purpose
sand from the beach at Point Mugu, Calif. was used. Briefly, the methods
used by Keeton to establish control densities were as follows:
1. Sand was dumped loosely into the box and tamped in 3-in. lifts with
a compressed air tamper having a 4-in. diameter tamper plate.
2. Sand was compressed with the compressed air tamper in 6-in. lifts.
3. Sand was hand tamped in 3-in. lifts with a Marshall Method compac-
tion hammer--10-1b weight, 18 in. fall, plate was 3.875 in. in diameter.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License
GRIFFIN ON ERRORS OF IN-PLACE DENSITY MEASUREMENTS 203

4. Sand was vibrated in 3-in. lifts with a Lazan Oscillator mounted on


a 2-in. thick wooden plate to fit the inside cross-section of the test box.
5. Same as above except 6-in. lifts were used.
6. Same as above except one lift overflowing the box was used. Different
fillings of the box were vibrated for different periods of time.
The density levels obtained by the different methods were not reproducible
for any given method.
The method used for measuring in-place densities in the control box was
the quart jar and sand funnel used in conjunction with a metal tube
attached to an annular supporting plate. Test results are shown in Fig. 2 as
control density versus test density. The effect of water is clearly pro-
nounced. For a given control density, the test density decreased as the
water content of the sand was increased. If the values in Fig. 2 were plotted
as percent error versus control density, the lines would slope downward
from left to right as for the dry Ottawa sand. The sand used in the quart jar
was dry Graded Ottawa sand.
I have conducted some tests similar to those described earlier using
Ottawa sand and a gallon jug with sand funnel in conjunction with a metal
tube attached to an annular supporting plate [4]. Dry Ottawa sand was
used in the jug, and the same type of Ottawa sand was used in the control
box as was used in the jug. Various densities of sand in the control box
were obtained by tamping the wet sand in varying multiple layers with a
metal disk 10 cm. in diameter attached to a handle.

9 ,

1.72~'--OTTe:tt:::rr:::ii::eWt;::~::: J O-[
- --IS
W
J WateCont
r entO S !~e ~ ~f~O0"1"~~
1.66
.E
1.60
>.
0
~, 1.54
0
o o

1.48 1.54 1.60 1.66 1.72 1.78 1.84 1.90


1.48

CCDry Control Density in Grams Per

FIG. 2--Test results for small sand ]unnel-tube in Point Mugu beach sand.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
204 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

Test results for Graded Ottawa sand are shown in Fig. 3. These results
and those of Keeton definitely show that wet sands respond differently
from dry sands when sampled for in-place density. These test results
together with other results for Ottawa 20-30 sand not reported herein,
indicate that each sand would show results similar to one another but
varying in magnitude of error.
Cohesionless Coarse Grain Base-Course Material--In-place density tests of
cohesionless coarse grain base-course material provided a great surprise [5].
The gallon sand jug and funnel with separate annular supporting plate was
used to measure volumes of cavities in the control box. Test results are
shown in Fig. 4 as control density versus percent deviation of measured
density from control density as a function of water content. Here the trend
lines slope upward from left to right, just the opposite of the results for
sands.
Two phenomena were observed. At water contents below 2.5 percent, the
base-course material is essentially cohesioniess. During the excavatiort of a
cavity, material from the cavity walls sloughed off, and the walls appeared
to expand into the cavity. Thus, volumes were measured too small and
densities were computed too high. At high water contents, the soil acted as
a plastic mass and, at about 8 percent water content, the soil mass was
almost fluid. When a cavity was excavated in the soil with such a high water

"~ 2
c ~--IL

9 W = 1.5 p e r c e n t -+ 9
o -I 9 W = 4 per cent +
9 W = 5.5 p e r c e n t _+
0 W = 7 per cent +
o

(b)
._o
:>

c I
I w ~. 9
o 0
~L
-I
(a)

1.32 1.36 1.40 1.44 1.48 1.52 1.56 1.60 L64 1.68 1.72
Control Density, g per cu cm

(a) Average of 4 tests in dry sand. (b) Individual tests in wet sand.

FIG. 3--Percent deviation of measured density versus control, graded Ottawa sand.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
GRIFFIN ON ERRORS OF IN-PLACE DENSITY MEASUREMENTS 2 0 5

32

~0 ] o/
/
28
/
26 Q AVERAGE OF 4 TEST~ PER BOX.
BROKEN LINES ARE APPROXIMATE LEAST SQUARES LINE~
Z 24
FOR PERCENTAGE DEVIATIOI OF TEST DENSITY FROM

22 CONTROL DENSITY ACCORDING TO WATER CONTENT.


NUMERALS ARE WATER CONTENTS IN PER CENT
/
/

8
18
W,7 T 0 8 % ~ ~
/
8 ,4

./ /
~' ,/'
/
/
W" 2 TO 3 % - ~ ,/
/
/
J ~s
o lj
/
~/, W=GTOT%-- Z /5(~

o..
WvSTOG% --
~'" ~'~ 0 3.3 ~-*
5.6
3.8 ~

-IC

5.7
-12

-14

o
-IE
1.50 1.54 1.58 1.62 t.66 1.70 L74 1.78 1.82 1.86 L90 1.94 196 2.02 2.06 2.10 2.14 2.18

CONTROL DENSITY (DRY), G PER CU CM

FIG. 4 - - T r e n d lines according to water content of percentage deviation of densities meas-


ured by the sand cone apparatus versus control densities in MonLalvo base course material.

content, plastic flow tended to fill the cavity, and again, measured volumes
were too small, giving computed densities that were too great. Between
these two extreme conditions, there were varying degrees of change in
volume caused by the effects of shear during sampling.
Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Ag
206 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

Relationship to Relative Density and Test Standards


Relative density of cohesionless soils is defined in A S T M Test for
Relative Density of Cohesionless Soils (D 2049-69), as the " . . . state of
compactness of a soil with respect to the loosest and densest states at
which it can be placed b y the laboratory procedures described in this
method." Ultimately, relative density can have significant meaning only in
real life situations. In-place density measured for soil as found naturally or
as remolded in the field may be expressed in terms of relative density.
Therefore, the in-place density measurements in the field must be as free
of error as possible. Freedom from unknown error may be achieved b y
means of developing density calibration curves as previously suggested.
The impact of studies reported herein on standard tests is that existing
A S T M standard tests for measuring in-place soil density should be revised
to recognize the type of errors involved. The engineer should be responsible
for reporting the degree of sensitivity to shear of a soil physically sampled
for in-place density.
Overall C o n c l u s i o n s
When a soil is physically sampled during the process of conducting an
in-place density measurement, a shearing action of the soil is unavoidable.
Cohesionless soils are sensitive to volume change during shear; dense sands
tend to expand and increase in volume; loose sands tend to contract and
decrease in volume. Thus, in general, measured in-place density of dense
sand is found to be relatively low and for loose sand, it is found to be
relatively high compared to control values.
A plot of measured values versus control values provides a means for
adjusting the value measured in the field more closely toward the true value.
All measuring methods tested require such correction plots. Moreover,
specific plots generally are not applicable to other sands of even slightly
different character or water content. Each method requires a separate plot
for each soil or soil moisture condition.
References
[I] Griffin, D. F., "Study and Development of Methods for Determining In-Place
Density of Soils," USN Bureau of Yards and Docks Contract NOy-73233, 23 Sept.
1953; Proceedings, American Society for Testing and Materials, Vol. 54, 1954, p. 1270.
[2] "A Study of In-Place Density Determinations for Soils," TM 3-415, U. S. Army
Corps of Engineers, Waterways Experiment Station, Vicksburg, Miss., Oct. 1955.
[3] Keeton, J. R., "A Study of Methods for Determining In-Situ Densities of Cohesion-
less Beach Soils," Thesis presented to the Faculty of the Graduate School, University
of Southern California, Los Angeles, June 1955.
[$] Griffin, D. F., "The Relationship Between Water Content and the Accuracy with
Which In-Place Densities of Sands May Be Measured," Bulletin 212, American
Society for Testing and Materials, Feb. 1956, p. 51.
[5] Griffin, D. F., "In-Place Density Tests of Montalvo Base Course Materials Under
Controlled Conditions," USN Bureau of Yards and Docks Contract NBy-3101,
30 June 1956; "In-Place Density Tests of Cohesionless Coarse-Grain Base-Course
Material," Bulletin 230, American Society for Testing and Materials, May 1958,
p. 31.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No furthe
R. J. Frost ~

Some Testing Experiences and Character-


istics of Boulder-Gravel Fill in Earth Dams

REFERENCE: Frost, R. J., "Some Testing Experiences and Character-


istics of Boulder-Gravel Fill in Earth Dams," Evaluation of.Relative Density
and Its Role in Geotechnical Projects Involving Cohesionless Soils, ASTM S T P
553, American Society for Testing and Materials, 1973, pp. 207-233.
ABSTRACT: This paper comprises a series of experiences and characteristics
of boulder-gravel fill used in earth dams, particularly that in the large Mangla
and Tarbela Dams of West Pakistan. The paper includes details of both field
and laboratory testing. On the former subject, the paper discusses the choice of
minimum sample weight, repeatability of gradation results, and the plastic
sheet errors in the fill in situ density test by the water replacement method.
In the case of the laboratory maximum density test, the effects of such vari-
ables as apparatus and procedures are discussed and examples given. It was
concluded that a variety of apparatus can give essentially the same test results.
The uses and limitations of the Humphres Laboratory Test Procedure are also
discussed. On the Mangla Dam Project, a correlation was established between
gradation and laboratory density, and this is described together with an in-
vestigation of its application to other dam sites. The problem of the relationship
between the fill prototype grading and the laboratory model grading is dis-
cussed, and various solutions, such as oversize corrections, extrapolation, and
modeling, together with examples, are given. The paper concludes with some
comments on fill density control and results.
KEY WORDS: cohesionless soils, density (mass/volume), earth fills, tests,
compacting, vibration, boulders

E a r t h a n d rockfill d a m s a r e b e i n g b u i l t t o e v e r i n c r e a s i n g h e i g h t s a n d
w i t h ever i n c r e a s i n g q u a n t i t i e s of fill. One r e s u l t of t h e l a t t e r s i t u a t i o n is
t h e u n a v o i d a b l e i n c r e a s i n g use of coarse g r a i n e d fill, i n c l u d i n g b o u l d e r -
gravel mixtures.
T h e A m e r i c a n S o c i e t y of Civil E n g i n e e r s C o m m i t t e e on E a r t h a n d
R o c k f i l l D a m s (1967), in c o m p i l i n g t h e p r o b l e m s concerning t h e d e s i g n

1 Formerly, soils engineer, Tippetts-Abbett-McCarthy-Stratton, Tarbela Dam Project,


West Pakistan; presently, resident engineer, Binnie and Partners, High Island West
Dam, Hong Kong.

207

Copyright by ASTM Int'l


Copyright9 by(all
ASTMrightsInternational
reserved); Fri Mar 11 16:13:06 EST 2016
www.astm.org
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
208 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

and construction of such dams, rated compaction methods in coarse gravels


and rock fills as of Priority three and rated control of compaction of such
materials as of Priority eight in their list of ten priority problems. The
years since 1967 have left many problems associated with coarse gravel
fill still unsolved, and the June 1972 American Society for Testing and
Materials (ASTM) Symposium, of which this paper forms a part, was a
further attempt to clarify the issues.
The author has been in charge of soil testing on two of the worlds largest
dams (namely, Mangla and Tarbela Dams in West Pakistan) and has been
associated with engineers who have worked on other large earth dams.
This paper represents a collection of the experiences of in situ and labora-
tory testing of b0ulder-gravel fill gained from those projects.

Definitions
The Unified Soil Classification System is adopted herein to describe the
various constituents which comprise a soil mass. The ranges of particle
sizes which apply to each constituent are as follows:

boulders or cobbles = particles greater than 3 in.


gravel = particles 3 in. to No. 4 sieve size
sand = particles No. 4 to 200 sieve size
silt = particles No. 200 sieve to 0.002 mm size
clay = particles less than 0.002 mm size

Fill Gradation Sampling


Almost all earthwork projects specify the gradation limits within which
the fill materials will lie. Testing for compliance with these specifications,
therefore, represents the major activity of field soil laboratories. The fol-
lowing describes the two important aspects of this testing.

Minimum Specimen Size


Field tests and observations indicate that the coarser the material, the
greater the segregation that occurs during placement, and the greater the
difficulty in selecting a representative specimen. This difficulty can be
overcome by selecting a specimen weight of sufficient size, so that it in-
cludes both segregated and unsegregated portions. Practical difficulties,
however, limit the maximum size that can be taken, and it is necessary to
determine an acceptable minimum size. On various projects, the author
has carried out trials to determine such a minimum weight by initially
taking a specimen from a small hole in the placed fill and determining
its gradation. The hole is then gradually enlarged in increments; the
excavated material being graded and added to the gradation of the previous

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licens
FROST O N BOULDER.GRAVEL FILL IN EARTH DAMS 209

GRAfN SIZE IN MILLIMETER


O O
o o o
{' I I 1 I I I I t

lO,OOO

BS -- 1377 .~

10
5

1,0
0.5,

o.11 . . . . . . . . ~ "~

MAXIMUM SIZE OF MATERIAL IN SUBSTANTIAL P R ~ T I O N (MORE THAN 2OPERCENT)

FIG. 1--Minimum specimen weight for fill gradation~density tests.

excavations. A specimen weight is reached at which further increase in


specimen size makes no significant difference to the gradation. This weight
is then considered to be the acceptable minimum specimen weight. In one
such exercise on boulder-gravel fill at Mangla Dam, the total sample
weight sieved was 150 000 lb.
These trials indicated, no unexpectedly, that the minimum specimen
weight increases as the maximum particle size increases, and the results of
such are presented in Fig. 1o The definition of maximum particle size is
taken from Ref 1 and is that present in "substantial" proportion (sub-
stantial being defined as more than 20 percent). Figure 1 indicates that
those determinations from 21/~ up to 24 in. max particle size are com-
patible with those from British Standard 1377 [1]2 for sizes No. 4 sieve to
2 ~ in.
The minimum specimen weights of Fig. 1 can apply to fill in situ density
tests as well as gradation tests, and commonly, both determinations are
made on the same specimen.

The italic numbers in brackets refer to the list of referencesappended to this paper.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreemen
TABLE 1--Details of materials. ~o

6
Particle Plasticity Oversize Correction Tests (Fig. 19)

D a m Site Zone Shape Specific Liquid Plastic- Compac- Oversize Added Matrix Parameter
o Gravity Limit ity Index tion
Method Size 7m a b
T1 Fig. 2 Tarbela core angle 2.70 22 2 G,F(24 - - 4 i n . to + a ~ i n . -- ~ in. 129 0.026 --1.03 o~
-4
in.) 0.085 --1.08 -<
T2 Fig. 2 Tarbela transition angle 2.70 N1~ NP ~ F(24in.), - 6 in. to + ~ i n . --a/~in. or =1=140 0.018 --0.45
Humphres or + N o . 4 --No. 4
T3 Fig. 2 Tarbela --1 in. core angle 2.70 22 2 ASTM --1 in. to + ~ in. - - 1 in. 138 0.033 --0.73 ~,-
D 1557 z
T4 Fig. 2 Tarbela - No. 4 angle 2.70 22 2 ASTM --1 in. ~o + N o . 4 --No. 4 134 0.028 --0.35 O
core D 1557
T5 Fig. 2 Tarbela foundation round 2.88 NP NP F(24in.) - - 6 i n . to + a ~ i n . = - ~ i n . or 136 NONE 8X
or + N o . 4 --No. 4 130
T6 Fig. 2Tarbela drainage round 2.88 NP NP F (24 in.) --6 in. to + 1 ~ in. --11/6 in. 151 0.036 --1.07 O
z
T7 Fig. 2Tarbela plant angle 2.70 NP NP F ( 2 4 i n . ) - - 6 i n . to + ~ i n . _3/~in. or • 0.018 -0.45 N
product H u m p h r e s or + N o . 4 --No. 4
T8 Fig. 2 Tarbela shoulder angle 2.70 NP NP . . . . . .
(sandy) _o
T9 Fig. 2 Tarbela shoulder angle 2.70 NP NP ...
O1 Fig. 5 Oroville core round 2.84 27 to 36 10 to 18 H No. 4 to "3/~'"in. -No: lig'to 0:036 "'"
11/6 in. or 4 in. 128 (0.026 --1.03
to 0.054)
02 Fig. 5 Oroville shoulder round 2.84 NP NP ...
Mil Fig. 5 Mica shoulder angle 2.75 NP NP E -3in. to + ~ i n . -a~in.
Mi2 Fig. 5 Mica shoulder round 2.76 NP NP . . . . . . . . .
M1 Fig. 5 Mangla shoulder round 2.78 NP NP . . . . . . . . .
J1 Fig. 5 Jari shoulder round 2.78 NP NP . . . . . . . . .
C1 Fig. 5 Cougar shoulder angle 2.63 NP NP . . . . . . . . .
(approx.)
a N P = Non Plastic.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
FROST ON BOULDER-GRAVEL FILL IN EARTH DAMS 211

Repeatability of Results
Significant differences between gradation tests can occur over short
distances within apparently uniform material, due to inherent natural vari-
ations within the material occurring during the borrowing, manufacturing
(blending), dumping, and spreading operations. In seven trials carried out
on Material T1, whose properties are described in Table 1 and Fig. 2, in
which a large number of specimens were taken within 2-ft spacing of each
other, differences from the average were found of up to 10 percent in the
amount passing any sieve (that is, for example, a range of from 20 to 40
percent passing No. 200 sieve). In the nondimensional terms of coefficient
of variation (CV) (equal to standard deviation divided by the average
value), these were equivalent to CV values of 4 percent on 11/~ in. sieve to
10 percent on No. 200 sieve. Other tests on Material T2 (Fig. 2) indicated
similar orders of CV values.
Recognizing that such variations must occur, it seems highly desirable
that specifications for boulder-gravel fill materials should provide two sets
of upper and lower gradation limits; one set being those within which all
gradations must fall, and the other between which some significant pro-
portion (say 80 percent) should fall--the material between the 100 percent
and 80 percent limits being provided to accommodate both some true
gradation departures and the unrepresentative gradations from sampling
errors.

BOULDERS J GRAVEL ] SAND JSILTAND CLAY


US STANDARD SIEVE SIZE

100 ~'

~ ~o I

20

I I

GRAIN S/ZE IN MILLEMETERS

FIG. 2--Gradation8of materials.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproduc
212 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

Fill In S i t u Density Tests


Need for Tests
Fill compaction may be assessed by settlement or by fill density tests.
The former avoids the possibility that the excavation of the hole results in
a reduction in the lateral stress imposed during compaction which leads to
a decrease in the hole volume and an incorrect higher density. In the
United States and Canada for example, a settlement of 3 to 5 percent is
often specified for a layer depth of 3 to 4 ft. However, such procedures are
comparative only and do not provide absolute values of density, and it is
believed that in situ density tests are essential to provide such data for
design and contractual purposes.
Procedure
The procedure most often used is the water replacement method which
determines the volume of the test pit hole by lining it with polythene sheet
and filling with a measured amount of water. A heavy steel ring is used to
contain the polythene sheet, and an initial volume reading to a reference

FIG. 3--Fill in situ density test procedure.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
FROST ON BOULDER.GRAVEL FILL IN EARTH DAM5 213

FIG. 3--Continued.

mark is made before the test hole is excavated. A typical procedure is shown
in Fig. 3.
Errors---In the author's experience, random and systematic errors intro-
duced in the determination of water volume and sample weight amount
to some 1 percent of the fill density. (The former can be largely eliminated
by using three water meters in series.) However, the greatest area of doubt
lies in the effect of lining the hole with polythene (plastic) sheet, and Fig. 4
presents the results of a series of tests carried out at Tarbela Dam on trial
holes lined with concrete (so that the actual volume was known) of various
sizes and degrees of roughness. Various thicknesses of plastic sheet in various
layers were used. The results were in agreement with expectations, in that
the errors in lining larger holes were smaller than in lining smaller holes,
that smooth holes produced smaller errors than rough holes, that thinner
sheets were more accurate than thicker sheets, and that two sheets resulted
in greater errors than one sheet of the same total thickness. As a result of
these trials, a standard test procedure using two sheets of 0.004-in. poly-
thene sheet was adopted. This produced errors of between 1 and 2 percent

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agre
214 RELATIVE DENSITY I N V O L V I N G COHESIONLESS SOILS

,4.0[ v
e = One sheet
x - Two sheets.
--= Smooth (Concrete finish)
/
..... Rough (One 3"--6"cobble /
half protruding
per sq, ft. of hole
/
I surface.)
~- 3,C
Z t
/ /

4
E
/'
/
/Y
O>2.C
z 30 hole (2.5 cu,ft.)
/ /
x 36 e ~ r i n g . "~I
3
J< O

B
48* ~ ring

~ 1.s
td
:.50"hole [16 c u . f t )
x. 6 ~ r ring

O ~
0 .002 .004 .006 .008 .010 .012 .014
TOTAL THICKNESS OF PLASTIC SHEET (INCHES)
( See distinction between 1 sheet and 2 sheets )

FIG. 4--Errors in lining field density holes with plastic sheet.

within the range of holes tested in the fill. This procedure was necessarily
a compromise between the greater puncture-proof strength of the thick
sheets and the flexible conforming to the shape of the hole of the thin
sheets. Two sheets reduce the possibility of undetected leak holes better
than one sheet.
Overall, it is believed that the maximum error in the fill density determi-
nation is of the order of 2 to 3 percent of the fill density. This does not
apply to those holes which contract as they are excavated; typical of which
are one-sized boulder or gravel filter type materials.
Laboratory M a x i m u m Density Tests on Boulder Gravel
Need for Test
The determination of the laboratory maximum dry density value for
boulder-gravel fill (hereafter referred to as Lab. MDD) requires a procedure

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
FROST ON BOULDER-GRAVEL FILL IN EARTH DAMS 215

which is more costly and elaborate than that required for testing fine
grained soils. For this reason, the test facility is usually only provided on
those earth dam projects containing a large quantity of bonlder-gravel fill.
Justification for the testing facility is usually an extension of that estab-
lishing the need for fill density tests, since it is desirable to know the com-
paction level of the fill material with relation to some standard laboratory
procedure.
Vibratory Compaction Apparatus
A comprehensive review of the factors influencing the vibratory com-
paction of soils is given in Ref. 2. These include resonant frequency of the
compactor-soil system, number of load cycles or frequency, shear strength
(both real and apparent) during vibration as it affects state of motion of
the soil particles, and the pressure and shear stresses (both static and
dynamic) generated during compaction.
Through the laws of physics, centrifugal force and frequency, amplitude
and weight, and compaction effort and weight are interrelated. Hence,
effective vibratory compaction can be achieved by judicious selection from
the large number of variables available. This is demonstrated in Table 2
which lists the procedures for determination of Lab. MDD at a number
of dam sites known to the author. The table indicates the nonstandard
nature of the test within a wide range of variables.
Dynamic and Kneading Compaction Apparatus
Laboratory determination of MDD need not be limited to vibratory
compaction procedures as the following two examples will illustrate.
Method G--At Tarbela Dam, laboratory compaction of gravel-sand-silt
material similar to grading T1 (Table 1 and Fig. 2) was carried out by a
kneading compaction effort consisting of a 30 deg steel foot (36.8 in. 2)
loaded to 550 psi pressure and applied to the soil in a 24-in. diameter by
24-in. high mold in 24 tamps per layer with 4 layers to fill the mold. The
density achieved with this apparatus was equal to that obtained in the
fill and was equivalent to 97 percent of that obtained by ASTM D1557-70
(Mod. AASHO effort).
Method H--Reference 3 describes the falling hammer (dynamic) pro-
cedures adopted at Oroville Dam for compacting clayey gravel similar to
grading 01 (Table 1 and Fig.~5) with up to 4-in. maximum particle size.
The largest mold used was 17 in. in diameter (2 ft 3) for which a hammer of
33.5 lb falling 18 in. in 265 blows on each of 3 layers was used.
Effect of Some Variables on Lab. MDD
The effect of the variables of compaction in various layers, of compaction
with various centrifugal forces, and of compaction for various times are
shown on Figs. 6, 7, and 8 from tests on Tarbela Materials T2, T5, T?,

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lice
O,

-4
.<

-4

r-
<
Z
o
T A B L E 2--Summary of vibratory compaction methods used at various dam sites.
o
Method No. A B C D E F (24 in.) F (36 in.)
D a m site Mangla Jari Oroville Cougar Mica Tarbela Tarbela o
z
Location Pakistan Pakistan USA USA Canada Pakistan Pakistan

Mold Diameter of mold 27.5 12 27 41 25.3 24 36 o


(in.)
Volume of mold 8 1 11.5 27 8.7 6.3 21.2
(ft9
Position of mold buried in con- bolted to floor suspended bolted to bolted to bolted to floor bolted to floor
crete from spring vibrating table vibrating table
yoke
Position of vibra- on soil surface on soil surface attached to attached to a t t a c h e d to on soil surface on soil surface
tor during com- b o t t o m of sus- vibrating table vibrating table
paction pended mold
Surcharge on soil ... ... 2 10 2
during compac-
tion (psi)

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
Vibrator Total weight of 383 25 1000 2100
surface applied
vibrator (lb)
Type of vibrator Cleveland electric kango Viber model Syntron elec- 2 No. vibro- 2 No. vibro-
pneumatic hammer Px-7 with tro mechanical verken Model verken Model
Type F (4 in. No. 8 eccen- rotary ER52 ER71
stroke, heavy tric
duty)
Frequency of vi- 1850 1800 8000 1200 1200 2850 1450
bration (vpm)
Centrifugal force ... 4400 7000
(Ib)
Amplitude (in.) ... ~ 0.2 0.023 0.06
Acceleration (xg) ... 5.3 3.6

Procedure No. of layers com- 3 1 5


o
pacted to fill mold z
Time of vibration 15 15 1
per layer (min) o
C

Maximum par- 6 :t:24 12


ticle size compact-
able (in.) o

F-

N}
-=l

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
218 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

BOULDERS I GRAVEL J SAND I SILTAND CLAY


U.S. STANDARD SIEVE SIZE ~ cJ
~ o ~ Q
lo4 ~ ~ ~ el-
i
e'lx- ~ ~
i
~,
i
~. ~
i
~

90

80
-NoA

7O

GO

5O

40

30

1r

i 1 1
O'

GRAIN SIZE IN MILLIMETERS

FIG. 5--Gradat/ons of mater/a/s.

and T8 whose properties are described in Table 1 and Fig. 2. Compaction


Method F (24 in.) as detailed in Table 2 was used. The results confirm
expectations that greater compactive effort produces the higher density.
However, care must be exercised in choice of procedure to ensure that
particle breakdown is not occurring which might result in a more favorable
particle packing and, hence, higher density. This was the situation with
Material T7 consisting of clean - 6 in. to +a/~ in. angular gravel (whose
density increased with compactive effort), since gradation tests at the end

106,

Totol vibr~tion
time by oaml:~ction ~ I
method F ( 2 4") /

QiCI
~EI~E
Q
r o 1 2 3
NUMBER OF" LAYERS

FIG. 6--Effect of compaction in layers upon Lab. MDD.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Ag
FROST ON BOULDER.GRAVEL FILL IN EARTH DAMS 219

9S

97

E 9E
Compaction method FC2zl)
_o
.~ 95 i i I i
0 500 1003 1500 2000 25O0
CENTRIFUGAL FORCE IN LBS.

FIG. 7--Effect of centrifugal force upon Lab. MDD.

of the test reported on Fig. 8 indicated a small (2 percent) amount of


_3/~ in. gravel originating from the angular corners of the material. An
assessment of the breakdown that can be tolerated during compaction must
depend on the individual case. In other tests at Tarbela, compaction of
Material T2 by Method F(24 in.) gave Lab. MDD values only 1 percent
lower than were obtained by the same method but with a total vibrator
weight of 1800 lb.
At Mangla, trial compactions on various materials within the gradation
envelope M1, shown on Fig. 5, indicated only a 2 percent increase in
density when the vibrator frequency was increased from 1450 to 1650 and
1850 vibrations per rain.
Moisture content of the specimen during compaction is only of im-
portance when the material is not free-draining; such material being char-
acterised by having more than 5 to 10 percent passing No. 200 sieve or

tO5,

~ loc
9
:E
LP

95
( T Compaction method F (2
o
9(:: I I I
0 5 10 15 20
TrME OF VIBRATION IN MINUTES

FIG. 8--Effect of compaction time upon Lab. MDD.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Ag
220 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

more than 20 percent passing No. 4 sieve. For such soils, compaction at
various moisture contents must be carried out, and the MDD determined
from the dry density-moisture content curve. For free-draining soils, com-
paction in either the dry or saturated condition gives similar density values.
Comparison of Various Apparatus
Figure 9 presents a comparison of Lab. MDD values obtained by
Methods F(24 in.) and F(36 in.) for Material T5 split on both No. 4 or
3/~-in. sieves and recombined in various proportions. The Lab. MDD for
the natural gradation T5 by these methods are similar (161 lbs/ft3). An-
other comparison (not shown) for Material T9 also gave similar Lab. MDD
values by these methods (149.0 versus 148.5 lbs/ft 3, respectively).
Figure 10 presents a comparison of Lab. MDD values obtained by
Methods F(24 in.) and G for Material T1 split on a/~-in, sieve and recom-
bined in various proportions. The data indicates that vibrating Method
F(24 in.) achieves lower Lab. MDD values as compared with kneading
Method G or dynamic method ASTM Tests for Moisture-Density Re-
lations of Soils, Using 10-1b Rammer and 18-In. Drop (D 1557-70).
Although Material T1 is compacted in the field by 10 ton vibrating
rollers or 100 ton pneumatic rollers to densities equal to those by lab-
oratory Method G, such densities could not be achieved in the laboratory
at any moisture content by vibratory procedures. This is believed to
be due to the relative impermeable nature of the material which causes
partial bulking and loss of compactive energy in the laboratory mold.

170 x \i r i i

Method "',.", Equation 1


F(31~) ~4) " " ' / Lab MDD for natural
(%~-~:~.4) " gradation ( T5 )
160 _ ~I.~/II'~,~,.. " . , shown t h u s : x
.'// ~.
rz
::5
!,, //
/,;//
z

t4C

13C

Deviation ( ) rorr
Equation I .
12( I I I i
20 40 60 80 100
% PASS No.4 OR a/o PASS 3/:

FIG. 9--Compaction of boulder-gravel-sand by various methods (grading TS).

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
FROST ON BOULDER.GRAVEL FILL IN EARTH DAMS 22 ]

160 T - - I ,

1
Equation 1. : *~T- p + 1 P

15(: / \ ,.,. ~-.

/ / , ~ ~\ ~ ~ Compaction by
k~ / ~ "",,,~: \ \ ~ kneading
/ \ \ ~, method-G
41.O
14C {ASTM
D1557
-70)
z
/ Ii ............... / \ -...'-. \
/ # / ".~\-...\
/// / / ~ .
hi
C3
,/// /

/// - b a s e d on m e t h o d F ( 2 4 )
~
12C
DD o f - 3 / 4 ~/ f r a c t i o n ,
#
11( , _ _ i - t - - - - ,
0 20 40 60 80 100
% PASS 3/4

FIG. lO--Compaction of gravel-sand-silt by various methods (grading T1).

Reference 4 shows that similar compacted densities of fine grained free


draining soils were achieved with vibrating tables (acceleration values above
3 g) with and without surcharge on the soil surface and with surface vi-
brating compactors. Furthermore, the discussion associated with Fig. 18
herein indicates a fair correlation between laboratory compacted densities
on different sites with different apparatus.
It therefore appears from the Tarbela, Mangla, and other data that the
maximum density value of nonplastic boulder-gravel mixtures containing
sand can be achieved fairly easily in the laboratory with the normal
varieties of vibratory compactors. It is probable that surface vibration is
somewhat more efficient than mold or table vibration, as the former im-
parts both pressure and motion to the soil particles.

Humphres Method

Procedure
This procedure is described in Ref 5, and only a brief outline is given
here. It consists of splitting the specimen into two portions; usually, but
not necessarily, into +No. 4 and - N o . 4 sieve sizes. On each fraction, a
solid (equal to specific gravity times the unit weight of water) and a
maximum and minimum (loose) density are determined in the laboratory.
The total densities of eight combinations of two of these six determinations
are then computed for various proportions of the fractions by means of the

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to
222 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

11C I I I I

CompQction method F(2zD.


t~ gradations between T2 & T7,
0(3
z<<_
~ 105 z~
~ MDD Hurnphres split on NO.4 ~ ~1 7 6 O0
MDD total sample c o m l ~ c t i o n - ~ 9 /"---
g.
,/o z; .
9 " ~ 2 "~;

-r -'~-/6"~ - - "~ ~ MDD Huml:~res sp~t on NO 4


9~ "/" 9 J~MDD Humphres spirt on 3/4"-
~~
o O_
f~

9(: I I I I
10 20 30 40 50
~ PASS NO. 4 SIEVE

FIG. l l---Comparison of Humphres Lab. M D D by splitting on ~ in. verus No. 4 sieve.

following equations:
1
~,T = (1)
P 1-P
(62.4Gp) ~m

where
1,T = theoretical dry density of total material (soil matrix and over-
size),
~ m - - dry density of the soil matrix,
~ = bulk specific gravity of oversize,
P = proportion (as decimal) of oversize to total material,
oversize = plus fraction at any convenient sieve size (usually No. 4 sieve),
and
matrix = minus fraction at any convenient sieve size (usually No. 4 sieve).
Equation 1, in effect, assumes that the matrix density of vm remains con-
stant while increasing amounts of oversize are added to it.

~ T = 9- ~/m
- or = -~p
- (2)
1--P P
where ~/p = d r y density of oversize.
Equation 2, in effect, assumes that the volume remains constant while
portions of one fraction are added to the voids of the other.
F r o m combined densities of fractions computed from Eqs 1 and 2,
Humphres adopts a somewhat empirical approach in graphically producing
a curve of mr versus matrix content.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to L
FROST O N BOULDER-GRAVEL FILL IN EARTH DAMS 223

Size of Fractions
To investigate the effect on the resultant Lab. M D D of the natural
gradation by the Humphres Method of splitting the specimen at different
particle sizes, a series of tests were carried out on materials of gradations
between the range of T2 to T7 in Fig. 2. A comparison of Lab. M D D of
natural gradation split on No. 4 sieve with that split on a/~ in. sieve is
shown on Fig. 11, together with a comparison between the former and the
Lab. M D D obtained by compacting the total material. Compaction Method
F(24 in.) was used for all tests.
The data of Fig. 11 indicates that splitting on No. 4 sieve gives M D D
values by Humphres Method which are lower than by splitting on ~ in.
sieve at sand contents less than 35 percent pass No. 4 sieve, but that the
position reverses for higher sand contents. A similar conclusion was found
with the relationship between M D D by Humphres Method split on No. 4
and M D D by compacting the total specimen.
While the low densities obtained by compacting total specimens of high
sand content agrees with the Mangla Dam experience, the discrepancy
between Humphres Method by splitting on either 3/~ in. or No. 4 sieve
suggests that a closer examination should be made of Humphres Method.

170

CONDITIONS :
(a) (Mclx~ivlJn.)Dry density of-6yr
=20 Ibs/cu.ft.
#
160 U
(b) Max. DD o f - N o 4 = 130 Ibs/cu.ft.
u_
(r (Mex,-Min.) D.D of-No 4 =40 Ibs/cu.ft
~tD
(d) Specific gravity . 2.78 I

For correction under other


150 ~)
conditions, see figures 13,14j15.

z
I
/
7 I
130
I
JS" ' t,.~p--~ Example
c~
.'t ~ / I
1120 9
I x
<
I
I I
I
i i i i I
45 8b
RATIO % PASS 3"/~ PASS 3/4aOF-6 " ~ PASS No. 4
TO + No. 4 FRACTION

FIG. 12--Mangla grading~Lab. MDD correlation--uncorrected MDD. of - 6 in. frac-


tion.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
224 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

Mangla Dam Grading/Lab. MDD Correlation


Description
Mangla Dam contains some 55 million cubic yards of boulder-gravel fill.
Initial laboratory tests by Method A (Table 2) indicated that the material,
whose contractual gradation limits are shown as M1 (Fig. 5), could be
described as essentially free-draining or nonfree-draining; the division being
below or above about 20 percent passing No. 4 sieve, respectively. Com-
paction of the latter material resulted in bleeding of the surface and Lab.
MDD values less than expected. Consequently, Humphres Method was
introduced with separation of the - 6 in. test material at the No. 4 sieve;
the plus product being compacted by Method A, the minus product by
ASTM D 1557-70 (Mod. AASHO effort).
After some 300 such determinations of Lab. MDD had been made, a
correlation was established between such values and the gradation (shown
on Figs. 12 to 15). Since Humphres Method is based on five variables
(namely, maximum and minimum densities of the plus and minus fractions
and the proportion of such fractions), a correlation was first established
between the laboratory densities and gradation parameters. For example,

% Pass No. 4 of total ~ampf~


o o ~ ~
8~
~ ~0 B-CorrectiOn for minimum dry density
/ I of-G*/+No. 4 fraction (to be added
to uncorrected M.D.D.from Fig.12)
in Ibs/cu.ft.

q
+
J/j z
O
A

/!

-30
J ~ \"\\\\\ \\\
'-- - - ,,,f-- q - b
I \\ \

Example
~
I

O 20 ,, 40 100
~ PASS 11/2 SIEVE OF - 6*/ + No.4 FRACTION

FIG. 13--Mangla grading~Lab. MDD correlation--correction for (maximum-minimum)


dry density of - 6 in./ T No. 4 fraction.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
FROST O N BOULDER-GRAVEL FILL IN EARTH D A M S 225
l i I 60 I I J ]
C - Correction for (max-rain)
d r y density o f - ~ . 4
fraction (to be added to ~ 2
uncorrectedMD.D of Fig J 2 ) -50
in I b s / c u . f t .

./8 /
L
~t

-30

-20

~I 7 ~ - - C (See note abOVe )


o ~'o ~b Go o 20 40
#
60 810 lOO
~ PASS NO. 2 0 0 OF ~No.4 FRACTION ~ PASS NO. 4 (-6 FRACTION)
FIG. 14--M angla grading/Lab. M D D correlation--correction for (maximum-minimum)
dry density of - N o . ~ fraction.

the MDD of the - 6 in. to -FNo. 4 fraction was found to be related to the
ratio of percent passing 3-in. sieve to that passing a/~-in, sieve as shown on
Fig. 12. The Humphres Curves of Fig. 12 were drawn with variable MDD
of - 6 in. to + N o . 4 fraction, but with the other three laboratory densities
held constant. Then, Figs. 13 to 15 provide the correction to be applied for

020 40 6 0 8 0 1 0 0

/ / /X~ r 135 r-
~L

\ 130zU

L ~255

i Examp~,._~.t \ \\ /t-120Z
~E
\1 130 ':50 6 0 40
~/o PASS NO 4
20
I
0

' 1
.I5 I I I
2 253 4
I I
5 6
I I I
g 10 12 1 51-1 i
-10
i
-5
Ii
0 +5
i
+10
RATIO % PASS NO25/% PASS No.2C0 CORRECTION (D) FOR MAX DRY DENSITY O F - N o 4
FRACTION ( ADD TO UNCORRECTED MDD FROM FIGURE12)
N LtBS/CU. FT

FIG. 15--Mangla Grading~Lab. M D D Correlation--correction for maximum dry density


of the --No. ~ fraction.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further
226 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

. . . . //
/./
T~sT~.
..
' ..r
160
5
<
d
Z 150 ::?t.
d

~ 14o

./" ~ ~ G e n
/ ..//.~ / \ o~ , s
il I scatter"
i~/c0.ft,
~
U
130
/~'/y.
/ 7 . e

/__1
f20 130 lJtO 150 1s 170
~o.~..~,o. ~. ~oo. ,, ~ s / ~ u . ~ .

FIG. 16--Mangla grading~Lab. MDD correlation--comparison of actual and computed


Mangla densities.

the actual laboratory densities; these corrections being added to the un-
corrected MDD value of Fig. 12.
A comparison of the actual laboratory density with that determined by
the correlation is presented on Fig. 16 and shows reasonable agreement

i
bunded)
ingladam(M1)

us
dams
~8 g unknown)

2 mvcrt~d to
~ ;.-"~
i
O 20 40 60 t50 1OO
~ PASS NO. 4 SIEVE

FIG. 17--Mangla grading/Lab. MDD correlation--comparison with data from other


dams.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No furt
FROST ON BOULDER.GRAVEL FItLIN EARTH DAMS 227

within a general scatter of 4-5 lbs/fP. This is believed to be the order of


repeatability of the Lab. M D D test.
The correlation of Figs. 12 to 15 is applicable to other materials if the
relationship between the grading parameters and the laboratory densities
is established and superimposed on these figures in place of the Mangla
relationships.
Comparison with Other Data
Figure 17 presents a comparison of the actual Lab. MDD values de-
termined at a number of dam sites including Mica, Jari, Oroville, Tarbela,
and Cougar with that determined by the Mangla Correlation from the
material gradation. All data is reduced to a common specific gravity basis
of 2.78.
In general, good agreement (within 5 percent) was achieved with the
correlation on most project data. Those large differences that occurred
with Cougar Dam material demonstrate that transposition of correlations
of this nature cannot be undertaken rashly.

140 i i 1 ~ i i i i i i i i

AIJ densities
common SG., 2.78
conwrtcd to ~on~o._t'~_.
qO'"" " -~'-~9~ J
J
135 //" .

,,." ~.-" I~.Mangla


." ~.../ ~oboog~=I
I
..~.. ,~ .i//" "
,o-
angu~ar
. ~ ."~ .y / gr~v~,,

7= /\ /"'" I / //

../ .--" /

11(

/ / Tarb~la angular //"/


.i gravel /

10= ~ ,/ k O/o Pass 3" of -G*


to+Nb 4 f r a c t i o n
/

1~ 12 -6 7 6 5 4.0 3.0 2.s -So 1.~ 1.o


RATIO ~ PASS 3"/% PASS 3 / 4 0 F - 6 ~ T O - P N o . 4 FRACTION

FIG. 18--Gradation variables effecting Lab. MDD of --6 in. to +No. 4 fraction.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
228 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

Effect of Particle Shape~Grading


The partial lack of agreement between Mangla Correlation Lab. MDD
values and those actually obtained on other projects, as shown on Fig. 17,
is believed to be mainly due to differences in the compaction equipment
and effort or to differences in particle shape or both (grading and specific
gravity being covered in the correlation). Since it was felt that the fraction
most sensitive to particle shape influences on Lab. MDD would be the
- 6 in. to +No. 4 fraction, a series of tests were performed at Tarbela on
this fraction. The materials tested were angular boulder-gravel and river
rounded boulder-gravel, and the results are shown on Fig. 18.
The Mangla correlation variable of ratio percent pass 3 in. over percent
pass a/~ in. was maintained (which for these materials is somewhat equiva-
lent to coefficient of uniformity), but a second variable of percent pass 3 in.
was introduced. Also shown on Fig. 18 are the Mangla Correlations from
Fig. 12 which have a range of percent pass 3 in. of 55 to 70.
The data indicates that despite the relative one constituent nature of
the - 6 in. to +No. 4 material, large differences in Lab. MDD resulted
from changes in grading and particle shape. However, the Mangla data
agreed with that from Tarbela with the same particle shape (rounded) and
percent pass 3-in. sieve (55-70) suggesting similar Lab. MDD for similar
material despite the large difference in compaction method.

Oversize Corrections
Equation 1 represents a theoretical attempt to compute the maximum
density of the total mixture when only the maximum density of some finer
fraction is available. This allowance for the coarser fraction omitted from
the test is known as the oversize correction. Previous experience (for
example, Ref 6) has shown that addition of oversize had a "disturbing"
effect on the matrix density and that the total density was less than that
given by Eq 1; significant deviations occurring with oversize contents in
excess of between 30 and 50 percent.
The results of a number of tests on various materials with various over-
size added are shown on Fig 19 in the form of deviation (D) of actual
Lab. MDD from the MDD by Eq 1 as a percentage of the latter. The
actual Lab. MDD were obtained by various methods including dynamic
(ASTM D 1557-70), vibratory, and kneading compaction as shown in
Table 1. The value of ~m used in Eq 1 is that obtained at zero oversize
content for the particular compaction method.
Generally, the data of Fig 19 follows a linear relationship, (although
Materials T1 and T5 are exceptions) of the form:

logD = a P + b, or
D = antilog(aP + b) (3)

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement.
FROST O N BOULDER-GRAVEL FILL IN EARTH DAMS 229

z
bJ
>
@

LU
0
W

~G
Q8

U.

0 10 20 30 40 GO 60 70 80
PERCENT OVERSIZE ADDED ( = P )

FIG. 19--Oversize corrections.

This form of presentation has been sdected as it most readily enables the
deviation to be used as a correction of E q 1, as follows:

~T(corrected) = (100~0 D ) 5"T (Eq 1) (4)

The constant a represents, in physical terms, the extent of disturbance


of the matrix by the oversize, which results in less than theoretical density.
Hence, the greater the sensitivity of the matrix-to disturbance, or the
greater the disturbing shape of the oversize, the higher the value of a
obtained. While the constant b should strictly be equal to zero as no dis-
turbance is possible at zero oversize content, the plots of Fig. 19 indicate
small values of b which result in small values of D (less than 0.4 percent).
Having regard to the complexity of the matrix-oversize interaction, it
is not possible to adopt general values for a and b, but some guidelines can
be obtained from the data of Table 1.
Model--Prototype Relationships
A major problem in laboratory compaction tests on boulder-gravel
material (the prototype) is the proper representation in the laboratory of
the particle oversize in the field which is necessarily omitted from the
laboratory test specimen (the model). Three possible solutions are (a) over-
size correction, (b) extrapolation, and (c)tnodeling.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licen
T A B L E 3--Model-prototype relationships illustrated by Lab. M D D for grading T1. ~o
r
o

ItemNo. Method Procedure Lab. M D D Nomenclature


ASTM
D 1557-70,
(lb/ft ~)

1 Oversize correction Lab. M D D on -- aA in. scalped fraction corrected by Eq 3 and 140.2 F = fraction to be replaced
4. f = replacement fraction
2 Extrapolation (from Lab. M D D on fractions - N o . 4 (131.4), - ~ in. (132.8), 141.6 P = % of F by weight
scalped fractions) _ a/~ in. (135.9), -- 1 in. (137.4) and -- 11~ in. (138.3) extrapo- p = % of f by weight 0r -
lated to --4 in. D = Average diameter of F
3a Lateral grading shift Lab. M D D on -- 1 in. grading computed graphically by shift- 136.8 d = Average diameter of f o
ing from 100% pass 4 in. to 100% pass 1 in. SG = specific gravity of F r
3b Weight replacement Lab. M D D on -- 1 in. material containing a weight of - 1 in. 137.8 sg = specific gravity of f o- r
to -]-~ in. replacement equal to weight of --4 in. to Ta/~ in.
replaced. o
Lab. M D D on - 1 in. material containing a weight of - 1 in. 138.9 Z
Y.,
to q-No. 4 replacement equal to weight of - 4 in. to q-No. 4
replaced.
3c Cross sectional area re- Lab. M D D on -- 1 in. material containing a weight of -- 1 in. 139.3 o_
placement to q- a/~ in. replacement of -- 4 in. to -{-a/~ in. replaced given by
equation
d SG
p . . . . . p
D sg
3d Surface area replacement --do-- 139.3
3e Volume replacement Lab. M D D on -- 1 in. material containing a weight of - 1 in. 137.8
to q- ~ in. replacement of --4 in. to W ~ in. replaced given by
equation
sg
p ='~'P

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reprod
FROST ON BOULDER-GRAVEL FILL IN EARTH DAMS 231

The oversize correction solution was shown to have a theoretical basis


with an adjustment factor (D) based on actual test data. The extrapolation
solution consists of performing compaction tests on scalped fractions (of
the prototype) of different maximum size within the range of sizes that
can be tested, and extrapolating the graphical plot of maximum size versus
Lab. M D D to the actual maximum size of the prototype. The third so-
lution, modeling, consists of adjusting the grading of the model so that it
behaves similarly to the prototype. The various methods of performing
this adjustment are (a) lateral grading shift, (b) weight replacement,
(c) cross sectional area replacement, (d) surface area replacement, and
(e) volume replacement.
The lateral shift method consists of shifting the gradation curve laterally
to the fine side by an amount equal to the difference (in linear measurement
such as inches, not in particle size) between the maximum particle size of
the prototype and of the model. This method ensures that the relative
distribution of particle sizes is the same (the gradation curves have the
same shape), but it is effectively equivalent to scalping. The replacement
methods consist of replacing the oversize fraction by an amount of finer,
but still coarse, fraction on the basis that the latter will have either the
same weight, the same cross-sectional area, the same surface area, or the
same volume as the coarse oversize fraction omitted.
These concepts are illustrated in Table 3 which provides details of the
transformation equations, together with actual Lab. M D D data obtained
from t e s t s o n Material T1 (Fig. 2). Equations under items 3c and 3e are
derived by geometry and the assumption that the coarse fractions have
similar degrees of lack of spherity.
The results suggest that oversize correction and cross sectional area
replacement are most similar and more like the expected Lab. M D D of
the prototype.

Relative Compaction Versus Relative Density


Fill in situ (FDD) and laboratory maximum and minimum dry densities
(Lab. MaxDD and Lab. MinDD) are usually compared in the dimension-
less terms of relative compaction and relative density for which the equa-
tions are as follows :--
FDD 100
Relative Compaction = X (5)
Lab. MaxDD 1

Lab. M a x D D ( F D D Lab. MinDD)100


- -

Relative Density = (6)


FDD(Lab. MaxDD - Lab. MinDD)

For boulder-gravel materials, it is the author's experience that the order


of error or repeatability for field and laboratory tests is up to 3 to 5 lbs/ft ~.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
232 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

Such errors in Eq 5 result in total errors of less than 7 percent in the worst
instance, and usually only a few percent.
However, since the difference between laboratory maximum and mini-
mum densities is only of the order of 30 lbs/ft a, the errors which are ac-
ceptable for Eq 5 are magnified in Eq 6, and errors in relative density of
up to 50 percent may be produced. Therefore, it is believed that relative
density concepts, which have generally been used for fine grained free-
draining soils, should not be used for boulder-gravel material, but instead,
relative compaction concepts should be used.
Fill Compaction
No general paper of this kind would be complete without some comments
on compacted fill densities. The large quantities of boulder-gravel fill in
Mangla (55 million cubic yards) and Tarbela (181 million cubic yards)
Dams were compacted in 2-ft layers with 4 passes of a 10-ton (static)
weight smooth vibrating roller; Allgemeine Baumaschinen Gesellschaft
rollers being used on Mangla and Dingler Rollers on Tarbela. The fill com-
pacted by these rollers reached densities which generally were 97 to 100
percent of the Lab. M D D values obtained by laboratory Methods A or
F(24 in.), respectively. Other projects of the author's knowledge reached
similar density levels.
It therefore appears that compaction of nonplastic boulder-gravel fill
containing up to 50 percent silt can be readily achieved with properly
chosen vibrating rollers.
Summary
Comments on the various aspects of the paper have been given in the
relevant sections. However, some overall comments as they specifically
apply to boulder-gravel material are as follows:
1. Fill gradation and density testing can be performed within acceptable
limits of accuracy.
2. Laboratory testing for determination of maximum density has been
carried out to date with a wide variety of apparatus and procedures. Al-
though it appears that these give essentially the same result, it is desirable
that a standard procedure be adopted. It is suggested that Method F (24 in.)
as described is worthy of consideration.
3. The problem of prototype-model gradation relationships is acute,
since the laboratory most often can only test a fraction of the fill material.
The oversize correction method described in the paper offers the most
promise of a solution.
Acknowledgments
The views expressed in this paper are those of the author and not neces-
sarily those of the various organizations by whom he has been employed.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License
FROST ON BOULDER-GRAVELFILL IN EARTH DAMS 233

Binnie and Partners of London and T i p p e t t s - A b b e t t - M c C a r t h y - S t r a t t o n


of New Y o r k were the consultants on the Mangla and Tarbela Dams,
respectively; the owner being the W a t e r and Power Development Authority
of West Pakistan. The author wishes to t h a n k those organizations for the
opportunity to have gained the experiences described in this paper.

References
[I] British Standard 1377, Methods of Testing Soils for Civil Engineering Purposes, 1967.
[2] Broms, B. and Forssblad, L. in Proceedings, Specialty Session 2, Soil Dynamics, 7th
International Conference, Soil Mechanics and Foundation Engineering, Mexico,
Aug. 1969.
[3] Gordon, B. B., Hammond, W. D., and Miller, R. K., "Effect of Rock Content on
Compaction Characteristics of Clayey Gravel," Compaction of Soils, ASTM STP
377, American Society for Testing and Materials, 1965.
[$] Forssblad, L., "Investigations of Soil Compaction by Vibration," Acta Polytechnica
Scandinavica, Vol. 624, No. 138, Stockholm, 1965, p. 22.
[5] Humphres, H. W., "A Method for Controlling Compaction of Granular Materials,"
Bulletin 159, Highway Research Board, 1957.
[6] Earth Manual, U. S. Department of the Interior, Bureau of Reclamation, 1960.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No furt
R. J. Stephenson ~

Relative Density Tests on Rock Fill at


Carters Dam

R E F E R E N C E : Stephenson, R. J., " R e l a t i v e D e n s i t y T e s t s o n R o c k F i l l


a t C a r t e r s Darn, ** Evaluation of Relative Density and Its Role in Geotechnical
Projects Involving Cohesionless Soils, A S T M S T P 523, American Society for
Testing and Materials, 1973, pp. 234-247.

A B S T R A C T : During the construction of Carters Dam, 126 relative density


tests were performed on specimens selected from over 200 in-place density tests
made throughout four zones of the compacted fill in the dam. Each specimen
was dug from the compacted fill and was approximately one cubic yard in
volume. The material varied from weathered rock to sound fresh quartzite
with up to 36-in. maximum particle size. The relative density of the compacted
fill was seldom less than 100 percent, and in many instances, it substantially
exceeded the maximum vibrated test density. All of the material in each speci-
men was graded to determine the grain size distribution. This series of tests
indicated further development in equipment and procedures is needed to obtain
laboratory densities for rock fill that are comparable to the densities achieved
in the field w i t h vibratory rollers. Valuable experience was gained through the
problems and utilization of large scale density tests of rock fill material.
K E Y W O R D S : cohesionless soils, construction, dams, field tests, earth fills,
rock fill dams, soil mechanics, tests, density (mass/volume)

In recent years the use of rock fill in dam construction has increased
considerably, but information is scarce concerning relative density tests on
broken rock that contains large percentages of material coarser than the
3-in. particle size. The U. S. Corps of Engineers, however, obtained a large
amount of density data on rock fill during construction of Carters Dam
in the Mobile District. At Carters Dam a series of large scale density and
gradation tests were made on the rock fill throughout construction of the
main dam from 1965 to 1969. This paper summarizes those tests where 126
maximum and minimum density tests were performed on specimens se-

x Director, South Atlantic Division Laboratory, U. S. Army Corps of Engineers,


Marietta, Ga. 30060.

234

Copyright by ASTM Int'l


Copyright9 by(all
ASTMrightsInternational
reserved); Fri Mar 11 16:13:06 EST 2016
www.astm.org
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
STEPHENSON ON RELATIVE DENSITY TESTS ON ROCK FILL 235

TABLE 1--Location and compaction specifications for rock types in Carters Dam.

Embankment Zone Type of Rock Specified Maximum Lift


Particle Size, Thickness, in.
in.

3A, best quality rock sound fresh quartzite with less than 36 36~
30% phyllite and argillite
3B, second best rock slightly weathered quartzite with 24 24~
less than 40% phyllite and argillite
3C, random rock moderately weathered quarry run 24 24~
rock
2, transition material spalls and weathered rock 8 12b

a Compacted by four passes with 10-ton vibratory roller at not over 1.5 mph or two
passes with 15-ton roller.
b Compacted by four passes with 50-ton rubber-tired roller at not over 5 mph.

lected from over 200 in-place density tests. A grain size analysis was also
performed on every specimen.
Carters Dam is located on the Coosawattee River about 60 miles north
of Atlanta, Ga. The main dam is 445 ft high (the highest earth-rock dam
east of the Mississippi River) and 2050 ft long. The total volume of the
dam, rock and earth, is about 15 million cubic yards with about 13 million
cubic yards of rock. The rock fill includes four zones that contain rock of
varying degrees of hardness and weathering. These zones are located up-
stream and downstream of the earth core to achieve the most efficient
design and utilization of the available material.

Material Tested
The rock fill is composed of a continuous series of rock types with quartz-
ites predominating. These rock types can generally be grouped, on the basis
of their physical characteristics, into quartzite, argillite, and phyllite cate-
gories. Other intermediate rock types exist, but their quantity is insignifi-
cant. The specific gravity of the rock varies from about 2.75 to 2.86.
The quartzites are the best quality rock in the area with excellent engi-
neering properties. When quarried they break into blocky shapes with
abrasive faces. The term "argillite" is used to define a group of rocks whose
properties are between the quartzites and phyllites. These rocks are hard
and resemble slate except that they do not cleave into thin slabs but
quarry more as a semi-slabby to massive rock. The phyllites are moderately
hard with a characteristic wavy foliation and silky sheen on the fracture
faces. Table 1 indicates the zones in the embankment and the compaction
specifications where the various rock types were placed. Table 2 summarizes
the basic characteristics of the vibratory compaction equipment.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lic
236 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

TABLE 2--Characteristics of vibratory compaction equipment used on rockfill.

Manufacturer and Modela Weighton Drum, Centrifugal Force, rpm


lb lb

Bros VP 20D (10-ton) 18 900 40 000 1300


Ferguson (10-ton) 22 000 41 500 1300
Vibro-Plus CTo60 (15-ton) 25 000 60 000 1200

a Weights shown in parentheses are the designationsused in this paper.

Test Equipment
Density Ring--A 6-ft diameter by 8-in. high ring made of structural
steel plate served as a template for the in-place density tests in the com-
pacted rock fill.
Specimen Container--The container used for the maximum and minimum
density measurements was specially fabricated from rolled steel stock. It
was 41 in. in diameter by 46 in. deep with a volume of just over one cubic
yard. The inside of the container was vertical for 24 in. and then rounded
into a spherical shaped bottom. Accessories for it included a special cover
plate and a bearing plate. The function of these accessories will be described
later.
Truck-Crane--East test specimen weighed over one ton, and with the
container, the combined weight was nearly two tons. Thus, a large truck-
crane was required for the heavy lifting and transporting of the specimen
and container. An integral part of the truck-crane was a water tank, pump,
and meter for measuring the water used in volume measurements.
Scales--Platform scales with a capacity of 5000 lb were specially built.
to provide the requisite capacity and accuracy in weighing.
Vibrating Table--A vibratory table actuated by an electromagnetic vi-
brator in accordance with ASTM Test for Relative Density of Cohesionless
Soils (D 2049-69) with sufficient load capacity was not available for this
series of tests. Instead, a Syntron RVP-9 rotary motor vibratory packer
was used. This device produced a horizontal elliptical vibratory motion
utilizing a 900 rpm rotary, eccentric, centrifugal force of approximately
6000 lb. The motion developed in the loaded specimen container can best
be described as a "rocking" motion in which the top and bottom of the
container pivoted in an elliptical path about the center which rotated in a
smaller circular path. The horizontal displacement at the top and bottom
of the container was about 2 in., while the vertical displacement was about
1 in. These displacements varied slightly depending on the weight of the
material in the container and the length of the moment arm from the input
force to the center of gravity of the vibrating mass.
Air Bag--One of the major components of the surcharge apparatus for
the maximum vibrated density test was an inflatable air bag. It was

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licen
STEPHENSON ON RELATIVE DENSITY TESTS ON ROCK FILL 237

fabricated of rubber, 40 in. in diameter to fit inside the specimen container.


It was pancake-shaped with an accordian fold that permitted expansion in
a vertical direction.
Air-Compressor--An air compressor was required to provide a source of
regulated air pressure for the surcharge apparatus. Compressed air was
also required to inflate the air rides (which act as springs) on the vibrating
table.
Electric Generator--The vibrating table required 220-V, 3-phase elec-
tricity which was not available at Carters. Therefore, a diesel powered
generator was an extra expense and problem at this particular location.
Screens--For the grain size analyses, heavy duty screens were specially
fabricated 3-ft square with square openings of 1, 3, 6, and 8 in. In addition,
a standard mechanical shaker for grading sizes smaller than 1 in. was used.
Field Laboratory B u i l d i n g - - A n open-air building with a concrete floor
was constructed to house the test equipment and provide a suitable area
for testing and processing the specimens. An important feature of the
structure was sufficient overhead clearance for heavy lifting with the truck-
crane. An overhead chain hoist mounted on a track running the length of
the building was also used for handling portions of the specimen separated
during the grain size analyses.

FIG. 1--The 6-ft diameter density ring positioned on surface of compacted rock fill for
in-place density test.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
238 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

FIG. 2--Carpenter's level and scale utilized as reference point while membrane-lined
density ring is filled with metered water during in-place density test.

Test Procedure

In-Place Density of the Compacted Rock Fill


The density of the compacted rock fill was determined in a manner
similar to the standard water or oil density procedure (ASTM Test for
Density of Soil in Place by the Rubber-Balloon Method (D 2167-66))
except on a much larger scale. The density ring was placed on top of the
rock fill at the test site as illustrated in Fig. 1. To calibrate the surface, a
thin plastic membrane was placed within the ring and filled with metered
water to a reference point as shown in Fig. 2. The water and membrane
were removed and the compacted rock fill carefully excavated from within
the ring with hand tools until the specimen container was full. The material
was shoveled directly into the container as it was excavated from the hole.
When rocks over 12 in. in size were encountered, they were set aside for
weighing and grading but not placed in the specimen container for the maxi-
mum and minimum density tests. The hole excavated within the ring was
lined with the plastic membrane and filled with metered water to the surface
calibration reference point. Thus, the difference in the two water volume
measurements equaled the in-place volume of the material removed from
the density hole. This volume was usually about three fourths of a cubic

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agree
STEPHENSON ON RELATIVE DENSITY TESTS ON ROCK FILL 2 3 9

yard. The total specimen was subsequently weighed at the field laboratory
and its compacted in-place density computed.
M i n i m u m Density Measurements
To avoid densification of the material inside the specimen container dur-
ing handling, its "loose" or minimum density volume was obtained before
moving from the test hole site. The volume of the container had been
calibrated earlier against its depth, so the loose volume of the specimen
was obtained by measuring its depth in the container. The minimum den-
sity was computed on the basis of this volume when the weight of the
material in the specimen container was determined later.
M a x i m u m Density Measurements
At the field laboratory the specimen and container were weighed and
the container bolted to the top of the vibratory packer as shown in Fig. 3.
The sample was then vibrated in the container for 15 min under a surcharge
of 10 psi. During vibration the surcharge was applied by the air bag which
was placed between the bearing plate resting on top of the specimen and
the cover plate bolted to the top of the container. As densification occurred
during vibration, the air bag expanded and forced the bearing plate against
the specimen, thereby maintaining the surcharge on the specimen. After

FIG. 3--Sample container positioned on vibrating table for maximum vibrated density
test.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
240 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

FIG. 4--Heavy-duty screens for gradation analyses on rock fill material.

vibrating, the surcharge apparatus was removed and the volume of the
specimen determined by measuring the depth of the material in the con-
tainer as described previously. This volume was then used to compute the
maximum vibrated test density.

Grain Size Analyses


All of the material in each specimen was graded over the 1, 3, 6, and
8-in. screens. These screens were placed in a wood frame forming an inclined
chute as shown in Fig. 4. The specimen was dumped at the top of the chute,
and the material raked by hand over the screens. Particles larger than 8 in.
were measured with a carpenter's rule. The material finer than 1 in. was
quartered, and a representative specimen of about 150 lb graded through
a mechanical set of screens. The material retained on the various sieves was
weighed and the grain size distribution data computed in the conventional
manner in accordance with ASTM Particle-Size Analysis of Soils (D
422-63).

Special Processing of Wet Samples


Occasionally, the rock fill was moist when the in-place density test was
performed, so the specimen had to be dried for the minimum and maximum
density tests. In those instances, the moist material was spread under heat
lamps in the field laboratory to air dry. When air dry, the material was

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agree
STEPHENSON ON RELATIVE DENSITY TESTS ON ROCK FILL 241

recombined, shoveled back into the specimen container, and the loose and
vibrated density volume measurements determined as previously described.

Moisture Content
Moisture content determined in accordance with A S T M Laboratory
Determination of Moisture Content of Soil (D 2216-66) for the specimens
in this test program was not considered significant, because the materials
were cohesionless with no plastic fines. Moreover, the contract specifi-
cations contained no moisture content requirements for compaction of the
rock fill. All density measurements were based on air d r y weights as dis-
cussed elsewhere in this paper.

Manpower Requirements
It is emphasized that considerable manual effort was required to conduct
these tests. Ordinarily, two technicians comprised the crew performing the
tests. When the weather was good and the material dry, two men could
perform a complete series of tests on one specimen, including the grain
size analysis, in a long workday.

Results

Density Tests
The average density test values are summarized for each embankment
zone and type of rock in Table 3. Note the compacted rock fill densities
exceeded the maximum vibrated test densities in each rock type. This is
contrary to the usual relationship where the compacted fill density of
cohesionless material is between the loose, or minimum, density and the
maximum vibrated test density. At Carters, the first few tests indicated
the compacted rock fill would exceed the vibrated test density most of the

TABLE 3--Suramary of density test results.

Type of Rock Average Dry Density, lb/ft 3


Compacted Vibrated Test, "Loose", or
Rock Fill Maximum Minimum

Sound fresh quartzite 137 122 109


(72)- (58)" (58)"
Slightly weathered quartzite 140 123 110
(11) (11) (11)
Moderately weathered, quarry run 142 123 108
rock (42) (36) (36)
Spalls and weathered rock 132 109 99
(62) (21) (21)

" Numbers in parentheses indicate the number of observations.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
242 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

140
LEGEND FOR
v. VIBRATING ROLLERS /
m X I 0 - TON
>: 9 15 - TON
/ ~ /
t30 / /
/
Z /
U.I

/
/ o
/x 9
x
xx,,'.:
~ " / /. , , ,
~ 120 ~
/ 9 o/
I--
/ /9 99XXx/
9 9 ~/ 9 9
../
/

U.I
I..- ./
I10 / :
/
/ / / -X--%OF VIBRATED
/ T E S T DENSITY
m // /
/
lOOi /
I10 120 130 140 150 160
MEASURED DRY D E N S I T Y - C O M P A C T E D FILL, I b / f t 3

FIG. 5---Sound fresh quartzite density data.

time. Indeed, this was true in all but four sets of tests during the program.
Therefore, relative density computations in the usual manner (ASTM
D 2049-69) were not made because relative density values above 100 per-
cent were not considered appropriate.
Since relative density values above 100 percent were undesirable, the
ratio of each compacted rockfill density to the corresponding vibrated test
density was expressed in percent. In general, the density of the compacted
fill was between 100 and 130 percent of the maximum vibrated test density.
This is shown graphically in Figs. 5, 6, 7, and 8 where the data is summarized

~ 140
LEGEND FOR
*9"
>.-
VIBRATING R O L L E R S
X I0 - TON
/ /' ,~
9 15 - TON
,- t30 / / /
t,,D
Z
hi
/ / ~.x x /
0 / / /.
>,- / /
Q: 120
Q / /x
I'.-
(/)
LIJ
//- / x/
/ / /
o I10 i
W
I-.-
/ / "Xr OF VIBRATED
(n
/ / T E S T DENSITY

> I00
I 0
/ 120 130
/
140 150
I
160
I
MEASURED DRY D E N S I T Y - COMPACTED FILL, I b / f t 3

FIG. 6----Slightly weathered quartzite density data.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further
STEPHENSON ON RELATIVE DENSITY TESTS ON ROCK FILL 243

i,q<_ 140
LEGEND FOR
VIBRATING ROLLERS
X I0 - TON /
I-- 150
9 15 - TON / v/ _i
Or) / x /o "" 9 9 9
f
Z
LIJ
E3 9 x X ~ ""
>-
n,* 120
/
/
/ 9 //. ~,@~
E3 / /
/ /
I--
oO / / "/ /" ./
U,.I / : /
I-- / / x/ 7" x
Q II0 /" /
ILl
/ / >. -)(.% OF VIBRATED
/ / /l I TEST DENSITY
El
I00 9 / / I
I 0 120 130 140 150 160
MEASURED DRY D E N S I T Y - COMPACTED F I L L , I b / f t 3

FIG. 7--Moderately weathered, quarry run rock density data.

for each type of material. These figures also show the range of the data,
and indicate each type of roller used for compacting the rock fill was
equally effective in obtaining densities above the vibrated maximum test
density.
Grain Size Analyses
The grain size distribution was determined for every specimen taken from
the compacted fill, whether or not the minimum and maximum vibrated

140
FILL COMPACTED BY 4
..Q
PASSES OF 5 0 - T O N /
B

1-- 130
RUBBER TIRE ROLLER / *S
O9 / /
Z
hl / /

rr 120
/. /
/
E:I
/ J /
9/ . 9 / /
LU
I'- / "/ -/
II0 //
LU
l-- / :/'i o'/
-)('-% OF VIBRATED
n" / TEST DENSITY
m
> Ioo / / / f
I10 120 130 140 150 160
M E A S U R E D DRY D E N S I T Y - C O M P A C T E D F I L L , I b / f t 3

FIG. 8---SpaUsand weathered rock density data.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreeme
244 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

I' COBBLES I GRAVEL I SAND ] illEl=l~


u.S. STANDARDSIEVEOPENINGS,IN U.S.STANDARDSIEVENUMBERS
I00_ ~ 12 6 3
I
I
I
I/2
I
4 I0
I
20
I
40
I
I00
=
200
I

1
I
I"- 8O zo
L9
:2=
(.9
1 - ~

>- 6 0
I 40 m
{1:
rr
W
Z
U-
~- 4 0
z
: \%. ,', -

60
~
0
o
w
t)

n
20
I I 80 ~

o I00
500 I00 50 I0 5 I 0.5 O. I 0.05 0.01
GRAIN SIZE, MILLIMETERS

FIG. 9--Sound fresh quartzite gradation curves.

I COBBLES I GRAVEL I SAND I FINES


U.S. STANDARD SIEVE OPENINGS, IN. U. S STANDARD SIEVE NUMBERS
12 6 :3 I I/2 4 I0 20 40 I00 200
, o o . . . , . . , , , , , ,
I
I
I
2o

- ~
z

60 ~_

n 20 80 ~.
0 IJ_LI IOO
500 I00 50 IO 5 I 0.5 0.1 0.05 o.ol
GRAIN SIZE, MILLIMETERS

FIG. lO--Slightly weathered quartzite gradation curves.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant t
STEPHENSON ON RELATIVE DENSITY TESTS ON ROCK FILL 245

I COBBLES I GRAVEL I SAND I ,FINES ]


U.S. STANDARD SIEVE OPENINGS,IN U.S. STANDARD SIEVE NUMBERS
12 6 3 I I/2 I0 20 40 I00 200
I00
' i , , , ,

l'--
8O 20
T (.9
(.9 ~J

>-
>-
tn 60 40 ,.n
(z:
tr bJ

14. 0
~ 40 60 ~
I--
z
w 1 z
~.)
w
I -
I1:
20 BO ~

0 I I I I00
500 I00 50 I0 5 I 0.5 0. I 0.05 0.01
GRAIN SIZE, MILLIMETERS

FIG. ll--Moderately weathered, quarry run rock gradation curves.

I COBBLES I GRAVEL I SAND I FINES I


U.S. STANDARD SIEVE OPENINGS,IN. U.S STANDARD SIEVE NUMBERS
12 6 3 l l/2 4 lO 20 40 IO0 2 0 0
I00 , , , , , , , 0

i.- 80 2o ~=
.-r (.9
(,9 W
>-
e>-
n 60 40 m

W
Z 11:

0
I- 40 60 ~
z F-
w Z
W

Q. {IE
W
20 I 8 0 a.
I
I I
I I
0 I I IIIII I I J I llllll~ll I I IIIIIII I I IIIIIIll I I I00
SO0 IOO 50 I0 .5 I 0.5 0.1 0.05 0.01
GRAIN SIZE, MILLIMETERS

FIG. 12--Sp~lls and weathered rock grada$ion curves.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
246 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

density tests were performed. Typical gradation curves for specimens on


which a complete series of density tests were performed are shown in
Figs. 9, 10, 11, and 12. These particular curves also represent the gradation
"band" obtained for alI the specimens taken from each embankment zone
during construction.

Discussion
Test Equipment
Since the vibrated test densities were less than the compacted rock fill
density, there is some question about the adequacy of the specimen con-
tainer and the vibrating table. During vibration, the material tended to
"rotate" down one side of the container and up the other instead of con-
solidating vertically under the surcharge. It has been suggested the spheri-
cal shaped bottom of the container caused this; however, the vibratory
motion of the vibrating table could have been the major cause. If a fiat
bottomed container were used, inaccuracy would probably arise from the
irregular surface effects in the inside corners of the container, particularly
when the specimen contains particle sizes as large as those involved in
these tests. A limited amount of data indicates no significant increase in
density would result with a larger container that would provide a higher
ratio of container diameter to maximum particle size. Duplicate tests were
performed on some of the specimens after removing the larger sizes to
evaluate this effect.
Higher vibrated test densities could probably have been obtained if the
frequency and amplitude of the vibrating table could have been varied.
Equipment with these features and the required load capacity was not
available for this test program. It was later determined that two rotary
eccentric vibrators operating in perpendicular directions on the vibrating
table would eliminate the elliptical motion and produce a sinusoidal motion.
This would compare more favorably with the motion produced by electro-
magnetic vibrators (ASTM D 2049-69) and probably increase the maxi-
mum vibrated densities obtainable.
Test Procedures
In the early tests a mortar of fire clay was prepared to smooth the
crevices in the density hole before lining it with the plastic membrane for
the volume measurements. This required careful measuring of the fire clay
and correcting for it in the computations. It was soon discovered the same
results could be obtained by smoothing the sides of the density hole as
much as possible during excavation and carefully placing the plastic mem-
brane to follow the remaining irregular surfaces. This noticeably reduced
the labor and the time required for the equipment to be set up on the
density hole site. Anything that shortened the time required for the equip-

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License
STEPHENSON ON RELATIVE DENSITY TESTS ON ROCK FILL 247

ment at the test site pleased the project and contractor personnel as well
as the technicians performing the tests.
For the minimum density tests, no "refinements" in the method of
placing the material in the specimen container were investigated to de-
termine the effect, if any, on the minimum density values. Due to the
nature of the rockfill material and the prevailing field conditions, this was
not considered feasible or justifiable. The material was consistently shoveled
into the container from the density hole or from the floor of the field
laboratory if it had required air drying as described previously. In the
latter instance, care was taken to avoid segregation of the particles with
respect to grain size.
Limited studies were made to determine the effects of increasing the
surcharge and length of vibration in the maximum vibrated density tests.
Duplicate tests were performed on the same specimen with increasing sur-
charges up to 25 psi. Another series of tests were run in which the length
of vibration was increased up to 60 min. Neither increasing the surcharge
nor increasing the length of vibration produced an increase in the vibrated
test density.
Conclusions
Although this test program at Carters Dam was not established for
construction control, some of the test results were very helpful for com-
parison with the design assumptions. The minimum density and maximum
vibrated test density results were of little value on this project. It is obvious
that in this test program the maximum vibrated density test apparatus
was not capable of producing densities comparable to those achieved by the
compaction equipment in the rock fill. These tests, however, do provide
background information for necessary improvements in equipment and
procedures for testing rock fill that contains material coarser than the 3-in.
particle size.
The most reliable and useful information derived from this program was
the gradation data. As often happens, the rock did not quarry with the
gradations anticipated and originally specified. Yet, the gradation tests
provided accurate information regularly for the material being placed in
each zone of the embankment.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lice
Correlation Between Relative Density and
Measured Performance or Properties of
Granular Soils

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement.
Yves L a c r o i x 1 a n d H . M . H o r n s

Direct Determination and Indirect Evaluation


of Relative Density and Its Use on
Earthwork Construction Projects

R E F E R E N C E : Lacroix, Yves and Horn, H. M., " D i r e c t D e t e r m i n a t i o n


and Indirect Evaluation of Relative Density and Its Use on Earthwork
C o n s t r u c t i o n P r o j e c t s , " Evaluation of Relative Density and Its Role in Geo-
technical Projects Involving Cohesionless Soils, A S T M S T P 523, American
Society for Testing and Materials, 1973, pp. 251-280.
ABSTRACT: Relative density is often used as the criterion for controlling the
quality of compacted granular fill. The direct and indirect methods ~vailable
to determine in situ relative density are considered along with the difficulties
associated with these methods. Data are presented which indicate the degree
of error that can be expected when making in situ determinations of relative
density. Correlations are presented which relate in situ data obtained by the
most commonly used field techniques employed in the determination of relative
density.
K E Y W O R D S : density (mass/volume), earthwork, soft compacting, earth
fills, construction, cohesionless soils

Compacted fill has widespread application in construction, such as con-


struction of dikes, dams, building subfoundations, and highway embank-
ments. The properties of compacted fill that are of most concern to the
designer are the engineering properties; that is, compressibility, shear
strength, and permeability. Direct in situ determination of these properties
is time-consuming and costly, and, consequently, their direct measurement
is not an approach generally used to control the quality of compacted fill.
In most cases, the engineering properties are related to the compaction
characteristics of the fill material by means of laboratory or field tests,
and some measurement of compactness is used as the criterion for quality
control of the compacted fill. When predominantly granular soils are em-

1 Director, Woodward-Clyde Consultants, New York, N. Y. 10001.


Associate, Woodward-Moorhouse & Associates, Inc., Clifton, N. J. 07012.

251

Copyright by ASTM Int'l


Copyright9 by(all
ASTMrightsInternational
reserved); Fri Mar 11 16:13:06 EST 2016
www.astm.org
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
252 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

TABLE 1--Definitions of relative density, density ratio, and degree of compaction.


Relative Density (Dr)
Dr emax -- en 'Ydmax ~ "Ydn - - "Ydmin
emax -- emin "Ydn "Ydmax - - "~dmln

Density Ratio (Rr)


Rr = ")'tin ~ *Ydmin

'Ydmax - - 'Ydmin

Degree of Compaction (Dc)


De = "Ydn, =~Rr
~a~x Dr
where: en, emax, emin are the in situ, maximum, and minimum void ratios,
respectively; and,
"[dn, " f d m a l , ~fdmin are the in situ, maximum, and minimumdry unit weights,
respectively.
NOTE~Dr and Dc are frequently expressed as percentages.

ployed as fill material, relative density is frequently specified as the basis


for compaction control. Other measurements of compactness that are
commonly used are "degree of compaction" for cohesive soils and "density
ratio" for cohesionless soils; definitions of the above are given in Table 1.
It should be recognized that the three methods of defining compaction
previously cited are quite different, although the terms are too often used
interchangeably.
The concept of relative density is being used successfully by engineering
firms to define and control the quality of compacted granular fills (for ex-
ample, Leafy and Woodward [I]3). Our experiences have taught us that
relative density, while simple in concept, is a tool, the application of which
requires considerable care and judgment to obtain the desired quality
without imposing overly stringent and sometimes unfair or arbitrary re-
strictions on the earthwork contractor. Difficulties with the use of relative
density as a control criterion are associated with both the determinations
of the maximum and minimum dry unit weights (the reference densities)
of the fill material, and with the determination of the in situ density of the
compacted fill. In some applications, the use of correlations between rela-
tive density and some indirect measurement, such as static or dynamic
penetration resistance, have been used successfully; but, again, such ap-
proaches involve difficulties that must be recognized and coped with.
The purposes of this paper are to indicate the direct and indirect methods
that are being used to determine in situ relative density; present correla-
tions between results obtained by several of these methods; and present
data which indicate the reliability of these methods. Only by recognizing
the degree of error involved with a measurement procedure, can a reasoned

3The italic numbers in brackets refer to the list of referencesappended to this paper.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreemen
LACROIX AND HORN ON EARTHWORK CONSTRUCTION PROJECTS 2 5 3

judgment be made concerning its applicability as a quality control device.


Descriptions of how relative density has been used to control compaction
of granular fill on major earthwork projects are presented elsewhere [1].
Factors Affecting Relative Density
The earliest research of the factors affecting maximum and minimum
dry unit weights of granular soils was carried out by Burmister [~]. Since
these unit weights, along with the in situ dry unit weight, define the rela-
tive density of a compacted soil, those factors affecting their determination
must be recognized before the difficulties involved in the direct determina-
tion of in situ relative density can be appreciated. Burmister found that
both maximum and minimum dry unit weights were functions of gradation
as well as particle shape. These factors will now be considered.

Gradation
15G Type
C
i
f
CD~L
D, SD
S

"~ 130 /~ .... I ES,ED


_- ~ .I [ E

-'= t20 I//-I

b
Curves ore for subongularsand and
E /#' waterworn grovel. Specific gravity
taken to be 2.67

Reference: Burrnister, 196 2


I00 /
0 I 2 3 4 5 6 7 B 9 I0 I'I 12
Effective Groin Size Range, Cr expressed by Mean Slope in Soil
Fractions above 010

'2~ ~Ullll ] I lllll]l i

,oo ~ } ! ! !! I I T !
:~ ~o_ lull ] - - ~
fllli ]lH-i->.~o'.~ll:? ~
~: 80 ~ } t l I~r~oT---'t't~bt-
_ fllll lill I
70 I1 IIII I]lllll I IN ll
20 6 0.6 0.2 0.06 0.02 0006
Effective Groin Size, Dsoin Millimeters

Reference; Burmister, 1962

FIG. 1--Correlations between maximum and minimum dry unit weights and grain-size
characteristics of granular soils.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
254 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

Influence of Gradation on Reference Densities


Maximum Dry Unit Weight--Burmister [~] proposed that maximum
dry weight (~d=,.) is primarily a function of the range of particle sizes and
the shape of the grain-size distribution curve. He presented curves which
correlated ~d.~= with combinations of range of particle size and the shape
of the grain-size distribution curve. These correlations, which are given in
Fig. 1, indicate that ~ . can vary from about 100 lb/ft 3 for relatively
fine, uniformly graded granular soil, to 146 lb/ft 3 for well-graded granular

135
[ I I I I I I

130

125

120

x~

9~ IJ5

~ 110
i" NOTE:
E
9 ,,p~, Data ploffed are from the Ludingloo
9$,,.. Project and Cooper Stofion and
9,l % ~ from the following sources:
105 D' Appolonio el ol~ 1969
Koerner ~ 1970

I00

95

i
9O I I I i ~ I I I
t 2 3 4 5 6 7 B
Coefficient ofUniformity. Cu

FIG. 2--Correlation between maximum dry unit weight and coe~clent of uniformity of
granular soils.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
LACROIX AND HORN O N EARTHWORK CONSTRUCTION PROJECTS 255

120 I I I I J i l l /

il5
c
E

E~IIO I. Soil is omedium to fine sand,


with non- plasticfines
o Z. "Fines"refers to particlespassing
No.200 sieve

Eo
E

I00 ~'d m i n ~ , ~

95 I I I I I I I I I I
0 5 IO
Fines Content, %

FIG. 3--Correlations between maximum and minimum dry unit weights and fines content
for a glacial sand at the Ludington Project.

soils having a wide range of particle sizes. Burmister's correlations apply


to soils composed of subangular sands or water-worn gravels.
Poulos and Hed [3] carried out an extensive investigation of the density
characteristics of hydraulically placed clean sand. They found that the
maximum dry unit weight correlated quite well with the coefficient of
uniformity (C~ = Ddo/D~o). This finding is in agreement with data which
have been collected by the authors on two projects; namely, Cooper Nu-
clear Station near Brownville, Neb. (alluvial medium to fine sand), and
Ludington Pumped Storage Project, near Ludington, Mich. (glacial out-
wash sand containing less than 7 percent fines, by weight). The data re-
ferred to are plotted in Fig. 2 along with the curves of Poulos and Hed
correlating ~ x with Ca. This figure also includes data obtained in studies
by D'Appolonia et al [~] and Koerner [5]. It should be noted that all of the
results referred to, with the exception of Koerner's, involved sands that
were deposited either by flowing water or by wind, and it is likely that the
grain-size distribution curves had similar, more-or-less, S-shapes. Conse-
quently, before using correlations such as shown in Fig. 2, the grain-size
distribution characteristics should be considered; Burmister's findings in-
dicate that the shape of the grain-size distribution curve has a large influ-
ence on ~a~x when the soil has a wide range of particle sizes.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. N
256 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

T A B L E 2---Influence of gradation on reference densities of clean sand at Cooper Station.

~'~,.**, lb/ft 3 ~'~mio, lb/ft a

(1) Total specimen 123.5 109.5


(2) Fraction retained on No. 20 sieve 115.4 102.5
(3) Fraction passing No. 20 sieve 116.0 100.5
(4) Average of values for both fractions 115.7 101.5
(5) Difference between (1) and (4) 7.8 8.0

Minimum Dry Unit Weight--Burmister [2] also investigated the relation


between minimum dry unit weight (~'d~in) and grain-size characteristics.
He found that ~dmincould be correlated with combinations of D~ and the
range of particle sizes (see Fig. 1). Poulos and Hed [3] found that the
difference ~ x - ~'~mlnwas essentially constant throughout the two de-
posits of hydraulically placed fill that they investigated. They pointed out
that although a relationship between maximum and minimum dry unit
weights may exist for a given soil deposit, the same relationship cannot be
expected to be obtained with all deposits.
It should be noted that a unique relationship between maximum and
minimum dry unit weights was not found in the case of the glacial sands at
the Ludington Project referred to earlier. For these sands, the difference
increased with increasing percentage of fines (see Fig. 3). Burmister's
studies suggest that the shape of the grain-size distribution curve, which is
governed by the depositional process, determines the relationship between
~d~x and Ydmi..
Relative Density of Natural Deposits--Many cohesionless natural de-
posits are reasonably uniform over distances of several tens of feet, but
may be composed of many different layers of sand or gravel. Each such
layer has its own in situ dry unit weight and its own values of ~'d~ and
~/gm~.. Mixing layers together during sampling or preparation of test speci-
mens can produce a soil having density characteristics which bear no rela-
tionship to those of any of the individual layers. This is illustrated in
Table 2, which is based on test data obtained at Cooper Station.
The previous example demonstrates the importance of sampling only
individual layers and obtaining enough material to determine not only the
in situ dry unit weight, but also enough for determinations of ~/d~. and
~'d~.. It should also be recognized that in some natural deposits, individual
layers may be only a few grains in thickness and that, in such cases, deter-
mination of relative density may be impossible by conventional methods.

Influence of Particle"Shape on Reference Densities


Burmister [2] recognized that particle shape influences the maximum
and minimum dry unit weights. His work indicated that granular soils

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
LACROIX AND HORN ON EARTHWORK CONSTRUCTION PROJECTS 257

composed of highly angular particles had values of ~dm~ that were 10


lb/ft a to 15 lb/ft ~ lower than those of soils made up of subgranular grains;
no quantitative estimate was made of the influence of angularity on ~/dmin-
tIolubec and D'Appolonia [6] have shown that increasing particle angu-
larity decreases ~m~ a relatively slight amount, but can decrease ~dmin
substantially. Equally as important, they point out that soils having the
same general range of particle sizes and the same relative density, but with
different particle shapes, can have drastically different compressibility and
strength characteristics. The influence of particle shape on the reference
densities should be recognized before using correlations such as those in
Fig. 1 and 2. For similar reasons, published correlations between angle of
shearing resistance, or compressibility, and relative density should be used
with caution. Results of small-scale tests by Holubec and D'Appolonia [6]
suggest that the results of dynamic penetration tests are also affected by
particle shape.
While the important influences of particle shape on the reference densi-
ties and engineering properties are recognized, direct methods of measuring
and defining particle shape are cumbersome. Holubec and D'Appolonia [6]
suggest the use of indirect methods, such as that based on permeability
measurements, as a means of arriving at a measure of particle shape.

Direct Determinations of In Situ Relative Density


In fact, there is no direct method of determining in situ relative density.
Direct determination, as used in this paper, refers to direct measurement
of in situ dry unit weight and computation of relative density from it and
the reference densities. Methods that are generally used to measure in situ
dry unit weight are the water balloon method, ASTM Test for Density of
Soil in Place by the Rubber-Balloon Method (D 2167-66), and the sand-
cone method, ASTM Test for Density of Soil in Place by the Sand-Cone
Method (D 1556-64) ; the Washington Densometer 4 method, is a commonly
used type of water balloon test. Our firm has used both the Washington
Densometer and the sand-cone methods on several major projects involving
compacted granular fill. Both methods give about the same result when
used properly (see Fig. 4). The Washington Densometer method has, how-
ever, significant advantages over the sand-cone method.
In thinly layered deposits, the Washington Densometer method is more
satisfactory than the sand-cone method. The reason for this is that, for the
same volume of hole, the Washington Densometer provides one that is
shallower than that required by the sand-cone method. Therefore, it is
easier to keep the hole in a single layer of the deposit. In an alluvial deposit
of the Missouri River, consisting of thinly layered medium to fine sand, the
relative density of the in situ soil, as determined by the sand-cone method,

4 Manufactured by D. G. Parrot & Son, Olympia, Washington, Models 15 and 30.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further r
258 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

155

150

125

120

-~ 11,5

o
9 Q~ Q 9
I10

105

.
I00 I I [ I P J I
I00 105 I10 115 120 125 150 135
Dry L~it Weight by Washington Densometer, Ib/ft 3

FIG. 4--Comparison of results of in situ dry unit weight determinations in compacted


granular fill by Washington densometer and sand cone methods.

was found to be lower than that determined by the Washington Densom-


eter method. The in situ dry unit weights were about the same by both
methods. However, both the maximum and minimum dry unit weights
determined on the soil excavated from the sand-cone hole were about 0.6
lb/ft a higher than those determined on the soil excavated from the Washing-
ton Densometer hole. Careful observation showed that the samples from
the sand-cone holes were relatively well graded because they consisted of
a mixture of several sand layers, whereas the samples from the Washington
Densometer holes were more uniformly graded.
On earthwork construction projects involving cohesionless soils that
were compacted with vibratory rollers, the sand-cone method has several
disadvantages. Vibrations caused by construction equipment are sufficient
to increase the unit weight of the calibrated sand when it is poured in the
sand-cone holes; this leads to calculated dry unit weights lower than the

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to L
LACROIX AND HORN ON EARTHWORK CONSTRUCTION PROJECTS 2 5 9

in situ values. Rainwater, or even condensation water, are sufficient to wet


the calibrated sand and give it a different unit weight than when dry. In
addition, a significant portion of the calibrated sand must be wasted at
each sand-cone test location; the cost of replacement of the sand is not
negligible.
Tests made with the Washington Densometer and by the sand-cone
method require about the same amount of time. Considerable care must be
exercised by the operator when using either of the methods. Our firm's
experience has shown that reading errors are somewhat more frequent with
the Washington Densometer method than with the sand-cone method.
Relaxation of horizontal stresses that occurs when a hole is dug induces
a small reduction in the volume of the hole. The sand-cone method cannot
correct for this reduction, whereas a reasonable correction can be made
when using the Washington Densometer method [1]. The correction is made
by increasing the water pressure in the balloon until it is equal to the value
of horizontal stress estimated to have existed prior to making the hole.
Indirect Evaluation of Relative Density by Means of the Standard
Penetration Test
The standard penetration test (SPT) consists of a boring and sampling
technique which allows measurement of the penetration resistance of a
standardized sampling spoon driven by a specified impact energy. The
standard penetration resistance (N) is equal to the number of blows re-
quired to drive a 2-in. outside diameter, 1-3/~-in. inside diameter, split
spoon 1 ft into the soil after an initial penetration of 6 in. The driving
energy is delivered by a 140-1b hammer falling freely a height of 30 in.
The test procedure is specified by ASTM Penetration Test and Split-
Barrel Sampling of Soils (D 1586-67). The use and abuse of the SPT and
a thorough analysis of its various aspects have recently been reported
[7, s].
Factors Influencing Standard Penetration Resistance (N)
Even when ASTM D 1586-67 is scrupulously followed, there are still a
number of factors not specified which have considerable effect on N. The
primary such factor is the "boring technique." Various methods of ad-
vancing the boring are commonly used; namely, wash boring (using a
chopping bit in combination with wash water pumped down the drill rod
and through the bit; 2.5-in. diameter casing being used as necessary);
hollow-stem auger; rotary drill using water or mud as the drilling fluid; etc.
Data obtained at Cooper Station, and presented in Fig. 5, demonstrates
the substantial effect that boring technique can have on the results of the
SPT. For projects where repeatable values of N are important, our firm
specifies that the borings be made with a rotary drill using drilling mud,
and that N-size rods be used for the drill stem. Although the type of drill

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agr
260 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

Standard PenetrationResistance N, bllft


0 50 IO0 150 200 250 300
0 [ I I I I I

% Cooper Station
Compacted medium to fine sand
1.5-3% fines, Cu=l.5-3
IO LEGEND:
s N Standard penetration resistance calculated from
relative density determined on undisturbed Denison
samples using Gibbs and Holtz's method
A Nn Solid flight augerboring
20 Nw Rotary drill bit and water
O Nb Rotary drill bit and bentonite-water slurry

c)
30

40

50 I l I I I ]

FIG. 5---Effect of boring technique on standard penetration resistance.

rod is not very significant for shallow holes (that is, less than 50 ft), it is
important for deeper holes.
In general, it can be said that modern techniques of advancing borings
lead to higher standard penetration resistances than those obtained when
using the original (wash boring) techniques [9]. This, of course, assumes
that improper methods are not used, such as using a drill bit with straight
downward discharge holes, or inadequate cleaning of the bottom of the
hole prior to the SPT.
Another major factor affecting results of the SPT is the "effective con-
fining stress," which is influenced by the unit weight of the soft, depth
below ground surface, stress history, and the groundwater level. It has been
suggested that N be corrected when obtained below the water table to
take into account the reduction in effective stress due to pore-water pres-
sure. The quantitative influence of vertical effective stress on SPT resist-
ance has been studied in the laboratory [10] and was found to be substantial.
The effect of a decrease in the vertical effective stress on SPT resistance
is illustrated by data obtained at Cooper Station (see Fig. 6). At that site,
standard penetration tests were made in 15 borings drilled from the original
ground surface. An excavation was then made, and when it had reached
a depth of 26 ft, standard penetration tests were made in nine borings

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreemen
LACROIX AND HORN ON EARTHWORK CONSTRUCTION PROJECTS 261

drilled from the bottom of the excavation. The average values of N are
plotted in Fig. 6 as functions of depth for conditions before and after ex-
cavation. Values of N determined after excavation were found to be about
10 blows/ft lower than those made at the same elevation prior to excava-
tion. Similar findings have been reported by Mansur and Kaufman [11],
wherein the removal of 50 ft of overburden reduced N by a half.
Rather than considering the influence of vertical effective stress at the
location of the SPT, the authors suggest that the horizontal effective stress
be considered as the primary stress factor influencing N. This allows taking
into account high horizontal stresses, such as those existing in overcon-
solidated deposits or heavily compacted fills.
Another major factor affecting the results of the SPT is the "type of
soil;" that is, the gradation and the shape of grains. For example, the pres-
ence of gravel will result in values of N that are higher than those that
would be measured in a soil having the same relative density but which
does not contain gravel. A high degree of angularity has the same effect
on N.
Another factor affecting N is the location of the "groundwater level." In
the case of compact silty fine sand, N measured below the water table is

Original ground surface

I
Cooper Station
Alluvial mediumto fine SAND
7" = 133 Ib/ft 3
WT 7'=7'0 tb/ft 3
tO %.-
NOTE:
Re lative densities~Dr~ indicated were
determined by Gibbs and Holtz's Method

2O
GrOundsurface
.T- r excavation

3O
.__

~=73 ~73
"~/---Avei of '5
~ k ~ borings before
40 , ~ \ exccvation
Ave of 9 borings / \
after excavationJ ~e73 74~1

I t I
500 I0 20 30 40
Standard Penetration Resistance N, bl/ft

FIG. 6---Effect of decrease of vertical effective stress on standard penetration resistance.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
262 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

greater than that which would be determined above the water table in an
otherwise similar soil, because of negative pore water pressures resulting
from dilatancy; a correction equation has been proposed [12] to account
for such a condition. Conversely, the measured N of a loose medium to
fine sand located below the water table will be less than if the same soil
had been located at the same depth, but above the ground-water level. The
reason for this is that such a sand will tend to liquefy under the influence
of the vibrations and stresses induced by driving and, thereby, lose much
of its shear strength.
Equations relating N to soil properties may be misleading if the factors
cited previously are not taken into account, as well as the effects of the
presence of substantial numbers of large particles (for example, gravel or
cobbles), a collapsible soil structure (for example, loess), and cementation
between soil grains. Charts have been proposed in which the influence of
gradation on the relationship between relative density and N is taken into
account [2].

Correlations Between Relative D e n s i t y and Standard Penetration Resistance


Indirect evaluation of the relative density of cohesionless soils can be
made reasonably well by means of the standard penetration test. Initially

%
o~

~0 I0 20 30 40 50 60 0 80
Standard Penefrotion Resistance N, bl/fl

FIG. 7--Correlation between relative density and standard penetration resistance in ac-
cordance with Gibbs and Holtz [10].

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reprodu
LACROIX AND HORN ON EARTHWORK CONSTRUCTION PROJECTS 2 6 3

uJ

~ v

0 I0 20 30 ~0 50 60 70 80
Stondord PenetrofionResistQnce N, b l / f t

FIG. 8--Comparisons of several correlations between relative density and standard pene-
tration resistance.

[13], relative density was correlated qualitatively with values of N ob-


tained in borings that had been generally made for conventional soil in-
vestigations and for the design of shallow foundations. Later, the influence
of the effective confining stress was recognized and a quantitative correla-
tion between relative density and N was established which took into ac-
count the vertical effective stress [10]. This correlation was based primarily
on the results of a laboratory investigation, but was checked on U.S.
Bureau of Reclamation (USBR) projects. ~ Figure 7 is a chart replotted
from the Gibbs and Holtz chart published by the USBR [1~] and includes
both interpolations and extrapolations.
Figure 6 illustrates that in a natural alluvial deposit consisting of medium
to fine sand (Cooper Station), the influence of verticM effective stress is
accounted for quite well by the method proposed by Gibbs and Holtz. At
an elevation corresponding to a depth of 35 ft before excavation, the aver-
age standard penetration resistance was N = 26 blows/ft prior to excava-
tion, and N - 15 blows/ft after excavation. The relative densities calcu-
lated by the Gibbs and Holtz method for conditions before and after
excavation were the same, which is as they should be.
Figure 8 shows a comparison between relative density charts prepared by

Personal communicationwith"H. J. Gibbs, 1968.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further rep
264 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

Stondord Penetration Resistance N, bl/ft


00 50 I00 150 200 250
A

CooperStatic__.n_
p ~ Compacted medium to fine SAND
I0 ~ ~ 1"5-3 ~176fines'Cu = I 5 - 3

~C~ LEGEND
~.~'--0. [] N Standardpenetration resistance
~ . z~ ~ calculated from relative density
E~ ~ determined on undisturbed Denison
20 [~ ~ samples using Gibbs and Holtz,s
rnr~ ~ method
~""--[]~ 0 NB Standard penetration resistance
~.~" . ~ calculated from relative
30 _~ v ~ densitydetermined on undisturbed
E~ ~ Denison samples using Bazarao,s
" ~ _ "~ method
C~.~..~ ~ ~'ghtouger bor'ng
40

50 { I I {

FIG. 9--Comparison of the Gibbs and Holtz, and Bazaraa correlations between relative
density and standard penetration resistances.

several investigators [9, 14, 15, 16]. Casagrande's correlation was prepared
for a relative density of 85 percent for a specific project, and considered the
lowest values of N obtained in conjunction with several drilling methods.
All of the correlations are reasonably consistent at low relative densities,
but diverge significantly at high relative densities.
Figure 9 presents a comparison between the Gibbs and Holtz method of
analysis and the Bazaraa method of analysis, using the data obtained at
Cooper Station, and reported in Fig. 5. The relative densities of sand sam-
ples obtained by means of a Denison sampler were determined directly.
The corresponding values of N were calculated using both the Gibbs and
ttoltz, and the Bazaraa methods of analysis. These standard penetration
resistances are compared with values measured in a boring drilled in the
dry with a solid flight auger.
Our experience is that the Gibbs and Holtz method of analysis yields
relative densities that are too high for heavily compacted fill; this has also
been noted by others [17]. The probable reason is that increases in values
of N are not essentially due to increases in vertical effective stress, but are,
rather, the result of an increase in the horizontal effective stress. If this

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Ag
LACROIX AND HORN ON EARTHWORK CONSTRUCTION PROJECTS 265

concept is correct, and if it is assumed that the coefficient of horizontal


earth pressure (K) was 0.4 in the laboratory investigation by Gibbs and
Holtz, and in the natural deposits where their method works well (for
example, Fig. 6), we conclude that the vertical effective stress should be
multiplied by the factor K/0.4 before entering the chart shown in Fig. 7.
The authors recommend the use of this modification when evaluating
relative densities by the Gibbs and Holtz method. For example, at a depth
of 15 ft in a dry natural sand deposit with a unit weight of 120 lb/ft 8, the
vertical effective stress is ~ = 1.8 kips/ft 2. If N is 35 blows/ft, the relative
density would be estimated to be 95 percent on the basis of Fig. 7. Con-
sider a second case, similar to that referred to above, except that it involves
a heavily compacted sand fill. Measurements [4] have shown that sand fills
compacted to high relative densities have high coefficients of horizontal
earth pressure; as great as 2 to 3, but with values typically being about
1.5. If the second deposit had a coefficient of horizontal earth pressure of
1.2, we would multiply the vertical effective stress by the factor 1.2/0.4 = 3
before entering Fig. 7. Using this approach, the resulting estimated relative
density i$ 71 percent. Conversely, if the estimated coefficient of horizontal
earth pressure had been 0.3, the vertical effective stress would have been
multiplied by 0.3/0.4 = 0.75 before entering the chart.
The unmodified Gibbs and Holtz relationship between N and vertical
effective stress for a relative density of 100 percent has been plotted in
Fig. 10, along with measured values of N obtained in a test program in-
volving the heavily compacted sand fill at Cooper Station. Also included is
the relationship proposed by Bazaraa. As part of that program, determina-
tions of in situ relative density were made on the basis of Washington
Densometer results as well as by means of tests carried out on undisturbed
samples obtained with a Denison sampler. An evaluation of the results of
these tests indicated that the fill had an average relative density of Mmost
100 percent; namely, 97 percent. Examination of Fig. 10 reveals that the
unmodified Gibbs and Holtz relationship greatly underestimates the values
of N corresponding to 100 percent relative density, or conversely, over-
estimates the relative densities associated with the set of N values. On the
other hand, Bazaraa's relationship is in good agreement with the N-value
data.
Average values of N have been computed for the Cooper Station data,
and are plotted in Fig. 11 as a function of verticM effective stress. Also
shown are the Bazaraa relationship and the unmodified Gibbs and Holtz
relationship for relative densities of 100 percent. The excellent agreement
between the average values of N and the Bazaraa relationship will be noted.
Figure 11 also contains the relationship between coefficient of horizontal
earth pressure and vertical effective stress required to obtain agreement
between the average values of N and the Gibbs and Holtz relationship

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License
266 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

modified as described previously. The coefficient of horizontal earth pres-


sure would have to decrease from a value of between 2.0 and 2.5 near the
surface, to a value of about 0.5 at a depth of several tens of feet. These
values, and the shape of the K versus/5, relationship in Fig. 11, are reason-
able. The higher values are in agreement with the measurements made by
D'Appolonia et al [4] at shallow depths in heavily compacted sand fill. The
decrease of K with depth, becoming asymptotic with a value of K corre-
sponding to an "at rest" condition, is consistent with what would be ex-
pected for a heavily compacted fill.

Standard Penetration Resistance N, blows/ft


50 I00 150 200 250 300
I I I I I

D w

-.-.
N
000

|
59 00 9
~>

o~ N

oo 09

IP Dr = 100%

8, H U n m o d i f i e d
Dr : 1 0 0 %

8 I I l ,I

FIG. lO--Standard penetration resistances of a heavily compacted sand fill at Cooper


~tation.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agree
LACROIX AND HORN ON EARTHWORK CONSTRUCTION PROJECTS 267

Calculated Coefficienfof Horizontal Earth Pressure(K)


0 05 1.0 1.5 2.0 2.5 3.0
I I I I i I l

N, bl/ft
0 50 100 150 200 250 300
0

O ,J+

/
+ /
2
/+
N 3
/
-4- Q

03
4 LEGEND
0 Average N
~" 4" C a l c u l a t e d v a l u e of
~ ( ~ K u s i n g M o d i f i e d Gibb.,
Q 81Holtzr e l a t i o n s h i p
5

/-- .Bazaraa
/ ~ Or=lOO%

7
~p-~G ~ H Unmodified

~ Dr : 100%

FIG. l l--Variation of calculated coe~cient of horizontal earth pressure required to obtain


agreement between modified Gibbs and Holtz analysis and data obtained at Cooper Station.

The relationship between K and #~ in Fig. 11 is not meant to apply to


all heavily compacted granular fills, but rather~ is presented to demonstrate
that the high horizontal stresses built into such fills provide the likely
explanation of why the Gibbs and Holtz relationship greatly overestimates
the relative densities of these fills.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
268 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

Correlation Between Nonstandard Penetration Resistance N1 and Standard


Penetration Resistance N
It is sometimes necessary to estimate standard penetration resistance
from a penetration resistance N1, obtained with a nonstandard split spoon
or a solid conical point; in addition, the driving energy and depth of pene-
tration may be nonstandard. In such cases, and when an approximate
correlation is acceptable, we believe that it is satisfactory to assume that
the number of blows (n) required to drive the split spoon or conical point
to a penetration depth (L) is directly proportional to the square of the out-
side diameter (D) of the split spoon or conical point and the depth of
penetration, and inversely proportional to the energy per blow (WH);
that is,
D~L
n ~-- - - (1)
WH
Therefore, N may be estimated from N1 in the following manner:
//2 in.~ ~ 12 in. W1 H~ 2 N~W~H~
N - - Nlk, D~ ] X ~ X 1401b X 30 in. - 175Dl~L~ (2)
where the following apply to the nonstandard test:
D~ = the outside diameter of the split spoon or conical point in
inches,
L~ = the depth of penetration in inches,
W1 = the weight of the hammer in pounds, and
H1 = the height of free fall of the hammer in inches.
It should be recognized that this is considered an approximate correla-
tion and that, when possible, correlations should be developed for the
particular nonstandard equipment or methods being used, and for the
particular soil deposit being investigated.
Example of Application of Correlation to Results of Tests Involving a Cone
Driven With a Lightweight Hammer
There is frequently need to evaluate the relative density of cohesionless
soils in situations where the use of standard equipment is either too ex-
pensive or cumbersome. Such an example occurs when it is desired to evalu-
ate the relative density of a deposit of natural sand located below the bot-
tom of a footing excavation before the footing concrete is placed; the depth
of interest is usually 5 to 15 ft. In such cases, the working area is often
constricted and the use of lightweight equipment is advantageous.
The authors have found that it is practical and economical to obtain
the dynamic cone penetration resistance of naturally deposited or com-
pacted cohesionless soils with lightweight equipment. The equipment we
have used is manufactured by the Acker Drilling Company (similar equip-
ment is manufactured by others). A 60-deg, 1-5/~6-in. outside diameter

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Li
LACROIX AND HORN ON EARTHWORK CONSTRUCTION PROJECTS 269

steel cone is screwed on to the lower end of an E-rod. An attachment is


placed on the upper end of the E-rod which allows driving of the rod with a
34-1b hammer falling two feet. The number of blows required to drive the
cone 4 in. is recorded. The test can be done manually. However, a small
power winch and a light tripod are advisable when many of these tests are
required.
Our firm has correlated the dynamic penetration resistance Nd (blows/4
in.) obtained with this lightweight equipment with the standard penetra-
tion resistance N of cohesionless soils at several construction sites. The
cohesionless soils ranged from silty sand to coarse sand, and included both
natural deposits and compacted fills.
The correlation we use is:
N = Nd, for Nd < 20 (a)
N = 20+0.5(Nd-- 20) f O r N d > 20 (b) (3)
Equation 2, given in the preceding subsection, yields a correlation N =
1.33 Na.

.o
z 40 N=135Nd(E q 2) I
z~ ~ f
A
t, z~ z1 J

.:_o
C
N=20*O.5(Nd-20)

NOTE :
Tesls were conducted
/ & in o very dense, heavily
compacted fill
~ N=Nd(Eq 3a}
0 I I I I I
0 10 20 30 40 50 60
Dynamic Cone Penetration Resistance Nd, bl/4in.

FIG. 12--Correlation between standard penetration resistance a n d a light weight dynamic


cone penetration resistance.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
270 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

The correlation represented by Eq 3 is considered to be more satisfactory


for loose to medium relative densities than for very high relative densities.
The lightweight equipment described should not be used for very dense
compacted fill or for depths exceeding about 15 ft because of excessive
friction on the E-rod. An improvement might involve using an expendable
cone with a diameter slightly greater than the E-rod [18]; our firm has had
success in using such an approach. If this is done, it is recommended that
a conical point be used which has a 60-deg angle and a base area of 10 cm 2,
which is the shape and area of the cone used in the standard static cone
penetration test.
Figure 12 shows a rather poor correlation between Eq 3 and penetration
test results obtained during the construction of a very dense heavily com-
pacted fill; the correlation represented by Eq 2 is also given. Better correla-
tions have been obtained with Eq 3 when dealing with natural deposits,
and when the relative density was less than about 70 percent.
Indirect Evaluation of Relative Density by Means of Static Cone
Penetration Tests

Standard Static Cone Penetration Test (SCPT)


Static cone penetration tests are used widely in Europe to evaluate in an
indirect manner the in situ properties of deposits of soils. Much of the
development of the equipment and procedures used in such tests has been
carried out in Holland, and when mention is made herein to the standard
static cone penetration test, we are referring to the test made with the
"Dutch friction jacket cone" shown in Fig. 13.
The standard static cone penetrometer is a sounding apparatus. A cylin-
drical rod with a conical point ~tt the lower end is pushed into the soil. The
cone has a 60-deg point and a base with an area of 10 cm 2. A sleeve a few
inches above the conical point can be advanced simultaneously with the
point, or left stationary. The resistance to penetration is read on a gage as
the cone is pushed into the soil at a rate of penetration of about 2 cm/s.
Another method that gives somewhat lower but more repeatable values
involves reading the gage as the rate of penetration is decreased from
2 cm/s to zero.
The standard cone penetration resistance R, is equal to the vertical
force required to push the cone divided by the cone area (namely, 10 cm2).
R, is usually expressed in kg/cm2, e
Correlation Between Standard Penetration Resistance and Standard Static
Cone Penetration Resistance
Standard penetration resistance N and standard static cone penetration
resistance R, have been correlated to one another by investigators in many

e In Western Europe this unit is now referred to as "bar".

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
LACROIX AND HORN ON EARTHWORK CONSTRUCTION PROJECTS 271

Push outer rod iPush inner rod Conhnue inner rod push

_1

Cone only odvonces


for beoring copocify

Cone and jacket both odvonce


for beoring capacity e friction

FIG. 13---Dutch friction jacket cone.

p a r t s of the world for m a n y years. Unfortunately, detailed information


concerning the procedures and equipment used in these tests are almost
always missing.
Our firm has studied correlations reported in the literature (for example,

TABLE 3--Correlation between standard penetration resistance N and standard static cone
penetration resistance Rs as a function of soil description, R. (kg/cm ~) = CN (blows /ft).

Soil Description Correlation Factor C

Clay 1- 2
Silt 2- 4
Sand 4- 6
Gravelly Sand 6- 8
Sandy Gravel 8-10

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agre
272 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

Penetration Resistance
0 I0 20 5 0 N bllft
0 I I I
s'o ,;o 200 RS, kg/cm z

i
Ludington Project
Glacial deposits consisting
of sand and silty sand

IO 0 Average N f o ; 152b::~:dgings

.=-

c~
15

20
Rs=6"25N

25 I I I
T I I

FIG. 14---Example of a site correlation between standard penetration resistance N and


standard static cone penetration resistance R~.

Sanglerat, [19]) and those developed by us for various soil deposits. Based
on these evaluations, we suggest the following correlation (Table 3).
The correlation factor C appears to increase with depth. We suggest
that this correlation be limited to a depth range of 10 to 50 ft. Furthermore,
the correlation should be used only as a guide. If warranted, a better cor-
relation can be generally developed for a specific project by means of a
testing program. For example, Fig. 14 shows a very good correlation ob-
tained in glacial deposits consisting of sand and silty sand at the site of
the Ludington Project; the correlation is R8 = 6.25 N.
The correlation factors given in Table 3 are somewhat greater than those
proposed by other American investigators. For example, the correlation
factors suggested by Schmertmann [20] have the following trend (Table 4).

TABLE 4--Correlation factors suggested by Schmertmann [20].

Soil Description Correlation Factor C

Silt, and sand-silt mixtures 2.0


Fine to mediumsand 3.5
Coarse sand and gravelly sand 5.0
Sandy gravel and gravel 6.0

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
LACROIX A N D HORN ON EARTHWORK CONSTRUCTION PROJECTS 273

Indirect Determination of Dry Unit Weight by Means of Nuclear


Devices
Nuclear devices can be used to make indirect determinations of in situ
total unit weight or water content. Gamma ray sources are used for deter-
mination of total unit weight and neutron sources for determination of
water content. Either direct transmission or backscatter techniques are
employed, although direct transmission techniques generally give better
results. The composition of the soil affects the determination of both total
unit weight and water content. The total unit weight also affects the de-
termination of water content. The air-gap technique theoretically elimi-
nates the effect of soil composition. I t consists of making a measurement at
the soil surface and another at a fixed distance, usually 1 or 2 in., above
the surface of the soil.
Nuclear surface devices are designed to be used on the soil surface or at

125 I I I I0 I
0
0

1 2 0 r[ 9 O O O

O Q go O O 9 0
D go 6) Q 9 O 9
I~ 9 00 ID 9 Oi D OO
OQ

~ ~ 9 9 ~ ooQ~ c8 o o 9
~" ~ ~176 o~ o~ o o ~ 9
_--= E 0(~ 0 0 0 0

- 0 0 O O
= Ee' 0 0 0 0 0
O
~'-~ I01 - Oo Oo O
"~ ~ oO Compacted MediumOto Fine SAND
9 I%fines, Dio=O. 17mm, Cu=l.6
O 0 3%fines, DIO= O. 33mm,Cu=2.8

IOO 0 O O
0
0
O NOTE: All water contents obtained by
means of ASTM Designation
95 I I 1 D 2216-66
95 I00 105 I lO 115 120 125
Dry Unit Weight, I b / f t 3
Nuclear Moisture-Density Meter

F I G . 15--Laclr of correlation between dry unit weights obtained with Washing~m densom-
eter and those obtained with a portable surface nuclear device.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement.
274 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

the bottom of a test pit. Nuclear probe devices are designed to be placed
or pushed into a hole, a few inches to about 2 ft below the soil surface.
Nuclear devices, portable by hand, have a small radioactive power source
of two to five millicuries (mc); larger nuclear devices are available that can
be mounted on a small panel truck and which have a radioactive power
source of 300 to 500 mc.
Our firm's experience with several types of portable nuclear surface
devices which operate on the backscatter principle has been so poor that
our use of them has been discontinued. Even calibration of such devices
for each type of soil at least twice every day did not significantly improve
the results. The effect of soil composition cannot be eliminated when using
nuclear devices having small power sources. In addition, the effect of sur-
face irregularities is very significant when using portable surface devices
because the depth interval of measurement is very small, about 2 in. for
total unit weight, and about 5 in. for water content determinations.
On several major earthwork projects requiring compaction of sand or
silty sand, the standard deviation was found to be • 10 lb/ft 3 for an aver-
age unit weight of 130 lb/ft 3 when portable nuclear surface devices were
used. Figure 15 shows the lack of correlation between dry unit weight
calculated from total unit weights determined by means of a portable
nuclear surface device, and those determined with the Washington Den-
someter. In both cases, water contents were determined in accordance

~,- rso

C~

"5 125
Z
5
N

120

N 115
7
/2"" I I

o 115 r20 125 150 f35


Total Unit Weight from Washington Densometer, Ib/ft 3

FIG. 16--Correlation between total unit weights obtained with the Troxler 2401 portable
probe nuclear device and those obtained with the Washington densometer.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement.
LACROIX AND HORN ON EARTHWORK CONSTRUCTION PROJECTS 275

GommomyRoyue,eclorsDetectors / f Oelectors~

v~u /J ~\~ h I

Probe ~ Nuctear Sources


lOrnc for garnrno roys~
I 300 mc for neutrons ]

FIG. 17--Sketch of the truck-mounted French A GS probe nuclear device.

with ASTM Laboratory Determination of Moisture Content of Soil


(D 2216-66).
Water content determinations with portable nuclear surface devices are
even more erratic than total unit weight determinations. For cohesionless
soil, the water content can be efficiently determined with sufficient ac-
curacy by means of either the Speedy Moisture Tester Model MC3207
(water contained in a small soil specimen is quickly absorbed by hydration
of calcium chloride) or by drying the soil specimen on a hot plate.
Extensive field testing conducted by our firm shows that portable nu-
clear probe devices, for example, the Troxler Model 2401 nuclc~ar probe, 8
are sufficiently reliable for the determination of total unit ~eight. The
standard deviation was found to be -4-2 lb/ft 3 (see Fig. 16).
For control of a cohesionless embankment compacted ,by modern meth-
ods, namely, heavy vibratory compactors, the in situ unit weight must be
determined at a depth of 1.5 to 2 ft below the surface of the fill. The exca-
vation of a test pit to that depth without disturbance of the bottom of
the pit takes 15 to 30 min. After completion of the test pit, a nuclear test
will take 15 min to make, whereas the Washington Densometer test will
take 30 min. Therefore, the overall saving of time using the nuclear device
is not very significant. In both cases, the water content must be determined
by direct methods.
Truck-mounted, more powerful nuclear surface or probe devices appear
promising. New York State Department of Public Works [21] reports
satisfactory use of the Road-Logger9 which is a nuclear surface device with

7 Manufactured by the Alpha-Lux Co., Inc., Philadelphia, Pa.


s Manufactured by Troxler Electronic Laboratories, Inc. Raleigh, N. C.
9 Developed by Lane-Wells, Houston, Texas.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License
-4

TABLE 5--Comparisons between water co.tent and dry unit weight determinations with nuclear probe devices and by direct methods at two sites.

Site and Equipment Soil Characteristics Water Dry Unit Standard Deviation No. of Tests
Content, Weight
~, % 7~, lb/ft s ~, % ~ , lb/ft s <
Z
O
Mt. Cenis Dam, France, 1968; AGS silty sandy gravel 3-9 126-144 0.4 1.6 230
Gage, depth of measurement --- 12 in. -r
S
Ludington Pumped Storage Proiect, medium to fine sand, 3-9 109-127 1.0 1.8 60
USA, 1970; Portable Troxler Gage, trace silt z
depth of measurement = 6 in. silty medium to fine 3-15 109-125 1.8 2.8 40
sand
O
NOTeS--
At Mr. Cenis, the direct method was similar to the sand-cone method; at Ludington, the direct method was the Washington Densometer
method.
The standard deviation is defined in terms of deviation of values determined by the nuclear method from the arithmetic mean of the values
determined by the direct method.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
LACROIX AND HORN ON EARTHWORK CONSTRUCTION PROJECTS 2 7 7

a 430 mc gamma-ray source. Repeatability of 4-0.8 lb/ft a is reported with


the Road-Logger. Electricite de France used a truck-mounted nuclear probe
device for compaction control of Mr. Cenis Dam. Figure 17 is a sketch
of the AGS device 1~used at Mr. Cenis Dam. Table 5 presents comparative
results obtained with the Troxler portable nuclear probe device and the
AGS device.

Indirect Evaluation of Relative Density by Means of Standard


Plate Load Test
The standard plate load test (SPLT) consists of loading, in increments,
a rigid steel plate, one foot square, located at the bottom of a test pit
which is at least five feet square. The settlement under each load is meas-
ured. The results represent the loading pressure-settlement characteristics
of the soil within a depth interval of approximately 1.5 ft below the plate.
The loading pressure-settlement characteristics have been correlated quali-
tatively with relative density [13] for use in settlement estimates required
in shallow foundation design.
Our firm has conducted an extensive field test program at several sites
in an attempt to correlate relative density with settlement of a one-foot-
square plate under a loading pressure of 3 tons/ft 2. Figure 18 shows a tenta-
tive correlation between relative density and settlement during a standard

12~ " - - /i/~~, -I-Ill


i lip
l [[ il' !:lr[I T-rr--,,:
i --Ij
I ~ ~ ~ _L ~ . . . . . . r J Slarldard plate isone-foot square
]00 -- --~[~,~-So.d(]ndgr(]vel -~ ~-- 2 Soiiunderplateisdryorraois!
|1 I I I II / | 3 Water table isat least several
4[ I r ~ r ,eat be,o, 0,o*e ,
I j I I ~ r

. 0.05 0.1 0.2 0.! [ I ~


Deflection t in

FIG. 18--Tentative correlation between relative density and standard plate load test de-
flection under 3 tons /ft 2.

10Manufacturedby Appareils Gamma Siar, 92 Levallois,France.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreeme
278 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

160 I l I I I I

LEGEND
AA
& 0 Reletive density determined by Woshington Densometer
r40 Two tests were mode approx. 3ft from each stondord
plote Iood test during test fill procjrom
A Relotive density determined by Washington Densometer
Test mode ot the side of stondord plete Iood test during
construction of etructurol fill

~20 A
A
A

100

~A 0 0

? ~o~-o~ , S Tentative Cot re let ice, See Fig.18


8o

60

40

20
o

0 I I I I I I
0 0.02 004 0.06 0.08 0.10 0.12 0.14
Oefrection of Stondord I-ft Plote Under 3 t / f t 2, in.

FIG. 19--Comparison between tentative correlation and results of standard plate load
tests made on compacted fills composed of uniformly graded sand.

plate load test under a loading pressure of 3 tons/ft ~. For the same relative
density, the correlation indicates a greater settlement for a sand and gravel
subsoil than for one composed of a uniform sand.
The correlation in Fig. 18 is based on an evaluation of the results of
SPLT at over 15 sites. We believe that it can be used to obtain a reasonably
reliable indication of relative density from the results of a SPLT. How-
ever, considerable variation from the correlation may be expected. This is
demonstrated in Fig. 19, wherein results obtained in a plate load test
study made in connection with fill composed of uniform sand are plotted
along with the tentative correlation for such a gradation.
We believe that the correlation in Fig. 18 is adequate to investigate

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License
LACROIX AND HORN ON EARTHWORK CONSTRUCTION PROJECTS 279

natural cohesionless deposits, or compacted fills with relative densities of


up to about 65 or 70 percent, particularly when the granulometry of the
soil is such that direct methods of measuring relative density or penetra-
tion tests are unacceptable (for example, heterogeneous deposits of sand
with gravel and cobbles). The SPLT method is unsatisfactory in cases in-
volving dense to very dense soil (Dr > 80 percent) because the settlements
are too small (that is, a few hundreds of an inch) to measure accurately.

Closing Comments
Relative density is being used successfully to control the construction of
heavily compacted granular fills. It is by no means a "cure all", and its
application as a control criterion has limitations which must be recognized.
One limitation concerns the degree of accuracy to which the relative density
of in situ soils can be determined. There are two aspects to this problem;
namely, the determination of the in situ dry unit weight, and the determi-
nation of the reference densities. In this paper, the authors have dealt only
with the difficulties and errors associated with the in situ determination;
however, errors involving determinations of maximum and minimum dry
unit weights can also be substantial, and are the subject of current research
(for example, Tavenas and LaRochelle [gP]).
In recent years, the use of relative density as a compaction control
criterion has been criticized because of the inaccuracies associated with its
determination. The authors recognize this problem, but feel that similar
inaccuracies are involved with the application of other compaction cri-
teria; for example, degree of compaction. In all such applications, an in
situ density is compared to a reference density or a range of reference
densities. The actual errors associated with determinations of a measure of
compactness are due to errors attributable to the determinations of the
in situ density and the density reference, and thus, are common to all of the
compaction criteria that are in use.
In addition to the difficulties involved with determination of relative
density, there are questions concerning the relationship between it and the
properties of a soil mass. For example, it has been proposed that in some
"unseasoned" fills which are less than two months old, there is little in-
ternal grain-to-grain shear strength. 11 In such cases, the values of N are
lower than those of an old "seasoned" fill, having the same relative density.
Burmister reasons that in time, say four to six months, minor internal ad-
justments in the grain structure may occur without any significant change
in relative density, but which increase grain-to-grain shear strength and
are thereby reflected by higher values of N. If this proposal is correct, then
there is a question concerning the validity of applying to recently coin-

~tPersonal communicationwith D. M. BurmisCer, 1967.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further rep
280 RELATIVEDENSITY INVOLVING COHESIONLESSSOILS

p a c t e d fills, c o r r e l a t i o n s b e t w e e n r e l a t i v e d e n s i t y a n d N t h a t were d e -
v e l o p e d for s e a s o n e d fills or n a t u r a l deposits.
D e s p i t e its s h o r t c o m i n g s , t h e a u t h o r s b e l i e v e t h a t r e l a t i v e d e n s i t y is t h e
m o s t effective c r i t e r i o n for c o n t r o l l i n g c o n s t r u c t i o n of c o m p a c t e d g r a n u l a r
fills. H o p e f u l l y , m a n y of t h e c u r r e n t p r o b l e m s a n d u n c e r t a i n t i e s a s s o c i a t e d
w i t h its a p p l i c a t i o n will b e r e s o l v e d b y c u r r e n t a n d f u t u r e research. T h e
p r a c t i c i n g engineers d e a l i n g w i t h design, q u a l i t y a s s u r a n c e , a n d e v a l u a t i o n
of p e r f o r m a n c e of c o m p a c t e d fills h a v e a m a j o r role to p l a y in such research.

References
[1] Leary, D. J. and Woodward, R. J., III, "Experience with Relative Density as a
Construction Control Criterion," included in this symposium.
[2] Burmister, D. M. in Field Testing of Soils, A S T M STP 322, American Society for
Testing and Materials, 1962, pp. 67-97.
[3] Poulos, S. J. and Hed, A., "Density Measurements in a Hydraulic Fill," included in
this symposium.
[4] D'Appolonia, D. J., Whitman, R. V., and D'Appolonia, E., Journal of the Soil
Mechanics and Foundations Division, American Society of Civil Engineers, Vol. 95,
No. SM 1, 1969, pp. 263-284.
[5] Koerner, R. M., Journal of the Soil Mechanics and Foundations Division, American
Society of Civil Engineers, Vol. 96, No. SM 4, 1970, pp. 1221-1234.
[6] Holubec, I. and D'Appolonia, E., "Effect of Particle Shape on the Engineering
Properties of Granular Soils," included in this symposium.
[7] Fletcher, G. A., Journal of the Soil Mechanics and Foundations Division, American
Society of Civil Engineers, Vol. 91, No. SM 4, 1965, p. 67-75.
[8] deMello, V. F. B., Proceedings, Fourth Panameriean Conference on Soil Mechanics
and Foundation Engineering, Vol. 1, 1971, pp. 1-86.
[9] Casagrande, A., "Influence of Effective Confining Pressure on N Values," American
Electric Power Preliminary Safety Analysis Report, D. C. Cook Nuclear Power
Station, Amend. 5, Fig. 8.4-6, 1968, p. G.57.
[10] Gibbs, H. J. and Holtz, W. G., Proceedings, Fourth International Conference on
Soil Mechanics and Foundation Engineering, Vol. 1, 1957, pp. 35-39.
[11] Mansur, C. I. and Kaufman, R. I., Transactions, American Society of Civil Engi-
neers, Vol. 123, 1958, pp. 715-743.
[12] Peck, R. B., Hanson, W. E., and Thornburn, T. H., Foundation Engineering, Wiley,
New York, 1953.
[13] Terzaghi, K. and Peck, R. B., Soil Mechanics in Engineering Practice, Wiley,
New York, 1948.
[14] Earth Manual, USDI, Bureau of Reclamation, 1960.
[15] Bazaraa, A. R. S., "Use of the Standard Penetration Test for Estimating Settle-
ments of Shallow Foundations on Sand," Ph.D. thesis, University of Illinois, Ur-
bona, 1967.
[16] Schultze, E. and Melzer, K. J., Proceedings, 6th International Conference on Soil
Mechanics and Foundation Engineering, Vol. 1, 1965, pp. 354-358,
[17] Peck, R. B. and Bazaraa, A. R., Proceedings, Soil Mechanics and Foundations
Division, American Society of Civil Engineers, Vol. 95, No. SM 3, 1969, pp. 905-
909.
[18] Peck, R. B., Proceedings, Soil Mechanics and Foundations Division, American So-
ciety of Civil Engineers, Vol. 79, Separate No. 326, 1953, pp. 1-14.
[19] Sanglerat, G., The Penetrometer and Soil Exploration, Elsevier Publishing Co.,
New York, 1972.
[~0] Sohmertmann, J. H., Proceedings, Soil Mechanics and Foundations Division, Ameri-
can Society of Civil Engineers, Vol. 96, No. SM 3, 1970, pp. 1011-1043.
[~1] Weaver, R. J. and Reball, P. M., "Determination of Embankment Density by the
Seismic Method," New York State Dept. of Public Works, Physical Research Re-
port RR 67-5, 1967.
[~2] Tavenas, F. and LaReehelle, P., Geatechnique, Vol. 22, 1972, pp. 549-562.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
D. H. Cornforth I

Prediction of Drained Strength of Sands


from Relative Density Measurements

REFERENCE: Cornforth, D. I-I., " P r e d i c t i o n of Drained Strength o f Sands


f r o m Relative Density M e a s u r e m e n t s , " Evaluation of Relative Density and
Its Role in Geotechnical Projects Involving Cohesionless Soils, ASTM S T P 523,
American Society for Testing and Materials, 1973, pp. 281-303.

A B S T R A C T : The peak strengths of sand in symmetric strain (for example,


triaxial compression) and plane strain converge at very loose densities towards
the common ultimate strength. By dividing the peak strengths by the constant
ultimate strength, a set of dimensionless curves is obtained which represent the
extra strength that can be mobilized in sand at relative densities above zero.
For the range of ultimate strengths normally encountered in natural sands
( r = 28 to 36 deg), computations show that these extra strengths c~an be
represented by single curves of density components ~ac (in degrees) which can
be added to the ultimate strength to give the total drained strength, ~a. The
ultimate strength ~br can be reliably estimated by performing a static angle of
repose test on the sand. The recommended techniques for measuring the angle
of repose and the maximum/minimum density of sand in the laboratory are
described. The method of predicting strengths has been compared with actual
strength tests performed on several sands of diverse origins, and the compari-
sons generally agree to within 1 deg on average. Finally, typical computations
are presented to aid the practicing engineer in the use of the method.

KEY WORDS: density (mass/volume), sands, angle of repose, triaxial


stresses, shear strength, tests, cohesionless soils

When a practicing soils engineer needs to know the drained strength


(~bd) of a sand in a design problem, he must choose one of two alternative
procedures. The most common procedure, adopted for almost all routine
studies, is to estimate the relative density of the sand from the Standard
Penetration Test (SPT) blow count [1]/with perhaps some modification
for the effect of overburden pressure on the results [~]. The strength of the
sand is then estimated from the relative density values using a published

Managing Director, Nuttall Geotochnieal Services Ltd., Colnbrook, Bucks, England.


2 The italic numbers in brackets refer to the list of references appended to this paper.

281

Copyright by ASTM Int'l


Copyright9 by(all
ASTMrightsInternational
reserved); Fri Mar 11 16:13:06 EST 2016
www.astm.org
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
282 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

correlation [3-5]. These correlations usually give ~b-values corresponding to


different relative densities and do not take into account the actual proper-
ties of the sand under study.
The alternative procedure is to obtain undisturbed specimens of the
sand and measure the relative density in the soils laboratory. A series of
drained triaxial compression tests is then performed on the sand to define
the strength-density relationship and thus estimate the strength of the
sand at the particular relative density. Since at least six tests are necessary
to define an adequate strength-density curve, the procedure is both costly
and time-consuming. It also takes considerable expertise in the field and
laboratory to obtain reliable data.
The purpose of this paper is to present a middle course to the preceding
alternatives. The proposed method permits the strength of sand to be pre-
dicted with reasonable accuracy from laboratory tests which can be per-
formed in one or two days. Furthermore, the natural strength properties
are represented in the analysis by the static angle of repose of the sand.
Experience with the method to date indicates that it does give good agree-
ment with actual strength measurements, and thus should prove to be a
more accurate method of predicting the strength of sand from relative
density measurements than previous correlations.

Original Research Study


The proposed method is based on the results of a research study into the
shear strength properties of Brasted sand. To understand the rationale of
the method, some of the conclusions and concepts put forward in the origi-
nal paper will be briefly summarized.
Consider first the stress-strain-volume change characteristics of a con-
solidated-drained shear test on dense sand, as represented by the triaxial
test shown on Fig. 1. As shearing progresses, the specimen starts to dilate
and the maximum rate of dilation, ~(Av/v)/~l, occurs at or near the peak
stress, point A on Fig. 1. If shearing is continued beyond the peak stress,
the stress slowly decreases until a failure plane develops (point B, Fig. 1).
Once a failure zone has been observed, the stress rapidly decreases to an
essentially constant value, commonly referred to as the ultimate strength
(point C). At the ultimate condition, the sand specimen shears at constant
volume and the corresponding friction angle is usually termed ~b~,.
The main results of the research study on Brasted sand are summarized
on Figs. 2, 3, and 4. The consolidation pressures and stress paths are not
differentiated on Fig. 2, to simplify the presentation. The term "initial"
on the abscissa of the graph refers to the conditions at placement. In the
computation of ~-values, it is assumed that the Mohr envelope for cohesion-
less sands is a straight line and passes through the origin.
The test program consisted of plane strain, triaxial compression, and

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant
CORNFORTH O N PREDICTION OF DRAINED ,STRENGTH OF SANDS 283

A
140
. ! -Sg ed:Sg0 ~
-38
" 120 -37
-36
-35
-- I00 -34 .e.__~C~ 33. ~
~-----~-33
-32
I
b" 80

4-
u~ 60
Cell pr(,ssure = j401b/sq, in
K o -- c,~nsolidat~td
D
"; 4o
C
a
. 0.
4-
w
20

4 6 8 10 12

Axial Strain eI ~

+4 -~
//
Max. sl ~ ("~'Y')
• +:3 ----

,+2
m
c
D

u +1

E
2
o 0
[
i

1 I
F I G . 1--Stress-strain-volume change characteristics of sand in a consolidated-drained tri-
axial compression test.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
284 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

48 LEGEND
Max. UIt.
46 = ~ 9 13 Plane strain
~ _ _ 9 o Tria~iaJcompression
44 (symmetric strain)

42
"~ / P I . . . . train strength (~d

~ 4o
-g ~'~. .

~ ~ Minimum density
/ =~.~ ~ for an intergranular
/ ~ ~ sand structure
36
Triaxial compression s t r e n g t h ( ~ ) d ~ ' ~ . ~ I
9~.34 o "~._ ,~ J
a
oo
% ~176 ao o o O o
o - ~a - *~' ~"' ~ ' -
32
e ultimate strength, (~)cv a
30
34 35 36 37 38 39 40 41 42 43 44 45
Initial Porosity ~

FIG. 2--Strength-density relationships, Brasted sand.

170 " i : ~ - LEGEND


160 ~
150 9 Plane strain test
o Triaxial compression
140
All tests performed at cell
130
pressures of 40 Ib/sq. ln.
" 120
==" 110
;;o F r , [ [ r(b)~L~a,~
.~ 100
90
, 80 i/ s
. . . . . 801-j'~ ~ ~ ~ + •

, i , i

50
2
9 40 Ko. co,sol.at|on o
I~ 30 l-stress . t *
9-
~0~
/~stress
ii
i i =
!~
i :

0 2 4 6 8 10 I'2 0 2 4 6 8 10 12 14 16 18 20
Axial Strain e= ~ Axial Strain tEI ~

FIG. 3--Stress-strain relationships, Bra~ted sand.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
CORNFORTH ON PREDICTION OF DRAINED STRENGTH OF SANDS 285

0.8
A 9 I

>lane Strain

'• 0.4
\

nsity
| 0.2 lular
0
sand s'~ruct ire

34 36 38 40 42 44
Initial Pg_rosltY ~ _
0.8~
Tri , x i a l ompres~ )n

W~
C
o
"~ 0.4
_o
o -

*'~ ee
Minimu densLty
~ o.2- "'~. fs~ in~g~t 'uelar
D
!
h
03, 36 38 40 42 44
Initial Porosity ~
FIG. 4---Rates of dilation, Brasted sand.

triaxial extension tests, the relevant stress-strain relationships being:


Plane strain ~1 > ~2 > ~a e~ = 0
Triaxial compression ~z > ~2 = ~a e~ = e8
Triaxial extension ~1 = ~ > ~a e~ = el
where al, as, and a3 are the major, intermediate, and minor principal
stresses, respectively, and Ez, e~, and e3 are the corresponding strains.
T h e triaxial compression and extension tests each h a v e axially s y m -
metric strain conditions, and the test results (strength-density plot) were
essentially the same in these tests [6], although the extension test results
are omitted from Fig. 2. T h e following general conclusions can be made:
1. The strength of the sand in plane strain is always higher than the
strength of the sand in symmetric strain at a given placement density. T h e

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursu
286 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

shape of the strength-density curves are similar, but the difference in


strength increases with increasing density. Thus, the strain condition im-
posed on the sand is a major factor influencing strength [6].
2. The ultimate strength of sand is approximately constant, irrespective
of the placement density, and is essentially the same in both plane strain
and symmetric strain 3 (Figs. 2 and 3).

Minimum Intergranular Density Concept


An important concept to emerge from this study is that of the minimum
intergranular density. It can be seen from Fig. 2 that the strength-density
curves in plane strain and symmetric strain are converging towards the
ultimate strength as the density decreases. There is a corresponding de-
crease in the rate of dilation. It follows that, if a sand specimen can be set
up at a particular density which is slightly lower than the loosest condition
achieved in this research study, the specimen could be sheared to failure
without dilating and attain directly (without a peak) the ultimate stre~Igth,
in both plane strain and symmetric strain tests.
Fortunately, it is a fairly simple matter to estimate this density from the
available data. One method is to plot the rates of dilation against initial
density and extrapolate to zero rate of dilation, as shown on Fig. 4. Another
method is to plot the difference between the peak and ultimate strengths
against the initial density and extrapolate to zero difference. In both cases,
and for each strain condition, the answer is essentially the same.
This unique placement density, at which the peak and ultimate strengths
coincide and there is no dilation, has been termed the "minimum inter-
granular density." It appears to be a fundamental property of sands, in
that it marks the transition from an intergranular structure (as typified by
the results shown on Fig. 2) to an unstable structure which collapses when
subjected to shear. This opinion is supported by research work at the
Norwegian Geotechnical Institute [7] where a technique was developed for
preparing sand specimens in a very loose condition. In subsequent drained
shear tests, the sand failed at very low strengths, apparently due to break-
down of a honeycombed sand structure.
The Norwegian technique has been modified to achieve the minimum
intergranular density in the tests on other sands referred to at a later stage
of this paper. A three-part split mold [8] was used to prepare the test speci-
men directly onto the base pedestal of the triaxial apparatus. Damp sand
was placed inside the mold and de-aired water was allowed to percolate
slowly upwards from the base. The rising water caused the sand to slowly
collapse, and more damp sand was added as necessary until the water level

3 The absolute validity of this statement does not affect the proposed method of pre-
dicting strength-density relationships; there is some experimental evidence which sug-
gests that ~c~ in the triaxial compression test increases slightly with increasing placement
density.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License
CORNFORTH ON PREDICTION OF DRAINED STRENGTH OF SANDS 287

reached the top of the mold. After sealing on the top cap and applying a
small negative pore-water pressure, the split mold was removed and the
triaxial cell was assembled for the test. The test specimen was then fully
saturated by the application of back pressures. This latter refinement puts
the test specimen at, or very close to, the minimum intergranular density,
and the sand specimen then shears directly to the ultimate strength without
a peak in the stress-strain curve and without dilating.
Since the concept of the minimum intergranular density is considered to
be a fundamental property of sands, it is suggested that it be adopted as
the definition of the "minimum density" of a sand.

Proposed Method for Estimating the Strength of Sands


The next item to be examined is the strength relationships of sands.
Geological factors influencing the measured strengths are the origin of the
sand and its environment of deposition. In engineering terms, the geo-
logical factors control the coefficient of interparticle friction, the grain
shape (sphericity, angularity), and the gradation of the sand. All these
factors affect the strengths measured at any density, including the ultimate
strength which is common to all the tests. Therefore, it should be possible
to consider the increase in strength with density as a dimensionless rela-
tionship by dividing the peak strength at any density by the constant
value of the ultimate strength. This effectively removes the geological
factors.
The resultant density factor is defined as follows:

/ ffl --if3 ff3~/maximum


density factor = (1)

0"1 O'~ O'3/ultimat e

where
al - major principal stress and
a3 = minor principal stress.
The computations of density factors provide dimensionless curves with the
same shapes as the original strength-density plots, but having a value of
1.0 at the minimum intergranular density and higher values at other densi-
ties above the minimum (Fig. 5). It is interesting to note that, in the re-
search study on which the present paper is based, the density factor in
plane strain was on average 70 percent greater than the corresponding
density factor in symmetric strain. This ratio was sensibly constant over
the entire range of densities.
Because density factors are dimensionless, they provide a possible way

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuan
288 RELATIVEDENSITY INVOLVING COHESIONLESSSOILS
2.8

2.6

2.4

2.2 L

II

aL

d~
~

o
a

Tria" , "n ~
1.2 ($ylmetrm train I.... ~

1.0
100 80 60 40 20 0
- . , ~-~min. %
Relative Dry Density (R.D.u.; ~mox.-}~min.
FIG. 5--Density factors, Brasted sand.

of relating the strength-density curves to other sands with different ulti-


mate strengths. These computations have been carried out for ultimate
strengths (r ranging from 28 to 36 deg, to provide the chart (Fig. 6).
A set of calculations is given in Table 1 for two density factors, 1.50 and
2.00, which are typical values from the range plotted on Fig. 5.
A remarkable result of these computations, as shown in the right-hand
column of Table 1, is that the increase in angle ~d above the ultimate
strength ~ , for a given density factor is practically independent of the
original ~b~o-value. The same conclusion can be made for similar computa-
tions over the full range of density factors. Therefore, as an alternative to
using the graphs of Fig. 6, it is possible to draw one set of curves which
will give the predicted increases in strength with density for sands having

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licen
CORNFORTH ON PREDICTIONOF DRAINEDSTRENGTHOF SANDS 289

52 ~ \\ LEGEND gl
51~ 0
o
50~ Plane strain curves
49~ Triaxial compression curies o

e~
O
o
In
m m
o E
e
L
42 ~
O
J
CII
c

38 ~
c o
O 37
3d
T.
la_
35 ~
34 ~
33~
32 ~ 12~
.0
31 ~
30'
29'o
28 .28 ~
100 90 80 70 60 50 40 30 20 10 0

Relative Dry Density (R.D.D.)- ~-~min %


~max-" "~min.

FIG. 6--Strength-density curves based on density factors.

TABLE 1--Computation of density component for two density factors.

Ultimate Strength Density Estimated Strength Density


Factor Component
of Strength
~~ deg (~1o~, (~)o,~ 0~' deg C ~ - ~-),
deg

28 1.770 1.50 2.655 34.8 -I-6.8


32 2.255 1.50 3.382 38.9 -I-6.9
36 2.852 1.50 4.278 43.0 -I-7.0
28 1.770 2.00 3.540 39.7 411.7
32 2.255 2.00 4.510 43.8 4}-11.8
36 2.852 2.00 5.704 47.7 ~-11.7

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agree
290 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

18

14{
12 strain
i

i ij[
6
.Triaxkd c ~ .

~ _.
lOO 9o ~ 7o 6O 5O 4O 3O 20 10 0
Relative
FIG. Dry - ~min. O/o
Density (R.D.D.) ~x~---~min.

7--Density components of strength.

ultimate strengths ~bc, within the range 28 to 36 deg. The strength attrib-
uted to changes in the density of the sand is termed the "density component
of strength, ~dc" (in degrees) and is shown on Fig. 7.
The predicted angle of shearing resistance ~ of a sand under drained
conditions is then given by:
r = r + r (2)
where
~cv = ultimate strength of sand, deg, and
~dc = density component of strength (Fig. 7), deg.
The ultimate strength of a sand (~c~) can be predicted with reasonable ac-
curacy by performing a static angle of repose test on the dry sand.
Before leaving the subject of strength, it should be mentioned that par-
ticle crushing can become an important factor affecting strength as the
confining pressure increases [9, 10]. There are no available guidelines for
defining the upper limit at which the strength of a particular sand is sig-
nificantly affected by particle crushing, and it must depend on the mineral
constituents and perhaps also on the grain size characteristics of angularity

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to
TABLE 2--Test data for comparison of Brasted sand to three other sands.
0

Brasted Sand Guinea Sand Portland Sand Limassol Sand


"1"

Test conditions saturated saturated saturated saturated O


Specimen diameter, in. 4 4 4 4 z
Type of consolidation Ko ambient ambient ambient
Cell pressures, psi 40/60 20/40/60 25 30
Friction eliminators used a yes yes no yes
Minimum density test rapid tilt rapid tilt rapid tilt rapid tilt/ASTM O
Maximum density test Kango hammer Kango hammer surcharged on vibrating Kango hammer z
table O
Test results:
Minimum density, l b/ f t 3 93.4 87.8 81 78.2
Maximum density, l b / f t 3 113.5 110.2 104 108.2 E"
Static angle of repose, deg 32.5 33.3 35.0 33.2
Measured avg ~c,-values, deg 32.6 36.1 34.4
Specific gravity, soil grains 2.68 "2".'68 2.72 2.73
Z
o
-1-
Rotating bush mechanisms, which effectively eliminate the vertical component of piston friction during shear test.
o

Z
U~

,O

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions auth
292 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

and sphericity. However, the test results presented later in this paper,
which were performed on predominantly quartz sands, indicate that par-
ticle crushing did not have any significant influence on the measured
strengths for consolidation stresses in the ranges quoted on Table 2. There-
fore, it is probable that the proposed strength-density correlation will be
valid for most sands at the stress levels normally encountered in practical
problems.

Correlation of Sand Densities


The range of intergranular densities that can be measured on different
sands depend on factors such as specific gravity, gradation, and grain
shape characteristics. It has been known for many years that these factors
can be rendered dimensionless by computing the relative density of sands.
The definition of relative density [1] is:
em~x -- e
relative density (Dd) -- (3)
emax -- emin

where emax, emi, are the void ratios of the sand at the density limits and
e is the void ratio of the sand in its natural state.
Although the relative density has been widely quoted in soil mechanics
literature, Eq 3 is cumbersome to use in practice. The definition itself is
confusing (minimum void ratio is the maximum density and vice versa),
and engineers often go through the laborious exercise of calculating the
void ratio when it is actually unnecessary, since the figure can be computed
directly from the dry densities:

(~/ - - ~/min) ~/rnax


D~ = (4)
(~max -- ~ m i . )
where
7 = dry density of sand in its natural state. 4
Working in reverse, the calculation of the dry density corresponding to
a given value of relative density is:
~/max ~ m i n
= (5)

A more practical parameter for engineering studies is the relative dry


density, defined by:

relative dry density (RDD) - 7 -- 7mln (6)


")'max - - "Ymin

4Dry density is usually written as ~'dto differentiateit from the wet (bulk) density 7.
In this paper, the suffix " d " has been omitted from the relative density equations.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement
CORNFORTH ON PREDICTION OF DRAINED STRENGTH OF SANDS 293

all densities in the term being dry densities. Comparing Eqs 4 and 6, the
relationship between relative dry density and relative density is:

RDD = D,z.( ? ~ (7)

Thus, the relative dry density is always slightly lower than the relative
density except at the density limits, where the two parameters are identical.
The differences are quite small and can be ignored in most practical appli-
cations. However, the relative dry density parameter permits density to be
calculated more quickly and easily from the measured density limits. 5

Laboratory Measurements
The proposed method of predicting the strength-density relationships of
sands thus involves the measurements of the maximum and minimum dry
densities of the sand and the static angle of repose. The recommended
laboratory techniques are outlined in the next few paragraphs.
Maximum Dry Density--The British Standard test: Determination of
the Dry Density/Moisture Content Relation of Granular Soil--Vibrating
Hammer Method (BS 1377: 1967, Test 13,) is satisfactory, and it is es-
sentially the same as that used in the research study on which this paper is
based [11]. An essential requirement of any maximum density test is that
the sand should be effectively confined during vibration to eliminate turbu-
lence [12], so the vibrating plate must be placed on the surface of the sand
and a downward pressure applied, as described in the British Standard
test.
The British Standard test recommends that water be added to the sand
in increments and that a series of vibrating compaction tests be carried
out to define a dry density/moisture content curve in a similar way to
other compaction tests. It has been my experience that the maximum dry
density is achieved only when the sand is fully saturated (Fig. 8).
A similar conclusion was reached by Kolbuszewskl [13]. In conventional
compaction tests on silts and clays, the increasing percentage of water in
the soil decreases the dry density above the optimum water content be-
cause of the volume occupied by the water in the nearly saturated soil.
In permeable sands, however, excess water can escape from the soil during
compaction and stand above the soil surface. Thus, the test behavior of
soils in the two types of compaction test is not comparable. Furthermore,
at the high densities achieved by vibrating compaction of sands, even small
bubbles of entrapped air can produce sponginess and thus reduce the dry
density achieved (Fig. 8). It is recommended, therefore, that the test be
performed on fully saturated sand.
5 The relative dry density is numerically equal to (and thus interchangeable with)
the relative porosity, which is sometimes quoted in research papers on the properties of
sands.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License
294 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

120
,~Saturatioi line (zelro air
9 voids) For G = - 6 8
(a) B i n , t e d Sard,,,~~ \ . \
~o~ ~ " Muration line for
c~ ' ~ ' ~ - e - --''"~ G 2 "73
W

m 100 .- -

J ~ ( b ) LimassoSand
E 90
0
a

,,,= 80

70
0 5 10 15 20 25 30 35

Water Content W ~

FIG. 8--Density-water content relationships for vibrating compaction with a Kango


hammer.

To ensure that the sand is saturated, the dry sand should be poured into
a tray of boiling water and boiled for 5 to 10 min. The sand should then be
left overnight to cool down to room temperature before the test is per-
formed. The base of the compaction mold is sealed to retain water, and
about 2 in. depth of water is placed in the mold. The saturated sand is
placed below the water in the mold and the compaction test is performed
by the method described in British Standard 1377: I967, Test 13 (see Fig.
9). The test should be repeated until the maximum dry density has been
consistently established.
A common problem in using the vibrating plate on saturated sand is
that sand flows up around the annulus between the circular plate and sides
of the mold. It can be prevented by putting a small pad of cloth immedi-
ately beneath the vibrating plate. Care should be taken to avoid creating
a suction when the hammer is removed from the sand surface.
It is also recommended that the specific gravity (G) of the sand grains
be measured to give an independent check on the maximum dry density.
Since the sand in the test mold is fully saturated, the dry density is given
by:

G--r~ (s)
~ = (1 + wG)

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductio
CORNFORTH ON PREDICTION OF DRAINED STRENGTH OF SANDS 295

where
~'w = unit weight of water, and
w = water content of the sand.
T h e results of the check test should agree to within a b o u t 0.3 l b / f t 3 of the
d r y density computed from weight and volume measurements.

FIG. 9--Vibrating comoaction to measure the maximum dry density.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agree
296 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

M i n i m u m dry density--The research study on Brasted sand showed


that the minimum intergranular density could be measured by the rapid
tilt test [14]. The test procedure is as follows: 1000 g of dry sand are weighed
out and placed in a 1000-ml glass measuring cylinder. The cylinder should
be graduated at 10-ml intervals and preferably should not have a pouring
lip. A piece of rubber membrane is stretched over the top of the cylinder
and is sealed by a rubber O-ring. The cylinder is inverted, and then re-
turned to its original upright position in a movement taking about 3 s.
The sand, in general, does not form a level horizontal surface when it re-
turns to the bottom of the cylinder. Experience has shown, however, that
slight additional tilting of the cylinder to level the sand can be carried
out without affecting the measured volume.
The glass measuring cylinder should be carefully placed on a level bench
and the average volume (that is, height) of sand estimated to the nearest
milliliter by eye using the 10-ml graduations as a guide. If the sand surface
is reasonably level, the result should be accurate to within • ml. The

FIG. lO--Measurement of the static angle of repose using cardboard templates.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
CORNFORTH ON PREDICTION OF DRAINED STRENGTH OF SANDS 297

test is repeated at least ten times. The average volume measured in the
series of tests is used to compute the minimum dry density.
An alternative test for the minimum density is the ASTM Test for Rela-
tive Density of Cohesionless Soils (D 2049-69). The test consists of pouring
dry sand into a mold using a standard technique. Tests on six samples of
Limassol sand showed that this test gave essentially the same results as
the rapid tilt test. The ASTM procedure has the advantage that small
amounts of gravel can be included in the test sample.
Static angle of repose test--A large flat glass plate is carefully levelled
by a spirit level. Dry sand is deposited through a funnel into a conical heap
approximately 3 in. high (Fig. 10). The heap is undercut by scraping away
sand grains from the bottom of the heap in very small quantities until the
sand grains ravel down the face of the slope. The equilibrium slope of the
heap is then measured by some means, for example, by using stiff card-
board templates cut to different angles, the angles differing by 1 deg (Fig.
10). Measurement should take place only after a surface wave movement
has occurred down the whole face of the slope. However, the extreme top
of the slope does not always move, and a slightly steeper slope angle at the
top of the heap should be ignored. Similarly, flatter slopes resulting from
more dynamic wave runs should not he measured.
The test is repeated 10 to 20 times, and the average value is computed to
give the static angle of repose. The result should approximate the ultimate
strength of the sand. Although individual readings are measured to the
nearest degree, it is recommended that the result be expressed to 0.1 deg,
and any "rounding off" be carried out after it has been added to the density
component of strength.

Applicability of Method to Other Sands


The general validity of the proposed method of predicting the strength-
density relationships of sands in triaxial compression has been checked on
three additional sands of widely divergent origins. The gradation curves of
the three sands are compared with Brasted sand on Fig. 11, and the test
conditions and properties of the sands are listed on Table 2.
The shear tests were performed in three different commercial labora-
tories as consolidated-drained triaxial compression tests, that is, a2 = a3,
symmetric strain. The measured strengths are plotted on Fig. 12. The
predicted strength-density curve, based on the laboratory density limits
and the static angle of repose, is also shown for each sand, and the agree-
ment is reasonably good. For a given placement density, the average di-
vergence of the measured strengths ~ from the predicted relationship are
as follows:
Guinea sand --0.6 deg
Portland sand ~-0.8 deg
Limassol sand -]-0.8 deg

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licen
298 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

HYDROMETER ANALYSIS] BRITISH STANDARD SIEVE SIZES ~,~


100
90
111iii1'i 73"6

80
70
60
50 IL I]ll
~ Ll/lll/I
~
I LI/IllL~I I/i~l
liil] il I I[l[llL I
40 Ii IIIII II 1- I IIH I I/~ IIIII l I I)IIIIII i l
30 ! 1[11~1 I ~/V 1lilt 11 I Ill]II I
20
10
0
0.002 ).006 0.02 0.06 0.2 0.6 2 6 20
ne MediumlCoarse Fine Mek~mlCoarse Fine Mediuml
SILT SAND j ~GRAVEL

A - - River sand from Brasted, Kent, England


B -- Marine sand from Guinea , West Africa

C m River sand from Portland , O r e g o n ( U.S. s i e v e sizes)


D m Marine sand from Limassol , Cyprus

FIG. l l - - G r a d a t i o n curves.

None of the tests, including the original research study on Brasted sand,
were performed on specimens with free end support. However, research on
this topic (for example, Ref 15) indicates that frictional end restraint has
no effect on the measured peak strength provided the specimen height:
diameter ratio is at least 2.0. This condition was satisfied on all the test
specimens referred to in the present paper.
Guniea Sand--These results are probably the most accurate of the three
sets of data. The tests were performed at three different cell pressures but,
as the cell pressure had no apparent effect on the measured 0~-values, they
have not been differentiated on the graph.
Portland Sand--As stated in Table 2, this is the only set of tests in which
the triaxial cell was not fitted with rotating bush friction eliminators [8].
This could increase the measured 0d-values, especially at the lower densi-
ties where the axial strains at failure are greater. The maximum dry density
of this sand was measured by a different technique to the other sands.
Limassol Sand--This very fine grained sand had a significant fraction
passing the No. 200 BS sieve (Fig. 11). For this sand, the static angle of
repose test was less satisfactory than for the other sands. The outer surface
of the Limassol sand heap had a tendency to slump dynamically instead of
the sand grains ravelling down the outside of the heap. Consequently, the
measured static angle of repose may be lower than the true value.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lic
CORNFORTH ON PREDICTION OF DRAINED STRENGTH OF SANDS 299

I have not been able to perform additional tests to check the validity of
the method for predicting the strength of sands in plane strain. However,
in the basic research study, the ratio of the strength of Brasted sand in
plane strain to that in symmetric strain was consistent at all placement
densities. It is felt, therefore, that publication of the predicted curve for
plane-strain conditions is justified, and recent research work by others
shows good agreement with predictions based on this curve.
Plane-strain tests performed at Manchester University [16] on Mersey
River sand reported by P. W. Rowe, are plotted on Fig. I3. Similarly,
plane-strain tests performed on specimens of various sizes at the Univer-
sity of California (Marachi et al [10]) have been replotted on Fig. 14 on the
basis of a r of 36.9 deg, this figure being derived from the published
stress-strain curves for the sand. The average divergence of the measured
strengths r from the predicted curves at a given placement density are as
follows:
Symmetric Strain Plane Strain, deg
(triaxial compression),
deg
Rowe (1969) [16] -0.8 -0.9
Marachi et al (1969) [10] -2.1 -t-0.4

For most practical engineering studies, it is anticipated that the predic-


tion of strength in symmetric strain (corresponding to the results obtained
in triaxial compression and extension tests) will be of more interest and use-

114 112 110 108 106 104 102 I00 98 96 94 92 90 88 86 84 82 80 78

Dry Density Ib /cu. ft.

FIG. 12--Comparison of measured and predicted strength-density relationships for three


sands in triaxial compression.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further
300 RELATIVEDENSITY INVOLVING COHESIONLESSSOILS

100 90 80 70 60 50 40 30 20 10 0
Relative Dry Oensity ~

FIG. 13--Comparison of measured and predicted strength-density relationships of River


Mersey sand in triaxial compression and plane strain (16].

fulness. Some examples are given below of how the prediction method can
be applied to such problems.

Typical Examples of Computations


Example 1--The simplest use of the method is where the relative density
of the sand has been estimated from penetrometer tests performed down
boreholes. Disturbed specimens of the sand are recovered for description.
Suppose the relative density is estimated to be 60 percent and angle of
repose tests on sand specimens at the bearing level give values in the range
33 to 34 deg. What is the estimated friction angle of the sand at this relative
density?
Relative density --- 60 percent = relative dry density approximately.
Density component (~c) for R D D of 60 percent in the triaxial test is
5.5 deg (Fig. 7).
Static angle of repose = r = 33/34 deg, say 33.5 deg.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agre
CORNFORTH ON PREDICTIONOF DRAINED STRENGTH OF SANDS 301

Therefore, estimated friction angle Eq 2


~ = ~ . -t- ~o
-- 39 deg
Alternatively, the result can be taken directly from the graph, Fig. 6, by
interpolating between the curves representing ~=~ -- 33 deg and ~ = 34
deg.
Example g--As for Example 1, except that thin-wall piston specimens
have been recovered from the bearing stratum. The natural densities and
water contents of the undisturbed sand specimens are measured, and
density limit tests are performed. A typical result is given below:
natural dry density -- 98.0 lb/ft a
maximum dry density 107.3 lb/ft 3
--

minimum dry density = 84.0 lb/ft 3


From Eq 6,
98.0 -- 84.0
relative dry density (RDD) - 107.3 -- 84.0' = 60 percent

181
I
\ LEGEND
Measured strength, plane strain After
Marac~
m
Q Measured strength triaxial compression et.al.
i9i l .
'~

#
14
.\ f " -- I I
.
I!jll l
(19A?)

12
\
,I=

~ 8
\ ~ \~./Pre, ficted cJrve, plane stl ~in
4) \ \
I:
0
6
\ \
a,
E Predi(ted cur ~,e, ~
uo
4
triaxi :d cam aressioi~/ 9
>,,
m 2
c
o
a
0
90 80 70 60 50 40 30 20 10 0
Relative Dry Density ~

F I G . 14--Comparison of measured and predicted strength-density relationships o/ Mon-


terey No. ~0 sand in triaxial compression and plane strain [10].

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License
302 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

C o m p u t a t i o n t h e n proceeds as described in E x a m p l e 1. As a check on t h e


a c c u r a c y of m e a s u r e m e n t it is advisable to measure t h e specific g r a v i t y of
the sand grains. F o r the a b o v e specimen, t h e results m i g h t be:
specific g r a v i t y of sand grains G = 2.70
n a t u r a l water c o n t e n t w = 26.8 percent
As t h e sand is below the water table, it m u s t be fully saturated.
Therefore, void ratio e = wG
= (0.268)(2.70)
= 0.723
G.~'w
c o m p u t e d n a t u r a l d r y density ~d - (1 q- e)

= 97.7 l b / f t a (checks)
Example 3 - - E s t i m a t e t h e s t r e n g t h - d e n s i t y relationship for Guinea sand
(see T a b l e 3) in the triaxial compression test.
m a x i m u m d r y density = 110.2 l b / f t 3
m i n i m u m d r y density = 87.8 l b / f t 3
density range = 22.4 l b / f t 3
rearranging E q 6,
d r y density ~/d = "~min -[- R D D (Vmax - - ~/min)

= 87.8 -b R D D (22.4)
static angle of repose = 33.3 deg

total drained s t r e n g t h 0d = r -k Ode

TABLE 3--Computation of strength~dry density relationship for Guinea sand in the


triaxial compression test.

Dry Density
Increment
RDD Density
- - (~max -- 7rain), Component Estimated
100 Dry Density r (Fig. 7), Friction Angle
RDD, % lb/ft a "~d,lb/ft a deg ~d, deg

0 0.00 87.8 0.0 33.3


10 2.24 90.0 0.6 33.9
20 4.48 92.3 1.3 34.6
30 6.72 94.5 2.1 35.4
40 8.96 96.8 3.1 36.4
50 11.20 99.0 4.2 37.5
60 13.44 101.2 5.5 38.8
70 15.68 103.5 6.9 40.2
80 17.92 105.7 8.5 41.8
90 20.16 108.0 10.2 43.5
100 22.40 110.2 12.1 45.4

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License
CORNFORTH ON PREDICTIONOF DRAINED STRENGTH OF SANDS 303

Accuracy of Computations
I t is not being suggested t h a t the estimation of strengths b y these meth-
ods warrants the computations of angle r to tenths of a degree. As stated
previously, however, it is recommended t h a t "rounding off" the answer be
left to the end of the calculation rather t h a n at the intermediate stage.
I t is also recommended t h a t relative density (D~) and relative dry density
( R D D ) be used interchangeably in a n y computations. T h e m a x i m u m
difference between the two parameters occurs at the middle of the range and
is typically of the order of 5 percentage points in relative density. T o com-
p u t e the difference, if required, the first step is to compute the natural d r y
density from Eq 5 and then compute the relative dry density from E q 7.

Acknowledgments
T h e research work on Brasted sand referred to in this paper was per-
formed during the period 1957/59 at Imperial College, London under the
general supervision of A. W. Bishop.

References
[1] Terzaghi, K. and Peck, R. B., Soil Mechanics in Engineering Practice, Wiley,
New York, 1948.
[2] Gibbs, H. J. and Holtz, W. G., Proceedings, 4th International Conference on Soil
Mechanics and Foundation Engineering, London, Vol. 1, 1957, pp. 35-39.
[3] Peck, R. B., Hanson, W. E., and Thornburn, T. H., Foundation Engineering, Wiley,
New York, 1953.
[4] Sowers, G. B. and Sowers, B. F., Introductory Soil Mechanics and Foundations,
MacMillan, New York, 1951.
[5] Hough, B. W., Basic Soils Engineering, Ronald Press, New York, 1957.
[6] Cornforth, D. H., Geotechnique, Vol. 14, 1964, pp. 143-167.
[7] Bjerrum, L., Kringstad, S., and Kummeneje, O., Proceedings, 5th International
Conference on Soil Mechanics and Foundation Engineering, Paris, Vol. 1, 1961,
pp. 29-37.
[8] Bishop, A. W. and Henkel, D. J., The Measurement of Soil Properties in the Triaxial
Test, Edward Arnold, London, 1957.
[9] Bishop, A. W., Geoteehnique, Vol. 16, No. 2, 1966, pp. 91-130.
[10] Marachi, N. D., Chan, C. K., Seed, H. B., and Duncan, J. M., "Strength and
deformation characteristics of rockfill materials," University of California Report
TE-69-5, Berkeley, Calif., 1969.
[11] Cornforth, D. H., "Plane Strain Failure Characteristics of a Saturated Sand,"
Ph.D. thesis, University of London, 1961.
[12] Mogami, T. and Kubo, K., Proceedings, 3rd International Conference on Soil
Mechanics and Foundation Engineering, Zurich, Vol. I, 1953, pp. 152-155.
[13] Kolbuszewski, J. J. in Proceedings, 2nd International Conference on Soil Mechanics
and Foundation Engineering, Rotterdam, Vol. 1, 1948, pp. 158-165.
[14~] Kolbuszewski, J. J., Proceedings, 2nd International .Conference on Soil Mechanics
and Foundation Engineering, Rotterdam, Vol. 7, 1948, pp. 47-49.
[15] Bishop, A, W. and Green, G. E., Geotechnique. Vol. 15, 1965, pp. 221-242.
[16] Rowe, P. W. in Geotechnique, Vol. 19, No. 1, 1969, pp. 75-86.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
I. Holubec 1 and E. D ' A p p o l o n i a ~

Effect of Particle Shape on the Engineering


Properties of Granular Soils

REFERENCE: Holubec, I. and D'Appolonia, E., "Effect of Particle S h a p e


on t h e E n g i n e e r i n g Properties o f Granular Soils," Evaluation of Relative
Density and Its Role in Geotechnical Projects Involving Cohesionless Soils, A S T M
S T P 5~8, American Society for Testing and Materia/s, 1973, pp. 304-318.
ABSTRACT: The use of relative density correlations based on an "average"
sand to predict soil behavior without considering the particle shape can result
in poor or misleading predictions. Experimental data show that the particle
shape has a pronounced effect on all engineering properties studied. Angularity
of the particles increases the maximum void ratio, strength, and deformability
of cohesionless soils. Variations in engineering properties due to particle shape
can be as large as variations associated with large differences in relative density.
Penetration tests in small containers with small rods suggest that the Standard
Penetration Test is influenced by both the angularity and density of cohesion-
less soils.
K E Y WORDS: cohesionless soils, particle shape, density (mass/volume),
strength, deformation, void ratio, tests

The effect of particle shape on the behavior of granular materials has


been observed by the foundry industry, particularly when forming sand
molds with automated equipment. The "static" compaction of angular
sands with high pressure machines produces much higher spring back on
the release of the compaction load than does static compaction of equi-
dimensional sands [1].3 Tests on granular base course materials used for
flexible pavements disclose that crushed stone with angular particles has
greater elastic and permanent deformations under repetitive loading condi-
tions than crushed gravel composed of rounded particles [2, 3, 4]. Koerner
[5] considered the effects of mineralogy, gradation, and angularity on the
shear strength of cohesionless soils and has shown that the angle of internal

1 Project engineer and president, respectively, E. D' Appolonia Consulting Engineers,


Inc., Pittsburgh, Pa. 15235.
2 The italic numbers in brackets refer to the list of references appended to this paper.

304

Copyright by ASTM Int'l


Copyright9 (all rights
by ASTM reserved); Fri www.astm.org
International Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement
HOLUBEC AND D'APPOLONIA ON PROPERTIES OF GRANULAR SOILS 305

friction:
1. varies inversely with sphericity and directly with angularity of the
particles,
2. varies inversely with the effective size of the soil specimen,
3. shows no appreciable change with the uniformity of gradation, and
4. is influenced by mineralogy through its influence on the physical
properties of the particles.

Recognizing the significance placed on the correlation of relative density


with engineering properties, the question arises: Do granular soils at the
same relative density have the same properties? Tests on medium to fine
sands with varying particle shapes indicate that granular soils at the same
relative density can have drastically different engineering properties.
Therefore, the use of relative density criteria in design, without considering
particle shape, can result in poor or misleading predictions of soil behavior.
In this paper, data is presented showing the effect of particle shape on soil
properties for materials of similar gradation compacted to different relative
densities.

Particle Shape
Recent data show that particle shape has considerable influence on the
engineering properties of granular soils. However, lack of a particle shape
parameter makes correlation and evaluation of particle shape on soil prop-
erties difficult and, at times, confusing. To correlate properties to particle
shape, a standard method of measuring and expressing particle shape has to
be developed.
Particle shape has been defined in terms of the sphericity and angularity
of the particles. Sphericity, as defined by Wadell [6], is the ratio of the sur-
face area of a sphere having the same volume as the particle to the surface
area of the particle. A sphericity value of one implies a perfect sphere, and
the sphericity value decreases with the irregularity of the particle. Angu-
larity is a measure of the curvature of the corners to the average curvature
of the particle. Even though these definitions of the sphericity and angu-
larity have been widely accepted, methods of measurement have not been
standardized because of the tedious task of making numerous readings.
Indirect methods have been developed to measure the particle shape of
sand and silt-sized particles using porosity, permeability, and sedimentation
velocity of the materials. For this paper, the particle shape of the sands
was measured by an indirect method based on permeability developed by
Hoffman [7]. Hoffman developed equipment which determines the specific
surface of a sand from its permeability, and the particle shape is defined by
a coefficient of angularity which is the ratio of the measured specific surface
of the sand particles to the specific surface of equivalent spheres. In this

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licen
306 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

U.S. STANDARD SIEVE SIZES


I I/2 3/4 3/R 4. I0 20 40 60 I00 140 200
I00
I
I
90 I I
I ,
8O

I i
P- 7O
"I-
C~
hl 6O
I t I ~ I ! I
>- i i
m 50
n~
W
Z

,6O ,
I
!
Il/'
i
II
i J
I
I I I I 11=Ii ,,
' i I I i 1 i I

' ' J r
0 GLASS BEADS i [~] /3~
" ~ O T T A W A SAND I I\ I~
F:'iSOUTHPORT SAND
OLIVINE S A N D S...
v

ICOBSLES I COARSE
GRAVE
[ LINE ~OARSEi MEDIUM
SAND I FINE I
FIG. 1--Grain size curves.

measure, the coefficient of angularity of a sphere is one. The coefficient


increases with the deviation of the particle from a true sphere.
In the discussion of the data herein, descriptive terminology such as
"angular" and "rounded" is used to describe the particle shape. Numerical
values of the particle shape are presented in the tables and figures.

TABLE 1--1ndex properties of glass beads and sands.

Material Coefficient of Specific Effective Coefficient of Coefficient of


Angularity, Gravity, Diameter, Uniformity, Curvature,
E~ SG D~o C~ C~

Glass beads 1.16 2.48 0.40 1.5 1.05


Ottawa sand 1.24 2.65 0.24 1.9 1.13
Southport sand 1.55 2.68 0.15 1.8 0.86
Olivine sand 1.64 3.25 0.38 1.5 1.10

E = 1 for spheres; increase of E indicates increase of angularity.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agre
HOLUBEC AND D'APPOLONIA ON PROPERTIES OF GRANULAR SOILS 307

Test Sands
Results of tests on four granular materials with particles in the medium
to fine sand range are presented. The materials include glass beads, and
Ottawa, Southport, and Olivine sands which are listed in order of increasing
particle angularity. Grain-size curves and index properties are shown in
Fig. 1 and are given in Table 1.

M i n i m u m a n d M a x i m u m Void R a t i o s
Methods other than the A S T M Test for Relative Density of Cohesion-
less Soils (D 2049-69) were used to establish the minimum and maximum
void ratios for observing the effect of particle shape on these void ratios.
The maximum void ratios were obtained b y a method suggested by Lucks
[8], which produces slightly lower and more reproducible densities than the
ASTM method. With this method, a 3-in. diameter tube is placed inside a
Standard Proctor mold, the tube is filled with a sufficient quantity of sand
to overfill the mold, and is raised quickly, allowing the sand to fill the mold.
The minimum void ratios were obtained b y two methods. In the first, a
tube was filled with sand and horizontally vibrated by tapping until no
further settlement was observed. The tube was continually filled during
tapping. In the second method, the sands were compacted using the Modi-
fied Proctor compaction method according to ASTM Tests for Moisture-
Density Relations of Soils, Using 10-Lb Rammer and 18-In. Drop
(D 1557-70).
The maximum and minimum void ratio determinations are given in
Table 2 and are plotted in Fig. 2 as void ratio versus particle shape. The
densification by vibration is shown by dashed lines, and by compaction in-
dicated by solid lines. The effect of particle shape is indicated b y lines
drawn through the maximum and minimum void ratios. The data show
that the least minimum void ratio for the glass beads with near spherical
particles is obtained by vibration, while for the Southport and Olivine
sands which have angular particles, the lowest minimum void ratio is pro-
duced by dynamic compaction. The two methods give practically the same
values for Ottawa sand which has subrounded particles. Furthermore, both

TABLE 2--Minimum void ratios by two methods of densification, vibration and compaction
maximum void ratio [8].

Glass Beads Ottawa Sand Southport Sand Olivine Sand


Viba Mpb Vib'~ MP b Viba MP b Vib'* MP b

em~x 0.672 0.672 0.710 0.710 0.880 0.880 1.070 1. 070


e~ia 0.522 0. 592 0.482 0.481 0. 624 0.530 0.730 0. 628
emax -- emin 0.150 0.080 0.228 0.229 0.256 0.350 0.340 0.442
Vib = vibration by horizontal tapping.
b MP = modified proctor compaction method.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
308 RELATIVE DENSITY I N V O L V I N G COHESIONLESS SOILS

1.40

Z
0 u3
Z
O3 Z
1.20
I.- >
o
Q. MAXIMUM
:m
ot-Dz ~ V O I D RATIO
(:3
1.00 Z
<~
a)

5
OE 0.80
C~

>
1 !1 ii
0.60
' ' ,NIMUM VOID RATIO

T
0.40 ~-..VIBRATION
I
..k
MODIFI ED PROCTOR

0.20 I I I I
1.0 1.2 1.4 1.6 1.8 2.0
ROUND ANGULAR

COEFFICIENT OF ANGULARITY , E
o
MINIMUM VOID RATIO BASED ON MODIFIED
PROCTOR COMPACTION TEST, EXCEPT FOR GLASS BEADS,

FIG. 2--Effect of particle shape on minimum and maximum void ratios.

void ratios increase with increase in angularity of the particles; however,


the increase in maximum void ratio is considerably more than that for the
minimum void ratio. Thus, the difference between the maximum and mini-
mum void ratios is also directly related to the particle shape. For calcula-
tions of relative density of the test specimens, the smaller of the minimum
void ratios produced by the two methods was employed.

Stress-Strain Characteristics
Stress-strain characteristics of the four sands were determined by drained
triaxial tests on saturated 2-in. diameter by 4-in. long specimens. The
specimens were tested at constant cell pressure and at constant diameter.
The constant diameter was maintained by loading the specimens at a con-

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License
HOLUBEC AND D'APPOLONIA O N PROPERTIES OF GRANULAR SOILS 309

Brant strain rate and increasing the cell pressure to prevent any diameter
change monitored by a lateral deformation transducer mounted at mid-
height of the specimen. The lateral deformation transducer consisted of a
beryllium copper clip instrumented with strain gages [9] and connected to
a strain indicator. The samples were prepared by placing saturated ma-
terial in a split mold and vibrating the mold or tapping the material until
the desired density was obtained.

Shear Strength
The shear strength of granular materials is given by the friction angle
which is frequently correlated with the relative density [10]. Recently,
Koerner [5] and Lucks [8] indicated that the particle shape has to be also
considered in any relative density-friction angle relationship. Friction
angles determined from constant cell pressure tests on the four test ma-
terials at various relative densities are shown in Fig. 3.
These data clearly show that each of the four sands has a separate and
distinct relative density-friction angle relationship. The glass beads have
the smallest friction angles and exhibit the least increase in friction angle
with increasing relative density. Ottawa sand with subrounded particles

{n 4o
w
w
n,-
(.9
Q
[3

_.1
Z

30
Z
0
I-
ne
U- 0 GLASS BEADS
OTTAWA SAND i,~----DR = 7 0 %
I
SOUTHPORT SAND J
I
OLIVINE SAND J
I I I I I
0 20 40 60 80 100

RELATIVE DENSITY, D R , %

FIG. 3--Effect of particle shape on angle of internal friction.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lic
3 I0 RELATIVE DENSITY I N V O L V I N G COHESIONLESS SOILS

50
LEGEND
U) "~-- DR:90%
LLI
LU DR= 70 %

C~ DR=40% R=70%
~40

UJ
.J
C9
Z

230
0
I-- E) GLASS BEADS
0
E rq OTTAWA SAND
h
A SOUTHPORT SAND
@ OLIVINE SAND
20 I I I I
1.00 1.20 1.40 1.60 1.80 2.00
ROUND ANGULAR
COEFFICIENT OF A N G U L A R I T Y , E

FIG. 4---Effect of particle shape on angle of internal friction.

DR-~70%
T3= 2 TSF
9 9 9 0 INDICATE Q MAXIMUM
LL 61=9.2
~)

I1~) J

II
0

I PERCENT STRAIN REQUliED


n~ TO REACH FAILURE~.
W

E) GLASS BEADS I.I


17 OTTAWA SAND 2.9
A SOUTHPORTSAND 5.7
OLIVINE SAND 9.2
0| I I
0 2 4

AXIAL S T R A I N , 61 , %

FIG. 5--Stress-strain curves from constant cell pressure tests.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
HOLUBEC AND D'APPOLONIA O N PROPERTIES OF GRANULAR SOILS 311

has intermediate friction angle values and shows the largest increase with
relative density. Finally, Southport and Olivine sands with angular par-
ticles have the highest friction angles with an intermediate increase with
increasing relative density.
To illustrate the effect of particle shape on the friction angle, the friction
angles at relative densities of 40, 70, and 90 percent have been plotted
against the coefficient of angularity in Fig. 4, which shows that the friction
angle is a function of both relative density and particle shape. Furthermore,
it is observed that equally large differences in friction angle are possible
with variations of particle shape as with changes in relative density. For
example, the friction angles of Ottawa and Southport sands at DR = 70
percent are 35 and 40 deg, respectively, representing a 5-deg difference.
Equally well, the friction angle of Southport sand increased 4 deg with
densification from 40 to 90 percent relative density.

15.0

C) GLASS BEADS

IQO
I--

LIJ
t--
co

x
<

I--
< 5.0
z
<

I--

._J
<

<
| Q
|

I I I I
20 40 60 80 I00

RELATIVE DENSITY, DR, %

FIG. 6---Effect of particle shape on axial sSrain at maximum strength.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
312 RELATIVEDENSITYINVOLVING COHESIONLESSSOILS

Stress-Strain Relations
Figures 5 to 9 give test data for constant confining pressure and constant
diameter (Ko) tests, illustrating the influence of the particle shape on the
stress-strain characteristics of sands. The stress-strain curves for constant
confining pressure tests on specimens at the same relative density in Fig. 5
show that not only is the maximum strength affected by the particle shape,
but also the magnitude of the axial strain at failure. The failure strain
plotted against relative density in Fig. 6 shows both the influence of angu-
larity and densification on the failure strain. The more angular the par-
ticles, the greater the failure strain for a given relative density.While the
failure strain of the Olivine sand can be decreased from 13 percent at DR =
20 percent to 5.5 percent at DR = 90 percent, glass beads with near spheri-
cal particles show no visible alteration of the failure strain with density
changes. The sizable effect of the particle shape on the failure strain is
better illustrated in Fig. 7,where the failure strains of specimens at three
relative densities have been plotted against the coefficient of angularity.

12.0
LEGEND
.~DR = 40%
DR=70 %
10.0 DR:90%
(F
22
Z
LU 8.0
n.-
F-
(/3
~E

X 6.0
=E

t/i
Z

n~ 4.0
I---

41

2.0 /
(~ GLASSBEADS
[ ] OTTAWASAND
,6 <~ SOUTHPORTSAND
OLIVINE SAND
0 I I I I
1.0 1.2 1.4 1.6 1.8 2.0
ROUND ANGULAR
COEFFICIENT OF ANGULARITY, E

FIG. 7--Effect of particle shape on axial strain at maximum strength.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No f
HOLUBEC AND D'APPOLONIA ON PROPERTIES OF GRANULAR SOILS 313

4.0
(~) GLASS BEADS

5.0

o~
(D
Z
2.O
n-"
I--
03
._1

<~
I.O

I0 20 50

AXIAL STRESS, Y'I , T S F

FIG. 8--Stress-strain curves from constant diameter (Ko) tests.

5~I 0[ ] GLASS BEADS


OTTAWA SAND
/k SOUTHPORT SAND
OLIVINE SAND
4.0
NOTE : STRArN FOR STRESS
~T INCREMENT 0 - 1 0 TSF
' " - D R = 40'~
(D
I.-
5.0 " - - DR= 7 0 %
0
i! DR=90%

z
nr
2.0 .~0%
I.-
09

LO

0 I I I I
I.O 1.2 1.4 1.6 1.8 2.0
ROUND ANGULAR

COEFFICIENT OF ANGULARITY, E

FIG. 9--Effect of particle shape on strain observed in constant diameter (Ko) t~sts.
Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License A
314 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

The effect of the particle shape is shown by the line drawn through the
failure strain points for specimens at DR = 70 percent. This figure also
shows that the range of magnitudes of failure strain increases with the
angularity of the particles. This means that deformations of sands with
angular particles can be decreased considerably by densification. However,
little benefit is obtained by densification of sands with near spherical
particles.
Similar observations apply also to the stress-strain results from the con-
stant diameter (Ko) tests shown in Figs. 8 and 9. The stress-strain curves
in Fig. 8 illustrate the influence of particle shape on one-dimensional com-
pression tests. The magnitude of this influence and the effect of densifica-
tion for three relative densities are shown in Fig. 9 as a plot of strain at a
particular axial stress level against particle shape.

--"~-WEIGHT FALLING 6IN.


i (0.50 lb.)

~GUIDE

l/lllll/lllllJ),

W
I
0
Z

W
0
2

0
~ i
po
I
INI

z'o
14
5O
' N ' VALUES, BLOWS / INCH

FIG. l O--Modet penetration apparatus and typical N value data.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
HOLUBEC AND D'APPOLONIA ON PROPERTIES OF GRANULAR SOILS 315

Model Standard Penetration Tests


A miniature penetration test, as shown in Fig. 10, was used to investigate
qualitatively the effect of particle shape on resistance to dynamic penetra-
tion. The sands were compacted to selected densities in a cylindrical con-
tainer, and a ~/~-in. diameter rod was driven into the sand b y a falling
weight. The rod was driven to a depth of 8 in., and the number of blows
per inch during each penetration was recorded. A linearly increasing re-
sistance with depth was observed in all tests. The driving resistance plotted
against depth is shown in Fig. 10. The N values at the intersection of the
resistance curve with the 8-in. depth were selected for correlation. The
measured N values for the four materials tested at several densities are
presented in Fig. 11. The relative densities are based on minimum void
ratios obtained from Modified Proctor compaction tests.
The data show amarked increase in the penetration resistance with in-
crease of angularity of the particles. At a Da = 70 percent, the Southport
and Olivine sands with angular particles registered N values which are
nearly twice the values recorded for the subrounded Ottawa sand. Con-
versely, a penetration resistance of N = 20 indicates a spread in relative
densities from 66 to 86 percent. Since the particle shape influences both the

SO
0 GLASS BEADS
[] OTTAWA SAND
SOUTHPORT SAND //
40 OLIVINE SAND
"r
0
Z

03
~o 50

hi
.J
> 20
Z

I0

0 I I I
20 4O 6O 80 I00

RELATIVE DENSITY,DR,%
FIG. ll--Penetration resistance measured in model test.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
316 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

50

Lu SPECIMENS AT DR=70%
ILl
n~

E3
40

hi
J
Q
Z
<~

Z
g 50
F- (~) GLASS BEADS
u. [] OTTAWA SAND
A SOUTHPORTSAND
OLIVINE SAND
20 , I I I
0 IO 20 50 40 50

'N' VALUE, BLOWS / INCH

F I G . 12--Relation between angle of internal friction and penetration resistance from


model test.

resistance N and the friction angle, it can be postulated that the penetra-
tion values are also a measure of the shear strength of the sand. This is
confirmed by the correlation of N values and friction angles shown in Fig.
12 for sands at De = 70 percent.
Discussion
The test results show that the particle shape has a pronounced effect on
the engineering properties of granular soils. Furthermore, larger variations
in the engineering properties were observed between sands of different
particle shape at the same relative density, than the variations found with
relative density changes alone. Hence, the unqualified use of correlations
of engineering properties solely to relative density can be misleading.
The particle shape influences various methods of measuring the minimum
void ratio (maximum density). The best method for densifying a uniformly
graded cohesionless soil with near spherical particles is vibration. However,
this method is not appropriate for cohesionless soils with angular particles.
Angular particles can be forced into tighter arrangements by methods
employing a dynamic compactive effort. The Modified Proctor compaction
method generally produced maximum densities greater than those produced
by vibration for all sands, except the glass beads. Since there are no natural
sands with spherical particles, a method based on compaction should be
considered for the determination of maximum density.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License
HOLUBEC AND D'APPOLONIA ON PROPERTIESOF GRANULAR SOILS 3 T7

The particle shape has minor influence on the minimum void ratio, but
has a major influence on the maximum void ratio. Since the maximum void
ratio increases considerably more with angularity than does the minimum
void ratio, the void ratio difference diverges markedly with increase in
angularity. This divergence in void ratio difference with angularity should
be considered in an evaluation of engineering properties.
Experimental data show that the shear strength and deformability, as
given by the strain at failure and strain produced in one-dimensional com-
pression, increase with the increase in angularity of the particles. These
effects of particle shape are as great as differences caused by large changes
in relative density. The tests showed that the deformability of sands hav-
ing the same gradation and relative density increased with increasing
angularity. Also, strength, as measured by friction angle, increased with
increasing angularity. This suggests that the common assumption that
cohesionless soils with high shear strength are less deformable is not always
valid.
The tests performed with the model penetration test apparatus suggest
that the Standard Penetration Test (SPT) is also affected by particle
shape. Driving resistance at a particular relative density increases with the
angularity of the particles. The increase in the N values, which is similar
to the increase in friction angle, indicates that the SPT is affected by rela-
tive density and shear strength which is highly dependent on angularity.
Therefore, SPT correlations with relative density obtained for a particular
sand are not necessarily applicable to sands with different particle shape.
Because angularity influences both the SPT and the deformability of
sand, the error in using the SPT to predict settlement may be compounded.
For example, consider a correlation between SPT and settlement developed
for an "average" sand that is used to predict the settlement of an angular
sand. First, the SPT values for the angular sand indicate a higher relative
density because of the larger driving resistance caused by the greater angu-
larity of the particles. Settlement will be computed for a material denser
than actually exists. Secondly, the angular sands will deform considerably
more than the average sand for which the correlation was obtained. Ac-
cordingly, the settlement prediction using correlations developed for an
average sand would lead to an underestimate of the magnitude of settle-
ment for the angular sand.
Conclusion
The test data show that the particle shape has a significant effect on the
engineering properties of cohesionless soils, and it should be considered as
an index property in correlations of properties of granular soils. A standard
procedure to measure the particle shape should be adopted. For this pur-
pose, an indirect method based on either the porosity or permeability of

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License A
318 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

t h e sand is suggested because of its simplicity. A n y m e t h o d involving


m e a s u r e m e n t s of individual particles is too cumbersome.
T h e variation of the engineering properties due to particle shape can be
of the same order of m a g n i t u d e as the variation of the properties due to
changes in relative density. Therefore, the use of existing correlations of
relative density with engineering properties to predict soil behavior should
be u n d e r t a k e n with caution a n d with t h e full u n d e r s t a n d i n g of the assump-
tions and limitations of the current published correlations for granular
soils.

References
[1] Dietert, H. W., Brewster, F. S., and Graham, A. L., Transactions, American
Foundrymen's Society, Vol. 26, 1965, pp. 89-98.
[2] Haynes, J. H., "Effect of Repeated Loading on Gravel and Crushed Stone Base
Course Materials used in AASHO Road Test," Joint Highway Research Project,
N. 15, Purdue University, Lafayette, Ind., 1966.
[3] Dunlap, W. A., "Deformation Characteristics of Granular Materials Subjected to
Rapid Repetitive Loading," Ph.D. thesis, Texas A&M University, May 1966.
[4] Holubec, I., "Cyclic Creep of Granular Materials," DHO Report No. 147, Depart-
ment of Highways, Ontario, Canada. 1969.
[5] Koerner, R. M., "The Behavior of Cohesionless Soils Formed from Various Min-
erals," Ph.D. thesis, Duke University, 1968.
[6] Wadell, H., Journal of Geology, Vol. 43, 1935, pp. 205-280.
[7] Hofmann, F., Modern Castings, Vol. 35, 1959, pp. 125-128.
[8] Lucks, A. S., "The Influence of Particle Shape on the Strength of Granular Ma-
terials," Ph.D. thesis, Massachusetts Institute of Technology, June 1970.
[9] Holubec, I. and Finn, P. J., Canadian Geotechnical Journal, Vol. 6, 1969, pp. 353-
356.
[10] Peck, R. B., Hanson, W. E., and Thornburn, T. H., Foundation Engineering, Wiley,
New York, 1953.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
G. N . Durham~ and F. C. Townsend ~

Effect of Relative Density on the Liquefaction


Susceptibility of a Fine Sand under
Controlled-Stress Loading

R E F E R E N C E : Durham, G. N. and Townsend, F. C., " E f f e c t o f R e l a t i v e


Density on the Liquefaction Susceptibility of a Fine Sand under
C o n t r o l l e d - S t r e s s L o a d i n g , " Evaluation of Relative Density and Its Role in
Geotechnical Projects Involving Cohesionless Soils, ASTM S T P 523, American
Society for Testing and Materials, 1973, pp. 319-331.

A B S T R A C T : This investigation was an examination of factors affecting the


liquefaction susceptibility of a representative point bar sand from the Missis-
sippi River. Anisotropically consolidated triaxial specimens were sheared under
undrained conditions by controlled-stress loading under effective confiIfing pres-
sures ranging from 1.0 to 9.0 kg/cmL Three distinct stress-strain behaviors,
designated as liquefaction, limited liquefaction, and dilation, were observed.
The test results disclosed that a limiting relative density exists at which a
liquefaction or limited liquefaction would occur for a specified effective con-
fining pressure. As the magnitude of the effective confining pressure increased,
the relative density of specimens which liquefied also increased. These findings
permit an assessment of liquefaction susceptibility of a sand at a given density
and confining pressure. A critical void ratio curve was fitted to the test results
in which liquefaction occurred. A brief discussion is included on the effects of
relative density determinations by various laboratory methods.

K E Y W O R D S : cohesionless soils, sands, density (mass/volume), triaxial


tests, stress strain diagrams, liquefaction, tests

Liquefaction of loose saturated fine sands in point bar deposits has


caused numerous troublesome flow slides along the lower reaches of the
Mississippi River. Since this region of the United States is one of negligible
earthquake activity, earthquake-induced loadings, as studied by other
investigators [1, 2, 3]2, are not considered to have been responsible for

1 Engineer, Soils Test Section, and research engineer, Laboratory Research Section,
respectively, Embankment and Foundation Branch, Soils Division, U. S. Army En-
gineer Waterways Experiment Station, Vicksburg, Miss. 39180.
The italic numbers in brackets refer to the list of references appended to this paper.

319

Copyright by ASTM Int'l


Copyright9 by(all
ASTMrightsInternational
reserved); Fri Mar 11 16:13:06 EST 2016
www.astm.org
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
320 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

initiating these observed flow slides. Instead, these slides are believed to
be the result of liquefaction of sand layers of low density caused by un-
dermining of banks by scour during high flows of the river.
In investigating the problems involving riverbank stabilization and
levee and channel maintenance created by these flow slides, a laboratory
study was initiated to examine the factors controlling liquefaction and to
reproduce in the laboratory liquefaction of a saturated sand. Stress con-
trolled anisotropically consolidated undrained triaxial compression tests
(ACU~n) were performed on 1.4-in. and 2.8-in. diameter test specimens
using procedures similar to those of Castro [4] to analyze the effects of
relative density, deviator stress, major and minor principal stresses, and
pore pressure response on the liquefaction susceptibility of a representa-
tive point bar sand.

Procedure

Material
The sand used in the testing program was obtained from Reid Bedford
Bend, located south of Vicksburg, Miss. This material is a typical uniform
fine sand which has been loosely deposited in the upper layers of point
bars. Indications are that flow slides in point bar deposits are generally
limited in depth to these upper layers. The plus No. 16 U.S. Standard
Sieve fraction (representing 1 percent by weight of the total) was re-

U.S. STD. SIEVE N U M B E R S


IO 15 2 0 3 0 4 0 5 0 70 I00 200
104
I STANDARDCOIVIPACTIONTEST i
2:5 BLOWS PER EACH OF 3 LAYERS, WITH - -
5 . 5 - L B RAMMER AND 12-IN. DROP. 4-1N.-
8 DIAM MOLD.
F-
-r I02 4 j
!
uJ I

>. 6
no
,=,5 100
z
E
f_ 4
z
uJ
\
~3
uJ
7
Q.
2
98
/
96
GRAIN S I Z E , MM 0 5 I0 IS 20
SAND WATER CONTENT, PERCENT DRY WEIGHT
a MEDIUM FINE

FIG. 1--Gradation and density characteristics of Reid-Bedford sand.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
DURHAM AND TOWNSEND ON LIQUEFACTION SUSCEPTIBILITY 321

TABLE 1--Index properties of Reid Bedford sand.

Properties Values

Specific gravity of solids 2.65


Coefficient of uniformity Cu = D60/DIO 1.52
Curvature coefficient Cc = (D3o)~/Dlo X D6o 1.08
0.125 mm
Maximum density (Method A)
dry unit weight 108.2 lb/ft 3
void ratio 0.529
Minimum density (Method A)
dry unit weight 91.1 lb/ft s
void ratio 0.816
Optimum water content 16.5%
Maximum dry density 99.7 lb/ft 3
Consistency nonplastic
Unified Soil Classification Symbol Fine Sand (SP)

moved; this was the only alteration of the natural materiM. The particle
size-distribution curve and index properties of the sand are presented in
Fig. 1 and Table 1, respectively. Microscopic examinations revealed that
the grains were predominantly subrounded, with some rounded and
subangular grains present.

Equipment and Instrumentation


Test equipment (shown schematically in Fig. 2) was designed and
fabricated by Research Engineering of San Pablo, Calif., for purposes of
cyclic triaxial testing as developed by Seed and Fead [5]. However, the
cyclic capabilities of the device were not utilized, and only the controlled-
stress compression mode was needed. The device consists of a pneumatically
operated control unit which regulates the air supply to a double-acting
loading piston with rolling diaphragms (only one direction was used in
this program) and to the chamber pressures. The triaxial cell was equipped
with low-friction end platens made of 316 stainless steel, with the surfaces
highly polished to reduce friction. In the center of the contact bearing
surface of the platen is a flush-mounted porous sintered stainless steel
disk, about one-sixth the platen diameter, to provide for pore pressure
measurements or saturation. A thin layer of silicone grease was smeared
on the platens' bearing surfaces and a 0.012-in. thick annular rubber
membrane placed between the platen and soil to reduce further any friction.
Electrical transducers were used to measure axial load, pore pressure,
chamber pressure, and axial deformation, in order to record electronically
the rapid rate of deformation, reduction in axial load, and buildup of pore
pressure which oeeur during a liquefaction failure. The axial load cell
(Alleghany Instruments, 35-133-DDD) with a force range of 0 to 1000 lb
was used for testing the 1.4-in. speeimens, while a 0 to 500-1b or a 0 to

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to
322 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

CONTROL REGULATORS
FOR CHAMBER PRESSURE
BACK PRESSURE
FOR MONITORIN~
iNSTRU~ENTAT~ON

LOAD CELL~ LVDT 1


PRESSURE
LOADING PISTON
TRANSDUCER

\ /

IA
n ~
I

itLc~LOAD ,~--~LOADINGFRAME

LVDT
-TRIAXIALCHAMBER

~CHAMBs
PORE AND
PRESSURE
TRANSDUCERS
I
NOTE: LVDT IS A LINEAR VARIAEILE DIFFERENTIAL TRANSFORMER.

FIG. 2--Schematic of test equipment.

3000-1b load cell (fabricated by Research Engineering) was used for test-
ing of the 2.8-in. diameter specimens. Axial deformation was measured
by a Linear Variable Differential Transformer LVDT (Collins, SS-109)
with a range of • in. Pore pressures and chamber pressures were moni-
tored at the bottom of the specimen and chamber base, respectively, by
externally mounted 200-psi pressure transducers (BLH-Type GTCG).
The output signals of these transducers were recorded on a high-speed
recording oscillograph (CEC 5-124). In addition, three microvolt meters
were used to provide visual monitoring of the electrical output signals
and enabled the operators to adjust chamber pressures to the nearest 0.1
psi, monitor pore pressures to the nearest 0.1 psi, and apply axial load to
the nearest 1.0 lb. A more detailed description of the equipment, instru-
mentation, and calibration procedures is given by Durham [6].

Relative Density Test Procedure


Three test procedures were used for determining laboratory maximum
and minimum density values. Minimum density determinations for
Methods A and B (comparable to ASTM Test for Relative Density of
Cohesionless Soils (D2049-64T)) were made by pouring oven-dried ma-
terials into a 0.1-ft3 mold from a height of 1 in, through a 4-in. funnel with
a l~-in, diameter spout. For maximum density determinations by Method

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant
DURHAM AND TOWNSEND ON LIQUEFACTION SUSCEPTIBILITY 323

A, the mold plus sand is fitted with a 2-psi surcharge and vibrated at
60 Hz with an amplitude of 0.049 in. for 8 min on a Syntron Model VP-80
vibratory table. Using Method A, which was used for the relative density
values reported in this study, the maximum and minimum void ratios
were determined to be 0.816 (91.1 lb/ft 3) and 0.529 (108.2 lb/ft3), re-
spectively. The maximum density for Method B was determined using a
2-in. diameter by 4-in. high cylindrical mold and compacting in four 100-g
layers, using 25 blows per layer of a 4-1b rammer falling 12 in. Two tests
were performed, indicating an average minimum void ratio of 0.562 (105.9
lb/ft 3) as opposed to the average minimum void ratio of 0.529 (108.2
]b/ft 3) using a vibrating table. The minimum density using Method C was
determined by pouring oven-dried sand through a funnel with a 1/~-in.
diameter spout, to which was attached by strings 1/~ in. below the tip of
the funnel spout a horizontal piece of cardboard from which the sand
spilled into the 2-in. diameter by 4-in. high mold. The cardboard was kept
just above the surface of the sand in the mold and a spiral motion was
scribed by the funnel in order to keep the sand surface level. The correpond-
ing maximum density in Method C was determined using the same mold
by hammering the sides after placing a 3.54-1b surcharge plate on each
of the three equal layers. Method C produced maximum and minimum
void ratios of 0.842 (89.8 lb/ft 3) and 0.589 (104.1 lb/fta), respectively.
The three laboratory procedures for determining relative densities as
described have been used by other investigators and reported elsewhere

VOID RATIO eu= N


0.5 0.8 0,7 0.8 0.85
IO0

F- 80
Z
w

-~ 8o
Q

Z
g 4o
LLI
>

-J
UA
a: 20

0 I
0.5 0.6 0.7 0.8 0.85
VOID RATIO eaA x

FIG. 3--Comparison of maximum and minimum void ratios.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
324 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

[4, 7, 8]. The purpose of reporting these values is to illustrate the wide
range of relative density values which can be obtained on the same ma-
terial. Figure 3 is a graphical comparison of the results of these three meth-
ods. One should remember that the values of maximum and minimum
void ratios do not necessarily represent the maximum and minimum void
ratios at which a material can exist in nature, but instead reflect the
laboratory procedures used.

Triaxial Test Procedure


In order to obtain the low densities required to create liquefaction, a
molding water content of 5 to 7 percent, which corresponds to the bulking
water content (Fig. lb), was added to the sand prior to compaction. ,The
triaxial test specimens were molded with the aid of a pneumatic hand
tamper [9] which utilizes air pressure reacting against a rolling diaphragm
(instead of a spring used in the Harvard miniature compactor) to provide
a kneading force. The tamping forces used for compaction varied from
2.5 to 15 lb depending upon the size of the specimen and the desired
density. The combination of using low kneading forces and sufficient
moisture to bulk the sand accounted for the low molding densities. Speci-
mens were compacted in ten layers, applying 20 to 25 tamps per layer and
scarifying the surface of each layer before placing the next layer. The
density after compaction was determined after removing the mold by
directly measuring the volume of the specimen which was supported by
a vacuum of 0.7 kg/cm 2 (10 psi) and using the difference in dry weight of
material before and after compaction.
The following de-airing and saturation procedure was found to be most
effective in obtaining complete saturation. After measurement of the
specimen volume, the triaxial chamber was assembled and a chamber
pressure of 1 kg/cm ~ was applied while the drainage lines remained closed.
De-aired water was then allowed to flow into the specimen from the bot-
tom porous disk until the internal vacuum was reduced to zero. A small
amount of water was then forced through the specimen under a small
gradient by opening the top drainage line. The bottom line was then
closed, and a vacuum reapplied to the top platen. This procedure was
continued until no air bubbles were observed in the top drainage line
when the vaccum was applied. Back-pressure saturation techniques, con-
sisting of simultaneously increasing the chamber pressure and back pres-
sure in 0.7 kg/cm 2 (10 psi) increments were used to further saturate the
specimen. Complete saturation was assumed when Skempton's "B"
value (~u/A~3) was 0.96 or greater. Generally, 6.0 to 7.0 kg/cm ~ of back
pressure were required to reach this condition. Volume changes caused
by these de-airing and saturation procedures were calculated from height
changes measured with a 0.001-in. dial gage mounted to the axial loading
piston.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lic
DURHAM AND TOWNSEND O N LIQUEFACTION SUSCEPTIBILITY 325

In an effort to duplicate field conditions, the specimens were incre-


mentally consolidated anistropically under a major principal stress (alc)
and a minor principal stress (a3~). The test specimens were subjected
simultaneously to an effective confining pressure and a deviator stress
(zd~ = ~ - ~3~) such that a stress ratio of 51~/53c = 2.0 was achieved.
Volume changes during consolidation were calculated from observed
burette readings of the pore fluid.
After consolidation under the desired confining pressure was completed,
the drainage values were closed and incremental axial loads applied at
1-min intervals. The magnitude of axial load increments varied from 2 to
20 lb, depending upon the anticipated maximum deviator stress and
specimen size.

Test Results
Table 2 summarizes the results obtained for the anisotropie consolidated
tests. Columns 2 and 3 list the consolidation stresses and anisotropic
stress ratios, respectively. Column 4 lists void ratios after consolidation
which were determined from height and volume measurements during
3 0"3c-2 43 KG./CM2"t s W- . . . . _~- o-o,

0 2 ~ ~ ~ 3 + , = 0 . 0 5 KG/CMZ Z [~-c=2.44KG/CM,2/
- .... ~,r ~'G/C~ .

0 L I I I I h I
0 2 4 6 8 iO 12 0 2 4 6 8 iO 12
AXIAL STRAINj PERCENT
c,. L I Q U E F A C T I O N b. L I M I T E D LIQUEFACTION
TEST 2-2.5-(I) TEST 2-2.5-8

~3 / G'3C=2.44 KG/CM 2

~2
~r KG/CM z

O(
12 14 ~6 18 20 2'2
AXIAL STRAIN, PERCENT
c. D I L A T I O N
TEST 2 - 2 . 5 - 1 4

2 G[=27.70

b,~ ~ . .

0 I 2 3 4 5 6
~,r
2 i KG/CM z

cl. S T R E S S P A T H S

FIG. 4--Typical stress-strain diagrams and stress paths.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further
r
bO
O~

TABLE 2--Summary of A C U triaxial tests.

Test No. ~ a~, k g / c m 2 alc/q~c er b Type c At Maximum Deviator Stress ~/~3 ~ ! d, k g / c m ~ a~I d, k g / c m 2
Failure
ad(max), k g / c m 2 E
-4
-r
2.8-in. Diameter Specimens
2-1-6 0.98 1.99 0.799 LL 1.77 7.1 3.31 0.71 2.36
2-1-10 0.98 1.99 0.786 D 4.53 22.5 3.22 1.42 4.62 o
t-
2-1.5-8 1.46 1.97 0. 792 LL 1.73 0.5 2.58 0.35 0.99
2-2.5-(1) 2.43 1.96 0.819 L 2.66 0.4 2.58 0.05 0.19 Y:
o
2-2.5-8 2.44 2.00 0. 793 LL 2.94 0.5 2.70 1.10 3.50 t~
2-2.5-10 2.44 1.99 0. 789 LL 2.98 2.7 2.96 1.19 3.59 o
2-2.5-14 2.44 2.00 0. 775 D 5.06 20.2 3.17 2.02 6.55
1.~-in. Diameter Specimens o
z
2-3-17 2.95 1.98 0. 768 D 5.95 13.05 3.35 2.34 8.03 7~
o~
2-4-4 3.94 1.99 0. 805 L 4.63 1.14 3.01 0.87 2.84
2-6-14 5.99 1.98 0. 777 D 8.50 12.73 3.28 3.72 12.23
_o
2-6-14 6.00 1.99 0. 775 D 8.19 10.96 3.36 3.47 11.64
2-6-19 6.0O 1.97 0. 760 D 11.68 7.23 3.29 5.10 16.75
2-9-14 8.98 1.97 0.778 LL 10.65 0.63 2.64 4.21 13.89
2-9-18 8.96 1.98 0. 764 D 12.29 11.25 3.34 5.26 17.55
2-9-18 8.99 1.98 0. 765 D 12.83 10.91 3.59 4.96 17.78

a First number is ~1~/~3c consolidation ratio, second is a3~ consolidation pressure, and third is relative density (percent).
b ec = computed from volume changes during consolidation.
c L = liquefaction failure, L L = limited liquefaction, D = dilative.
Stresses a t minimum ad for L and L L failures and at maximum ~1/h3 for D type failure.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
DURHAM AND TOWNSEND ON LIQUEFACTION SUSCEPTIBILITY 327

consolidation. Column 5 indicates the type of behavior observed during


shear; L denotes liquefaction failure, L L limited liquefaction, and D
dilative response. Columns 6 through 8 present the test' conditions at
maximum deviator stress. Columns 9 and 10 are the stresses at the mini-
mum deviator stress for liquefaction (L) and limited liquefaction (LL)
failures and at ~1/~8, and at maximum ~1/~3 for dilative responses (D).
Deviator stress and induced pore water stress-strain curves and stress-
path curves for three anisotropically consolidated ACU tests are illus-
trated in Fig. 4. The three test specimens were prepared at three different
initial densities and consolidated under effective confining pressure of 2
kg/cm ~ and an effective principal stress ratio of 2.

Liquefaction Failure
Figure 4a shows the results of test 2-2.5-(1) in which complete liquefaction
occurred. The plot of the deviator stress versus axial strain shows that the
specimen developed its peak strength at an axial strain of about 0.4 per-
cent. This peak strength was obtained in about 6 min by increasing the
axial load, during which time the pore pressures increased to about 30
percent of the effective confining pressure. The pore pressure began to
increase after each load increment, but became stable after about 20 or
25 s. With the application of the last load increment, the rate of deforma-
tion began to increase and the rise in pore pressure became more pro-

FIG. 5--Photographic sequence of liquefaction failure similar to tests ~-~.5-(1).

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
328 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

nounced. The test specimen then suddenly failed by liquefaction and


deformed to about 11 percent axial strain in a period of 1 s. At this point,
the specimen lost contact with the loading piston and measurement of
further axial strain was not possible. The induced pore pressure increased
rapidly during loading to an axial strain of about 4 percent and, following
a slight decrease, slowly increased to a final value of 2.38 kg/cm ~, which
resulted in a final effective minor principal stress of 0.05 kg/cm 2, desig-
nated as ~3f. As liquefaction occurred, the test was no longer stress con-
trolled, rather, a reduction in deviator stress occurred because the loading
ram could not maintain a constant load due to the weakness of the speci-
men. The measurement of the sharp decrease in deviator stress to a final
value of 0.14 kg/cm 2 was only possible by using an electronic load cell and
high speed recording equipment. Figure 5 shows a photographic sequence
of a liquefaction failure similar to test 2-2.5-(1). The time of zero second
corresponds to the end of the previous load increment just prior to addition
of the load increment which caused liquefaction. The rapid deformation
and complete collapse of the specimen upon liquefaction is evident.

Partial Liquefaction Failure


Figure 4b shows results of test 2-2.8-8 in which limited liquefaction oc-
curred. The deviator stress increased with incremental loading up to about
0.4 percent axial strain, and correspondingly, the induced pore pressures
increased to 27 percent of the effective confining pressure. Axial strain
progressed from 0.4 to 11.4 percent in 11 s, and the induced pore pressures
became essentially constant after about 9 percent strain, and equivalent
to about 55 percent of the effective confining pressure. The failure be-
havior can best be described as intermediate between that observed
during liquefaction and dilative responses. However, it is interesting to
note that the stress vector path (Fig. 4d) moved downward along the
failure envelope in a manner analogous to a liquefaction response.

Dilative Response
Figure 4c shows the results from test 2-2.5-14 in which a dilative response
was observed. It can be seen that the pore pressure reached a maximum at
about 1.0 percent strain and then decreased with further straining, and
the deviator stress continued to increase in strength over the entire range
of deformation. The stress vector path (Fig. 4d) approached the failure
line and then moved upward along the failure envelope.

D i s c u s s i o n o f Results
In Figure 6 the relationships between void ratio after consolidation (ec)
and effective confining pressure (~3c) are shown. The L line delineates the
limiting relative density with increasing consolidation pressure for which
a liquefaction failure will occur, while the LL line delineates the limiting

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to L
DURHAM AND TOWNSEND O N LIQUEFACTION SUSCEPTIBILITY 329

relative density between limited liquefaction and dilative responses. These


lines indicate the relative density and corresponding confining pressure
after consolidation, which will lead to a liquefaction response when the
deviator stress is increased. Relative densities below the L L line will ex-
hibit a dilative response (will not liquefy) with an increase in deviator
stress. These relationships represent an important relationship, and indi-
cate that liquefaction can occur at a higher relative density by Increasing
the effective consolidation pressure. Conversely, other investigators [1, 2]
have reported that under cyclic loading conditions, lower chamber pres-
sures favor liquefaction. Solid data points in Fig. 6 are for 2.8-in. diameter
specimens, and open symbols are the results of 1.4-in. diameter specimens.
It is apparent that results of the two specimen sizes are compatible, and
it appears that a reduction in specimen diameter has negligible effects on
these relationships. It is interesting to note that limited liquefaction was
achieved only at relative densities less than 15 percent (based on maxi-
mum-minimum value from Method A). These values are somewhat lower
than the results of the ACU triaxial tests by Castro [4] on similar sands.
The difference is attributed to the difference in procedures used for de-
termining maximum and minimum densities.
Figure 7 presents the relationship between the effective confining pres-
sure at failure (5al) and the void ratio after consolidation (ec). The el line
has been referred to as the critical void ratio curve for cases in which
0.84
LEGEN6 I
2.0-1N. 1.4-1N.
Q O LIQUEFACTION
II O LIMITED LIQUEFACTION
A A DILATION
0.82
# *
0
z
9
c~

'~ o.ao i" ~


z ~---L u~
u w

LL LINE ~ " >~


o.78 ~,, ~- -~
o ~

& _>
9
..J
0.70 Lu
20

0.74 25
2 4 O 8 10
EFFECTIVE C O N F I N I N G PRESSURE
e3C , KG/CM 2

F I G . 6--Effect of confining pressures (e3) and void ratio (ec) on type of failure.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreeme
330 RELATIVE DENSITY I N V O L V I N G COHESIONLESS SOILS

CONSOLIDATION PRESSURE
~3c, K G / C M z
I 3 5 IO
0.84

Z 0.82
0 0
I-
p-
o
_7
0
e'u~'>E~ . o -o Z
(2: uJ
0.80
Z b_ bJ
0 <o.
0
1o~_
I- 0
0.78
LEGEND "4} ~,~ ~
0 2.8-IN. 1.4-1N, ~ -
t-
0-0 LIQUEFACTION
LIMITED LIQUEFACTION
o 0.76 - -
5 I 0
U

25
0.74
O,OI 0.03 0.05 0.1 0.3 0,5 I.O 3 5 10
EFFECTIVE MINOR PRINCIPAL STRESS AT FAILURE (~3.F, KG/CMz

FIG. 7--Critical void ratio line from liquefaction tests.

liquefaction developed, but not for cases in which a dilative behavior


existed [4]. Castro [4] found t h a t this critical void ratio line is different
than that obtained from drained triaxial tests; hence, it might be more
appropriate to refer to this ei line as the liquefaction void ratio curve.
The dashed lines indicate the manner in which the effective minor principal
stress will decrease continuously to a value designated b y the eI line for
cases in which a liquefaction failure occurred. The distance from the es
curve to a plotted point representing an in situ stress condition and relative
density above the L L line is a measure of the liquefaction susceptibility of
a given sand.

Conclusions
The following conclusions can be made based upon the material and
testing procedures employed in this study:
(a) The test results disclosed that a limiting relative density exists at
which liquefaction or limited liquefaction failure will occur for a specified
effective confining pressure. This relationship permits an assessment of
the liquefaction susceptibility of a sand at a given in situ confining pressure
and relative density. Generally, low relative densities (less than 15 per-
cent) were required to produce liquefaction.
(b) As the magnitude of the effective confining pressure increased, the
relative densities of specimens which liquefied also increased. For very

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursua
DURHAM AND TOWNSEND ON LIQUEFACTION SUSCEPTIBILITY 331

loose test specimens, liquefaction occurred at axial strains of less than 3


percent and for anisotropically consolidated specimens only small in-
creases in deviator stress were required to produce these strains. This sug-
gests t h a t only small changes in field conditions are required to trigger a
liquefaction flow slide.
(c) A critical void ratio curve (el), as suggested b y Castro, can be fitted
to the test results in which liquefaction occurred.
(d) Tests on both 1.4-in. and 2.8-in. diameter specimens gave similar
results.
(e) The stated value of relative density for liquefaction susceptibility
is a function of the methods used in determining maximum and minimum
densities. Void ratios or dry density values should accompany such re-
sults for consideration when comparing various data.

Acknowledgments

The tests described and the resulting data presented herein, unless
otherwise noted, were obtained from research conducted at the U.S. Army
Engineer Waterways Experiment Station under sponsorship of the Lower
Mississippi Valley Division, U.S. Army Corps of Engineers. Permission
was granted b y the Chief of Engineers to publish this information.
We wish to express our special thanks to W. E. Strohm, Jr., chief,
Engineering Studies Section, Soils and Pavements Laboratory, for his
assistance during the investigation. Assisting in the laboratory program
were T. V. McEwen and F. G. A. Hess.

References
[1] Seed, H. B. and Lee, K. L., Journal of Soil Mechanics and Foundations Division,
American Society of Civil Engineers, No. SM6, Nov. 1966.
[2] Lee, K. L. and Seed, H. B., Journal of Soil Mechanics and Foundations Division,
American Society of Civil Engineers, Vol. 93, No. SM1, Jan. 1967.
[3] Finn, W. D. Liam, Pickering, D. J., and Bransky, P. L., Journal of Soil Mechanics
and Foundations Division, American Society of Civil Engineers, Vol. 97, No. SM4,
April 1971.
[4] Castro, G., "Liquefaction of Sands," Harvard Soil Mechanics Series No. 81, Jan. 1969.
[5] Seed, H. B. and Feed, J. W. N., "Apparatus for Repeated Load Tests on Soils,"
Papers on Soils--1959 Meetings, A S T M STP 254, American Society for Testing and
Materials, 1959.
[6] Durham, G. N., "A Study of the Liquefaction Phenomena of a Fine Sand Utilizing
the Consolidated-Undrained Triaxial Compression Test Under Controlled Stress
Loading," unpublished Master's thesis, Mississippi State University, State College,
Miss., Aug. 1971.
[7] Laboratory Soils Testing, Engineer Manual (EM) 1110-2-1906, U. S. Army Engineer
Waterways Experiment Station, Vicksburg, Miss., Nov. 1970.
[8] "Density Changes of Sand Caused by Sampling and Testing," Potamology Investiga-
tions Report No. 12-1, U. S. Army Engineer Waterways Experiment Station, Vicks-
burg, Miss., June 1952.
[9] MacIver, B. N. and Donaghe, R. T., "Modified Berkeley Pneumatic Tamper for
Compacting Test Specimens of Cohesive Soils," MP 3-478, U. S. Army Engineer
Waterways Experiment Station, Vicksburg, Miss. June 1971.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No fu
M. M. A1-Hussaini ~

Influence of Relative Density on the Strength


and Deformation of Sand under Plane
Strain Conditions

R E F E R E N C E : A1-Hussaini, M. M., " I n f l u e n c e o f R e l a t i v e D e n s i t y o n


the Strength and Deformation of Sand under Plane Strain Conditions,"
Evaluation of Relative Density and Its Role in Geotechnical Projects Involvin 9
Cohesionless Soils, A S T M S T P 523, American Society for Testing and Ma-
terials, 1973, pp. 332-347.

A B S T R A C T : Plane strain and triaxial compression tests were conducted on


saturated Chattahoochee River sand to demonstrate the effect of relative
density and strain conditions on the strength and stress-strain characteristics
of sand. Specimens were prepared at relative densities ranging from 30 to 100
percent and consolidated isotropically to an effective confining pressure of 70
psi before shear.
On the basis of experimental data presented in this study, it appears that
the increase in relative density of the sand tested decreases its compressibility
during consolidation and increases the initial slope of the stress-strain curve, the
tendency for volumetric expansion in drained shear, and the tendency for nega-
tive pore pressure in undrained shear for both plane strain and triaxiai com-
pression tests. The data also indicated that the angle of internal friction for
the tested sand was higher, the volumetric strain during shear was more com-
pressional, and the axial strain at failure was lower in plane strain tests than
in triaxial compression tests of sand with nearly the same relative density.

K E Y W O R D S : cohesionless soils, soil tests, shear tests, density (mass/volume),


sands, soil properties, compression tests, stress strain diagrams

Among the many variables that affect the strength and deformation
characteristics of soils are the degree of compaction and the strain condi-
tions during shear. The influence of the degree of the compaction, which for
cohesionless soil is defined by the relative density, was recognized by early
work of Casagrande [i],2 who showed the full dependency of the angle of
t Research civil engineer, Laboratory Research Section, Embankment and Foundation
Branch, Soils and Pavements Laboratory, U. S. Army Engineer Waterways Experiment
Station, Vicksburg, Miss. 39180.
2 The italic numbers in brackets refer to the list of references appended to this paper.

332

Copyright by ASTM Int'l


Copyright9 by(all
ASTMrightsInternational
reserved); Fri Mar 11 16:13:06 EST 2016
www.astm.org
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
AL-HUSSAINI ON SAND UNDER PLANE STRAIN CONDITIONS 333

internal friction of sand on the void ratio. However, more recently, the
irdiuence of strain conditions, especially the axially symmetric case associ-
ated with the conventional triaxia! test and the plane strain condition that
simulates many shear deformations in the field, has been recognized, but
has received little attention.
To date most relative density studies with regard to strength and stress-
strain characteristics of cohesionless soils have been conducted under
axially symmetric stress conditions. However, the deformations associated
with many field situations, such as deformation of embankments, strip
footings, retaining walls, and dams, are closer to plane strain problems.
This lack of consistency may be attributed to the limited number of labor-
atory apparatus that can impose plane strain conditions on soils during
shear, to the difficulties involved in performing plane strain tests in com-
parison to conventional triaxial tests, and to the assumption that triaxial
tests are slightly conse~Tative. Consequently, the influence of relative
density on the engineering properties of granular soils under plane strain
conditions is not well defined.
The purpose of this study was to provide additional data to show the
variation of strength and deformation characteristics of sand due to the
combined effect of relative density and strain conditions during shear.
Properties of Testing Material
The sand used in the testing program was obtained from a site along the
Chattahoochee River near Atlanta, Ga. It consists of uniform subangular

SAND
- - FINE
I00

80

60 I
ii
BEFORE T E S T - ~ . ~ ,~----AFTER DRAINED SHEARI I r

40

20

% I 0 I
PARTICLE SIZE,, MM

FIG. 1--Gradation curve for Chattahoochee sand.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
334 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

quartz particles and some mica that is mostly muscovite. The physical
properties of the sand as determined by ASTM standard test procedures
are given below; the average grain-size distribution curve before testing
and after drained shear is shown in Fig. 1.
Specific gravity (ASTM Test for Specific Gravity of Soils
(D 854-58)) 2.66
Maximum void ratio (ASTM Test for Relative Density of
Cohesionless Soils (D 2049-69)) 1.09
Minimum void ratio (ASTM, D 2049-69) 0.59
Coefficient of uniformity (ASTM Definition of Terms and
Symbols Relating to Soil, and Rock Mechanics (D 653-67)) 2.00
Mica content 0.01%
Unified soil classification SP

Testing Equipment
Two shear testing devices were used in the study: a conventional triaxial
apparatus and a plane strain shear device. The triaxial apparatus was
similar in many respects to that described by Bishop and Henkel [2] and
was used to test a cylindrical specimen 2.8 in. in diameter and 6 in. high.
The plane strain apparatus used in the test (see Fig. 2) has been described
in detail elsewhere [3]; and the description presented herein is limited to
the most important features of the apparatus.
The plane strain apparatus is designed to test a prismatic specimen 16
in. long and 2 in. wide with a height of 4 to 41/~ in. The soil specimen is
encased in a rubber membrane and is bounded at the top and bottom by
two stainless steel platens. The upper platen is connected to the loading
cap, while the lower platen is attached to the base of the apparatus.
Plane strain shear is accomplished by increasing the axial load to the
specimen through a loading ram while the length of the specimen is held
constant by means of the longitudinal loading system. The longitudinal
loading system, which is responsible for maintaining plane strain condi-
tions, consists of a hydraulic jack for applying the intermediate principal
stress a2 on one end and a load cell for measuring the stress on the opposite
end. The hydraulic jack and the load cell are attached to two identical
steel end plates connected by four tie rods to form a cage around the speci-
men. In the longitudinal direction, the soil is confined between two highly
polished stainless steel plates which are joined together by two frames in-
strumented with strain gages and called the longitudinal strain sensor. The
whole assembly surrounding the specimen is placed inside a cylindrical
pressure chamber capable of applying confining pressure up to 5000 psi.
The hydraulic system, as shown schematically in Fig. 3, consists of three
separate pressure control systems. The first is the system to fill the com-
pression chamber with fluid and apply the desired confining pressure. The

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lic
I CASE*HARDENED STEEL RAM
2 TENSION CONNECTOR
3 LOADING CAP
4 LATERAL STRAIN S~NSOR
5 LONGITUDINAL STRAIN SENSOR
6 SPECIMEN
7 ADJUSTABLE CLAMp
S MOVABLE PLATE
9 TIE RODS
10 SADDLE PLATE
II SIJDING PLATE
12 O-2 ELECTRIC LOAD CELL
13 FRONT AND SACK END PLATES
14 PORE PRESSURE LEAD -r
15 LATERAL SENSOR LEAD c
~S LONGITUOfNAL SENSOR LEAOS
20 G"z JACK

r/"T1
o
Z
r

Z
r
9
i . S Z

E
r162
~T| I s _~ s 12 I 13 z

g I o
Z
,SIDE VIEW

SCALE IN INCHES
Z
i 0 i 2 3 4

Co
FIG. 2--General features of the plane strain apparatus, t~

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
p~ssuat a~a~
r
AIR aa~SSUR~S U P P L r - - - - ~ ~ O~
aa~ssua~ aecu~Aroa~

"=4
*<
Y.
<
0e=
<
0
0, - r

0
z
~-vAcuu~ CA~E 7n
o~
o
7~
VACUUMPUMP

VOLUMF- CHANGE
MEASURING SYSTEM

CHAMBs PRESSURE SUPPLY SYSTEM

FIG. 3--Schematic drawing of hydraulic systems.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
AL-HI.ISSAINI O N SAND UNDER PLANE STRAIN CONDITIONS 337

second system operates the hydraulic jack inside the chamber for applying
the intermediate principal stress. The third system includes the volume
measuring devices for the drained test and a pore pressure transducer that
can be attached to permit measurement of pore water pressure.
Preparation of Soil Specimens
Due to difficulties in preparing a uniform specimen of predetermined
density, major attention was given to obtaining a uniform specimen with
either a dense, medium dense, or loose packing. All specimens were pre-
pared under water after the required amount of sand was boiled for about
5 min and cooled to room temperature. Dense specimens were prepared by
spreading the sand evenly in layers of about 1 in. thick and vigorously
vibrating each layer with a small mechanical vibrator. Specimens of me-
dium density were prepared by rodding with a 1/~-iu. diameter rod. Loose
specimens were prepared by allowing the sand to drop through water
evenly inside the rubber membrane. This procedure was found to be effec-
tive in providing uniform sand specimens with relative densities varying
from 30 to 100 percent.
After placing the sand in the mold, the top surface was leveled, the upper
platen placed in its proper position, and the rubber membrane sealed with
the loading cap. The soil specimen was then connected to a volume measur-
ing device and deaired water was circulated through the system. A vacuum
of 2.0 psi was next applied to the specimen to make the specimen self-

70

Dr = 9 5 %

/.
~L 60

b
bJ 50

(,3
bJ
Q.

z
40 O--E

/
z 3 0 - -

z
0
M
w 20

W
w
Y
f
O
0
P L A N E S T R A I N TEST
TRIAXIAL COMPRESSION TEST
o I I I I
o 2 3
VOLUMETRIC STRAIN 6 V / V c ~

FIG. 4--Variation of volume change of Chattahoochee River sand under hydrostatic


pressure.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreemen
338 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

_v
I
F- I
I
~2 N i
>o I
w

D PLANE STRAIN
o.
O TRIAXlAL COMPRESSION
0":~ = 7 0 PSI I
0 "l I I J
30 40 50 60 70 80 90 100
INITIAL RELATIVE D E N S I T Y DF, ~

FIG. 5--Relationship between volumetric strain at the end of consolidation and the initial
relative density.

supporting; the specimen former was then removed, and measurements of


specimen dimensions were taken. Finally, the longitudinal loading system
was secured along the specimen, and the soil specimen and surrounding
equipment were placed inside the pressure chamber.

Compressibility of Sand
Before consolidating each specimen, the degree of saturation was checked
by determining the pore pressure parameter (B) and back pressure was
then applied to bring the sand to saturation. All plane strain and triaxial
specimens were consolidated isotropically by increasing the effective con-
fining pressure incrementally from 2 to 70 psi; the corresponding volume
change was recorded when the level of water in the burette of the volume
change derive reached equilibrium.
Typical stress-strain curves during isotropic consolidation for Chatta-
hoochee River sand are shown in Fig. 4. Squares refer to results of the plane
strain tests (prismatic specimens), and circles refer to results of the triaxial
compression tests (cylindrical specimens). It is apparent that there is no
difference in the compressibility of the prismatic and cylindrical specimens
tested under hydrostatic stress conditions. The relationship between the
volumetric strain at the end of consolidation and the initial relative density
(D,) for all specimens tested is shown in Fig. 5. This figure indicates that
the volumetric strain decreased at a decreasing rate with increasing initial
relative density.
Drained Shear Test
At the end of consolidation, the cell pressure was held constant at ~'3
equal to 70 psi and the stress difference (al - a3) was increased at an axial
deformation rate of 0.003 in. per min. Plane strain conditions were main-

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to L
AL-HUSSAINI ON SAND UNDER PLANE STRAIN CONDITIONS 339

rained during shear by adjusting the intermediate principal stress (a2)


such that the length of the specimen was kept constant until failure oc-
curred. The soil specimen was considered to have failed when the differ-
ence between the major and minor principal stresses reached a maximum
value. Continuous recordings of stress, strain, and volume change were ob-
tained throughout the test. Triaxial compression tests were conducted in
the standard manner under testing conditions comparable to those of plane
strain tests.
The principal stress difference (al - a3) and (as -- ~) for typical plane
strain specimens are plotted as functions of the axial strain (el) in Fig. 6.
The figure shows that there is a similarity in the variation of ~'1 and ~'2
with respect to el in that both increased to reach maximum value at about
the same time and both decreased beyond failure. However, the ratio of
r which is represented for convenience by ~'~/(~'i ~'~), was not - -

constant throughout plane strain shear. This variation is illustrated in Fig.


7 which shows that the ratio a'3/(a'l Jr a'a) defined two regions: a transi-
tional region in which a'~/(a'l + ~'~) dropped sharply from a value equal
to 0.5, representing the hydrostatic state of stress to reach state of perfect

280

240 I/,.~ Dr=88%

-'
200
/
LU 1 6 0

.
ul
j
~. 120

,/
7, 0"3 = 70 PSI

; ~o I /
Y
4O

O0 2 4 6 8 IO 12 14 16

8O

0 0- 2
S 4 6 8 10 12 t6
AXIAL STRAIN EI~

FIG. 6--Typical stress-strain curves of plane strain tests.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to L
340 RELATIVE DENSITY I N V O L V I N G COHESIONLESS SOILS

0.8

.t-
"~0.6

0 0.4 L O O S E Or : 4 2 %
I
Q: DENSE Dr : 88 ~

~W 0.2
I1:
(T31= 70 PSI
m
0
0 2 4 6 8 10 12 14
A X I A L STRAIN E l , %

FIG. 7--Relationship between ~2'/~1' -t- aa' and axial strain during plane strain shear.

plane strain condition at an axial strain of 2 percent, and a second region,


which extended from 2 percent axial strain until failure in which the ratio
a's/(a'l ~- a'3) remained practically constant and perfect plane strain con-
dition was maintained. The constant value of ar2/(a'l -Jr a'3) is equivalent
to Poisson's ratio (~) for isotropic material and, as depicted in Fig. 7, it is
higher for loose sand than dense sand.
Stress-Strain Characteristics During Shear
The stress-strain plots for specimens sheared at ~'3 of 70 psi (Figs. 6
and 8), reveal several interesting aspects regarding the influence of relative
density and strain conditions on the stress-strain characteristics of the
sand tested. It is clearly evident that the increase in the initial relative
density increased the initial slope of the stress difference (al - a3) versus
the axial strain (~), and also increased the strength of the material. How-
ever, the axial strain at failure was found to decrease with increasing rela-
tive density for both plane strain and triaxial compression tests. The axial
strain at failure for all specimens tested, plotted in Fig. 9, indicates that
sand under triaxial compression exhibited slightly larger axial strain at
failure than under plane strain shear. The difference is obscured by the
scatter of triaxial test data; however, the average difference for the same
initial relative density did not exceed 1.5 percent.
The effect of the initial relative density on the volumetric behavior of
sand (Fig. 10), indicates that both plane strain and triaxial compression
specimens exhibited compressional volumetric strain at the early stages of
shear. However, specimens with low relative densities continued to com-
press, while those with high relative densities expanded with increasing
axial strain. Specimens with intermediate relative density indicated inter-
mediate volumetric strain during shear. The figure also shows that plane
strain specimens exhibited more compressional volumetric strain than those

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lice
AL-HUSSAINI ON SAND UNDER PLANE STRAIN CONDITIONS 341

280

2 4C
Dr =.~_._
f
~ 200
I

~j 160
/, 36%
U
Z
w J

4o/j
CE
LL 120 /
J
LL

~,
w 80
o:
0"3r = 701 PSI

0(~ 2 .4 6 8 10 12 14 16
AXIAL STRAIN El, 96

F I G . 8--Typical stress-strain curves for triaxial compression tests.

tested under triaxial compression, which agrees qualitatively with the Finn,
Wade, and Lee [4] theory on spheres with regular packing.
Early studies by Casagrande [1, 5] demonstrated the interrelation be-
tween the volumetric strain and the effective angle of internal friction (@')
of sand and its initial compaction. He showed that dense sand expands

16
C T 3 : 7 0 PSI

9 14-0 \

I- 8
O
_1
<
x
< 6

4
30 40 50 60 70 80 gO I O0
INITIAL RELATIVE DENSITY Dr,%

F I G . 9--Failure strain relative density curve far Chattahoochee River sand.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
342 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

I
o f
J
J

~R - 3 42%

<1 - 4 0 2 4 6 8 I0 12 14 16
Z
<C o.. P L A N E STRAIN TESTS
0c+2

_U
jJ
oO~u f
J 9 7__.~%
7

J
J
j J

-2.

-3

~ 3 ' = 7 0 PSI
t
-4
o 2 6 8 I0 12 14
AXIAL S T R A I N El, ~,
b. T R I A X I A L COMPRESSION TESTS

FIG. lO--Typical volume change characteristics of sand under triaxial compression and
plane strain shear.

during shear and exhibits low strength. Casagrande also demonstrated


that there is a certain void ratio that corresponds to a state of density at
which the volume of sand remains constant during shear; this void ratio
was called critical void ratio.
The relationship between the void ratio at the end of consolidation (ec)
and the volumetric strain at failure is depicted in Fig. 11. This figure shows
that the sand sheared under plane strain conditions exhibited a slightly
lower critical void ratio than that under axially symmetric stress condition.
These results confirm the results suggested by Lee [6] on anisotropically
consolidated undrained Antioch sand. Because the critical void ratio is less
in plane strain tests, one may assume that the critical confining pressure,
as defined by Lee and Seed [7], will be lower under axially symmetric stress
conditions.
Since sand tends to dilate more under triaxial compression than under

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Ag
AL-HUSSAINI ON SAND UNDER PLANE STRAIN CONDITIONS 343

I.O l

~r u
O.g L 0 7.
~J =70 PSI

~ o.8
I
og
2m
~ o.~ ~
0.6

I
COMPRESSION 4 ~ : EXPANSION
o.% -4 -3 -
I
-I
] I
0
I
"FI +2
VOLUMETRIC STRAIN AT FAILURE ( A V / V c ) , 915

FIG. l l - - V o i d ratio at the end of consolidation e~, versus volumetric strain at failure.

plane strain conditions during drained shear, it can be assumed that higher
positive pore pressure may occur during undrained shear under plane strain
conditions than under comparable triaxial compression. This hypothesis
was verified by several consolidated undrained triaxial compression and
plane strain tests, the results of which were presented in Fig. 12. The
figure shows that the pore pressure at failure was about 8 psi higher under
plane strain conditions than for comparable triaxial compression tests,

3O
I I I
0 PLANE STRAIN
0 TRIAXIAL C O M P R E S S I O N

a zo

.J

I-
.r

~ 0

o.

0
g.

-Z0
50 60 70 0 90 ~00
INITIAL RELATIVE DENSITY D r , 9~

FIG. 12--Relationship between pore pressure at failure and the initial relative density.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License A
344 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

Tests on compacted glaciated till reported by Bishop [8] are in general


agreement with this hypothesis.

S t r e n g t h Characteristics
The Mohr-Coulomb theory was used in evaluating the strength param-
eter (@') for both plane strain and triaxial compression tests on the as-
sumption that the failure envelope is a straight line passing through the
origin and that the theory is applicable to soil under plane strain shear
deformation. Thus, the effective angle of internal friction may be defined as
O-I1 - - 0-/3
r = sin -I O-t1
- -
-J[- O-t3
(1)

The effective angle of internal friction for the sand tested was correlated
with the initial relative density for both consolidated drained plane strain
and triaxial compression tests, and the results are depicted in Fig. 13. The
figure clearly shows that the angle @' increased almost linearly with in-
creasing relative density in both plane strain and triaxial compression
tests. The figure also shows that, within the range of relative densities
tested, the angle of internal friction for plane strain is always higher than
comparable triaxial compression tests; the difference is about 1 deg for
loose sand and increases to about 3 deg for dense sand.
It may be debated that the difference in the value of @' between plane
strain and triaxial compression is due not only to the results of the strain

42
(~3' = 70 PSI
/
(:3
.,40

Z PLANESTRAIN~ J
0
I- 3 8
U
oc
L,.

~ 38
/ J
Z
< 34

uL. 3 2
u.
S CO"PRESSlON

303 0 40 50 60 70 80 go I00
INITIAL RELATIVE DENSITY Dr~ %

FIG. 13~Effective angle of internal friction and initial relative density for Chattahoochee
sand in drained shear.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
AL-HUSSAINI O N SAND UNDER PLANE STRAIN CONDITIONS 345

42

t~
hJ
J
-G-
z
O
~- 38
cx.
U
L.
:o/"-c.~" .L~ v-

~e v
Z
'~ 3 4

U
~ 32
t~
w

30
3O 40 50 60 70 80 90 I O0
I N I T I A / RELATIVE DENSITY Drt %

FIG. 14--Comparison between corrected and uncorrected using Bishop's energy correction.

conditions imposed on the sand during shear, but also is due to the amount
of energy absorbed or generated during volume change. This energy con-
cept was first suggested by Taylor [9] for direct shear tests and later was
adopted by Skempton and Bishop [10], who derived an expression for cor-
recting observed triaxial compression test data. The Skempton-Bishop
energy correction, which is applicable to both triaxial compression and
plane strain shear, may be stated as follows:
~'3 d(AV)
~(~1- ~) - (2)
Vc d~l
where
A(al -- a3) = amount of deviation in the observed stress difference,
Vc = initial volume,
AV = change in volume, and
el = axial strain.
When the Skempton-Bishop energy correction was applied to the ob-
served plane strain and triaxial test data (see Fig. 14), 4' for plane strain
tests was still higher than that of triaxial compression tests.
It should be mentioned that procedures for correcting the observed r
for sand suggested by Poorooshash and Rosco [11] and Rowe [12] were
investigated, but they did not show any closer agreement than the Skemp-
ton-Bishop equation.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lic
346 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

Summary
The plane strain apparatus described in this paper appears well suited
for testing drained and undrained sand under plane strain conditions.
Probably the most unique aspect of the apparatus is that the soil specimen
can be consolidated isotropically before plane strain shear; however, aniso-
tropic consolidation is also possible with this apparatus.
It has been demonstrated that the relative density influences the strength
and the stress-strain characteristics in many ways, and the effect is more
pronounced under plane strain than axially symmetric stress conditions.
On the basis of isotropically consolidated triaxial compression and plane
strain tests on Chattahoochee River sand sheared at ~'3 equal to 70 psi, it
appears reasonable to conclude the following:

(a) The compressibility of sand tested under hydrostatic pressure de-


creases ~4th increasing relative density, and it was independent of the
shape of specimens tested.
(b) An increase in relative density increases the initial slope of the
stress-strain curve, the maximum stress difference, the tendency for volu-
metric expansion in drained shear, and the tendency for negative pore
pressure in undrained shear; an increase in relative density decreases the
axial strain at failure for both triaxial compression and plane strain tests.
(c) The data confirm the general understanding that the angle of in-
ternal friction for sand is higher under plane strain conditions than under
comparable triaxial compression test, even though the pore pressure de-
veloped in undrained shear is higher in plane strain tests.
(d) The volumetric strain is more compressional and the axial strain at
failure is slightly lower in plane strain tests than in triaxial compression
tests of sand with comparable relative density.
(e) With the exception of the transitional stage from the hydrostatic
state of stress to perfect plane strain condition, the ratio of z~2/(arl - at3)
remains practically constant during shear, and it is slightly higher for loose
sand than for dense sand.

Acknowledgments
The experimental investigation was conducted at the Georgia Institute
of Technology, Atlanta, Ga. The writer wishes to express his appreciation
to Professor N. H. Wade and regent professor G. F. Sowers of the Georgia
Institute of Technology for their encouragement during the execution of
the experimental work. Special thanks to J. R. Compton, chief, Embank-
ment and Foundation Branch, and S. J. Johnson, special assistant, Soils
and Pavements Laboratory, U.S. Army Engineer Waterways Experiment
Station, for their helpful comments on this paper.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License
AL-HUSSAINI ON SAND UNDER PLANE STRAIN CONDITIONS 347

References
[1] Casagrande, A., "Characteristics of Cohesionless Soil Affecting the Stability of
Slopes and Earth Fills," contribution to Soil Mechanics, 1925-1940, Boston Society
of Civil Engineers, Oct. 1940.
[2] Bishop, A. W. and Henkel, D. J., The Measurement of Soil Properties in Triaxial
Test, Edward Arnold, London, 1957.
[3] A1-Hussaini, M., "The Behavior of Sand Under Plane Strain Conditions," PhD
thesis, Georgia Institute of Technology, Atlanta, Ga., 1968.
[4] Liam Finn, W. D., Wade, N. H., and Lee, K. L., Journal of Soil Mechanics and
Foundation Division, American Society of Civil Engineers, Vol. 93, No. SM6, Nov.
1967, pp. 297-308.
[5] Casagrande, A., in Proceedings, Soil and Foundation Conference of the U. S. Engi-
neer Department, June 1938.
[6] Lee, K. L., Journal of Soil Mechanics and Foundation Division, American Society
of Civil Engineers, Vol. 96, No. SM3, May 1970, pp. 901-921.
[7] Lee, K. L. and Seed, B. H., Journal of Soil Mechanics and Foundation Division,
American Society of Civil Engineers, Vol. 93, No. SM6, Nov. 1967, pp. 117-141.
[8] Bishop, A. W., Proceedings, Fifth International Conference of Soil Mechanics and
Foundation Engineering, Paris, Vol. III, 1961, pp. 135-157.
[9] Taylor, D. W., Foundamentals of Soil Mechanics, Wiley, New York, 1948, pp. 342-
347.
[10] Skempton, A. W. and Bishop, A. W. in Building Material, Their Elasticity and In-
elasticity, North-Holland Publication Company, Amsterdam, 1954, p. 467.
[11] Poorooshasb, H. B. and Roscoe, K. H., Proceedings, Fifth International Conference
on Soil Mechanics and Foundation Engineering, Paris, Vol. I, 1961, pp. 297-303.
[12] Rowe, P. W., Journal of Soil Mechanics and Foundation Division, American Society
of Civil Engineers, Vol. 89, No. SM3, May 1963, pp. 37-42.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
F. C. Townsend ~

Comparisons of Vibrated Density and


Standard Compaction Tests on Sands with
Varying Amounts of Fines

R E F E R E N C E : Townsend, F. C., "Comparisons of V i b r a t e d D e n s i t y a n d


Standard Compaction Tests on Sands with Varying Amounts of Fines,"
Evaluation of Relative Density and Its Role in Geotechnical Projects Involving
Cohesionless Soils, A S T M S T P 523, American Society for Testing and Ma-
terials, 1973, pp. 348-363.
ABSTRACT: The effects of gradation, percentage and plasticity of fines, and
moisture on vibratory and impact compaction of granular soils were investi-
gated by adding measured percentages of low plasticity (ML) and medium
plasticity (CL) tines to a poorly graded (SP) and nearly well graded (SW-SP)
sand. Results indicated that more fines can be added to a uniform sand and
that a uniform sand densities by vibration more effectively than a well graded
sand. The same densities axe produced by impact and vibratory compaction
at higher percentage of tines added to the well graded sand compared to the
percent fines added to the uniform sand. Moisture and plasticity are inter-
related factors which greatly affect compaction. Saturation facilitated vibratory
compaction of low plasticity mixtures; however, for more plastic mixtures, ad-
hesion of the tines to the sand grains restricted vibratory densification. Current
compaction test selection criteria, which ignore plasticity and moisture effects
by comparing vibratory densities of oven-dry materials with those determined
by standard compaction, can lead to the untenable conclusion that vibratory
compaction should be used for sands containing in excess of 20 percent tines.
KEY WORDS: sands, density (mass/volume), tests, cohesionless soils,
compacting

Proper compaction is an important criterion in the placement of em-


bankment materials. Commonly, two laboratory test methods of compac-
tion are utilized as standards for comparing placing densities. For cohesive
materials, the standard compaction test (impact) is used, and for cohesion-
less soils, the relative density test is used. However, in the case of sandy
1 Research engineer, Laboratory Research Section, Embankment and Foundation
Branch, Soils and Pavements Laboratory, U. S. Army Corps of Engineers, Waterways
Experiment Station, Vicksburg, Miss. 39180.

Copyright by ASTM Int'l


Copyright9 (all rights
by ASTM reserved); Fri www.astm.org
International Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreemen
TOWNSEND ON SANDS WITH VARYING AMOUNTS OF FINES 349

soils containing fines (that is, particles smaller than No. 200 sieve), current
guidance is not clear cut. The Corps of Engineers suggests the standard
compaction test may be more applicable than the relative density test for
cohesionless soils having more than about 5 percent by weight finer than
the No. 200 sieve, depending upon the particle size distribution [1].2 An
alternate suggestion is that if 98 percent of the maximum density from the
standard compaction test is higher than 85 percent relative density, the
standard compaction test should be used [2]. ASTM Test for Relative
Density of Cohesionless Soils (D 2049-69) suggests that 12 percent fines be
considered as a basis of selecting between standard Proctor and relative
density tests. Guidelines, specified by the U.S. Bureau of Reclamation
(USBR), classify soils suitable for vibratory compaction into two groups:
(a) suitable and (b) borderline. Borderline soils may contain up to 12 per-
cent fines, but control is based upon 95 percent of Proctor maximum
density or 70 percent relative density, whichever produces the greatest
unit weight.
The objectives of this study were to investigate the effects of various
factors (gradation, percentage and plasticity of fines, and moisture) on
densities obtained by relative density and standard compaction tests on
sands with varying amounts of fines. The test results would then be ana-
lyzed to establish criteria for selecting which test method, vibratory or
impact, should be used for compaction control of cohesionless materials.
The testing program included testing of a uniform (SP) and a nearly
well graded (SW-SP) sand to which were added various percentages of low
plasticity (ML) or medium plasticity (CL) fines. Maximum density tests
using a vibratory table were performed on both saturated and oven-dry
soils. Minimum density tests were performed on oven-dry soils, and stand-
ard compaction tests were performed on soils at various water contents.

Previous Investigations
It appears that the concept of free drainage soil type, according to the
Unified Soil Classification System, is the basis for the current guidelines
used for selecting impact or vibratory compaction methods. For example,
SW and SP sands are considered as being free draining and by definition
contain less than 5 percent fines, while SC and SM sands are considered
impervious and are defined as containing in excess of 12 percent fines. Al-
though free drainage would seem to be an important criterion in selecting
the compaction test method, tests by the USBR [3] have shown that there
is poor correlation between permeability and effectiveness of vibratory
compaction.
The influence of gradation, which involves both percentage of fines and
grain-size distribution, on maximum and minimum densities of sands has

The italic numbers in brackets refer to the list of references appended to this paper.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
350 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

been studied by several investigators [8, 4, 5]. Generally, a uniform sand


contains more voids than a well graded sand. The greater void space of the
uniform sand thereby allows a higher percentage of fines to be added and
still densities by vibration satisfactorily.
Plasticity of fines also contributes to the ineffectiveness of vibratory com-
paction on saturated or moist mixtures, in that the more plastic tines tend
to restrict particle movement into denser configurations by adhering to the
sand grains and creating "bridging" between grains. Obviously, plasticity
would be an insignificant factor if the vibrated density test were conducted
on oven-dry material.
Kolbuszewski [6] demonstrated that vibratory densification of sand was
more effective when the material was compacted under water rather than
in the air-dried condition. Hutchinson and Townsend [4] confirmed this
observation for uniform sands, but showed that well-graded sands produced
higher densities in the air-dried state. Felt [7] indicated that for sands of
varying gradation and percentage of fines (0 to 15 percent), the difference
between maximum density produced by vibrating wet versus dry sand was
insignificant, but at moisture contents between the wet and dry extremes,
lower densities were obtained. Pettibone and Hardin [8] indicated that for
the sands they tested, no reliable relationship was available between grada-
tion and the appropriate moisture condition, wet or dry, for indicating
which condition produced the greater density.
In summary, previous investigators have reported that the maximum
and minimum densities of cohesionless soils are influenced by gradation,
percentage and plasticity of fines, moisture, and particle shape.

Investigative Procedures
The natural sands utilized in this study were (a) subangular to sub-
rounded concrete mortar sand (nearly well graded, SW-SP), and (b) a
local subangular to angular sand termed Campbell Swamp sand (uniform,
SP). These sands were processed by sieving to remove all natural fines.
Standard Corps of Engineers' soil samples of ML and CL material passing
the No. 200 sieve were used as fines to be added to the sands. The proper-
ties of these fines and the grain-size distribution curves of the sand plus
fines mixtures are presented in Figs. 1 and 2.
Standard Corps of Engineers' test procedures were followed for standard
compaction and relative density determinations, respectively [1]. The
standard compaction procedure, which is generally comparable to ASTM
Tests for Moisture-Density Relations of Soils, Using 5.5 Lb Rammer and
12-In. Drop (D 698-70), involves impact compaction of three layers of soil
into a 4-in. mold by 25 blows on each layer of a 51/~-lb hammer falling 12 in.
Water content determinations are made after compaction. The maximum
density test procedure for use in relative density determinations involves
vibrating oven-dried soil in a 0.1-ft 3 (6-in. diameter) mold on a vibratory

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to L
TOWNSEND ON SANDS WITH VARYING AMOUNTS OF FINES 351

" " u.S. S T A N ~ I O SII~VIE OPENING IN INCHES U.$. $TANDAI~ ~ NU~II~ HYDROMETER
3

I0

20

!'_~i ~I~: I. l -!- ----~:::: ----:!IIi17 r -m~li


i,,, -- +....... ~+~!_Ts . . . . . . . .
~ f --tfif ~ ; fl-:, ,~i tlr!,b---~-l#ft[~t -- If!Hi
I I i ' ~- l ] % FINES

~4----I0o
:1--:-- T{-T-[ .... T +~-TT :l ~-7-~,LI~!;'77-~II4H
0.001

I ~ I--~-<<---~~-~--t<o.,,, f .... "N"l ' I ......... 1


{~ICI~ON mi

S••ND
ML FI~ES 20 ~ s
CL FINES

(SPI - C = 1.e. c~ =o.s7

FIG. 1--Grain-size distribution of Campbell Swamp sand (SP) plus fines.

U.$, STkNDAItD $i[~/I~ O ~ N I N G IN K H E $ U.S. S T A N 0 k I ~ $1EV]E NU/AIilER$ HYDROME~R


0
r _ __L . . . . . . . . . . . . . . . . . . . . .

S0 . . . . . T-

,o01-~- 2 2 12

= H ,0 ; , ~o/~ o, 02, 001 0015 0001

oeavet ~
SILTOR CLAY ]

CtAS~W~CA~O~ ~ W'~ a m
ML FINES 2e 23 5
C L FINES 34 22 12

S A N D ($P-~W) - C = 4 , 9 , Cr = o,e~

FIG. 2----~ain-~ize distr~ution of Concrete mortar sand (SP-SW) plus fines.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
t~

r-
T A B L E 1--Summary of density $est results. <_
Z
Q
Material Maximum Vibrated Density Minimum 2/d a t D d =
Density 85%% 0-i"
Percent Standard Compaction -ya(VD), ~(VS), -~d(VO), ib/fts
Fines lb/fta lb/ft 3 lb/ft3 ~'d (rain), o
~d (dry), ~ (opt), w at 95% 98% lb/ft 3 z
lb/ft a l b / f t 3 9 ~/d (opt), ~/d (opt), ~ (opt),
(opt w, %) lb/ft 3 lb/ft 3
S
Campbell Swamp Sand (SP)
No fines
0 102.0 94.8 NA b 93.5 96.4 106.1 102.8 90.0 103.3

With ML fines
9.1 110.2 107.2 12.0 101.9 105.1 110.0 108.9 94.8 107.4
16.7 116.2 115.3 10.1 109.6 113.0 116.4 116.4 98.4 113.3
23.1 ... 116.2 10.2 110.4 113.9 115.2 121.4 94.6 111.6
With CL fines
9.1 110.0 107.2 12.0 101.7 105.0 109.2 107.8 95.2 106.8
16.7 116.0 116.1 9.5 110.2 113.8 115.4 113.7 98.2 112.4
23,1 119,0 122.7 10.0 116.6 120.2 117,6 117,2 96,4 113.8

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
Concrete Mortar Sand ( S W - S P )
No fines
0 121.3 114.5 NA 108.8 112.2 125.4 125.0 ... 111.8 123.1
With M L fines
9.1 128.4 121.7 8.1 115.6 119,3 130.3 129.2 124.3 113.7 127.5
16.7 131.8 128.3 7.0 121.9 125.8 132.7 134.9 130,9 112.3 129,2
23.1 131.0 130.5 7.2 124.0 127.9 129.0 132.8 129.9 108.6 125.5
With CL fines
9.1 126.9 124.2 8.0 118.0 121.7 133.9 128.3 123.7 113.5 130.4
16.7 129.5 130.0 7.5 124.0 127.9 136.0 128.6 124.8 112.9 131.9
23.1 128.2 131.1 7.8 124.5 128.5 130.7 123.5 119.6 108.9 126.9

Using vd(VD). 0
Z
b N A = n o t applicable.
Notes--Standard compaction Vibratory table compaction z
~'d (dry) = dry density using oven-dry soil "yd(VD) = maximum dry density using oven-dry soil
7d (opt) = maximum dry density of soil at optimum ~d(VS) = maximum dry density using saturated soil
water content -y~(VO) = maximum dry density with soil at standard compaction -q
Z
optimum water content

z
o

0e.:
z
--I

0
..n
z

CO
CO

CopyrightbyASTMInt'l(allrightsreserved);FriMar1116:13:06EST2016
Downloaded/printedby
(UFPE)UniversidadeFederaldePernambuco((UFPE)UniversidadeFederaldePernambuco)pursuanttoLicenseAgreement.Nofurtherreproductionsauthorized.
354 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

table at a frequency of 3600 vibrations per min for 8 min under a sur-
charge of 2 psi. The amplitude of vibration used for these tests was 0.019
in. This procedure is comparable to ASTM D 2049-69. Maximum vibrated
density was also determined on saturated material by placing the wet
sand in the mold as it was vibrating at a low frequency (placement time
not exceeding 6 min) while maintaining a small amount of free water above
the soil. After the mold was filled, the surcharge was placed, and the mold
vibrated for 8 min. In addition to the above tests, specimens of SW-SP
sand plus ML and CL fines were vibrated after being brought to water
contents corresponding to the optimum water contents previously de~er-
mined by the standard compaction tests on the same mixtures. Minimum
density of oven-dry material was determined by pouring the material
through a funnel device into the mold.

Test Results
Results of relative density and standard compaction tests on the two
sands and on the sand-plus-fines mixtures are summarized in Table 1.
Definitions of the notations used in this table and in subsequent figures to
identify the various density determinations are given in the footnotes of
Table 1.

~',(vs)-.~
125

120
% F=/7. 5-".-.-~1

a. ~Q/~ F = 20
I10

pr;

I05

I00
NOTATIONS DFFINED ON TABLE I

95
0 5 10 I5 20 25
PERCENT FINES, %

F I G . 3--Comparisons between vibratory density and maximum standard compaction far


Campbell ~wamp sand (SP) plus ML fines.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
TOWNSEND ON SANDS WITH VARYING AMOUNTS OF FINES 355

125

120

.s O,or:,5.z-, ~ ~ / ~

i % F=8----b.
o

i ,
Jos / j /

g5
0 5 10 15 20 25
PERCENT FINES, ~

FIG. 4--Comparisons between vibrated density and maximum standard compactionfor


Campbell Swamp sand (SP) plus CL fines.

Vibrated densities of oven-dry material and saturated material are


plotted against percent fines in Figs. 3 through 6, together with standard
compaction densities of oven-dry material and of material at water con-
tents producing the highest density. Figures 5 and 6 also show the densi-
ties, designated "Yd(VO), obtained in vibrating the SW-SP sand plus ML
and CL fines, respectively, with the mixtures at standard optimum water
contents.

D i s c u s s i o n o f Test Results

Comparison of Densities Obtained on Oven-Dry Versus Wet Material


Figure 7 graphically shows the differences in dry density obtained in
vibrated tests on oven-dry versus saturated material. These data indicate
that higher vibrated densities are obtained for a uniform sand without
fines when the materials are oven-dry. Apparently, in a well-graded sand,
the vibratory movement of the smaller particles into the voids between
the larger particles is unhampered by saturation. On the other hand, in a
uniform sand the absence of smaller particles to fit in the voids and the
presence of water reduce the density slightly. Conversely, Hutchinson and
Townsend [4] observed that saturated uniform sands produced higher rela-

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License A
356 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

140

% F = 22-~.I~

o/o F = 19
13,5

130

120
/

11.5 /
NOTATION,5 DEFINEDON TABLE I

II0
O 5 10 1,5 20 25
PERCENT FINE,5~ %

FIG. 51Comparisons between vibrated density and maximum standard compaction for
concrete mortar sand (SW-SP) plus ML fines.

tire densities than the same sands dry. However, other investigators [7, 8]
have shown that the maximum vibrated densities of wet or dry sands with
fines are rather independent of gradation, a conclusion which is supported
by the data in Fig. 7.
Obviously, plasticity and moisture are interrelated for the sand-plus-
fines mixtures since the plasticity characteristics of the fines would not be
exhibited in compaction of oven-dry material. The test results in Fig. 7
show that in the case of the sands with the more plastic fines (CL), the
maximum vibrated density of the dry material was higher than the satu-
rated material for both the well-graded and uniform sands. Similarly, for
both the uniform and well-graded sands plus ML fines mixtures, the dry
material initially produced the higher maximum vibrated density. However,
as the percentage of fines increased, the saturated mixtures with ML fines
ultimately produced the higher maximum vibrated density. These vibrated
density tests on saturated materials indicate that when moisture is present
the more plastic fines adhere to the sand grains and thereby restrict the
vibratory shifting of the sand particles into denser configurations. In the
case of the sands with less plastic fines (ML), adhesion was not as great
and the water present assisted in filling the voids with these fines, thereby
generating higher densities. Based upon these observations, it appears that

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuan
TOWNSEND ON SANDS WITH VARYING AMOUNTS OF FINES 357

i4o I
"Y~(v u j ~ % F =23 ~ '

13s ~'~ -"~ ,J i

1-
125 ~ --'-V,. ,

-%F:a ~(VOJ----.'X

-- ~ (oPr)
tl5

N O T A T I O N DEFINED O N T A B L E I

It0
0 5 10 15 20 2'5
P E R C E N T FINES, ~

FIG. 6---Comparisons between vibrated density and maximum standard compaction for
concrete mortar sand (SW-SP) plus CL fines.

25
IFINES

2O

)1, /
o~
Z
b.
I-
Z
ulO
LEGEND
bJ
o,. SAND
O S P + M L FINES
g S P + C L FINES
S W - S P i* M L
FfNES
0 SW-SP"P CL
FINES

0
I
-I0 -5 0 l0
")"a(VD,}-"Xa(VS), PCF

FIG. 7--Effect of fines on the difference between vibrated density of oven-dry and saturated
sand.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
358 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

the plasticity of the fine fractions rather than gradation is the primary
factor in governing whether oven-dry or saturated sand with fines will
produce the highest density. Vibration of well-graded sand with fines at
standard optimum water content produced the lowest values of vibrated
densities. As shown in Figs. 5 and 6, the density-percent fines curves were
below and generally parallel to the curves for saturated material.
For both the well-graded and uniform sands, the mixtures with the more
plastic fines (CL) densified more by impact compaction at optimum mois-
ture content than did the same material with low plastic fines (ML).
Comparison of Vibratory Versus Impact Compaction
Figure 8 graphically compares the differences between 85 percent rela-
tive densities of oven-dry materials and 98 percent of standard compaction
densities versus the percentage of fines. These results indicate that if cus-
tomary guidelines are followed, then compaction control by relative density
rather than standard impact compaction would be used for sand-plus-fines
mixtures with fines contents ranging from 13 to 22 percent. However, it
may be inappropriate to base compaction test selection on densities ob-
tained on oven-dry materials to those determined on moist materials. Also,
it is pertinent to note that for uniform sands with fines, the differences in
densities produced by the two compaction methods became less than 2
lb/ft 3for fines in excess of 10 percent. Conversely, for the well-graded sands
with fines, the difference in densities produced by the two methods remains
larger than 2 lb/ft 8 for beyond 20 percent fines content. It appears that the

2O

f,-
Z
w
~m
'xX
SAND
0 SPtML FINES
5-1:1 SP+CL FINES
A
FINES
V SW-SP+CL
FINES
0
J
-8 -4 0 4 8 12
Yd (VD) AT 85 ~ D~ - 98~176
~ (OPT)

FIG. 8---E.~ect of Fines on the difference between 85 percent relative density of oven-dry
sand and 98 percent maximum standard companion.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
TOWNSEND ON SANDS WITH VARYING AMOUNTS OF FINES 359

/o
2(:

~15
Z
b.
p-
Z

LEGEND
SAND
0 SP+ML FINES
O SP+CL FINES
A SW-SP-I-M L
FINES
V SW-SP+ CL
FINES

-8 -4 0 4 8 12
% (vs)- ~,,(o e'r)

FIG. 9---Effect of fines on the difference between vibrated density of saturated =and and
maximum standard compaction.

compaction of well-graded sands with fines is more affected by factors


which lead to bridging between particles, that is, apparent cohesion and
plastic fines, than-that of uniform sands with fines. Apparently, uniform
sand having more void space than well-graded sand can accommodate
more fines in its structure, while the various sized particles of the well-
graded sand restrict movement of the fines and are more conducive to
bridging. Because moisture has such significant effects on the compaction
characteristics of the mixtures, it would seem preferable to compare maxi-
mum densities at optimum water content with those obtained by vibratory
tests on saturated materials. These comparisons are presented graphically
in Fig. 9. The test results indicate that materials containing up to 10.5 to
15 percent of medium plastic fines (CL) will densify more by vibration
when saturated than by impact methods at optimum water content. How-
ever, in the case of lower plastic fines (ML), higher densities were produced
by vibration in a saturated condition for fines contents in excess of 23 per-
cent. Apparently, the wet ML fines do not exhibit sufficient plasticity to be
restricted in their mobility when vibrated as do the CL fines. As a result,
the relative density increases with increasing fines until all the voids are
filled by these fines. Beyond this condition, additional fines tend to segre-
gate and move around the surcharge plate and out of the mold, as well as
begin to alter the sand-fines structure. For the well-graded sand (SW-SP),
it appears that due to the lower void ratio, the mobility of the ML fines is
more restricted and alteration of the structure occurs rather than segrega-
tion, causing an increase in volume and lower densities. However, for the

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to
360 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

uniform sand mixtures segregation evidently occurs, and the ML fines


move out from beneath the surcharge plate with results indicating an un-
reasonable conclusion that vibratory compaction is better than impact
compaction for unlimited quantities of fines. Although material consisting
of 100 percent ML fines was not saturated and vibrated, it would be antici-
pated that the surcharge plate would merely settle to the bottom of the
mold; this is based on the fact that in tests on sand with 23.1 percent fines,
it was noted that considerable fines were present on top of the surcharge
plate at the conclusion of the test.
Similarly, for impact compaction the density will increase correspond-
ingly with the amount of fines until a certain fines content is reached, after
which the addition of fines will reduce the density to that corresponding to
100 percent fines (approximately 106 lb/ft 3 for ML fines and 109 lb/ft ~
for CL fines). The fines content required to achieve the maximum density
is obviously quite different for impact and vibratory compaction due to the
different sand-fines structure created by the two methods.
Previous investigators [4, 9] have obtained good correlations between
maximum densities of sands and Bagnold's [10] grading parameters. Be-
cause vibratory techniques used by these investigators were different and
the sands investigated contained little or no fines, these correlations were
not applied directly in the oresent study. Therefore, independent correla-
tions were made based on Bagnold's grading parameters (obtained as
illustrated in Fig. 10) and percent fines.
The following correlations were obtained relating standard maximum
density at optimum water content and maximum density obtained by vi-
brating oven-dry material.

~/d(opt) = 112.78 + 0.95 (% F) -- 2.70(Sc)

~,~(VD) = 124.89 -~- 0.47 (% F) -- 3.80(Sc)


where

~'d = dry density, lb/ft 8,


% F = percent fines, and
Sc = slope of frequency curve for coarse sizes (Bagnold's distribution).

The coefficients of multiple regression are r 2 -- 0.921 and r 2 = 0.831, re-


spectively. By equating these correlations, limiting values of Bagnold's
parameter (Sc) and percent fines can be established to indicate which of t h e
two compaction methods would produce the higher density for a given sand
with fines. Unfortunately, this is not a realistic criterion since the correla-
tions, by ignoring the obvious effects of plasticity of fines and water con-
tent, lead to the untenable conclusion that vibratory compaction should
be used for sands containing as much as 25 percent fines.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
TOWNSEND ON SANDS WITH VARYING AMOUNTS O F FINES 361

GRAIN SIZE~ iviM


5 1.0 0.5 0.2 0.1 0.05 0.02
I00
2
~ --Dp: 0/29M
8M

=
20

I0
Dw=OI31SMMx/-

-oil
I~
\\
'A l
--cAI I'~

~ s

==7 II _g___% fcoMcRET-E


-" I I I '~ F " '~oRrAR
~. z /I I , \s.N~

0.5

0,2 /_~_5.32

0.~ I I I I I I I I I I
6 I0 Ifl 20 30 4 0 5 0 7 0 I00 140 2 0 0
U.S. SIEVE N U M B E R

FIG. lO--Bagnold's grain-size distribution for Campbell Swamp (SP) and concrete
mortar (SW-SP) sands.

R e c o m m e n d a t i o n s and C o n c l u s i o n s
The guidance provided by the testing of the two sands with fines of
different plasticity characteristics is not clear cut, as there are different
relations of vibrated versus standard compaction densities with increase in
fines depending upon sand gradation, moisture conditions, and plasticity
of fines. However, it may not be pract!cal to specify the method of com-
paction control on the basis of these parameters. Therefore, considering
that field experience has demonstrated that vibratory compaction is best
for essentially cohesionless soils and since oven-dry conditions seldom occur
in nature, it is recommended that if the fines are somewhat plastic, the
relative density method should be limited to sands containing 12 percent
or less, as recommended by ASTM D 2049-69. If the fines are essentially
nonplastic, then this limitation can be relaxed.
Based upon the materials and testing procedures employed in this study,

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lice
362 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

the following conclusions can be made:

1. The maximum densities of sands as determined by vibratory and im-


pact procedures are sensitive to the gradation and percentage of fines pres-
ent in the material. More fines can be present in a uniform sand and still
densify more effectively than for a well-graded sand.
2. Moisture and plasticity of fines are interrelated factors which greatly
influence the compaction characteristics. For vibratory compaction when
low plasticity fines were present, saturation facilitated densification. Con-
versely, for the more plastic fines, adhesion to the sand grains restricted
vibratory densification.
3. Compaction of well-graded sand with fines is more affected b y mois-
ture than a uniform sand with fines. As a result, differences between maxi-
mum densities obtained b y impact and vibratory methods of sand with
fines are greater for well-graded sands.
4. Current guidelines, which ignore the effects of moisture and plasticity
of fines by comparing vibratory densities of oven-dry material with those
of standard compaction, may be misleading and lead to untenable conclu-
sions t h a t vibratory compaction should be used for sands containing as
much as 20 percent fines.
5. A correlation exists between Bagnold's grading parameters, percent-
age of fines, and maximum denisty for both relative density of oven-dry
material and standard compaction.

Acknowledgments
The tests described and the resulting data presented herein, unless
otherwise noted, were obtained from research conducted b y the Waterways
Experiment Station for the Engineering Studies Program sponsored b y the
Office, Chief of Engineers. Permission was granted b y the Chief of Engi-
neers to publish this information. The author wishes to express his special
thanks to J. R. Compton, chief, E m b a n k m e n t and Foundation Branch,
Soils Division, for his assistance throughout the investigation. Assisting in
the laboratory program was F. G. A. Hess.

References
[1] "Laboratory Soils Testing," Engineer Manual EM 1110o2-1906, U. S. Army Corps
of Engineers, Waterways Experiment Station, Vicksburg, Miss., Nov. 1970.
[2] "Earth and Rock-Fill Dams, General Design and Construction Considerations,"
Engineer Manual, EM 1110-2-2300, U. S. Army Corps of Engineers, Waterways
Experiment Station, Vicksburg, Miss.
[3] "Research Tests to Investigate Criteria for Selection Between Vibratory or Impact
Compaction Methods," USBR Earth Laboratory Report No. EM 441, U. S. Bureau
of Reclamation, 1955.
[4] Hutchison, B. and Townsend, D., Proceedings, 5th International Conference on Soil
Mechanics and Foundation Engineering, Vol. 1, 1961, pp. 159-163.
[5] Burmister, D. M. in Field Testing of Soils, ASTM STP 32~, American Society for
Testing and Materials, 1962, pp. 67-97.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
TOWNSEND ON SANDS WITH VARYING AMOUNTS OF FINES 363

[6] Kolbuszewski, J. J., Proceedings, 2nd International Conference on Soil Mechanics


and Foundation Engineering, Vol. 1, 1948, p. 158.
[7] Felt, E. J. in Symposium on Application of Soil Testing in Highway Design and Con-
struction, A S T M S T P 239, American Society for Testing and Materials, 1959, pp.
89-108.
[8] Pettibone, H. C. and Hardin, J. in Compaction of Soils, A S T M S T P 377, American
Society for Testing and Materials, 1965, pp. 3-19.
[9] Shockley, W. G. and Garber, P. K., Proceedings, 3rd International Conference on
Soil Mechanics and Foundation Engineering, Vol. 1, 1953, p. 21.
[10] Bagnold, R. A., The Physics of Blown Sand and Desert Dunes, Morrow and Co.,
New York, 1943.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
J . O. Osterberg 1 a n d S e r g e V a r a k s i n ~

Determination of Relative Density of Sand


Below Groundwater Table

REFERENCE: Osterberg, J. O. and Varaksin, Serge, " D e t e r m i n a t i o n of


R d a t i v e D e n s i t y o f Sand B e l o w G r o u n d w a t e r Table," Evaluation of
Relative Density and Its Role in Geotechnical Projects Involving Cohesionless
Soils, ASTM STP 523, American Society for Testing and Materials, 1973,
pp. 364-378.
ABSTRACT: In-place densities of a fine sand placed as dumped fill partly
below and above water in Lake Michigan were determined by the sand cone
method. In order to excavate below the water table, a ring wall was frozen to
below the 21 ft of sand into the underlying clay. Cores were taken from the
frozen wall for comparison with sand cone densities. In addition, tube samples
were taken outside the freeze wall with a hydraulic piston sampler, and stand-
ard penetration tests were made in the holes drilled for the freeze pipes. Den-
sities obtained were converted to relative densities. Results from the frozen sand
cores, when corrected for water expansion due to freezing, agreed with the
densities from the sand cone tests. The tube sample densities agreed reasonably
well with the sand cone densities. The relative density varied from 40 percent
at the water table to 90 percent at the bottom of the sand.
K E Y WORDS: cohesionless soils, sands, earth fills, tests

T h e p u r p o s e of t h e i n v e s t i g a t i o n d e s c r i b e d in this p a p e r is to d e t e r m i n e
t h e i n - p l a c e d e n s i t y of a b e a c h s a n d d u m p e d in w a t e r w i t h o u t c o m p a c t i o n .
I n o r d e r t o do this, i t was n e c e s s a r y to freeze a n a n n u l a r r i n g p e n e t r a t i n g
t h r o u g h t h e s a n d i n t o t h e c l a y b e l o w t o s h u t off w a t e r so t h a t t h e u n f r o z e n
s a n d inside t h e r i n g could b e e x c a v a t e d a n d t h e i n - p l a c e d e n s i t y d e t e r -
m i n e d b y t h e s a n d cone m e t h o d a t f r e q u e n t i n t e r v a l s . A n o t h e r , a n d e q u a l l y
i m p o r t a n t , p u r p o s e of t h e i n v e s t i g a t i o n is to d e t e r m i n e densities o b t a i n e d
by various other sampling techniques and to compare them with the densi-
ties o b t a i n e d b y t h e s a n d cone m e t h o d . A n o v e r a l l o b j e c t i v e of t h e i n v e s t i -
g a t i o n was t o d e t e r m i n e w h a t r e l a t i v e d e n s i t y could b e o b t a i n e d for a
t y p i c a l L a k e M i c h i g a n s a n d p l a c e d in w a t e r w i t h o u t c o m p a c t i o n .

1 Professor and chairman, and former graduate student, respectively, Department of


Civil Engineering, Northwestern University, Evanston, Ill. 60201.

364

Copyright by ASTM Int'l


Copyright9 by(all
ASTMrightsInternational
reserved); Fri Mar 11 16:13:06 EST 2016
www.astm.org
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
u. s. ~ Slew Ol~*k~l ~. immll
! 3 4 9 I I0 20 30 40 SOIO?O I00 140 200
]00 I
I 0
' 'lll'[I '' Itlll
90 tll lltll IS
I IIII
80~; i II II IIIll 2o
. . L ,, , ~ ,
~o '. II]ll II ! IlllI 3o 0
i

I '"~ i IIIII Z
II II 40 ~
Illll II ,., i Iltll IIIII
II1[ I "~'; i Illll Illll u
b~ 501 ; ; , IIIll
11[I ' ;i i 111tl
~ ..]
0
g.
tiill
Illll Z
ill ll :. i:', :' i Itlll 7O r
lllll Illll Z
~ ~,~. III I : r ; lllll
~']
,il IIII 'i ', g
IIIll " [% ~, II
0
oi:i
100 50
Illil
10 5 1 0,5 0.1 0,05
t[
0.01 0.005
t~'
Grain Size in Millimeters C
Z
I GRAVEL. SANO
Co=rse I MediH 'i Fine J C..... i/~4.dlu. J Fine ] SILT or CLAY
I

Fine Sand, trace silt, light brown: (SP) Minimum density: 90.5 pcf (ASTM D2049-64T)

Specific Grevity: 2.67 Maximum density: 109.0 pcf


t~
O~
FIG. l--Prop~tiesofthesand investigated.
Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
366 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

Sand Fill
The investigation was performed on the Northwestern University lake
fill, consisting of 78 acres of reclaimed land made by filling an area adjacent
to the shoreline of the existing campus. The fill consists of a fine beach
sand taken from the south end of the lake in Indiana. A grain size analysis
and description of the sand is given in Fig. 1. This sand is very typical of
much of the sand found along the shore at the south end of Lake Michigan
in Illinois and Indiana. At the location of the project, the depth of the fill
is 21.5 ft. Below the fill is a deep layer of soft to medium blue silty clay.
Less than 6 in. of sand existed on the bottom before the fill was placed.
At the time of the investigation the ground water level was 9 ft below
ground surface and within a few inches of the level of Lake Michigan.
During the summer of 1964 when the sand fill was placed, the elevation of
Lake Michigan was 576.0 ft, and in February 1970 when this investigation
was performed, it was 578.5 ft, or 2.5 ft higher.
The lake fill was constructed by first building a stone breakwater parallel
to the shore, enabling work to be performed in calm water. The sand was
hauled to the site on flat-topped barges, and the sand was pushed off the
flat-top into the water by a dozer. A dragline crane standing on previously
placed fill scooped up the sand from under the water to finish grade 7 ft
above the water level. Six inches of topsoil were added later. It is believed
that at the project location there has been no significant vehicular traffic.

FIG. 2---Plan of borings and freeze wall-$reeze pipe installeA at each boring location.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further
OSTERBERG AND VARAKSIN ON SAND BELOW GROUNDWATER TABLE 367

I
\, I

"i
~4Nt" ~ANO
,elL J.

CLAY

,~E6E,'VD
.... ~APIk' ~'E D ,.,"N~'T

ACTUAL F~EEXE WELL

FIG. 3--Vertical section through center of frozen ground.

Plan of Investigation
To form a continuous freeze wall enclosing unfrozen sand, freeze pipes
were installed at equal distances at eight locations on a 6-ft diameter circle.
Before installing the freeze pipes, a boring was made at each location. The
types of sampling and locations are shown in Fig. 2. A 6-in. trieone roller
rock bit was used to advance the holes, and drilling mud was used to stabil-
ize the holes and as a medium to bring up the cuttings. Borings (B-la)
were performed using a 2-in. outside diameter split barrel sampling spoon.
The equipment used and the standard penetration resistance were obtained
by following the AST1V[ Penetration Test and Split-Barrel Sampling of
Soils (D 1587-67) for split spoon sampling. Borings (B-2a) and (B-5a)
were performed using the same procedure but substituting a 60 deg conical
probe of the same end diameter as the drill rod ( 1 ~ in.) for the split barrel

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licen
368 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

sampler. Borings (B-3a) and (B4a) were performed according to the ASTM
D 1587-67 methods using 2-in. and 3-in. shelby tubes. Only one sample in
each boring could be recovered. Borings (B-7a) and (B-9a) were performed
using a 3-in. Osterberg hydraulic piston sampler [1].2 Samples could not be
obtained from boring (B-7a) because of the difficulty in extruding the
sampling tube from the piston due to freezing in the 5~ to 10~ F air temper-
ature. The hole was drilled without sampling and almost continuous sam-
pling was obtained in boring (B-9a) several weeks later during more favor-
able weather. One hundred percent recovery was obtained for all samples.
In each of the eight bore holes located on a 6-ft diameter circle, 4-in.
diameter, 27-ft long freeze pipes were installed to freeze an estimated
radius of 1.5 ft, thus forming eight intersecting frozen cylinders (Fig. 2).
To insure a proper seal against seepage when the hole would be excavated,
the freeze pipes penetrated approximately 4 ft into the clay (Fig. 3). Chilled
brine was circulated through the pipes using a 10-ton capacity freezing
machine of 30 kilovoltamperes. The brine was circulated at a rate of 800
gal per min at a temperature ranging from 8 ~ to 12~ F. Three 1-in. diameter
pipes 26 ft long were also installed at locations shown in Fig. 2 as thermis-
tor pipes. Temperature readings throughout the project were taken by
lowering thermistors in the oil filled pipes and recording the temperature
at 5 ft depth intervals.

Testing Procedure
When the temperature in the pipes indicated that a continuous freeze
wall was formed, a 3-ft diameter shaft was excavated inside the wall in
1-ft increments by means of a bucket type caisson rig. Due to the smaller
mass of soil to freeze above the water table, the sand was totally frozen
from 0 to 8 ft (Fig. 3). It was therefore necessary to drill a 36-in. hole
through frozen soil to a depth of 8 ft (1 ft above the water table).
In order to obtain information in the upper 8 ft, an 8.5 ft deep auxiliary
shaft was excavated alongside the freeze wall using the same procedure and
performing the same tests as described (Fig. 3). From 8 ft to the under-
laying clay in the main shaft, a sand cylinder 24- to 26-in. diameter re-
mained unfrozen, allowing the bucket, when drilling, to be supported on
frozen sand around its perimeter and thus not pound the unfrozen part of
the sand below the bucket. After having excavated each foot, 6 in. of sand
was carefully removed with a spade to insure testing of the undisturbed
sand. In-place density determinations were then made using the sand cone
method and taking cylindrical samples of frozen sand by coring for volume
and weight measurements.
Sand Cone Method (ASTM Test for Density in Soil in Place by the Sand-
Cone Method (D 1556-64 Reapproved 1968))--In this method, a hole of ap-

2The italic numbers in brackets refer to the list of referencesappended to this paper.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreemen
OSTERBERG AND VARAKSIN ON SAND BELOW GROUNDWATER TABLE 369

proximately 0.12 ft 3 is carefully excavated through a base plate, and the


dry weight of the excavated soil is determined. From a bottle with a cone
that fitted on the base plate, calibrated ovendried silica sand was poured
inside the hold, and the volume of the hole was determined. The dry density
of the silica sand as poured by the bottle and cone in the hole was deter-
mined by cutting a plexiglass cylinder of approximately the same volume
as the excavated holes (0.12 ft 3) and pouring the sand from the cone into
the cylinder of predetermined dimensions.
The obtained densities of the silica sand were 1.53 4- 0.008 gr/cm 3
(95.4 4-0.5 lb/ft 3) based on ten tests. This density corresponded exactly
to the minimum density (~, rain) of this silica sand (ASTM Test for Rela-
tive Density of Cohesionless Soils D 2049-69. The dry weight of the exca-
vated soil, divided by the volume, yielded the density.
Since trapped water was present in the shaft, the water level had to be
lowered below the hole excavated for the sand cone test in order to prevent
any volume change due to caving of the sides. Therefore, a 6-in. diameter,
1-ft long aluminum casing with a wall thickness of a/~2 in. was pushed
carefully in the soil, the sand was excavated, and the water was scooped
out. This method was adopted since it was felt that all noncased excavation
or insertion of a well point could cause sloughing or densification, respec-
tively, and in this small area could result in serious density modifications
of the saturated sand.
Maximum and Minimum Density Determinations--The minimum den-
sity of the sand was determined according to ASTM D 2049-69. The maxi-
mum density was also determined by ASTM D 2049-69 (dry method)
with the only exception being that the apparatus was not functioning cor-
rectly and the amplitude was less than required (the frequency met stand-
ards). However, after the density was determined the cylinder was vigor-
ously tapped to see if any increase in density resulted. Any increases were
not measurable.
Coring of Frozen Soil--An air drill with l~/~-in, core barrel was used to
take cores of the frozen sand at elevations corresponding to the sand cone
test elevations. At each elevation two cores were taken on the north side
of the freeze wall and two cores on the south side of the freeze wall. The
cores were carefully wrapped~ marked, and stored in an ice chest filled
with dry .ice.
Density Determination of the Cores--Each sample was weighed, waxed
carefully, and reweighed after waxing. The submerged weight of the waxed
sample was then obtained in water at 20~ C. The wax was then removed,
the sample was ovendried for 24 h, and the dry density was computed.
In the computation, the specific gravity of the wax was taken as 0.895
(previously determined to 4-0.002).
Thin Wall Tube Sample Density Determination--The samples obtained
in the tubes were trimmed square inside the tube at both ends and m e a s -

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to
370 RELATIVEDENSITY INVOLVING COHESIONLESS ,SOILS

ured accurately immediately after sampling in the field to avoid any density
change due to shocks or vibrations during transportation. The sand in the
tubes was saved in a sealed container and brought to the laboratory for
oven drying.

Test R e s u l t s
Sand cone densities are shown in Fig. 4, frozen core densities in Fig. 5,
and densities obtained from the piston samples in Fig. 6. It can be seen
that in all the figures the densities increase with depth as shown by the
average line in each figure. Also, each figure shows there is considerable
scatter from the average, about 4-3 lb/ft a for sand cone and frozen core
methods and about 4-1.5 lb/ft a by the piston sample method. Examining
the frozen core test restflts (Fig. 5), it is seen that where duplicate cores
were taken at the same depth and locations (six duplicates) they do not
differ by more than 0.50 lb/ft s from each other, indicating that the scatter

,~E,vS/79" PCF'.

0
,95 t?o /05"
, I

g 1o ' ~ '

.
/0

\
\

/5

s z O,
25 ! I I I
FIG. 4--Densities by Sand Cone Method.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
OSTERBERG AND VARAKSIN ON SAND BELOW GROUNDWATER TABLE 371

~EHS/TY PCF.

0 .9o ,95 ~'oo /o5


I a i i
/P~ATIVE D~-N~r/z'Y
2o ~10 6I0 8O

iX
S /alpine MPING

A
.~..~_G..W. v
IO m

/S

2o - \\-

2S 1 I I ! I

FIG. 5--Densities from frozen cares taken from freeze wall.

is due to actual variation in sand densities. For the pistor~ sampler densities,
since the sample tested is 30 in. long, a much larger sample is obtained
and scatter is less. Thus, at any given depth there appears to be a variation
of about • lb/ft 3 in the density within a few feet in a horizontal direction.
In the process of freezing, it is possible that water in the voids of the
soil is expelled from the frozen zone due to the expansion of the water
upon freezing. It is also possible that all the water is actually frozen in situ
if the freezing occurs rapidly enough. In that case the water in the voids
should expand 8.15 percent when frozen. If the volumes of the frozen sand
cores are corrected for the expansion, the unfrozen density will be about
4.0 lb/ft 8 more than the frozen if the dry density is 100 lb/ft 3 and about
3.0 lb/ft 8 if the dry density is 105 lb/ft 3. This correction has been made
and is shown for the average line in Fig. 5.
Figure 7 shows a comparison of all the data below the water table for the
sand cone densities, frozen core densities corrected for the expansion, and

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lice
372 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

,~)fNSlTY PCF.

O .,,
@~ .95 /oo /05
J I I I
R~z~r/vE ~FN$/ry
0 ' go 2o ~o s ,o"

10--

/5--

2o

25 I I I I

FIG. 6---Densitics from 3-in. Osterberg Hydraulic Piston samples.

the piston sampler densities. It is seen on this figure that the average line
for the corrected frozen densities is almost identical to the average line
for the sand cone densities. If we assume that the sand cone densities are
the correct densities, then it appears that in the freezing the water is not
expelled from the voids. The average line of the piston samples is about 2
Ib/ft a less than for the sand cone average line.
It is also seen from Fig. 7 that the relative density of the sand varies
from 40 percent just below the water table to 90 percent at the bottom of
the sand 13 ft below the water table. There is no explanation why the in-
crease with depth occurs. However, it must have something to do with
the method of placement of the sand. It is, nevertheless, noteworthy that
the relative densities are so large for a sand placed in water with no attempt
made to compact it.
Figure 8 shows a statistical comparison of the test data. It is seen t h a t
above the water table the average relative density is 52 percent and below
the water table 73 percent. However, these figures taken alone can be de-
Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licen
OSTERBERG AND VARAKSIN O N SAND BELOW G R O U N D W A T E R TABLE 373

L)#//S/7-2" .~CF.

0 ~O ,95 /oo /o.5"


1

I
.,5:'e'z,47-/ v E
I l
Z)E~/ r F
I 11
I
20 4 0' 60 ~o /0

!
5 0 r~'OZEN CO~Ze-S ('CO,~ECrE~ ---J

OD x,A~'\ 0

is- -~.~\
"~\~ O -

<>A
A oO

25 I ! I I
FIG, 7--Comparison of densities by different methods.

UNBIASED ESTIMATE
OF STANDARD DEVIATION STANDARD E ~ O R OF
OF THE POPULATION THE MEAN
N I ~ E R OF SAMPLES
D ~ S I T Y TESTS TESTED "

%
PCF RELATIVE DENSITY PCF PCF

Sand Cone - All TeStS 18 101.8 63 4.08 0.96

Piston Sampler * All Testa 7 100.6 57.5 3.00 1.13

Below Water Table Only

Sand Cone 10 103.7 73 3.78 1.20

Fr Core* - U n c o r r e c t e d 18 99.5 52 3,55 1.18

Fozen Cores - Corrected 18 I03,5 73 3.55 1.18

Piston Sampler 4 I0i.2 60.5 2.92 1.46

Above Water Table Only

S a n d Cone 5 99.4 52 4.27 1.91

Piston Sample~ 3 99.8 53 3.56 2.03

FIG. 8--Statistical comparison of test data.


Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
374 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

ceiving since the average relative density varies from 40 percent at the
water table to 90 percent at the bottom of the sand. The table also shows
that the standard deviation of the results for the sand cone and frozen
core densities are about the same, and the piston sampler deviation some-
what less, again indicating this is a natural scatter in the actual densities
and not an inaccuracy in the method.
Figure 9 shows the standard penetration test results and Fig. 10 the cone
penetration results. Both indicate a variation with depth similar to the
measured densities. However, the cone penetration results (Fig. 10) appear
to have about ten blows less than the standard penetration tests with the
split spoon, indicating variations from the dimensions, shape, and other
features can cause considerable differences in the penetration test results.
The correlation curves of Gibbs and Holtz [~] were used to convert the
standard penetration tests to relative density. Using their curves for satu-
rated fine sand and taking into account the effective overburden pressure,
a penetration of 18 blows (average at the water table) corresponds to over
95 percent relative density, and 35 blows (at the bottom of the sand) was
off the chart well in excess of 100 percent. For a coarse saturated sand the

~0 JO dO ,50 60 7"O
0
I ) I I I I

/0

l /5

2O

25 I I I I
FIG. 9~--S~rd P ~ a t ~ Testa.

CopyrightbyASTMInt'l(allrightsreserved);FriMar1116:13:06EST2016
Downloaded/printedby
(UFPE)UniversidadeFederaldePernambuco((UFPE)UniversidadeFederaldePernambuco)pursuanttoLicenseAgreement.Nofurtherreproductionsauthorized.
OSTERBERG AND VARAKSIN ON SAND BELOW GROUNDWATER TABLE 375
Co~ P,-,vEr,vAr/oN (8~o~.~/FT.)
0 0 /0 .20 ,,~0 .~0 ,5"0
J I I

.(8-2A•
_ i

i
2o _ _

~5 I ! t I
FIG. lO---Conepenetration tests.

relative density for 18 blows is 85 percent and for 35 blows is off the chart.
Therefore, the relative densities obtained from the standard penetration
tests using the Gibbs and Holtz correlation had no relationship to the
actual relative densities.

Conclusions
1. The relative density found by the sand cone method for the fine sand
deposited below the water without compaction varied from 40 percent at
the water table to 90 percent at the bottom of the sand, with an average
of 73 percent. The average relative density above the water table was 52
percent.
2. The densities obtained from the frozen sand cores agreed almost
perfectly with the sand cone densities when corrected for the expansion of
the water in the voids on freezing.
3. Densities obtained with the 3-in. piston sampler agreed reasonably
well with the sand cone densities, but were about 2 lb/ft 3 low on the
average.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License
376 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

4. There is a variation of about =t=3 lb/ft s at any given depth in the


density within short distances.
5. Using the Gibbs and Holtz correlation for converting the standard
penetration tests to relative density was not valid for the sand tested.

Acknowledgments
The freezing was done with equipment and services donated by Lake
States Engineering Corporation and their associate, Mile High Drilling
Company, Wehat Ridge, Colo. The soil borings were made by the donated
services of Soil Testing Services, Northbrook, Ill.

References
[1] Osterberg, J. O., Engineering News Record, 24 April 1972, pp. 77-78.
[~] Gibbs, H. J. and Holtz, W. G., Proceedings, 4th International Conference on Soil
Mechanics and Foundation Engineering, London, Vol. 2, 1957, p. 38.

DISCUSSION

W. G. Holtz 1 (panel discussion)--The question here is how good is the


Standard Spoon Penetration Test for determining the relative density of
sands? First, I think that everyone should recognize that the Standard
Spoon Penetration Test is a relatively crude test and no one should expect
to determine the relative density of sands to the nearest one percent or
anything like that. When Mr. Gibbs and I developed a set of correlations
to take into account the effect of overburden pressures, we never indicated
that the sets of curves developed at that time were necessarily applicable
to all cohesionless soils under all conditions. Second, we always stressed
the relative density trends indicated by the Standard Penetration Test
(SPT) values rather than the specific individual values. Third, I wanted
to point out that Mr. Gibbs and I are not particular "promoters" of the
SPT, although we think it is useful for certain types of foundation investi-
gations, and at certain stages of investigation.
At the Toronto symposium on soil sampling there was considerable dis-
cussion on securing good undisturbed sand samples. Even if such samples
can be obtained, without changing the natural density, the determination
of true relative density values from them may not be possible. In many

1Wheat Ridge, Colo. 80033.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No furthe
DISCUSSION ON SAND BELOW GROUNDWATER TABLE 377

sand deposits the materials are badly lensed, the lenses often being quite
thin. If the sample, which contains layers of sands having different grada-
tions, is mixed and used for the relative density test, the resulting grada-
tion and relative density situation is quite different from that which exists
in place. This then brings us back to some method of inplace testing.
In the paper prepared for this session by Dr. Osterberg and Mr. Varaksin,
they showed a comparison of the relative densities obtained by Osterberg
Piston Samplers, in-place sand cone density tests (with necessary labora-
tory minimum and maximum density tests), and as estimated from " N "
penetration values analyzed by means of the so-called "Gibbs-Holtz" cri-
teria. They stated that "Relative density based on blow counts corrected
for overburden using curves based on the work by Gibbs and Holtz did
not correlate with their test results."
In comparing these test results we should first note that the material
tested was a fine uniform sand with the range of minimum to maximum
density (90.5 to 109.0 lb/ft 3) being only 18.5 lb/ft a. The reproducibility of
sand cone density tests are not closer than 2 percent or 2 lb/ft 3, which, in
this case, would be equivalent to about 10 percent relative density. If we
look at the trends which are evident in the soil below the groundwater
table, we can note the following: the mean density determined by the piston
sampler method and the mean density determined by the sand cone method
were 101.2 and 103.7 lb/ft 3, respectively, a difference of 2.5 lb/ft 8. The
mean relative density value determined by the SPT was about 83 percent
or 106 lb/ft 8. Thus, the difference between the mean sand cone test density
values and the mean density as estimated from the SPT is approximately
2 + lb/ft 3, or about the same as obtained with the piston sampling method.
It therefore appears that the SPT relative density estimates were exceed-
ingly good.
With respect to the SPT of sands of high density (~-90 percent relative
density), I would like to quote from the discussion Mr. Gibbs and I pre-
pared for the Journal of the American Society of Civil Engineers, Soil
Mechanics and Foundations Division, May 1969.

The discussion by Peck and Bazaraa suggests that the indicated R.D. = 100% by the
Gibbs-Holtz relation is at lower densities than shown by their field observations for dense
sand. It was our feeling that this high R.D. would not be appropriate to judge the entire
relationship of the entire criteria. If R.D. = 100% is viewed as the "ultimate" density
at which sand is becoming impenetrable, unless dilated, penetration resistance would be
highly irregular and correlativity would become questionable at this level. Therefore, we
felt that any penetration resistance criteria would be most meaningful in the R.D. range
of 90% or less, and we were sure that most soils engineers would judge soils having higher
" N " values to be in the very dense range and to be non-critical.

J. O. Osterberg and Serge Varaksin (authors' closure)--The authors want


to emphasize that no recommendation is made in the paper for methods of
measuring in-place densities. It is'recognized that the freeze wall method is

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agr
378 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

an expensive and impractical method of obtaining in-place densities of


sands in the ordinary job situation. However, since the object of the re-
search was to determine the in-place density of sand below the water table,
and the sand cone method was the only proven reliable method, it was
necessary to freeze a wall to cut off the water so a hole could be excavated
in order to determine the in-place density. Since the freeze wall was needed
anyway, cores were taken to make comparisons between frozen and un-
frozen densities. As stated in the conclusion, excellent agreement was ob-
tained between the sand cores and the sand cone densities which in turn
agreed well with densities taken with the 3-in. piston sampler.
No claim is made that the relative densities obtained are typical of sands
placed in water. However, for dune sand in the lower Lake Michigan area,
it is believed this is typical of sand dumped from barges and handled under
water, with no effort at compaction.
The remarks of Mr. Holtz are well taken. He points out that the spread
between minimum and maximum densities is small, and that one cannot
expect a reproducibility of better than 2 lb/ft 3, which is equivalent to 10
percent relative density. The authors also agree with Holtz that the Stand-
ard Penetration Test is a relatively crude test and shows only trends when
used to determine relative density.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to
Use of Relative Density in Geotechnical
Projects

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reprod
D. J. Leary 1 and R. J. Woodward 1111

Experience with Relative Density as a


Construction Control Criterion

R E F E R E N C E : Leary, D. J. and Woodward III, R. J., " E x p e r i e n c e w i t h


R e l a t i v e D e n s i t y as a C o n s t r u c t i o n C o n t r o l C r i t e r i o n , " Evaluation of
Relative Density and Its Role in Geotechnical Projects Involving Cohesionless
Soils, A S T M S T P 523, American Society for Testing and Materials, 1973,
pp. 381-401.
ABSTRACT: Limitations of several, shallow and deep, direct and indirect
methods to obtain relative densities are discussed, and opinions are given
concerning the use of these methods for construction control purposes. Experi-
ence from two earthwork projects are given for: (1) direct methods--conven-
tional sand cone and water balloon, cutting cylinder, manually excavated pit,
and Denison sampler; and (2) indirect methods--nuclear, standard plate
load test, standard penetration test, and static cone penetration test. Some of
these methods proved satisfactory as a means of obtaining relative densities,
others did not. The water-balloon method was found more suitable than the
sand-cone method. Reference curves were effective in some cases. Use of 6-in.
diameter cutting cylinders to obtain field dry unit weights was found unsatis-
factory for sand containing gravel. Relative densities obtained from measure-
ments of individual layers from a manually excavated pit were found to be
greater than those obtained from the water-balloon method. Dry unit weights
obtained from Denison samples gave reasonable relative density values. Rela-
tive densities from nuclear methods are only approximate. Use of the standard
plate load test to obtain relative densities was not successful. High standard
penetration resistances resulting from residual lateral stresses were obtained
in sand fill compacted in layers by vibratory compactors and lead to very high
inferred relative densities. Static cone penetration resistances give qualitative
measures of relative density of completed earthwork and are useful in evalu-
ating uniformity of compaction.
K E Y W O R D S : cohesionless soils, unit weight (weight/volume), compacting,
earthwork, lateral pressure, sampling, sands, vibratory compaction

T h i s p a p e r describes experiences using different m e t h o d s to o b t a i n rela-


t i v e d e n s i t y (Dr) d u r i n g c o n s t r u c t i o n of two large e a r t h w o r k projects. T h e

1Associate and project engineer, respectively, Woodward-Moorhouse & Associates,


Inc., Clifton, N. J. 07012.

381

Copyright by ASTM Int'l


Copyright9 by (all
ASTMrights reserved); Fri www.astm.org
International Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement
382 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

methods were used to obtain Dr at relatively shallow depths representing


small volumes of earthwork and at greater depths representing large vol-
umes of earthwork. They are catagorized with respect to direct and in-
direct, and shallow and deep methods as follows:

Direct Indirect
Conventional l Nuclear
Cutting Cylinder Shallow Standard Plate Load Test ~ SShall~
Manually Excavated P i t J Standard Penetration Test [.-.
9 ~)eep
Denison Sampler }Deep Static Cone Penetration T e s t ;

Relative density is an index property frequently used during earthwork


construction to indicate the looseness or denseness of compacted cohesion-
less sand9 It is defined
eo - - e ~dmax "~d - - *Ydmin
Dr- -- X (1)
eo - - emin "/d ")'d m a x -- 'Yd m i n

where
eo and ~d rain void ratio and dry unit weight, respectively, of the
sand in the loosest stable state that can be obtained
in the laboratory,
emin and ~d max = void ratio and dry unit weight of the sand in the
densest state that can be obtained in the laboratory,
and
e and Td = void ratio and dry unit weight of the sand in-situ.
Relative density is used as a construction control criterion in end-result
specifications, and to determine whether or not the desired compaction is
being achieved in method specifications.
Relative densities can be obtained directly from in-situ and laboratory
determinations of dry unit weights, and indirectly by inference from em-
pirical relationships between Dr and the results of measurements that re-
flect the compactness of the sand. In either case, it is essential that Dr for
a given soil be related to the magnitude of the significant engineering prop-
erties (strength, compressibility, and permeability) established during de-
sign and that the same procedures and equipment used to establish the
relationships be used during construction. It is important to describe ade-
quately the grain-shape and grain-size characteristics of the sand and the
field and laboratory procedures used.
If the sands have a wide range in grain-shape or grain-size characteristics,
it is generally more satisfactory to obtain Dr using direct rather than
indirect methods in order to account for the influence of variations in these
characteristics. On the other hand, if these characteristics are reasonably
uniform, sufficiently accurate Dr may be obtained by correlating the results

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licen
LEARY AND WOODWARD ON CONSTRUCTION CONTROL 383

of other tests that reflect the compactness of the sand. If such correlations
do not already exist, they should be established in the design stages of the
project and used during construction. The advantages of this approach are
particularly significant when large volumes of earthwork are constructed
within short construction periods.

Projects
The two projects where one or more of the previously named methods
to determine Dr were used are the structural fill foundation for Cooper
Nuclear Station (CNS) and the upper reservoir embankment for the
Ludington Pumped Storage Project (LPSP). Some of the methods proved
satisfactory for earthwork construction control, others did not.
Cooper Nuclear Station is an 800-MW electric nuclear generating station
located near Brownville, Neb., on the west bank of the Missouri River.
The station is supported by a 1 000 000 yd 3 (765 000 m 3) zoned compacted
sand structural fill. Construction of the fill required excavation of a 15-ft
(4.6-m) thick stratum of recent alluvial silts and soft clays and the under-
lying 45 ft (13.7 m) of loose to medium dense alluvial sand. The 60-ft
(18.3-m) deep excavation was backfilled with a zoned compacted sand
structural fill consisting of a 25-ft (7.6-m) thick zone of sand compacted
to an average Dr of 80 percent and an overlying 48-ft (14.6-m) thicl~ zone
of sand compacted to an average Dr of 85 percent. The upper zone extends
13 ft (4.0 m) above original ground surface. The sand was dredged from
the Missouri River. A typical grain-size distribution curve is given by
Curve A in Fig. 1. The sand was placed in 12-in. (30.5-cm) loose lifts and
compacted at natural water content with eight or nine coverages of a
Vibro-Plus CH-43 vibrating smooth drum roller towed by a crawler tractor
at 1.5 mph (0.67 m/s) or less. Laboratory maximum dry unit weights
~/d ~a~, were determined using a vibrating table 2 to densify the oven-dried
sand in a 6-in. diameter (15.2-cm) mold under a dead weight surcharge of
20 psf (957 Pa) at a frequency of 34.2 Hz and an amplitude of 0.0125 in.
(0.0318 cm) for 30 min. Minimum dry unit weights ~d rain were determined
in accordance with ASTM Test for Relative Density of Cohesionless Soils
(D 2049-69).
The Ludington Pumped Storage Project is an 1872 MW hydroelectric
generating station located on the east shore of Lake Michigan four miles
south of Ludington, Mich. The upper reservoir will be enclosed by a clay
and asphalt lined, 39 000 000 yd a (29 800 000 m3), 6-mile (9.7-km) long
multiple zoned earth embankment. The embankment will have a maximum
height of 180 ft (54.9 m) and is constructed primarily of glacial outwash
sands excavated from within the perimeter of the embankment. The sand

2Manufactured by All American Tool and Manufacturing Co., 8027N. Lawndale Ave.,
Skokie, Ill., Model 100-VP-D.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further
/Vo./O .,V~ s A~. #.O No. ~,0 A/~. / 0 0 M~ s
I

Y O - - 1 ) t ]
I I I
I .!
~o~ I------ I----
Z
I I
--I
70 I f-
f
.~o I J ..I I I ~
I !\\\; I I
z
I o
I -- c'~Pve ~ I
I !\\k --9"" A~
k 40 I 8"1"
1
I_ I i \g, , o
I
!Curve C! \ I I
"~,~oli -
I I
I
I
I_
,,30 I0 Z / 0.-.~ 0.2- 0.1 O. o,~r"

[- ' c,,r
L ,ci_~e ( ,c,'n,e tSILT ' I
C~ - A / o . 2 0 0 J / ~ ve
~9,m&o/ Avg R=t,79e AvcJ Sovrce

9.0 2 /~I.~ % 5% Sd~ Se,Td LPSP


Z./ 0.5/,,.r ~ !C~,ir~~#a,,~Sond31anketLPSP

I~IG. 1--Typical grain-size distribution curves.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
LEARY AND WOODWARD ON CONSTRUCTION CONTROL 385

zones are site sand (upstream of vertical chimney drain), chimney drain,
and sand blanket (below and downstream of chimney drain). These zones
are compacted to a minimum Dr of 80 percent. Typical grain-size distribu-
tion curves for the sands in these zones are given by Curves B and C in
Fig. 1. The sands were placed in 8-in. (20.3-cm) loose lifts and compacted
at natural water content with six coverages of a triplex arrangement of
Vibro-Plus CF-43 vibrating sheepsfoot rollers towed by a rubber-tired
tractor at 3 mph (1.3 m/s) or less. Laboratory ~,~ max and ~,~ mi. were deter-
mined for sands having 7 percent or less fines passing by washing on a No.
200 sieve in accordance with ASTM D 2049-69.

Direct Methods
Direct methods to obtain Dr require field sampling of the constructed
earthwork for determination of 7d, and calculation of Dr using the values
of 7~ max and 7d man obtained in the laboratory. Direct methods also require
determination of in-situ sample volume. The direct methods discussed
here are conventional sand cone and water balloon, cutting cylinder, man-
ually excavated pit, and Denison sampler.

Conventional Methods
Conventional methods of obtaining Dr of earthwork require the deter-
mination of "rd. The test hole volume for the sand-cone and water-balloon
methods is approximately 0.1 ft a (0.003 mS). The ~'d max and ~ mill of the
sand obtained from the in-situ density test are determined in the labora-
tory, and the Dr value is calculated using Eq 1. Procedures for the sand-
cone and water-balloon methods are given by ASTM Test for Density of
Soil in Place by the Sand-Cone Method (D 1556-64) and ASTM Test for
Density of Soil in Place by the Rubber-Balloon Method (D 2167-66),
respectively.
The most often used laboratory procedure for determining ~d m~x and
~'~ mill is given by ASTM D 2049-69. There are several disadvantages to
this procedure; the acceleration of the vibrations is much too high and
particles segregate when the specimen contains coarse sand and gravel.
Relative densities obtained at LPSP are high and occasionally greater than
100 percent. The reason for this is that the ASTM procedure results in
lower "In max than can be obtained using other laboratory methods, and the
construction method occasionally results in higher unit weights than the
ASTM procedure ~/~ . . . . It is important that the designer know the sig-
nificance with respect to engineering properties of the results obtained by a
particular procedure, including such disadvantages as segregation and poor
reproducibility, rather than whether or not the ~ ma~ and ~d rain are indeed
the maximum and minimum values that can be achi~eved in the laboratory.
Although it is likely that the sand cone has been used more extensively
than the water balloon, the water-balloon method is preferred when con-

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to
386 RELAI'IVE DENSITY INVOLVING COHESIONLESS SOILS

struction involves the use of vibratory compaction equipment, because the


loose sand used in the sand-cone method is easily densified by vibrations
from nearby equipment causing an erroneous determination of test hole
volume. The water-balloon method was used at CNS and LPSP. The
apparatus used was the Washington Dens-O-Meter3
Some factors to be considered in using conventional methods are ac-
curacy of the tests, representativeness of tests made near unconfined
boundaries, meaningfulness of Dr when heterogeneous soils are encountered,
and use of testing expedients. The accuracy of either the sand-cone or
water-balloon methods to measure correctly the in-situ unit weights de-
pends upon: (1) the amount of closure of the hole due so relaxation of
lateral stresses as the hole is excavated, and (2) the ability of the sand or
balloon to conform to the shape of the test hole. The measured unit weight
varies with test depth and the distance from the test location to an uncon-
fined boundary. Relative density has no meaning when heterogeneous soil
conditions are encountered. If the grain characteristics of the sand are uni-
form, reference curves can be used to obtain Dr.
Test Hole Closure--The amount of hole closure depends mainly on the
value of ~d and the lateral stresses in the sand. A larger amount of closure
will occur for small ~ than for large values because both compressibility
and rebound due to relaxation of lateral stresses are greater for small ~d.
Measurements of LPSP in medium to fine sand compacted in layers and
having 1 to 2 percent fines and a uniformity coefficient (Cu) of about two
indicate that the volume change due to lateral stress relaxation is less than
1 percent of the test hole volume for the range of Dr generally required
for earthwork construction control. The water balloon has the advantage
of greater accuracy, because the pressure of the water in the balloon can be
controlled and an increase in pressure can improve the conformance of the
balloon to the shape of the hole and tends to restore the hole to its original
position. The hydrostatic pressure in the center of a Washington Dens-O-
Meter water balloon is approximately 0.9 psi (6205 Pa). When both closure
of the hole due to lateral stress relaxation and expansion due to the pres-
sure applied by the water balloon are taken into account, the net volume
change of the hole is very small. Consequently, the water-balloon method
offers the added advantage of a more accurate test hole volume
measurement.
Test Depth--In-situ dry unit weights increase with depth below the
compactor working surface because of surficial loosening effects of com-
pactor vibrations and the additional densification resulting from the com-
paction of overlying lifts. Therefore, tests made too near the surface give
low and often unrepresentative results. The results of measurements in the
sands described previously showed that Dr obtained 2 ft (0.6 m) below the

8Manufacturedby D. G. Parrott & Son, Olympia,Wash., Models 15 and 30.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further r
LEARY A N D WOODWARD ON CONSTRUCTION CONTROL 387

compactor working surface are 5 to 10 percentage points lower than Dr at


a depth of 3 ft (1 m). Therefore, the significance of test depth on Dr should
be determined, and the test depth established at the start of construction
to ensure representative results.
Unconfined Boundaries--Because of the lack of confinement near uncon-
fined boundaries of earthwork, ~(d is lower near the boundary and increases
with distance away from it. To obtain representative results the minimum
horizontal distance required from an unconfined boundary to a test loca-
tion should be established on the basis of test results st the start of con-
struction. Tests made closer than this distance should be so identified.
Heterogeneity of So/l--While removing material from the density test
hole, the technician should visually identify the soil and examine the wall
and bottom of the completed hole. If the excavated soil is heterogeneous
(that is, if it consists of lenses of sands of different gradation or mixtures
of sand and clay), the test should be abandoned. If such heterogeneous soil
is thoroughly mixed in the laboratory, it is a different soil than the in-situ
soil. Rather than measuring the volume of such a hole, a new test should
be made in homogeneous soil in the vicinity of the abandoned hole.
Reference Curves--Reference curves relating ~/~ to Dr and uniformity
coefficient Cu to ~d max can be used if the grain-shape and grain-size char-
acteristics of the sand vary within narrow limits. Reference curves relating
~d to Dr are used to obtain Dr values in the sand blanket and chimney
drain zones st LPSP. Four out of five Dr values are obtained using refer-
ence curves if ~(d is greater than a limit value. Relative densities are ob-
tained on the basis of ~/d m~ and "vdrain: (1) for one out of every five Dr

/.~O-- r ...................... ~. . . .
9

/tO __.~.P..~e r e nee..~ur ve --r-.. . . . .


9,co," c h , ' ~ n e y dJ,.,~',~ G ~ ' ~ . .. O i .*~"
5 = n d b/,~n,~eP ~ "* - * /
IlO

9 'Li ":2(" "-" ; -

9 9 I o9 9 9
C5~o
9 9 Smoo ~ o e e e 9 1 4 9e ~ ~ o9 9 9
9 ~ele 9
el 9 el 9
...........
9 9 .o v9
9 " 9' l = .
Pr.80~
9 9 ~ ~ $a, 9 9

-1.s..~, * o o

6O- - @
9 L't~i" I)r~j Uni'[" Wei?hi', !09 Ib/ff ~

,o,
so I r i ~;o i t l I I i i I I
I15" /20
I,-S~tvD,~ Un;f ~&~ghf a'd, Ib//ft ~
FIG. Y,---Relativedensity versus in situ dry unit weight referencecur~--LPSP.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agree
388 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

ma

,g 9
| .'."

.~ ~/S ~ '

"~" ol1

.Se~ curve A in F i l l For


I/0 9mie.zjle di,~tr

..*.,
io 9
ip
I0~ 9
2 3 4
Un,'fo,-,~,My C ~ e ~ ; o e . t ,Cu

FIG. 3--Correlation between maximum dry unit weight and uniformity coe~ieient--C N S.

determinations, and (2) when ~d is less than or equal to the limit value.
The reference curve and limit value of ~/~ used at L P S P are given in Fig. 2.
The D r values obtained using laboratory ~d ~,ax and ~'d rain are used t o
confirm or adjust the reference curve. In some cases D r can be estimated
or confirmed using reference curves relating ~'d max and ~'d mi, to Cu. At
C i S '~/d m a x correlated well with Cu. The correlation shown in Fig. 3 was
used at CNS to ensure that ?d max was reasonable when D r was very differ-
ent from what would be expected based on previous experience or
observation.
Satisfactory measures of D r can be obtained using conventional methods
when consideration is given to the above factors. In some cases, reference
curves can be used effectively to estimate Dr.
C u t t i n g Cylinders
I n - s i t u dry unit weights of compacted sand were obtained at L P S P b y
means of thin-wall cutting cylinders and compared to ~'d obtained using
the Washington Dens-O-Meter within 2.5 ft (0.76 m) of the cylinder sample
locations. A typical grain-size distribution curve for the sand is given b y
Curve C in Fig. 1; the sand contains fine gravel. The samples were taken
at a depth of 2 ft (0.61 m) below the compactor working surface. The thin-

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant t
LEARY AND WOODWARD ON CONSTRUCTION CONTROL 389

wall cutting cylinders consisted of 6-in. diameter, 6-in. long (15.2-cm),


16-gage steel tubes. The bottom of the cylinder was beveled to form a
cutting edge. Using slight pressure, the cylinder was gently pushed into
the sand while the sand around the cylinder was excavated about 2 in.
from the cylinder to a depth not greater than 1.5 in. (3.8 cm) below the
cutting edge.
Dry unit weight of the sand in the cylinder is plotted versus ~d obtained
from the Washington Dens-O-Meter test in Fig. 4. These data show that
in five out of six cases, ~,d obtained with the Washington Dens-O:Meter are
higher than those obtained with the cutting cylinders. This lack of correla-
tion was found to be due mainly to the pushing ahead of gravel by the
cutting edge of the cylinder which resulted in voids along the sides of the
specimen as the gravel scored the specimen entering the cylinder. In addi-
tion, the sand dilated as result of shear stresses caused by the penetration
of the cylinder into the compacted sand.
The use of 6-in. diameter (15.2-cm) cutting cylinders to obtain ~ of
dense sand containing gravel from which Dr may be calculated was unsat-
isfactory because of the influence of the gravel and sampling shear stresses.
It may be possible to reduce the influence of the gravel by using larger
diameter cutting cylinders having smaller perimeter sample area rations,
although the size of the cylinder will depend upon the size and amount of
gravel in the sand and the difficulties in handling a large specimen.
Manually Excavated Pit
Relative densities were obtained for each' approximately 9-in. (0.23-m)
thick layer of sand excavated from a 3.6-ft long (1.l-m), 2.7-ft wide
(0.82-m), 4.2-ft deep (1.3-m) pit in the sand blanket at LPSP and com-

9 J S e e curve C i, Fi9. / For


9 J 9rain-s]e,s d,3/ri,6,,/:bn

~ //~

"k

//2.

%
~ I/0
/
J
I I I
/10 I1~ I1~ I1~
Dry Un,~ Me4gh,~ From Cu/l~mg C~,/:nder • , Ib/:f"

FIG. 4---Comparison of dry unit weights from Washington Dens-O-Meter and cutting
cylinder--LPSP.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agree
o
,o
9 I I o
~/~/oo/ ro//~r ; e ~ o P F e n e t / . w ~ g ~ p p r o x 0 7 F t
9 W~/~D~n~-O-/~e~ ~m
m
).
-i
Z.I 2.0 #.6 ~k /~ <

o
x
(,,
1.7 2.2 3.3 ~ ,.,I
.4
'~'\
1.7 2.2-i 2.~ ~\ ~'~ /
/
!
z
,t {)
~.7 2.5 4.1 /
/ rl
0-!.
ig
N~
2.0 2.2 /.i.6 o
z
F,
IVO~5: r

s
E,

~ " W " " q~J ~ 9 r c,--,t;"/'on


ProF,'leoF~z~Pi~

FIG. 5---Summary of re~ul~ from manually excavated pi$--LPSP.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions aut
tEARY AND WOODWARD ON CONSTRUCTION CONTROL 391

pared with the results from Washington Dens-O-Meter tests in each layer.
A typical grain-size distribution curve for the sand is given by Curve C in
Fig. 1. The volume, weight, and water content were determined for each
layer. Laboratory ~d m~x and ~d mi~ were determined from a carefully
selected representative specimen from each layer. The layer volumes ranged
from 5.2 ft 8 to 7.4 ft ~ (0.15 m 3 to 0.21 m 3) and the Washington Dens-O-
Meter test hole volumes ranged from 0.11 ft 3 to 0.12 ft 3 (0.0031 m s to
0.0034 m3). The pit was excavated in six layers, but measurements of the
top layer of the pit were not made because of the sheepsfeet impressions
in this layer. The bottom of the sand blanket and pit consisted of calcareous
silty sand having 23 percent fines. The ~'d and Dr and index properties for
each layer are given in Fig. 5. These data show that, with the exception of
the second and bottom layers, ~d and Dr of each layer are greater than t h e
values obtained with the Washington Dens-O-Meter. The ~'d and Dr of
individual layers are believed to represent more closely the actual condi-
tions, because the volumes of sand in the layers are 48 to 60 times the vol-
umes of the Washington Dens-O-Meter test holes.
Determinations of ~'d and Dr in this manner have a number of disad-
vantages which include the destruction of a large volume of completed
earthwork, and the large amount of time required to make the excavation
and measurements and to restore the area to its original conditions. How-
ever, if the grain characteristics and index properties of the sand are essen-
tially uniform, Dr obtained from individual layers in a few pits can be
compared to the results from the more expedient conventional methods.
Such comparisons can indicate the extent to which the results from con-
ventional methods are representative of the completed earthwork.

Denison Sampler
A 6-in. diameter (15.2-cm) Denison sampler was used with bentonite
slurry as drilling fluid to obtain samples of very dense sand at CNS. In
addition, Standard Penetration Tests (SPT) (ASTM Penetration Test and
Split-Barrel Sampling of Soils (D 1586-67)) were made in borings drilled
using a bentonite slurry as drilling fluid. The borings with SPT were lo-
cated 5 ft from the Denison borings.
Relative densities of the Denison samples are compared in Fig. 6 with
Dr inferred using the standard penetration resistances (N) and the rela-
tionships given on p. 314 of the Earth Manual [1].4 The very high Dr values
were inferred from an extrapolation of these relationships. These data show
considerably lower Dr from the Denison samples than those obtained from
N. The high N values reflect the influence of the residual lateral stresses
developed during vibratory compaction of the sand fill. Measurements of
lateral stresses in vibratory compacted sand fill have been made by

4 The italic numbers in brackets refer to the list of references appended to this paper.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agree
,0

Relofive Z~en.s/~.~ Dr, %


~o ~0 80 I00 I~0 140 l&O 180 ZOO
o ! ! >
I
I
Leo e rid." <.,
- 9 ~,,i,,.~CZ4"l(X 'olu.~f
r3
9 " " C 2~J ( ~ e n / $ o n ) ~.
T,
I0 -4
-<
" . . . . I C2~O(D.,,,~o,,',~
r..
<
Z
2O 0

o
2:

o
Z
ul
u)
ul
o
/ f
4o

_ Notes:
I.See c u r v * ~ A i n /:/9./[Or9r~/n-aize dis~crihuDion
2. 0 6 P a . , ' n e ~ u s i n 9 r~JaP/o~sh;ps ~ v c ~ o n p. 314c o F ~E~r~h /~tQnv~l"

IX9
FIG. 6--Relative density from N-values and Denison samples versus depth---CNS.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
tEARY AND WOODWARD ON CONSTRUCTION CONTROL 393

D'Appolonia et al [2], who found that lateral stresses increased with the
number of compactor coverages and the operating frequency. Further evi-
dence of the effect of these stresses on N is given in a subsequent section,
"Standard Penetration Test."
The mechanics of sampling with the Denison sampler require the ad-
vancement of a rotating outer barrel ahead of an inner barrel into which the
sand sample enters. The outer barrel is equipped with cutting teeth at the
lower end. The drilling fluid flows through the annular space between the
two barrels and the teeth and removes sand as the outer barrel is rotated.
This action causes release of the high residual lateral stresses that were
developed during compaction of the sand. The Dr obtained by means of
Denison samples are more representative of the actual conditions than Dr
inferred from N when high lateral stress conditions exist. The Denison
sample Dr values in Fig. 6 are approximately equal to the average Dr ob-
tained during construction with the Washington Dens-O-Meter at depths
of 1.8 ft (0.55 m) below the compactor working surface.
The use of bentonite slurry drilling fluid assists in retaining the sample
in the inner core barrel. There remains the possibility, however, that re-
covery of samples from below the water table will be difficult. When quanti-
tative values of Dr are required and Dr inferred from N appear very high,
the use of the Denison sampler is recommended.

Indirect Methods
Indirect methods do not require field sampling of the constructed earth-
work; Dr is inferred from empirical observation and experience. The in-
direct methods discussed here are nuclear method, standard plate load test,
standard penetration test, and static cone penetration test.

Nuclear Methods
In-situ wet unit weights can be obtained using nuclear methods. The
procedures for these methods are given in ASTM Tests for Density of Soil
and Soil Aggregate in Place by Nuclear Methods (Shallow Depth) (D 2922-
71). Although these methods are nondestructive, the upper foot of fill must
be destroyed to test at the proper depth and obtain a sample to determine
"y~ma~ and -y~mln. A comparison of wet unit weights obtained using the
Washington Dens-O-Meter and the Troxler Model 2401 nuclear gage5 is
given in Fig. 7a. A typical grain-size distribution curve for the soil is given
by Curve B in Fig. 1. These data show that the results are comparable
and that there is a fair correlation between the wet unit weights with a
possible error of :t:2 lb/ft 3 (314 N / m 3) with respect to the values obtained
with the Washington Dens-O-Meter. However, since ~d values are required

Manufactured by Troxler Electronic Laboratories, Inc. P.O. Box 5997, Raleigh,


N.C., Model 2401

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No f
394 RELAllVE DENSITY INVOLVING COHESIONLESS SOILS

t3C

:/.
Z
oE
12o
t~
(o,
IIS

See Curve. B in Fi~. I Tot


,di~'t,I r i b u t i o n I qratin-sis
HO
I10 li5 I;~o IZS 130 t35
J! ~ / e t Unit. Weight1: From W~tshiflgton Dens-O-Meter, ib/ft 5

9 /
el D Q
9

9 [I ~| 9 indicates nu:iee,r qat~e ar~ r tins


9~[b~[D
) ~'~0 ~ ~._. "_ary,n~
. Speedy HoistureTesl
~ ~' w~en a.nd

i
;-/ /~c~
I J
See curve ~ i n I:iq.I for ~jr~in-si~.e
/ v ~, d i s t r i bur
~1/ o I I
3 5 7 q tl i"S
W a t e r Content. From Oven Drvino.

FIG. 7--Compar/sons of wet unit weights and water contents--LPBP.

it is necessary to determine water contents with which to calculate them.


Manufacturers of nuclear gages report that water contents can be obtained
using nuclear methods, and a procedure is given by ASTM Test for Mois-
ture Content of Soil and Soil-Aggregate in Place by Nuclear Methods
(Shallow Depth) (D 3017-72). A comparison of the water contents ob-
tained by oven drying and by the Troxler nuclear gage for the same soil
are given in Fig. 7b. These data have considerable scatter and significantly
higher water contents are obtained with the nuclear gage. Consequently,
9 d calculated using water contents obtained from nuclear methods would
have considerable scatter and be erroneously low. Data showing oven-dry
water contents versus Speedy Moisture Tester e water contents are also

s Manufactured by The A1pha-Lux Co., Inc., Trenton Ave. and Somerset St., Phila-
delphia, Pa., Model MC 320.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
LEARY AND WOODWARD ON CONSTRUCTION CONTROL 39,5

included in Fig. 7b. These data show a small amount of scatter with water
contents obtained with the Speedy about 0.5 percent lower than those ob-
tained by oven drying. The Speedy can be effectively used with a small
correction to obtain proper water contents. Estimates of Dr which may be
satisfactory for some earthwork construction purposes can be obtained
from in-situ wet unit weights obtained using nuclear methods, corrected
water contents from the Speedy Moisture Tester or by oven. drying, and
laboratory ~d m~ and ~ mi, values or reference curves. However, one must
consider that the results obtained with portable nuclear gages involve a
small volume of sand, approximately 0.05 ft s (0.0014 mS), and are influ-
enced by variations in chemical composition and heterogeneity of the
sand, and irregularities of the test surface.

Standard Plate-Load Test


The standard plate-load [3] consists of a 1-ft square, 2-in. thick steel
plate to which incremental vertical loads are applied, and the settlement
for each load is measured. The results represent the load-settlement char-
acteristics of the soil within approximately 1.5 ft (0.46 m) below the plate.
A correlation between Dr and settlement at a particular load can be ob-
tained by making conventional Dr determinations within a few feet of and
immediately below the level of the plate. Such a correlation was attempted
at CNS before and at the start of construction in the hope of being able to
infer Dr from the results of standard plate-load tests made at a depth of
2.5 ft (0.76 m). A typical grain-size distribution curve for the sand is given
by Curve A in Fig. 1. Settlement versus Dr is plotted for a contact stress
of 6 kips per ft 2 (ksf) (287 280 Pa) in Fig. 8 together with the correlation
curve obtained from the data. Although these tests were made in an essen-
tially uniform sand, the ~ata show a considerable amount of scatter. At-
tempts to reduce the scatter by correlating Dr wish settlement for other
stresses, 1 to 12 ksf (47 880 to 574 560 Pa) were unsuccessful. Consequently,
Dr inferred from such correlations were questionable, and the Washington
Dens-O-Meter was used at CNS as the primary means of obtaining Dr.
The settlement of a standard load plate where the reaction is provided
by earthwork construction equipment is small for De values greater than
approximately 65 percent which is often the lowest value used as con-
struction criterion. The small settlements are due to the residual lateral
stresses which result in a less compressible and stronger sand without
changing Dr. The small settlements and the accuracy with which ~a can be
measured are reflected by the scatter of data which results in uncertainty
as to the proper Dr to be inferred from such correlations.
In addition so the difficulty in obtaining accurate Dr values wish the
plate-load test, other disadvantages must be considered: (1) construction
equipment to provide the reaction must be taken from earthwork con-
struction activities or the tests must be made between construction shifts;

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to L
396 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

/,Fo
~ee curve A ~n Fi I. I ~ r 0 9 (m I /0

".j
/ZO

I00
~s
~11~ o n b a s i : ~ o~ ~e*ull'.~ oF &Fac~d~rd ~o~d
~ t ~ 8 f~O A Fpo~e 'Te~,~,~hi .rid p~ck (Iq~e'l~
A

40

0 Ord~er~i/~eal ~/'~h Waah/nq~on ~-0-I~.Tr/o~r#a/a m~de

fee/F/I/p~o~v~.

I I I I I I
O.O~ o.0r ~ 06 0.08 0.10 O~R 0. /~
.5eiM~nf of 5 ~ n ( ~ /'~ P/a~ ~/nder 6 0 0 0 11~ 9 in.

FIG. 8--Correlation between relative density and deflection of standard 1-ft2 plate under
60O0 lb load--CNS.

(2) two or three persons including the equipment operator are required to
make the test, whereas conventional methods require only one technician;
and (3) approximately 1 to 2 h are needed to obtain reaction equipment,
set-up, and perform the test, whereas 1/~ to 1 h is required by conventional
methods.
If the load-settlement characteristic of the sand is basic to design as-
sumptions, it may be appropriate to employ standard plate-load test results
rather than D r as construction control criteria. The load-settlement curves
given by Terzaghi and Peck [3] qualitatively correlated with D r and in-
tended for estimating settlements are included in Fig. 8. A comparison of
these curves with the CNS results shows that for the same applied load and
Dr, the corresponding settlement estimate is higher using the Terzaghi and
Peck curves. It is suggested that when standard load plate settlements
versus D r values are required for construction control of earthwork having
high residual lateral stresses, that the relationship be established for the
project either during the design stage or at the start of construction.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lic
LEARY AND WOODWARD ON CONSTRUCTION CONTROL 397

Standard Penetration Test


The standard penetration resistance N was first related in a qualitative
way to Dr by Ter~aghi and Peck [3]. A quantitative correlation between
N and Dr which includes the influence of effective overburden pressure was
developed by Gibbs and Holtz [4] and is included in the Bureau of Reclama-
tion's Earth Manual. Bazaraa [5] developed a quantitative correlation be-
tween N and Dr which gives lower Dr values than the Earth Manual for
equal values of N and effective overburden pressures.
A difficulty with using N to infer Dr is that a large volume of earthwork
of sufficient thickness is required to justify the time and expense involved
in obtaining N. In addition, she inferred Dr is approximate and based on
empirical observation and experience primarily from sands which have

5[~ndard Pene/ro ~ion /~.~.,.S,r ,/[J, blows/fl/


0 SO /00 150 ZOO 25"0
0
[ I f
Z eqe,,Td:
@ ,,~,o'.,~a/~ o,,or,,j~ Z).~ gO %
9 . . . . . g5%

I : Noh~ ,

~tere o~za/n~d.
(Z) See c~rce A / n ~/9./ f o r g r a i n ' ~ i ~ -
a/,f,.i , ~t-,o,., 1

9 Ip
~ 3
9 eO
9 9
e
@ e

0
~ * II @

~6--- \
0
D~=/O0 100%

7
Z \ "

FIG. 9---Results of standard penetration tests in structural fill zones--CNS.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to
398 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

relatively low lateral stresses. On the other hand, as stated earlier, sand
placed and compacted by vibration in layers has high lateral stresses.
These stresses can be greater than the overburden stress and result in high
N values that lead to very high apparent Dr values.
Standard penetration tests were made in borings drilled at different
stages of construction at CNS using a bentonite slurry as drilling fluid.
The N values are plotted in Fig. 9. The symbols indicate the average Dr of
80 and 85 percent that were obtained from the results of many Washington
Dens-O-Meter tests made during construction of the zones in which SPT
were made. Also shown in Fig. 9 are N corresponding to 100 percent Dr
according to the Earth Manual and Bazaraa. These data show that N
corresponds to much greater Dr volues than indicated by the Earth Manual
and Bazaraa or the results obtained with the Washington Dens-O-Meter.
These high N values are the result of the residual lateral stresses developed
during vibratory compaction of the fill. There is no doubt that the fill is
very dense; however, there is difficulty associated with a quantitative in-
ference of Dr using N.
Static Cone-Penetration Test
The static cone-penetration resistance (R~) from static cone-penetration
tests (Dutch cone) have not yet been correlated to relative density to the

0.0

0.6

-
g
A rect Y'epresen~s sand
cornpacfed, be~een Dr
>~ o.e valuesof 60%md74%.
._ Adapted from Sh, ltze
-~ ami Muhs, ,"/67

i.o imtir~tes Dr val,,e from


~ V,'~hlf,~ton D-O-M tests
in pits LPSP. See c.rve
-

C in Fi9. I ~r ~ain-slze
distribution.
1.9. J t I
0 tOO 2OO 300 400

~'l'a'l'ic Cone Pene§247 Re_~~stance_ Rp, H%//c-m"


FIG. lO---Relative density versus static cone penetration resistance.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
LEARu AND WOODWARD ON CONSTRUCTION CONTROL 399

5~ttfiC Cone P~nefrafion Resishncc Rp, k3~mt~


,, ~. 5oo 700

_c?s.

L~ I0_
See curve B in Fit~.l
~or 3rain-si}a c~is}r~bu~ion
N
p'_
4
14_

FIG. 11--Examples of Static cone penetration results in compacted cohesionless embank-


ment--LPSP.

extent that there is general acceptance of a correlation. The broad correla-


tion by Schultze and Muhs [6] appears reasonable and is given in Fig. 10.
Static cone-penetration resistance measurements made at LPSP in the
chimney drain and site sand zone are included in Fig. 10 together with Dr
obtained from Washington Dens-O-Meter tests in test pits excavated at
approximately the same locatbns and after the static cone penetration
tests were made. These data show that R~ corresponds to higher Dr than
one would infer from the Schultze and Muhs correlation. The higher Dr
values are due primarily to ~ ~ax obtained using the ASTM procedure.
Because rough qualitative indications of the denseness of completed earth-
work can be obtained from static cone-penetration resistances and because
a large number of Rv values can be obtained rapidly, the uniformity of the
work can be evaluated readily during construction. This is particularly
useful in evaluating the consistency of application of method specifications.
An example of results obtained in the LPSP site sand zone is given in Fig.
11.

Conclusions
Relative density (Dr) is frequently used during earthwork construction
to indicate the looseness or denseness of compacted sand. Experience gained
using different methods to obtain Dr on two large earthwork projects leads
to the following conclusions.
1. Satisfactory measures of Dr can be obtained using the conventional

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licen
400 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

water-balloon method. In some cases, reference curves can be effectively


used to estimate Dr or confirm Yd . . . .
2. Use of 6-in. diameter cutting cylinders to obtain ~/d of sand containing
gravel is unsatisfactory because of the effects of the gravel and sampling
shear stresses. It may be possible to reduce the influence of the gravel by
using larger diameter cutting cylinders having smaller perimeter sample-
area ratios.
3. Relative densities obtained from measurements of individual layers
from a manually excavated pit are more representative of the actual condi-
tious, because the volumes of sand in the layers are considerably greater
than the volumes of sand in conventional test holes. Determination of ~/~
and Dr in this manner has a number of disadvantages which include; de-
struction of a large volume of completed earthwork, and the large amount
of time required to make the excavation and measurements and to restore
the earthwork to its original conditions. If the grain characteristics and
index properties of the sand are relatively uniform, Dr obtained from indi-
vidual layers in a few pits can be compared to the results from conventional
methods. The results of such comparisons can indicate the extent to which
the results from conventional methods represent the completed earthwork.
4. Denison samplers can be used to obtain samples from which satis-
factory Dr can be obtained. There is the possibility that recovery of samples
from below the water table will be difficult. Use of the Denison sampler is
recommended in situations where quantitative values of Dr are required
and Dr inferred from N are extraordinarily high.
5. Approximate Dr values can be determined from in-situ wet unit
weights obtained using nuclear methods, corrected water contents from the
Speedy Moisture Tester or by oven drying, and Yd max and y~ rain or refer-
ence curves.
6. The use of correlations based on standard plate-loud test results is not
an expedient means of obtaining quantitative Dr during earthwork con-
struction. It is, however, a means of qualitatively estimating relative den-
sity. It is suggested that when standard load plate settlements versus Dr
are required for earthwork construction control, the relationship be estab-
lished for the project either during the design stage or at the start of
construction.
7. Very high N values are obtained in compacted sands because of both
the high Dr and residual lateral stresses caused by vibratory compaction of
the sand in layers. There is considerable difficulty associated with a quanti-
tative inference of Dr using N in such cases.
8. Static cone penetration resistance (Rp) can be used to obtain qualita-
tive measures of relative density of completed earthwork. Because a large
number of Rp values can be obtained rapidly, the uniformity of the work
can be evaluated readily during construction. This is particularly useful in
evaluating the consistency of the results of method specifications.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License
LEARY AND WOODWARD ON CONSTRUCTION CONTROL 401

Acknowledgments
The authors gratefully acknowledge Nebraska Public Power C o m p a n y
(Cooper Nuclear Station) and Consumers Power Company, The Detroit
Edison Company, and Ebasco Engineering Corporation (Ludington
P u m p e d Storage Project) for permission to publish the d a t a from their
projects.

References
[1] Earth Manual, U.S. Department of the Interior, Bureau of Reclamation, Washington,
1968.
[2] D'Appolonia, D. J., Whitman, R. V. and D'Appolonia, E., "Sand Compaction with
Vibratory Rollers" Journal of Soil Mechanics and Foundations, Vol. 95, SM1, Jan.
1969.
[3] Terzaghi, K. and Peck, R. B. in Soil Mechanics in Engineering Practice, 1st ed.,
1948 and 2nd ed., 1967, Wiley, New York.
[$] Gibbs, H. J. and Holtz, W. G., "Research on Determining the Density of Sands by
Spoon Penetration Testing," Fourth International Conference on Soil Mechanics and
Foundation Engineering, Vol. 1, London 1957, p. 35.
[5] Bazaraa, A. R. S. S., "Use of the Standard Penetration Test for Estimating Settle-
meats of Shallow Foundations on Sand", PhD thesis, University of Illinois, Urbana,
Ill., 1967.
[6] Sehultze, E. and Muhs, H., Bodenuntersuchungen fiir Ingenieurbauten, 2nd ed.,
Springer, 1967.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No furt
S. J. Poulos 1 and A. Hed ~

Density Measurements in a Hydraulic Fill

REFERENCE: Poulos, S. J. and Hed, Alexander, " D e n s i t y M e a s u r e m e n t s


i n a H y d r a u l i c Fill," Evaluation of Relative Density and Its Role in Geo-
technical Projects Involving Cohesionless Soils, ASTM STP 5~3, American So-
ciety for Testing and Materials, 1973, pp. 402-424.

A B S T R A C T : Field density measurements and compaction tests on uniform


sand placed hydraulically and without subsequent compaction showed a linear
correlation between percent compaction and relative density, the mean values
for the fill being 89.3 :t= 3 percent and 50 -v 16 percent, respectively.
It was found that the maximum density obtained in compaction tests is
strongly affected by apparently minor variations in grain size distribution.
Hence, to obtain reliable estimates of percent compaction, it was found neces-
sary to perform a one-point compaction test on the soil excavated from each
field density test hole, rather than applying one compaction test to a whole series
of field density test results.
Some advantages of percent compaction over relative density, as a measure
of degree of compaction, are enumerated.

K E Y W O R D S : cohesionless soils, density (mass/volume), sands, soil com-


paction, compaction tests, percent compaction, field density, hydraulic fill

The hydraulic placement of sand fill for a section of the New Jersey
Turnpike was monitored b y numerous, Carefully performed field-density
tests, compaction tests, and grain-size determinations. The field density
and compaction tests show the distributions of density and percent com-
paction that are achieved in hydraulic fill composed of a clean, uniform,
medium sand and not subjected to mechanical compaction. The test results
are used to show (1) the relationships t h a t exist between relative density,
percent compaction, and maximum density, (2) the effect of slight fluctua-
tions in grain size on the maximum dry density, and (3) the advantages of
percent compaction over relative density as a measure of degree of dense-
ness for this sand.

1President, Geotechnical Engineers, Inc., Winchester, Mass. 01890.


8Soils engineer, Praeger-Kavanagh, Waterbury, N. Y. 07010.

402

Copyright by ASTM Int'l


Copyright9 by(all rights International
ASTM reserved); Fri Marwww.astm.org
11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No fur
POULOS AND HED ON HYDRAULIC FILLS 403

Description of the Hydraulic Fills


The data presented in this paper were obtained from two different por-
tions of the Turnpike, which were let in two separate contracts. For con-
venience, these two fills are designated Fill I and Fill II. The lower portion
of each fill is referred to as Level A, and the upper portion as Level B.

Fill I
Transportation and Placement--The sand for Fill I was dredged from a
deposit of very silty sand on the continental shelf in Raritan Bay, New
Jersey. The material was dredged hydraulically and pumped into scows.
The water that overflowed the gunwhales of the scows carried away most
of the fines. The filled scows were towed to a creek which crossed the job-
site, and their contents were dumped into a hole t h a t had been excavated
in the b o t t o m of the creek. More fines were lost in this process, as the slowly
settling fines were carried away b y currents. Another dredge was used to
pump the sand from this underwater stockpile through 2000 to 5000 ft of
27-in. pipe to the site of filling.
The fill was placed across a marsh in a 150 to 300-ft wide trench t h a t had
been excavated through peat to Elevation (EL) - 1 0 to - 1 5 (based on
USGS mean sea level), at which elevation firmer material was encountered.
Brackish tidal water from the Hackensaek River filled this excavated zone.
The tidal range was 4 to 5 ft. The sand fill was placed in two levels as
follows:
Level A (bottom of excavation to E L + 6 ) - - F i l l IA was constructed in one
lift b y placing the invert of the 27-in. diameter hydraulic-fill pipe at E L
+ 6 and pumping until a fan-shaped deposit was formed with a slope of

FIG. 1--Photograph showing dischargefrom ~7-in. diameter hydraulic-fill pipe and bull-
dozers moving sand from front of pipe.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
404 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

about 1 vertical on 20 horizontal. Pumping was then stopped, a bulldozer


pushed sand in front of the pipe to form a level surface at EL + 6, another
section of pipe was connected, and the filling operation advanced. The
bulldozer was also used during pumping to spread the sand that accumu-
lated near the discharge point.
Level B ( E L + 6 to E L + 20)--Fill IB was placed in lifts with thicknesses
of 8, 3, and 3 ft. Dikes were first constructed around the zone to be filled
using the uppermost sand from the previous lift. The dikes were 1 or 2 ft
higher than the desired top of the lift. At locations opposite to the dis-
charge point drain pipes were placed'through the dikes. The remainder of
the operation was as described for Fill IA.
The photograph (Fig. 1) shows the sand-water suspension discharging
from the hydraulic-fill pipe and bulldozers moving accumulations of sand
from near the discharge point. Cross dikes just behind the discharge point
were built up to prevent backflow of water onto the already placed fill.
Probably not more than 15 percent of the total volume of sand placed
was handled by the bulldozers during the hydraulic-fill operation for Fill
IB.
Description of Sand in F i l l / - - T h e grain size distributions of 212 field-
density specimens taken from Fill IB are summarized in Fig. 2. The soil
ranges from a very uniform, medium to fine sand, to coarse sand with a
wide range of grain sizes.
The sand contained no more than 3 percent fines. The median grain size

FIG. 2--Grain size distributions for ~12 field-density specimens from Fill IB. Sand from
Raritan Bay.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
POULOS AND HED O N HYDRAULIC FILLS 405

2" j" 3~4" ~/i' ~ " 4" s%~ 3o* 5o ~ s o= 20r


I00
' I'
90

80

70

30 l
'; ifill 1~
2O ! I
10

0 I0 2 I
5O
GRAIN SIZE, mm

FIG. 3--Median grain size curvesfor field-density specimens from Fills IB and II.

curve for the specimens from Fill IB in Fig. 2 is shown in Fig. 3. The size
for which 10 percent of the particles are finer (the D10 size) is 0.31 mm, and
the coefficient of uniformity (C~ = Deo/Dlo) is 3.1. The process of moving
the sand from Raritan Bay to the fill caused removal of practically all fines
and transformed the deposit from one that is not well suited to underwater
placement to one that is practically ideal for that purpose.
The sand that was placed in the fill is composed of bulky, angular to
subangular grains that are chiefly quartz, with considerable feldspar, and
some augite and mica. It also contains a variety of seashells, which were
present in practically all specimens and comprised about 2 percent of their
weight.
Fill I I
Transportation and Placement--The sand for Fill II was dredged from a
deposit of very silty sand on the continental shelf off Staten Island, New
York. The procedure for transporting the soil to the fill site and the de-
scription of the site are essentially the same as for Fill I. In this case, the
sand was pumped directly from the scow to the fill through a 20-in. pipe.
Other details of the fill operation are the same as before, except as noted
below.
Level A (bottom of excavation to E L + 6)--No bulldozer was used at any
time.
Level B (EL + 6 to E L +20)--Lifts with thicknesses of 8, 3, and 3 ft
made up this fill. One front-end loader was used for handling the sand.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lic
406 RELATIVEDENSITY INVOLVING COHESIOHLE$$ SOILS

Description of Sand in F i l l / / - - T h e grain-size distributions of 207 field-


density specimens taken from Fill II are summarized in Fig. 4. The soil
ranges from a very uniform, medium sand to a somewhat more widely-
graded, coarse sand. No specimen contained more than 2.5 percent fines.
The coefficient of uniformity (C~) ranged from about 2 to 7. The median
grain size curve for 80 percent of the specimens is shown in Fig. 3, where
D10 is seen to be 0.23 mm, and C~ is 2.4. The sand in this fill is slightly
finer and slightly more uniform than that in Fill I.

Nonhomogeneities in the Fill


The range of sizes in the fill material, the variations in slope of the fill,
and velocity of flow over it during filling resulted in considerable stratifica-
tion and lateral variation in the fill. Upon close inspection, such as when
making a field-density test, one could distinguish layers as thin as 1/~6 in.
and, in some cases, even thinner layers could be found. On a larger scale,
layers 2 to 14 in. thick could be found.
Local high velocities, such as near the point where the discharge im-
pinges on the fill, form layers of coarse particles. At one location a thin
layer of gravel was discovered at EL + 10. The grain size curve for this
gravel is shown in Fig. 2, and a photograph of it is shown in Fig. 5. These
large grains were subrounded to subangular. The finest soils in the fill were
layers of reddish-brown clayey silt up to 2 in. thick which were deposited

FIG. 4---Grain size disSributions/or $07 field-density specimens/ram Fill II. Sand from
bay oH S ~ is~nd.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License
POULOS AND HED ON HYDRAULIC FILLS 407

FIG. 5--Photograph of $-in. thick seam of gravel from Fill IB. Largest particle size is
~in.

on the fill surface when water from abandoned ponds evaporated and
drained through the fill. At some locations isolated boulders as large as 16
in. diameter were found on the site.
Other nonuniformities in the fill arose from the construction procedure.
The dikes were built with sand at water contents that caused bulking.
Therefore, the edges of the fill were very loose. In a number of cases the
transverse dikes were not removed and recompacted, so that loose spots
occurred there also. Such unusual zones were normally avoided when the
locations of the field density tests were selected.

Test Procedures

Field-Density Test
Field densities were determined by means of the sand-cone method. The
tests were made at the bottom of 3 by 3 ft pits with depths of 1 to 7 ft.
Two to three tests were made at each level.
In view of the stratified nature of the fill, it was extremely important to
obtain the density specimen from one layer only and to take sufficient soil
from that same layer to perform all desired tests, including maximum and
minimum density, compaction, and grain size. If the specimen comprises
material from two or more layers, the maximum density of the mixture is
usually greater (but is sometimes less) than the maximum density of the
individual layers. Thus, the percent compaction or relative density would
be meaningless if layers are mixed in the field. Much care was taken to
ensure that there was practically no visible stratification within the zone
from which the field-density specimens were taken.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agree
408 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

Modified AASHO Test


Normally the modified AASHO compaction curve shows a peak at an
optimum water content corresponding to a degree of saturation of 92 to 95
percent. However, for very clean, uniform sands, the optimum water con-
tent corresponds to 100 percent saturation, a fact that has been known for
some time [1].a The modified AASHO compaction curve for such a soil is
concave upward, as shown in Figs. 6 through 8. The oven-dry sand has a
higher dry density than partially saturated sand, because it is not acted
on by capillary stresses which inhibit densification. When the soil is fully
saturated, such that no capillary stresses exist, this sand reaches a slightly
higher dry density than is attained for the oven-dry soil.
After it was observed that the modified AASHO compaction test yields
a U-shaped compaction curve for this sand, it became possible to determine
the maximum dry density simply by compacting duplicate specimens at
full saturation. There was no longer any need to produce the entire curve.
Just enough water was added to the density specimens to produce a layer
of water on the surface (after compaction of each layer) not more than
1/~6 in. thick.

115
I00 --~IO0

IlO

95U 75

I05
0
901_ 5O ~
e ~ " t : ~0o.9 pc, Z
IO0
Q
85 CL ~~:
~l"J

- - ~n~ ~ 9 3 3 pcf ~-~0


J 80
5 IO ZO
WATER CONTENT~ %

ioo

5
8O
\
,.=-6o Cu=2 3 ~~ ~ " O.15rnm

~ 4o
~ 2o
~ Olo=O.O65mm
0 i I , 4 9
~00 I0 1.0 0,1 .01
GRAIN SIZE, mm

FIG. 6---Compaction tests, grain size curve, and field density test for Specimen 178.

a The italic numbers in brackets refer to the list of references appended to this paper.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agree
POULOS AND HEO ON HYDRAULIC FILLS 409

I00 00

115 ~e
?'= 114.1 ~r
orion
9s
,o L

90 o~
f..

IOO ~ Y m i n : I 0 0 . 3 p~t ss ~ 0

95 i i
5
WATER
I0
CONTENT, %
,~ ~o BO

I00--

~40 Cu - 3 . 7 .
\
\. O.l~m
0 ~ -- I I ~I~.-~I0
I00 I0 l.O 0,1 .01
GRAIN SIZE, mm

FIG. 7--Compaction tests, grain size curve, and field density test for Specimen 179.

It should be noted, however, that sands with only 10 percent fines may
have the usual peak in the compaction curve. No assumption should be
made about the shape of compaction curves until complete compaction
tests have been performed.

Minimum Dry Density Test


Two procedures were used to measure the minimum dry density, de-
pending on whether or not the specimen contained fine gravel sizes. If no
fine gravel was present, then the oven-dry sand was poured through a
funnel that was held I/~ in. above the sand surface. The spout of the funnel
was s/~ in. diameter. If the sample contained fine gravel sizes, the soil was
placed by pouring it from a spoon very carefully from a height of about
1/~ in. A 4-in. diameter compaction mold was used as the container in both
c~es.

Maximum Dry Density of Oven-Dry Sand


The maximum density of oven-dry sand was measured using the modified
AASttO test procedure.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License
410 RELATIVEDENSITY INVOLVING COHESIONLES5 SOILS

120

~0 IOO
115 - ~ l" ~

* "_-.~-'~-. 100% SATURATION r


95 i 75 ~_
I10

.o!
z

~ 105
85 25 ~
IO0
,--- ~fmln= 97 2Pcf 0
I I
5 I0 15 2O
WATER CONTENT, %

,7 60 ~: D60 =o.2Bmm

40 Cu=34 \
20
\
i~Olo .0.083 mm
0 i I
I00 I0 1.0 O. I .01
GRAIN SIZE, mm

FIG. 8--Campaction tests, grain size curve, and field density tests for composite of Speci-
mens 300, 301, and 302.

Definitions of Measures of Denseness


Percent Compaction--The percent compaction is defined as 100 times the
ratio of the field dry density, ~I, to the maximum dry density measured ia
a modified AASHO compaction test, ~a:

e ( % ) = 100 ~__f (1)


"YA

Relative Density--Relative density is defined as:

R~(%) = 100 (-emax___-~__6_L.~ (2)


\emax -- emia/

where:

ema~ = maximum void ratio (oven-dry sand),


emi, = minimum void ratio (oven-dry sand), and
el -- field void ratio.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agre
POULOS AND HED ON HYDRAULIC FILLS ,41 1

Relative density can also be computed from the measured dry densities as
follows:

= too ( - mio (3)


k'~max -- Tmin/ "Y:
where:
~/max ---- maximum dry density (corresponding to emin) and
~min = minimum dry density (corresponding to emax).
Note that relative density is normally based on the maximum dry density
of the oven-dry sand (~/m~x), whereas percent compaction is based on the
maximum dry density obtained using the modified AASttO compaction
procedure, irrespective of the water content at which it is achieved. For
the data in this paper, both relative density and percent compaction were
computed based on the latter maximum dry density. For the sands tested,
the modified AASHO maximum occurred in practically all cases when the
sand was fully saturated during compaction, as shown in Fig. 9.

l i i I i

150 ~..
+

+ 9
o,

125 " ~ o " "


~- .o +
9 § ~-*. ~ 9

+ . -
~-++ .~..~.1. ..
t20
§ ", +4- §

++r

IIC

9 fill I

I05

I00

,~o ,~ ,Io ,I~ ,~'o ,~


"asm. T OVEN DRY, pcf

FIG. 9--Comparison of maximum densities obtained on saturated and oven-dry specimens


using Modified AASHO compaction effort.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further
412 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

I00 r

90 "~.' ' ' ' ' ' ' ' ' ' '

.o

70 OUPLtCATE MAXIMUM DENSITY TESTS,a

~< so
$o "\\...
~
z 40 .\
,,r 30 ~ ~'-DUPLICATE MINIMUM DENSITY TESTS,b

a. 20 "~2~:~:~.
I0 :~':'~:~'~ ~ ~ ~:~_==~.. ~
0 I I I I I I I [ I ~" ] I
1.0 2.0 S.O 4.0 5.0 6.0 7.0 80
VARIATION IN DENSITY, pcf
FIG. lO---Reproducibility of maximum and minimum dry density test results, Fills I
and I I .

Laboratory Test Results

Reproducibility of Density Test Results


Figure 10 shows the reproducibility of the maximum and minimum
density test results for all specimens from Fills I and II. Curve (a) shows
that the maximum dry densities, measured with the modified AASHO
test procedure on duplicate specimens of fully saturated soil, differed by
1.2 lb/ft 3 in 50 percent of the cases. For 90 percent of the cases the differ-
ence was less than 4.0 lb/ft 3. Curve (b) shows that two measurements of
milfimum dry density on the same specimen differed by 1.0 lb/ft 3 for 50

130

>. 125

!
=E 120
Z 'Y" / : i:'"

It5

I10 I
/ I I I I
85 90 95 I00 105 I10
MINIMUM DRY DENSITY, pcf

FIG. ll--Maximum versus minimum dry density for Fill IB.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
POULOS AND HED ON HYDRAULIC FILLS 413

percent of the cases. For 90 percent of the cases, the variation in minimum
dry density was less than 4.0 lb/ft 3.
Relation between M a x i m u m and M i n i m u m Dry Density
Figures 11 and 12 are plots of the maximum dry density versus minimum
dry density for Fills IB and IIB, respectively. These two densities are
linearly related by the equations:
Fill IB ~m,x = ~m~n + 24 (:i=2%) (4)
Fill IIB ~m~x = ~(m~n+ 24 (=t=1.5%) (5)
The two dry densities differ only by a constant.
Normally the maximum and minimum dry densities are used together
with a measured field density to compute the relative density of the soil
in situ. It is usually assumed that the behavior of the fill is correlated with
the relative density. Since for this soil the two density values from which
relative density is computed are not independent values, only one of them
is needed.
Although for any given deposit of soil there may be a relation between
minimum and maximum density, it cannot be expected that the same rela-
tion will exist for all soils. Hence, it could be that relative density has merit
for comparing the engineering properties of two different soils. That is, it
may be (although the authors do not consider it probable) that different
soils at the same relative density may have similar engineering properties.

135

13C
/ .' -

125
o

_=

120

: : ' "!-i.:" " / .


115

Iio I I I I I
85 90 95 I00 105 !10
MINIMUM DRY DENSETY, pcf

FIG. 12--Maximum versus minimum dry density for Fill IIB.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Li
414 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

135 i 1 ~ I i i

130

125

~ 12o!
|
H5

I ,~l I I 1 I I I I
I10 I 3 4 5 6 7 8 9 I0
COEFFICIENT OF UNIFORMITY, Cu = D 6 0
D tO

FIG. 13--Maximum dry density versus coe#~cientof uniformity for Fill I.

Relation between M a x i m u m Dry Density and Grain Size


Figures 13 and 14 are plots of maximum dry density versus coefficient
of uniformity, C~, for all samples for which the grain-size distribution was
determined in Fills I and II, respectively. These figures illustrate that the
more widely graded sand (higher C=) could be compacted to a greater
maximum density, as might be expected, because the finer particles fill the
voids between larger particles more completely. Increases of C~ above
about 6 seem to have no further influence on the maximum dry density.
For this particular soil a relation also exists between C= and Ds0 as shown

135

~
t25

o 120

115

I10 I I I I I I I
2 3 4 5 6 7 8 9 I0
COEFFICIENT OF UNIFORMITY, Cu 9 0 6 0
D I0

FIG. 14---Maximum dry density uersus coe~cient of uniformiSy for Fill II.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lic
POULOS AND HED ON HYDRAULIC FILLS 415

3,.0

2.5

20
** 9

9 ~
A
E
s
. . . .
9 "~. . 9 o~176
i " : "
.... "-:.: :. ..
:. "?:~
.:?-::.::. ~ .
-.Oo. ,
9 . , ,

::::: o! ~176
9 o

.5 "!i.';.:".: :
1.0 ' '" "2 '. ' ' ! I0.0

C0EFFICIEINT OF UNIFORMITY, Cu

FIG. 15--D60 versus coe~cient of uniformity for Fill I specimens.

in Fig. 15 for Fill I. The method of transportation and placement of this


fill caused the specimens with smaller De0 also to be more uniformly graded.
(It seems that the percent passing the No. 200 mesh remained relatively
constant due to the method of transportation and placement, while the left
side of the grain-size distribution curve oscillated in accordance with the
content of coarse grains in the original deposit.) Because there is a relation
between D60 and C=, it is not possible to determine for this soil whether
the changes in uniformity or in average particle size have larger effect on
maximum density. However, it is expected that the degree of uniformity
controls, since only minor effects on density would be anticipated for such
a small change in average size.

Density Measurements
Field Density, Percent Compaction, Relative Density
The field densities varied from 93 to 125 lb/ft ~, with a median value of
approximately 110 lb/ft 8. Fig. 16 shows the reproducibility of the results
of the field density tests. Duplicate samples from the same layer in the fill
differed in dry density by 1.4 lb/ft 3 for 50 percent of the tests, and in 90
percent of the tests the difference was 4.4 lb/ft 3 or less.
Figures 17 and 18 show the cumulative frequency distribution of values

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licen
416 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

IO0 r

80

~ 60

~_ 4 0

~ ~o
2o

I I I I I I " ~
0 0 I 2 3 4 5 6 7
VARIATION IN FIELD DRY DENSITY , pcf

FIG. 16---Variaticn between measured values of field dry density for duplicate tests in
Fills I and II.

I00 i J =.A._.~; , , , ~ ~ , , , ,

90 ~ -~ ~..\

fifl ]] \\
50 ~ ""~'~

ol_ ,,,
i fl. ~A-~, ~, \
0 I" . . . . . . . . . . ~" " , - --~-
~ '~ ~ q ~
80 I 2 3 4 5 6 7 8 9 90 I 2 3 4 5 6 7 8 9 ]00
PERCENT COMPACTION

FIG. 17--Percent compaction of as-placed hydraulic fill. Cumulatiue frequency distribu-


tion curves for Fills I and II.

tO0
90 x. .\

80

5O
~, 40

20
IO

I0 20 50 40 50 60 70 80
RELATIVE DENSITY, PERCENT

FIG. 18--Relative density of as-placed hydraulic fill. Cumulative frequency distribution


curves for Fills I and II.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to
POULOS AND HED ON HYDRAULIC FILLS 417

of percent compaction and relative density that were obtained for Fills
IA, IB, IIA, and IIB. The results are summarized in the following table.

Fill No. Median Values Degree of Denseness, %

C. D~o, m m 90%GreaterThan Median 90%SmallerThan

P Rd P Rd P Ra

IA 3.1 0.31 87.3 34 90.0 53 94.0 73


IB 3.1 0.31 86.3 36 89.6 53 93.1 68
IIA 2.4 0.23 86.3 27 88.8 43 91.0 56
IIB 2.4 0.23 86.3 34 88.8 50.5 92.8 66

For 90 percent of the cases for both fills the median percent compaction
averages 89.3 • 3, which corresponds to a relative density, in percent, of
50 • 16. These values are considerably lower than those that would usually
be considered acceptable for a fill.
It is difficult to assess the physical meaning of the values of relative
density, but the values of percent compaction have a direct physical mean-
ing. The precent settlement (namely, the vertical strain) of a given layer
of fill in one-dimensional compression is given by the difference between
the final (which may exceed 100 percent) and original percent compaction,

P ---- p r
~(%) = 100 X ~ -- P (6)

Assuming that traffic vibrations may cause an increase in percent compac-


tion from the value achieved in these fills to a value of, say, 95 percent of
the maximum, the percent settlement would be:

9,(%) = 95 -- 89.3 (• = 2.7 to 8.7% (7)


Assuming further that a 3-ft thick layer is dens• to 95 percent of maxi-
mum, the total settlement would be about 2 • i in. Thus, the distribution
of values of percent compaction can be used to make estimates of the
maximum total and differential settlements that might develop subse-
quent to placement. Study of the variations in density in specific limited
zones would be required to judge the seriousness of these potential differ-
ential settlements.

Effect on Density
Figure 19 shows a cumulative frequency distribution of relative density
values determined in Fill IB at three different depth ranges below the fill
surface. The lower values observed in the top 3 ft may be attributed, at

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant
418 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

tOO

9O
"~" ~\ ,~ f ! "-D~Sft ( l ' g s o r t i e s )
80

mTO
\ ' '\\ ' ~ ~9

,;4o
. . . . . D < 3ft (69sample D >Sft (63 somples)

r
0
0 ,o zo 30 4o .o so 70 80
.~-AnVe OCNSrrY.~.C~.T
FIG. 19--Effect of depth on relative density of as-placed hydraulic tilt.

IOC ; I f ! f I f I f f f f I --

90

80

70

r 60

>_-
1-
50 I
~J
c~

40

/
20

1 I I I [ I I J i I I I
7 8 9 90 I 2 $ 4 5 6 7 8
PERCENT COMPACTION

FIG. 20---Relative density versus percent compaction for Fill I.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agree
POULOS AND HED ON HYDRAULIC FILLS 419

least in some instances, to disturbance caused by the spreading and level-


ing equipment. The similar distributions of Rd values in the two lower
depth ranges suggest that depth has practically no influence on the density.
The densities measured at depths below 3 ft probably can be considered
representative of the entire thickness of a hydraulic fill. Thus, for all prac-
tical purposes, the data from all depths may be considered as a unit when
studying other variables that affect the results.

Relation between Relative Density and Percent Compaction


The correlation between relative density and percent compaction for this
soil is shown in Figs. 20 and 21 for Fills I and II, respectively9 For practical
purposes, the relationship is a straight line. The existence of these correla-
tions shows directly that relative density and percent compaction may be
used interchangeably for this soil.

I00
i I i I i i ~

90

80
o ."

70

9 ......'.: .:
s ~o

....-:El.. :.':..-
9 .. . . . .

so
9 . . . . . : . .
9. .9
9 . .': . -. : -
... : :.....-.:-":: ..'.
40

. .- .,'-.

f .

9 .

20

o I i i I I I I I I I I I
3 4 5 6 ";' 9 90 I 2 3 ,~ 5
PERCENT COMPACTION

FIG. 21--Relative density versus percent compaction for Fill II.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No furthe
420 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

Relation between Maximum Density and Percent Compaction


Figure 22 is a cumulative frequency distribution of the percent compac-
tion in situ for samples that fell in three ranges of maximum dry density.
A clear increase in percent compaction with maximum density can be dis-
tingulshed. Not only did the actual density rise with maximum density, as
would be expected, but even the percent of maximum density increased.
Because the maximum density increases with both C~ and D~0 for this soil
(see section on relation between maximum dry density and grain size), it is
not possible to determine which of these parameters governs the percent
compaction achieved by hydraulic filling. However, intuition would suggest
that the percent compaction should decrease as C, increases. By elimination
it would appear then that De0 may control the percent compaction achieved,
and that the effect of C~ is small.
Based on the preceding and the section on the relation between maxi-
mum dry density and grain size, it is suggested that the maximum density
of this soil is largely controlled by the grain size distribution (C~), whereas
the percent of maximum density achieved by hydraulic filling is controlled
chiefly by the grain size (D60). Field tests on soils with carefully chosen
grain size curves are needed to confirm this tentative conclusion.

Discussion

Comparison of Relative Density and Percent Compaction


The comparative utility of relative density and percent compaction
should be judged on the basis of whether one or the other is better correlated
with such properties as strength and compressibility. The maximum and
minimum density used to compute Rd and the maximum density used to
compute P are all index properties. They are valuable only if they can aid
in judging the engineering properties of soils.
Figures 20 and 21 show that percent compaction and relative density
are not independent for this soil. They are linearly related in the range of
interest. Thus, neither parameter can be said to be better correlated than
the other with engineering properties of this soil. Since neither parameter
has any technical superiority over the other, one can choose between them
simply on the basis of their economy, ease of performance, and precision.
Relative density is more cumbersome in practice for several reasons.
Most important, one can use this parameter only for clean uniform sands
and gravels with not more than a trace of dry strength. If the soil is non~
uniform, one cannot obtain a proper maximum density due to segregation
(which is a drawback for both measures of denseness), and, if the soil is
too fine, the minimum density cannot be measured accurately. Thus, rela-
tive density is, at best, applicable only to a very limited range of soil types.
Secondly, two index tests must be performed to obtain relative density,
and inaccuracies occur in both measurements. Thus, the accuracy of rela-

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to L
P O U L O S A N D HED O N HYDRAULIC FILLS 421

rive density values suffers. For example, the variation in percent compac-
tion between duplicate specimens taken at the same elevation, in the same
test pit, in similar material was 0.75 percent or lower in 50 percent of the
cases, and 1.6 percent or lower in 90 percent of the cases for Fill II. The
corresponding variations in relative density were 4.8 and 12.2 percent.
Using the correlation in Fig. 22 between percent compaction and relative
density, these later two values correspond to 1 and 2.5 percent in terms of
percent compaction. Thus the measured variation in percent compaction
at the 50 percent level was 25 percent smaller, and at the 90 percent level
40 percent smaller than the measured variation in relative density.
Thirdly, relative density suffers in that it has no direct physical meaning
with respect to the behavior of a fill. On the other hand, percent compaction
is related to the potential settlements of a fill. The difference between any
two values of percent compaction is equal to the percent settlement re-
quired to cause that change in percent compaction.

9o, , -~__.\
80 - \
', \
!
70 \~ ~'mox ~ll5pcf (Ssomples)
, \

60 \\ ~'m~ <120pcf (27samples)

u) 50 ~ ~ >120pcf U77somples)

t
\
\
t

20

I0 ~

',,,
80 t 2 3 4 5 6 ? 8 9 90 I 2 3 4 5 6 7 8 9 1(30
PERCENT COMPACTION

FIG. 22--Relation between maximum dry density and percent compaction achieved in
hydraulic fill without subsequent mechanical compaction.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
422 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

In summary, it appears to the authors that percent compaction is a


considerably more useful parameter than relative density for measuring
degree of denseness.

Common Errors in Measurements of Degree of Compaction


Frequently, when controlling a fill, the result of one compaction test is
used for computing percent compaction for a large number of field-density
tests. If this practice had been followed for controlling this fill a completely
erroneous picture of the percent compaction of the fill would have been
obtained. The maximum density is very sensitive even to minor changes in
grain size distribution. Hence, it is essential that a compaction test be per-
formed on the same material as that of the field-density test.
Proper attention is also not given, in many cases, to the presence of
layering in the field density specimen. If a soil is stratified, the result of a
compaction test on a mixture from the several layers will bear no relation
to the result that would be obtained for the individual layers. Extreme care
is essential when trying to establish the degree of denseness of any stratified
soil.
It has been known for many years that the modified AASHO compaction
curve for clean, uniform sands is U-shaped, and that the maximum density
occurs at, or very near, 100 percent saturation. Nevertheless, one often sees
reported results for such soils that display the conventional peaked com-
paction curve. Because such a soil has to be very wet (saturated) to attain
its maximum density, the person performing the test may hesitate to add
the needed amount of water. To add so much water is contrary to experi-
ence with practically all other soil types. The fact that full saturation is
needed in such soils to achieve maximum density offers one distinct ad-
vantage; with little experience one can obtain the maximum dry density
by obtaining only one or two points on the compaction curve.

Compaction of Clean Sand in the Field


The fact that 100 percent saturated or oven-dry sand yields the maximum
density in the laboratory indicates that clean sand should be compacted
in situ either very dry or in the presence of sufficient water to cause full
saturation during compaction. The presence of capillary stresses in the
ground inhibits compaction. Field tests are required to determine whether
full saturation will significantly increase compaction when modern vibra-
tory rollers are used. The increase in compaction efficiency must be suffi-
cient to warrant the use of the expensive watering procedures that may be
needed to obtain the desired high water content at the moment the roller
passes over a given point. Whether the addition of water is appropriate is
dependent on the densities required and on the economics of the compaction
operation.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lic
POULOS AND HED ON HYDRAULIC FILLS 423

Discrepancy between Density Criteria for Rolled and Hydraulic Fills


Specifications for rolled sand fill usually provide for compaction to 70
percent relative density or greater. In the present case, only 50 percent or
less relative density (90 percent or less compaction) was achieved by hy-
draulic filling. Yet, based on previous experience, the fill was considered
satisfactory for a first-class highway. The difference between the criteria
for fill placed by scrapers and dozers, as compared to hydraulic fill, is at
first surprising. One would expect the same criterion to be used for a given
soil regardless of method of placement. The difference may be related to
the fact that hydraulic fills have traditionally consisted of clean sands,
whereas the specifications for rolled fills apply to much poorer materials in
general. But there is one justification for use of more stringent criteria for
rolled fill even for the same clean sand. A rolled fill is more likely to have
nonuniform density, because it is usually placed at water contents that
cause bulking. Unless the rolling operation is thorough, these initial non-
uniformities cannot be satisfactorily eliminated. The hydraulically-placed
clean sand is less likely to contain local nonuniformities that lead to differ-
ential settlements.

Summary
Density measurements of a clean uniform sand placed hydraulically and
without subsequent compaction yielded a median percent compaction of
89.3 • 3.0 percent of modified AASHO maximum dry unit weight, which
corresponds in these sands to 50 =t= 16 percent relative density.
It was shown that for the sand used in this fill (1) the maximum and
minimum densities are linearly related, (2) percent compaction is linearly
related to relative density for the range of values of interest, (3) a change in
coefficient of uniformity from 2.5 to 6 (with simultaneous increase in D60
from about 0.5 to 2.2 mm) caused the modified AASHO maximum dry
density to increase from 114 to 131 lb/ft 3, and (4) this wide range of maxi-
mum density made it imperative that a compaction test be performed on
each field-density specimen to obtain a proper measure of percent compac-
tion in each case. Because the variations in this soil were relatively small,
it appears that this procedure, or a simplified substitute, should be used
for control of practically all fills.
It was pointed out that percent compaction has the following advantages
over relative density as a measure of degree of denseness: (1) percent com-
paction can be used for practically any soil, whereas relative density is use-
ful only for clean, uniform sands; (2) two tests are needed to obtain rela-
tive density, but the two results are not independent for this soil, therefore,
one of the tests (with its associated errors) can be discarded; and (3) a
change in percent compaction of 2 percent (say 90 to 92 percent) is numeri-

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licens
424 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

cally equal to the percent settlement needed to cause the change. Relative
aenslty has no such direct physical significance.

Reference
[I] Leonards, G. A., Foundation Engineering, McGraw Hill, New York, 1962, p. 1011.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No fu
W . I . L o w 1 a n d C. S e n e r ~

Field and Laboratory Determination of


Maximum Density in Coarse Sands and
Gravels for Mica Dam

R E F E R E N C E : Low, W. I. and Sener, C., " F i e l d a n d L a b o r a t o r y D e t e r m i -


n a t i o n o f M a x i m u m D e n s i t y i n Coarse S a n d s a n d G r a v e l s for M i c a
D a m , " Evaluation of Relative Density and Its Role in Geotechnical Projects
Involving Cohesionless Soils, A S T M S T P 523, American Society for Testing
and Materials, 1973, pp. 425-443.

A B S T R A C T : Checks on the placement of cohesionless embankment shell ma-


terials for the 800-ft high Mica Dam by field and laboratory density tests
were required to ensure that the properties of the placed earth fill were com-
patible with design assumptions. As a result of studies made in the design of a
quality control system for the dam, it is believed that the data obtained from
large field specimens can be considered representative of the fill material, but
only if the comparative laboratory data are also obtained on large specimens.
In addition, it appears that the representative test results can be obtained
only by sampling the total (or near total) material and testing without a
major reduction in the maximum particle size. The resulting specimen sizes
necessitated some modifications in the conventional test methods and develop-
ment of large apparatus to accommodate a wide range of materials. The
equipment and procedures adopted or developed for density testing on coarse
sands and gravels in the field and in the laboratory at Mica Dam are described.
Methods for predicting the laboratory densities and the relative compaction,
based on three years of extensive testing, axe presented.

K E Y WORDS'- cohesionless soils, density (mass/volume), sands, gravel dams,


earth fills, tests, compacting

The quality control needed during construction of earth dams requires


rapid and accurate assessment of the properties of the compacted fill.
Construction of larger dams, use of larger capacity equipment, and higher
placing rates has required a review and redesign of testing methods,
equipment, and methods of fill quality evaluation.
The principal parameters in the design of a dam embankment are shear
1 Area engineer and soils engineer, respectively, Mica Dam, CASECO Consultants,
Mica Creek, B. C., Canada.

425

Copyright by ASTM Int'l


Copyright9 by(all
ASTMrightsInternational
reserved); Fri Mar 11 16:13:06 EST 2016
www.astm.org
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
426 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

strength, permeability, and compressibility of the earth fill. It is not feasible


to test these parameters directly in the field, because direct quality control
is based primarily upon control of gradation, placement moisture content,
and density of the fill material. The relationships between the two variables
of placement moisture content and gradation, and the maximum laboratory
dry density were investigated at Mica Dam by laboratory testing, and a
correlation between gradation and the laboratory maximum dry density
was established.
Formerly, the concept of relative density was frequently used as a means
of assessing the quality of granular fill materials. However, over the past
eight to ten years, fill materials used on large dams have tended to contain
large particles, typically ranging from 6 to 24 in.
The engineer designing a quality control system for such a dam is thus
faced with the following problems:
1. What are the size and frequency of field density tests required to
provide an accurate picture of the in situ density of the fill?
2. What are the size and frequency of laboratory tests required to
provide an accurate comparison with field tests?
3. Are results of laboratory tests on a portion of the total field sample
of coarse materials valid as a comparison with field tests?
The senior author's experience in the use of the relative density test on
sand and gravel materials containing particles of 6 to 12-in. maximum size
at the W. A. C. Bennett (formerly Portage Mountain) Dam [1]~ indicated
that determination of minimum density was impractical in such coarse
materials, and as a result, the concept of relative (or percentage) compac-
tion, based on comparison with maximum density only, was substituted as
a fill quality criterion in place of relative density. In addition, it was found
necessary to carry out a laboratory determination of maximum density for
each field density determination in order to obtain a reliable measure of
compaction performance.
The granular materials used in Mica Dam [2, 3] were even coarser than
those in the Bennett Dam, with maximum particle sizes ranging from 6 to
18 in. In the quality control system designed for this project, emphasis was
placed on very large field specimens and laboratory testing on the entire
material. As with the Bennett and Oroville [4] Dams, the relative density
concept was not used, and the density criterion was based solely on relative
compaction.
D e t e r m i n a t i o n of Soil Control Parameters
General
The methods of sampling and testing for quality control of the materials
for Mica Dam were selected after a thorough review of both past and
2 The italic numbers in brackets refer to the list of references appended to this paper.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to
LOW AND SENER ON COARSE SANDS AND GRAVELS 427

current practice and of the nature of the materials expected to be used in


the dam [5-9].

Field Density Determination


In-place density tests of large diameters and volumes were considered
essential for the accurate determination of the density of coarse grained
embankment materials.
Maximum specimen sizes (namely, weights) were selected on the basis of
judgement after consideration of typical borrow area material gradations
and the individual weights of the largest particles anticipated in various
zones of the dam.
The size of the field density test pit, and consequently, the amount of
specimen, was related to this maximum particle size and controlled by
placing a steel template ring on the fill surface before excavation.
A pit of the prescribed dimensions was then excavated within the ring,
and the volume of the pit obtained by water replacement after lining the
pit with plastic film.
Specimen and test pit sizes used at Mica Dam for sand and gravel shell
materials are presented in Table 1.
Specially equipped field sampling vehicles were required because of the
large specimen sizes and the rapid placement rate of materials. The vehicle
used for testing was a 3-ton four-wheel drive fiat deck truck, equipped with
a 150-gal calibrated water tank with gage glass for water volume measure-
ment, hoses, a water pump to remove the water and anti-freeze solution
(needed in the spring and fall months) from the test hole for reuse, a hoist
for lifting the test ring, and specimen barrels. Space was also provided on
the vehicle for transporting loaded barrels containing the excavated
specimen to the laboratory.
Gradation and Moisture Content Determination
Numerous mechanical devices were provided for the project laboratory to
reduce manual labor and to accelerate testing. Experience elsewhere had
indicated that the gradation testing of large specimens of coarse materials
could not be successfully accomplished by conventional laboratory equip-
ment, since such apparatus is incapable of handling large rocks and would
be subject to mechanical breakdowns under continuous operation. There-
fore, in addition to conventional size testing apparatus, the laboratory was
provided with special equipment to accommodate the large specimens
obtained from the field.
A vibratory screening plant with four 2 by 6-ft decks was installed to
determine the gradation of the coarse fraction (from 3/~ to 6 in.) of the soil.
The gradation of the -3/~ in. fraction was obtained in standard laboratory
sieving apparatus by standard American Society for Testing and Mate-
rials (ASTM) method (including drying the specimen) after reducing

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License
<
m

.<

T A B L E 1--Sample and test pit sizes.


<
or -
Zones Maximum Maximum Specimen Test Pit Size Approximate Cross _<
Z
Specified Layer Particle Size, Weight, lb Section of Test Pit o
Thickness, in. in. Diameter, in. Volume, ft 3
0-I-

Shells, transition areas 6 6 900 54 6.30 o


z

Shells, general areas 18 18 1600 54 10.60


tn
o
Shells, outer parts 24 24 2200 54 14. O0

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
LOW AND SENER ON COARSE SANDS AND GRAVELS 429

and splitting the specimen to a lesser weight. Material larger than 6-in.
maximum size was hand picked from the specimen and sized and
weighed separately.
Drying sands and gravels in large quantities was not practical due to the
coarse nature of the material, although the normal procedure is to base
gradation calculations on dry weights. Therefore, to obtain the coarse
(+~/~ in.) gradation of the specimen without drying, comparative tests
were performed on dried material and material screened at field moisture
content. The following rapid methods of gradation control were developed:
(a) A correction of 10 percent by weight was assumed for the material
from + a/~ in. to -- 6 in. to allow for absorbed water and coating of - 3/~ in.
material on the + a~ in. rock. That portion of the correction factor which
relates to coating is approximately 1 percent, and typically would amount
to about 4 to 7 lb in a specimen of 400 to 800 lb. Because the accuracy of
weighing the total specimen is approximately • lb, no further correction
is applied to the - a/~-in, fraction for loss in coating retained or the + a/~-in.
fraction.
(b) No assumption was applied for + 6-in. material. Measured weights
were used as dry weights.
Handling of large specimens in the laboratory was performed by two
1-ton electrical hoists on a monorail over the vibratory shaker and extending
over an area where the maximum density test mold and table were located.
Because of the coarse nature of the materials, the moisture control was
based on that portion of the specimen finer than a 3/~-in. U. S. Standard
Sieve and was determined by quick pan drying on six large propane burners.
The total specimen moisture content was obtained directly from standard
charts using a family of correction curves, which were developed from the
fundamental correction equation which is based on the empirical approxi-
mation of 1 percent moisture content for the material retained on the
a~-in, sieve [7].

Laboratory Maximum Density Determination


It was considered necessary to carry out laboratory tests on a very large
part of the field specimen in order to avoid use of excessively large theoreti-
cal correction factors for oversize content. It has been stated [8] that
correction factors in excess of 25 percent produce increasing errors in
determining maximum density. Examination of the gradation data on
typical borrow area materials indicated that removal of material larger
than 8 in. would result in a loss of not more than 15 percent of the total
material, and this size was therefore chosen as a convenient and suitable
limit for laboratory test specimens.
No valid or acceptable method is known for determining minimum
density in materials with maximum sizes up to 8 in., that is, particles

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lic
430 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

weighing up to 50 lb. Therefore, no attempt was made to base construction


control on the relative density concept, and relative compaction was used
as a comparison criterion.
The maximum density was obtained by vibrating the -8-in. portion of
the specimen in a mold 25.3 in. in diameter and 30 in. deep.
The ratio (3 to 1) between diameter of the mold and maximum particle
size tested was considered a controlling factor in the selection of mold
diameter. It was also considered that the depth of the specimen should be
at least twice the maximum particle size tested. Therefore, for a maximum
particle size of 8 in., a diameter of 24 in, was the minimum mold size
required. To accommodate about 850 lb of material, and to get a convenient
whole number factor for volume calculations, a mold 25.3 in. in diameter
was selected. The mold was mountable on a 4 by 4-ft, 4000 lb capacity
vibrating table equipped with an electromagnetic vibrator operating at a
frequency of 1200 rpm (the same frequency as the field rollers), and an
amplitude of approximately 0.025 in. Initially, a static surcharge weight of
1000 lb was provided, but it was found that the high acceleration (in excess
of 1 g) caused this weight to bounce and crush the surface soil particles. The
surcharge load was subsequently applied to a light steel surcharge plate by
pneumatic jacks. Density testing was performed under 2 psi surcharge
pressure, since it had been shown that the maximum density could be
achieved between 1.85 and 3.70 psi [7].
Approximately 850 lb of moist material obtained from the compacted
embankment, after field density testing, was used to conduct the maximum
laboratory density test. Filling of the density mold was carried out carefully
and slowly to ensure that the specimen in the mold would be as uniform as
possible, and so that no coarse material would be permitted to segregate
against the side and top of the mold. A sufficient amount of water was
added while filling the mold to obtain maximum density at or close to 100
percent saturation. In fact, Figs. 1, 2, and 3 show that the lines of optimum
moisture contents found in the laboratory indicate about 90 percent
saturation. Experience showed that the materials were sufficiently free-
draining to shed excess water from the top of the specimens during vibra-
tion. The correct amount of water was estimated by judgement and by
experimentation with the material. The rate of water addition was limited
to prevent migration of fines at the time of compaction under vibration. (In
the field, no close control on moisture content was exercised, but the
specification provided for addition of water to the shell zones within broad
limits when it was considered necessary. This was required from time to
time. In addition, a maximum field moisture content based on 90 percent
saturation was required for materials which were less free-draining than
average.)
After filling, the mold surface was leveled and the surcharge plate
positioned close to the surface using spacers of various lengths to bolt the

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursu
LOW AND SENER ON COARSE SANDS AND GRAVELS 431

jack base plate to the top of the mold. After completion of the assemblage,
the mold was vibrated under 2 psi surcharge load for a total of 15 min, and
settlement measurements made approximately every 5 min by means of two
depth gages positioned on opposite sides of the surcharge base plate. It has
been found that variation of the surcharge load between 2 and 0.7 psi
during the last 3 min of vibration results in higher density values. The final
volume and density of the specimen were then computed. During testing it
was noted that the decrease in depth of the specimen, consequently the
increase in density, was not significant for periods greater than 15 rain.
After finishing the density determination, the entire material was dis-
charged into the vibratory shaker without prior drying, to obtain the total
specimen gradation.
The correction of the total specimen field-density measurement for coarse
aggregate content was necessary before calculating the relative compaction,
since the total specimen gradation was different from that tested in the
laboratory maximum density test by the amount of + 8-in. cobbles that
were removed by hand. This is obtained by using a family of standard
correction curves [7].

Method for Prediction of Laboratory Maximum Density


It was speculated that, when comparing the results of field compaction
with those obtained in the laboratory, the principal variables affecting the
density of the fill material (other than differences in compaction procedures)
would be, in order of importance, gradation and placement field moisture
content. Also, any definable relationship between the three variables of
laboratory optimum moisture content and gradation and laboratory
maximum dry density was expected to reduce the amount of laboratory
testing required for direct field control of compaction and to significantly
improve the effectiveness of control testing. Therefore, a considerable
number of laboratory maximum density test results were analysed,
representing varying gradations of each borrow area material. A correlation
between the laboratory maximum density, moisture content, and the
cumulative sum of the percentages of material passing U.S. Standard Sieves
3 in., 8/~ in., Nos. 4, 20, 60, and 200 was established for various materials to
be placed in the embankment. Grain size distribution curves and laboratory
maximum density prediction charts for each borrow area materials are
shown on Figures 1, 2, 3, and 4.
It was found that as both the maximum particle size and coarse material
content increased, the maximum dry density increased, and the optimum
moisture content decreased. The average line drawn through the optimum
moisture contents represents the highest dry densities that might be
expected for a range of gradation from coarse sand and gravel to fine silty
sand and gravel. This line is referred to as the "Line of Optimum Moisture
Contents from Laboratory Compaction Tests."

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lic
432 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

FIG. 1--Laboratory maximum density prediction chart for Material A.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agree
LOW AND SENERON COARSE SANDS AND GRAVELS 433

FIG. 2---Labarat~ry maximum density prediction chart for Material B.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agre
43,4 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

FIG. 3---Laboratory mazintum density prediction chart for Material C.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
UNIFIED SOIL CLASSIFICJ,TIOfl SYITEN
SILT SAND GRAVEL
I COIILEII,. ! -'.~"-".OiRII
FINE 1 .mum Ir FINE 1 COARSl
U . S, STAllING SIEVE " SIZE SCREEN SQUARE OPININI
! . . . ,, : l
ioe

Jill I I . ~ ~' IN) 6=E


I i/~','.,d?l'
6"17.# aO
+ z

?0
I#" b" / l // E
I # " I,~'..'l d
/ .~.~*'~ / ' : , ! ! & 0
z

0
40
/ .X ~- \~ .,i'~ ./ m

.'"l'~< A ~" _..~"J"\ .t


.,-" i ' \ ~_---"'" _--'.,~J,
./...~.-I"
.,' -','~'---~-
--"~" 20
z
I ,o 0
: ' I il~-"_~-~r"~-H "- I i I III111 1 111 11
i i I 1 ~ 1 ~ ~ 1 IIIIII 1 I I IIIIII 1 III II llll
t I-tllll l
* . . , ". -; ' " i 2 i
GRAIN SIZE IN MILLIMETERS : I
FIG. 4--~rc.iation of lnaler~ils, i.~

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
436 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

To obtain the laboratory maximum dry density from the prediction


chart, the following steps are used:
(a) The total specimen gradation of the excavated sample is recalculated
on the basis of -8-in. material.
(b) The predicted optimum moisture content and maximum dry density
for the -8-in. fraction are read from the appropriate chart by using the
cumulative percent finer by weight figures for each sieve size 3 in., 3/~ in.,
Nos. 4, 20, 60, and 200.
(c) To obtain an average, the individual predicted optimum moisture
content and laboratory maximum dry dei~ity values are plotted on the top
of the chart and examined by using the saturation lines. Those density
values which plot over 100 percent or under 80 percent saturation lines are
excluded as erroneous, and only the values between 100 and 80 percent
saturation lines are included in average calculations.
(d) The calculated average of the predicted density values is assumed as
the most probable laboratory maximum density for the material thus
analyzed, and the most probable optimum moisture content is obtained
from the appropriate chart by using the Line of Optimum Moisture
Contents from Laboratory Compaction Tests, shown on Figs. 2, 3, and 4.
Analysis of Results by Statistical Approach
The statistical significance of the difference between the actual laboratory
test values and the predicted mean values was examined by calculating the
standard error of the differences for the three variables (Table 2). It was
noted that for Materials A and B, the observed differences were less than
one standard error, and for Material C, were equal or close to two standard
errors of difference. This indicated that the differences between actual and
predicted mean values were not statistically significant. It was also noted
that for maximum dry density and optimum moisture content, the standard
deviation of predicted results was less than the standard deviation of actual
test results, while for relative compaction these were approximately equal.
Frequency distributions of the variations for actual and predicted
maximum dry density, optimum moisture content, and relative compaction
are shown for the three materials in Figs. 5, 6, and 7, respectively. No
explanation, other than uncontrolled (random) variation, was found for the
abnormal distributions. The maximum discrepancy ranges are equal or
close to two standard deviations of average discrepancies (Table 3), which
indicate that the prediction method was sufficiently accurate to be used as
a basis for compaction control, namely, for obtaining relative compaction
values.
Since actual maximum dry density results were compared with predicted
maximum dry densities, it was also necessary to examine whether or not
there was a significant difference between the end results of the two
procedures. This was done by testing the significance using Student's

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to L
T A B L E 2~Statistical summary.

Material A ~ Material B~ Material C"

--8-in. Laboratory M a x i m u m Actual test results Arithmetic mean (2~) 145.95 146.24 150.10
Maximum Dry Density, Standard deviation (al) 4.12 3.04 2.47
lb/fta Predicted test results Arithmetic mean (~2) 145.93 145.95 149.46
Standard deviation ( ~ ) 3.40 2.35 2.01
Difference between actual and predicted arithmetic means
0.02 0.29 0.64
Standard error of difference between arithmetic means

%/var. (~1 -- x2) = ~ a l 2 + a2~n2 0.50 0.38 0.27

- 8 - i n . Laboratory Optimum Actual test results Arithmetic mean (~) 5.70 5.40 4.64
Moisture Content, percent Standard deviation (at) 1.15 0.91 0.66
Predicted test results Arithmetic mean (~2) 5.74 5.64 4.81
Standard deviation (~2) 0.94 0.64 0.50 z
Difference between actual and predicted arithmetic means
0.04 0.24 0.17
Standard error of difference between arithmetic means
O
0.14 0.13 0.08 Z
O
Relative Compaction, percent Actual test results Arithmetic mean (21) 99.24 98.66 98.97
Standard deviation (al) 2.18 2.14 1.99 ~
Predicted test results Arithmetic mean (~2) 99.21 98.77 99.34
Standard deviation (a~) 2.20 2.22 2.27 z
Difference between actual and predicted arithmetic means
0.03 0.II 0.37
Standard error of difference between arithmetic means
O
0.29 0.30 0.26 ~
N
N u m b e r of applicable tests Actual (n~) = Predicted (n~) 114 102 139

a Materials are listed in the increasing degree of coarseness. "4

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
MATERIAL MATERIAL MATERIAL
nCll nBII llAU m
m
40

0
30 immmliijjHjjHjj) ..4

I0
!'!'!'!'i~176176176176176176176
~ z

0
I00 r~
BBNNNNNNNNNBIONPNaNNNNNN
Z ilillllimiOiliff~iiliilli 0
ililllllgilmlP~ililillli z
i- lminilliimlmDHUilmlmOllO
l- illiliUliill~lllllillinl ~n
(n iliiliUimll~nnimolimiimm
u) lmimmmmmmmwlmmmmimmmmmmm
~J mmmmmmmmmmAmmmmmmmmmmmmm o_
immmimmimumimmmmmmmmm|mm
" 50 mmmmmmmm~mmmmmmmmmmmmmmm
l- mmmmmmmw~mmmmmmmmmmmmmmi
Z |mmmmmm~mmmmmmmmmmmmmmmm
W mummmm~mmmmmmmmmmmmmmmmm
O mmmmmNmmmsmmmmmmmmmmmmm
|mmmm~mmmmmmmmmmmmmmmmmm
mmmm~mmmimmmmmmmmmmmmmmm
mmm~immmmmmmmmmmmmmmmmmm
mi~aimimimmmmmmmmmmmmmmm
,=immmmmmmmmmmmmmmmmmmmm
u-6-5-4-3-2-I 0 I 2 3 4 5 6 7 o.5-:~-~-~-T-o--,--s -s163163 7 6 7
Variation of differences between Variation of differences between Variation of differences between
predicted and actual minus 8in. predicted and actual minus 8in- predicted and actual minus 8In.
maximum dry densities, pcf maximum dry densities~ p c f maximum dry densitles~ p c f

FIG. 5---Frequency distribution of density variations.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
pIFHI;ENT LP..~3 I HAN NO' OF T E S T S
;U
|IIIIIIIIIIIIIIIllIIIIIllIIlllI IIIIIIIIIIUIIIIIIIIIIIIUIIIII
|IIIIIIIIIIIIIIIIIIIIIIUIIUIIIIIIIIIIIIIUUIIIII IIIIIIIIIII
iIIIIIlIIIlIIIIIIlIIIIIInUIII iiiininiiiiiiniiiiniiiiiiii
iiIIIIIlIIIIllIIIIIIIIIIIIIUIIIIIIIIInnIuIIUII iiiiiiniiI
IIIIIIIllIIIlllIIIIIlIIIIIInIIIIIIIIUIInIIUIIII IIIIIIIIIII
iiiiiliiiiilililiiiliiinllnUlUlnllnUlUlUlllllllllllli
l,llIllllIIIllIlllllll lllllllnlnllllllllnlllllllllllllI
iiIilllillllilllllllll IIIIIIIIIIIIlUlIIIIIIIII InllllnlII
3 a il,~llllllllllllllllll IIIIIIIIIIIIIIIIIII IIIIIIIIII II
o =p Illllllllllllllllllll Unlllnllllllllll iiiiIiiiilII
r r ]imm~-~qilUml'n'nnimmulnu mnuui
ilninmml'.;-"qUlaimlmmmmmlU Inumi
]iianliamnaazqnllinlnl
o~ 9, - nnnnnnnnnnunnn nn-,---.,, u n n nnnnnnu,
IIIlnillIllllUi~illl IIIll
InIlIlmmmIlUIIllI--'iI IlllI
9,,nnnnn,,nn nnnn n n a n a n , , , , m nn,,,,,,nn,,,,~ ,,,,,,,,,,,,-,,,,,,---,,,-,,,m
iumulilnmulnmimuuum,, nlulnlnnunumninmunnnmunnnmi
inninllnuiilimillillUllllUllllllllllllUllllllln iiiiiinlli
i nllnnliulnnillglminnnnllllllnllllnnllllllnlUll iiiiiiiiili
uninmlnnmulnlilnlinlimlllllllllllllllllllllllnllUll ilnllllllli
imlnnllilnlilUilmllllllUllllllllllmnllllUlllUnl iiiiiinllUi
ilnlillllllilllilnninllllllnlnllllUllnlnnllll iiiiiiiilli
ilinmlluililinmllUillllllUlllllnllllnllllnllllll ilUlllllll
l PENGENT LESS ;HAN NO' OF T E S T S
r r r - -
iiiiiiiiiiiilllllllillllnlllllnlnllllllllllllllllllllllll
IIIllllllllllllllllllllllllnllUlllllllnllllllllllnlUlll
E|-'-'-"--'|-'-'|'-|-'7,-'-'7,-'-'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII
Ilnlmmll #lmmlnllmlmmlmm iiiiinlllllnnnlnllllnllllllllUll
I n n n m m m n m l m m m m m m l i m m I mmI I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I
lllllilllllilllllllll IllllIlIllUlUlUlIIIIIIIIIInlIlllII
Itllllllllllll/llllll illllnlllUlnlllllllllllllllllllllll
I~llllllllllllllllll ilnnnlllnlnlllnllnnl, E
g- II~lllllllllllllllll IIIIIIIIIIInllnllllllllllll
Ill~lillllllllllllll IIIIIIIIIIIIIInlllnllll
Illli~lllllllllllill Inllllllllllllllllllllll ~m
IIIllli~llllllllllll IIIIIIIIIlUlIII
IIIllllll--'llllllllll IIIIIIIIIIIIIIII
Illlllllllli~lllllll IIIIIIIIIII
111lllllllllll-7-~llll IIIIIIIIIII r
l l l l l l l l l l l l l l l l l ~ l l IIInlllllllllllllllllllll
llllllllllllllllll~l iiiiinlnnlUllllllnlll
Illllllllllllllllll,I UllllllllUUlllllnllUllUlUlllll
Illllllllllllllllll|l IIIIIIiiiiiiiIInllllllllllUlUlllll
Illlllllllllllllllli nlllllnllllllllllllllllllllllllllnll
I l l l l l l l l l l l l l l l l l l l nlnllUlllllnlllllllllllllllllllllll
~ ~-" Illlllllllllllllllllnllllnllnllllllllllllllllllllllllllll
IIIIIIIIIIIIllllllllllllllllllllllllllllllllnnllllllllllll
aEHGENT LF..S5 THAN NQ" OF TE,~TS
O
...--
r
9 3
~ ~]~/~YaO O N V 5GNV5 3StlVQ3 NO ~N~S QNY /~O1
Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agree
4~
0
MATERIAL MATERIAL MATERIAL
Ilcll IIBII IIAII

l l l l l l l l l l l , 40 . . . . . . . . . . . . . . . . . . . . . . . . . 40 >
,3.-.-..-.-.-ll.-E--.-=-m.--ml.-._.-E__-- m m l i U m l i u m mma gn an immmmmmii mmm
:. . . . . _-.i_._.. _--.._. ..... = 9 I m mmN i l 1 iNn 1 l n l i l m mlmmm
::: ..................... mlmmmmm mmmm immmmm 1Rlm ammmm m mmm
!: :!: !: :!: i: :i :i:i: !: :!: i: :i :i:i: i: :!:i: i: :i i i i i i !
I I N m l l l l l m l l l l l l m l l l m l l 9 I mm II mmmuun m mmnmmm m l n m n
l l l l 1 m l m l l l l I . i N l i N I I N
: : : : : : : : : : : : : : : : : : : : : : : : : 9 m u mmm m mmm u m n I l l m m n m amu im n mmm u m
i l l l l B i l l i O n I l l
9lmmmlulummm m m m ummmn m u m l m n l l m m mmm
30 l l l m 30 . . . . . . . . . . . . . . . . . . . . . . . ~ 3o
m m m l I m P l l l I l l I l l l l l l m I l l immmm u a m n m n l l i m m m R m e mmm mmm m l l a mm
r l i n e I i l l l l m n l a B I- z
l l m l l l l l
m_--==_- = l U l l -, l m -- l m m_ mlmmm l l e l l mmmm n m
(n l i e n I I
o n e i N m I l l i n 1 l m i n I N m i l l 1,1 m i Iml nil mmI m m n m l l m mmn mmmmmmlm
&~l I l l l I l l l m l m l l l l l l l l l ::::::::::::::::::::::: .mlm
mi l l i N i n l m l l l l l m l l |lm Ul lllm mmmmllmmm a m l l m m l
. . um
. n.n.l l ml l um _.__..m.
(- i B I D e m m i l l
I- m . i l l i
mmlmnm lmnm m m l l m mmm m n u ~m
2O I I m l l m N e l l ~ ~0
u. 11 mm lull II m m l l I l l l l l l llm n m 9 m mlmmm mm uu imm mmml
I 1 0 9 all m e Ill m n lllml n II m m m
0 I 1 I I I n m l l l l l 9 iN m Im u m m u m
m l l l l m l Ill i n 1 m I l 9 mmm u m m m m m l m l l m m m
l
l
l
m l l l l
l l l l m
l l l l m
I l l l l l l l
I I I N I l l l
I l l l l l l l
)|ll-
| !'! 9 mmm i m m
9im m n l m
9m m l m n
mmmm
lm
m l
m mmmlm
mmm n I i n
m n l m m n
)
p.
Z 9 mmm m e n m mmmmm n m mmm m <
IO ! ) ~ - - . . . . . . . I .N 1. I N . mm~mm IO , ~l mi ~m
m m lmm m II m u m
| m l l l m i l l Ill l 9 mm z
1 l m l l m l l l m mmmmm ul m m m l m m llm
im
m
lm
I

I I
l
1 m
l l l l
i N
l l m I ~ 1
l l l l N l
II l m l
l m lmlll
i i
II m
~ l
'~ )i~)
9==-' I N '=-)|iii) 9a m m
m~mm
mmm
mmm
m m m m m m m m m m m
m e m l m e
mmillmllll
II m

I I m l mlm II m e mm m e
m l m l ~ m
iO0 Illlmmmmmnmmllmm~lumlmm I00 ~ m n n n n n m n m m l u i m n ~ ' d Ioo ;;mmmmmmmmmmmmmm~diiiiii
Imllmlmlmllmllm~lmllmlll I l l m l l m l l m l l l ~ mlmmmmmmmmmmmmm,,mmmmmmmmm
mmlmmmmmlmmmllm,~lmllmmlm I l l m l l l l m l m l l 4 i e I z mmmmmmmmmmmmmmr=mmmmmmmmm
Illlllllllllmlr, mllllmlml i m l l m l m l m m m m ~ m I J mmmmmmmmmmmmmmmrmmmmmmmmmm 0
-r z
i- Imlllmlmllllll~lmmmmllll I I I l I I I I I I I F l l I I mmmmmmmmmmmmm,Ammmmmmmmmm
illlnllmmllmmrJlllmlmlll I m l l l l m l m l W J l l I I I- uummmmmmmmmr, nmmmmmmmmmm
I,-
mlllmlllmmllm,#llmllmmllm il | m m l m l m l m m l l m l I[ i
o1
mlmmmmmmmmlm'llmlmmmmmmmm Ln
{e illmmlmmmlll~lmmllmlllll I l l m l l m l l l h l l l I I i (/) ,mmmmmmlmmmmmv, m m m l m m m m m m m m
ol Illimlllllllr~llllllm mmlm i m l m m l l m l l A l l l I l , mmlmmmmmmmmmammmmmmmmmmmm U~
u.I
Illmlmmllml~.llmmmlmmllll I m m l l l m l l ~ l m l l mmmmmmmmmm,~mmmmmmmmmmmmm o_
mllmlmlmll)~llmllmlmlmlll I l l l l l l l ) m l l l m I 1 " 50 mmmmmmmmmmmrJmmmmmmmmmmmmm
illmllmmll'~mllmmllmllllm I l l l l m l ~ l m l l m m I I ' I- mmmmmmmmm, A m m m m m m m m m m m m m
Ilmlllllmr~lmmllmllmmlllm i m l m m m ~ n m l m l m l z mmmmmmmm,~mmmmmmmmmmmmmmm
illmmmlmv, mlmmmlmmmlmmllm i m l l m ~ m m l l m m m l
Ill mmmmmmmm~mmmmmmmmmmmmmmmm
o
illliml~r I l l l l i l l l l l l l l (E ,mmmmmm,~mmmmmmmmmmmmmmmmm
ul lUlllmmrdllmllmlmlmllmlll I l l l ~ m l l m m l m l l
I| I I mlmmmm~mmmmmmmmmmmmmmmmm
G. illlml~,mmmlllmmmllmlllll i l I B l l l I I I I I I I mmmmmm~mmmmmmmmmmmmmmmmm
illlmw-mlmlmlmlllmlmmmll I m l ~ l l m l l l l l m m I~II, mommm~mmmmnmmmmmmmmmmmmm
Imll~mlmllmlmllllmlmllll I I I I V I Ill Ill lmnl~imlinBiiulmmuimnmni
mm-.mlmlnmlmlmlllmllmllll_.- OI I L~ I I [ I I I I II mmewmmmmmmlmmmmmmmmmmmmm
o O
-5-4-5-2-I 0 I 2 5 4 5 6 7 -5-4 -3 -2 -I 0 5 4 5 6 7 - 5 - 4 - 3 - 2 - I 0 I 2 3 4 5 6 7

variation of d i f f e r e n c e s between Variation of d i f f e r e n c e s between Variation of differences between


predicted and actual relative F)redicted and actual r e l a t i v e predicted and acfuot relative

compoctione ~ p e r c e n t compacfions, percent compacfions, percent

FIG. 7--Frequency distribution of relative compaction variations.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
TABLE 3--Statistical summary.

Material A9 Material B~ Material Ca

- 8-in. Laboratory Maximum Dry Den- Average discrepancy between actual and (~) 5
sity, lb/ft s predicted values -- 0.006 --0.29 --0.30
Standard deviation of the discrepancies (S) 2.18 2.30 2.65 z
Maximum discrepancy range +5.0/-5.4 +4.7/--4.5 +5.2/-5.o
- 8 - i n . Laboratory Optimum Moisture Average discrepancy between actual and (~)
Content, percent predicted values +0.05 +O.34 +0.16
Standard deviation of the discrepancies (s) o. 59 0.80 0.50 0
z
Maximum discrepancy range +1.3/--1.5 +1.8/-1.5 +1.2/--1.3
0
Relative Compaction, percent Average discrepancy between actual and (~)
predicted values --0.03 +0.05 +0.37
Standard deviation of the discrepancies (S) 1.48 1.41 1.46
Maximum discrepancy range +3.0/--3.0 +3.0/--3.0 +3.0/-4.0 z

Materials are listed in the increasing degree of coarseness.


O

4~
$ffi

CopyrightbyASTMInt'l(allrightsreserved);FriMar1116:13:06EST2016
Downloaded/printedby
(UFPE)UniversidadeFederaldePernambuco((UFPE)UniversidadeFederaldePernambuco)pursuanttoLicenseAgreement.Nofurtherreproductionsauthorized.
4a

T A B L E 4---Statistical summary [I0].

Material A~ Material B a Material C ~

--8-in. Laboratory M a x i m u m Sum of the squares of discrepancies between actual and pre- -4
Dry Density l b / f t a dicted 545.6 552.4 846.5
(~y~)
Algebraic sum of the discrepancies between actual and pre-
r--
dicted -0.7 -29.8 -41.7
(~ y ) Z
Average discrepancy o
f~
~y 0
--0.006 -0.29 -0.30
n

Variance
Z

v = s~ = ~Y--~' - ~ 4.79 5.33 6.99 O~


n

Standard deviation of discrepancies (S) 2.18 2.30 2.65


t distribution factor b

tffi
1P'-.~ I v Y - 1 0.03 1.26 1.30
8

N u m b e r of applicable tests (n) 114 102 139

a Materials are listed in the increasing degree of coarseness.


b Value of the group's mean discrepancy assumed as ~" = O.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
LOW AND SENER ON COARSE SANDS AND GRAVELS 4 4 3

t-distribution [10], Table 4. The difference between the two procedures was
not significant as the value of t did not exceed the value given b y the
5 percent probability level curve.
Conclusions
After a comprehensive study of the compaction characteristics of coarse
sand and gravel materials used for the shell zone's of the Mica Dam, it was
concluded t h a t the relationship between gradation, optimum moisture
content, and the laboratory maximum dry density forms a basis of direct
control of compaction on the embankment.
The prediction method is intended to provide a family of curves for
determining the maximum laboratory density and optimum moisture
content to calculate the relative compaction value. The method is based on
analysis of extensive experimental procedures, and it is suggested t h a t this
method be used as the basis of quality control for compacted fills.
For any project, establishment of a good correlation between density,
moisture content, and gradation, for different available borrow area
materials, will reduce testing to infrequent checking of fill materials having
significantly different gradations from the average of those used in the dam.
Some adjustments in plotted charts may be necessary after opening up of
the borrow areas and the beginning of fill placement.
Acknowledgments
The authors are indebted to British Columbia Hydro and Power Author-
ity and to CASECO Consultants for permission to publish this paper, and to
their colleagues, H. Nussbaum and K. S. Khilnani, for their helpful review.
References
[1] Low, W. I. and Lyell, A. P., Canadian GeotechnicalJournal, Vol. 4, No. 2, May 1967.
[~] Webster, J. L., Transactions, 10th International Congress on Large Darns, Vol. 1,
1970.
[3] Low, W. I., "Mica Dam Construction Methods and Control," Canadian National
Committee International Commission on Large Dams, CANCOLD/USCOLD
Joint Technical Meeting, Oct. 1971.
[~] Gordon, B. C. and Miller, R. K., Journal of Soil Mechanics and Foundation Engineer-
ing Division, American Society of Civil Engineers, Vol. 92, SM3, May 1966.
[5] Wilson, S. D. and Squier, R., "Earth and Rockfill Dams," State of the Art Volume,
Seventh International Conference on Soil Mechanics and Foundation Engineering,
Mexico, 1969.
[6] "Soil Density Control Methods," Highway Research Board Bulletin 159, Publica-
tion No. 498, National Research Council, Washington, D. C.
[7] Johnson, A. W. and Sallberg, J. R., "Factors That Influence Field Compaction of
Soils," Highway Research Board Bulletin 272, Publication No. 810, National Re-
search Council, Washington, D. C., 1960.
[8] Gordon, B. B., Hammond, D. W., and Miller, R. K., "Effect of Rock Content on
Compaction Characteristics of Clayey Gravel," Compadion of Soils, A S T M S T P
377, American Society for Testing and Materials, June 1964.
[9] A S T M Manual on Quality Control of Materials, A S T M S T P 15C, American Society
for Testing and Materials, Jan. 1951.
[10] Moroney, M. J., "Graphs of Students t," Facts from Fiaures, Penguin Books Ltd.,
Baltimore, pp. 216-236 (Fig. 81).

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
H. M . Reitz ~

Correlation Between Gradationai Param-


eters and Limiting Densities for Cohesion-
less Materials Placed Hydraulically

REFERENCE: Reitz, H. M., "Correlation Between Gradational Param-


eters and L i m i t i n g Densities for Cohesionless Materials Placed Hy-
draulically," Evaluation of Relative Density and Its Role in Geotechnical
Projects Involving Cohesionless Soils, A S T M S T P 5~3, American Society for
Testing and Materials, 1973, pp. 444-454.
ABSTRACT: Nonreproducible relative density test results for an hydraulic fill
of clean river sands, suggested errom either in assumed control densities or
methods or techniques of field testing. Control densities had been assumed to
vary only slightly because sample gradations and sizes appeared quite similar.
To verify this assumption, separate maximum and minimum density (ASTM
Test for Relative Density of Cohesionless Soils, D 2049-69) and gradational
parameters were determined for each field density test. No dependable corre-
lation of maximum or minimum densities or range between them with gra-
dational characteristics was apparent for results reported.
KEY WORDS: density (mass/volume), eohesionless soils, coefficient of corre-
lation, earth fills

A new fossil fuel electric generating plant is under construction on the


west bank of the Mississippi River approximately 55 miles below its con-
fluence with the Missouri River. The site selected is in the alluvial plain of
the Mississippi River. Originally, the plant fill was to be sand dredged from
the river. The utility company ultimately selected an alternate suggested
by the contractor to supply the plant fill by dredging from a flood plain
location immediately downstream from the plant with the volume of the
dredged borrow pit becoming available for ash disposal. This paper sum-
marizes the results of the maximum and minimum densities of about the
first 50 test series on this project. A period of 35 calendar days was involved
during which time approximately 230 000 yd 3of fill were placed. The testing
of maximum and minimum densities for each field density test of the hy-
l President, Reitz & Jens, Inc., St. Louis, Mo. 63105.

1,44

Copyright by ASTM Int'l


Copyright9 (all rights
by ASTM reserved); Fri www.astm.org
International Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreemen
REITZ ON HYDRAULICALLY PLACED MATERIALS 4 4 5

draulic fill material placed hydraulically gave a unique opportunity to


study the suitability of correlations reported herein.

Site C o n d i t i o n s a n d C o n s t r u c t i o n Procedures
Prior experience at two sites along the lower 150 miles of the Missouri
River, had indicated that a fill for warehousing and storage could be placed
hydraulically from sandy borrows. Considerations necessary for satisfactory
end results are: (1) attention to the discharge conditions from the dredge
pipe, especially the interim slopes, (2) appropriate escape velocities for
transporting only the finer grain materials away from the area of deposition,
and (3) these velocities also aid dynamically in packing more densely the
materials settling out.
The contract signed and the specification accepted were neither as com-
petitively bid nor as originally prepared by the design consultant and pro-
ject manager. These documents were modified to incorporate recommenda-
tions of an experienced dredging contractor who had suggested the changes
in construction methods and borrow location, and was awarded the con-
tract. As a condition of the proposal, the dredging contractor would
furnish and make available to the owner and its representatives the services
of a consulting firm experienced in soil mechanics and hydraulic fills
specifically. This offer of services was as much to expedite the filling which
was being done under a very tight time schedule regardless of weather, as
to assure compliance with the quantitative aspects of the specification
which were being continually checked both by the owner's representative
and the consultant to the contractor.
During the construction operation, the dredge pond is completely isolated
from the Mississippi River. The sequence of dredging and pumping through
a discharge pipe with return of dredge water from the fill flowing through
wide, relatively fiat swales into the dredge pond constitutes a completely
closed system which will not carry silts and clays in suspension to the
Mississippi River.
The soil profiles from the surface in the dredge pond are a mixture of
fine grain alluvial materials (silts and silty clays) that were as thin as 15 ft
and as thick as 35 ft above coarser cohesionless deposits that extended to
depths of 80 to 130 ft. The contractor proposed to dredge in a single pass
using a ladder that could be lowered to 55 ft beneath water surface. The
water surface in the dredge pond was held 15 to 25 ft below the natural
ground surface in the area. The purpose of the single pass was to pick up
the sands from depth and enable the mix in the dredged discharge to in-
clude some of the fine grain overburden material that went into suspension
and was picked up by the suction head. Much of the fine grain material
going into suspension in the dredge pond and during the actual cutting
operation, falls into the water and ultimately settles into areas previously
dredged behind the cutter.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Li
~ <

,i I
i ~
~ z

~~ 8=
~ o

~ 0

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
REITZ ON HYDRAULICALLY PLACED MATERIALS 447

TABLE 2--Comparison of Maximum Dry Densities, A S T M D ~0~9..69.

C~ Percent Finer Dry Method, Wet Method,


#200 Sieve lb/ft 3 lh/ft 3

2.4 1.8 112.8 114.6


3.6 1.2 117.8 120.5
2.0 2.0 114.4 113.7
2.0 1,1 112.4 111.8
1.9 2.0 109.5 109.3
2.0 3.9 116.3 113.1
1.9 2.5 113.4 112.5

The proposed volume of the borrow was approximately 41/~ million yd*
with the area of the borrow (ash disposal pond) 47 acres. The specifications
were properly set up on the relative density basis. The success of this
operation was dependent upon obtaining in-place fill densities of at least
60 percent relative density as determined by ASTM Test for Relative
Density of Cohesionless Soils (D 2049-69).
Testing Programs
Quantification of controls based on relative density as per AST~I D
2049-69 due to greater variability has proved to be more difficult than for

130--

125..~

llC

l 12o,-Z
~y Q
eo 9 9 9

eq 9

9 e ~
iiO--

Illlll'llllJ|lllll|Jlllll[
.05 .10 .15 .20 .25 .:30

D m (mn~

FIG. 1--Maximum density versus effective diameter.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
448 RELATIVE DENSITY I N V O L V I N G COHESIONLESS SOILS

f ,~0"~ J I I ] i I ! i

,25: ,
v

?.
-~llwllT,I ITl,, II l l l l l l , l , l l l l l II,,llllll'
.20 30 .40 .50 .60 .70 .80 90 1.00

Dso(mm ) + Cu = 2.0 or more

v Cu = I.r or less

BEST CORRELATION

FIG. 2 - - M a x i m u m density versus D6o.

compaction methods referenced to either ASTM Tests for Moisture-


Density Relations of Soils, Using 5.5-1b Rammer and 12-in. Drop (D 698-
70) or ASTM Tests for Moisture-Density Relations of Soils, Using 10-1b
Rammer and 18-in. Drop (D 1557-70). Therefore, at the start of the con-
struction operation, not only was the field density of the fill tested, but for
every field density test a maximum and a minimum density as per ASTM
D 2049-69 and a gradation by sieve or hydrometer methods or both were
run. The laborious nature of this testing requiring not only the fill density
as-placed, but also testing of samples taken at that time for density controls,
indicated great utility for correlations of maximum and minimum densities
with index properties which could give generally reliable quantitative re-
sults. The reduction of the time required for establishing controls in the field
laboratory could be spent more advantageously on the actual fill.
Approximately 50 sets of test results are used for this paper. No samples
with more than 10 percent by weight finer than 0.074 mm. (passing the No.
200 sieve) were encountered in the testing program. The results of these

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lice
REITZ ON HYDRAULICALLY PLACED MATERIALS 449

laboratory tests are shown in Table 1. Many more test sets ultimately will
be available. Determination of the maximum and minimum densities was
according to ASTM D 2049-69 with the reported maximum densities in all
cases being by the dry method.
As another comparison, the maximum density was also determined by the
wet method as well as dry methods ASTM D 2049-69 in at least seven test
sets. These maximum densities by wet and dry methods are summarized
in Table 2. A comparison of densities from the two methods shows that
while there was not complete numerical agreement, the difference varied
from 2.7 lb/ft 3 more to 3.2 lb/ft 3 less for wet method compared to dry. The
dry method averaged about 0.2 lb/ft 3 higher than the wet method. The
effect of method in determining maximum density should be investigated
further, especially for C~ > 2.0. The minimum density was determined
as prescribed for the fine grain soils.
Correlations were attempted betweer~ the densities and some aspects of
the grain size or the grain size distribution characteristics of the samples.
Specifically graphed were the maximum dry density, the minimum dry
density, and the differences between maximum and minimum as dependent
variables, using the "Effective Diameter" Dz0, the De0 size, or the coef-
ficient of uniformity as independent variables. Figures 1 through 9 are

I I I u I v

~=~
14.
25--
a.

E t2o--
e9 9
7
o9

Ios.~ ~e eQ e9
~ Oe e
oe 9 9

jlo-~
D 9

llllllllllulllllllljlllll III
150 2.00 2.50 3.00 5.rio 4.00

0to

FIG. 3--Maximum density versus cae~ient of uniformity.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreeme
450 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

II0- t I 1 i T I

i05--

,oo.-?.
-i ~ | I
95~ tit ee i

. i* *: .
9o~ 9 t
9 .;:. 9
Q 9

85~ e

eO 9

80 ~
lililll I I ! j I I I I I I I I Ill i I'l i|
9~ .10 .15 .2O .25 .~10

Djo (ram.)
FIG. 4.--Minimum density versus e~ective diameter.

graphs of combinations of variables to show relative correlations. Figure 10


graphically shows the limiting gradations of all soils tested but does not
show any one specimen.

Discussion
For this compilation of data and analysis, uniformity of conditions be-
tween successive samples was considered to be unusually good. This being
an actual construction project, there was no effort made to have uniformity
of samples; however, the nature of the deposition of material out of a dredge
discharge pipe gave a relatively uniform method of deposition. Natural
soil variations which existed in the profile beneath which the cutter head
was working were minimized to some degree by the fact that the cutter
head was held in the coarser sands at depth and that the material tested in
each of the samples was not the suspension as it existed in the discharge
pipe, but was the coarser portion of it as deposited on the fore slopes of the
fill. The personnel doing the sampling and testing were the same through-

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License A
REITZ O N HYDRAULICALLY PLACED MATERIALS 45 |

105-- I I I ! I I ! ! |'

I10--
I,r
e,
v
9 |
95~

9 ~0 00
90-- D $
D DO
Mgo -
85-
9 %
QO

Ill20 II I~i~ I I I1 I 4I I 0I II~)I I I~1t ~0111 I(~1 119~I il&

D6o (ram.)
FIG. 5--Minimum density versus Ds0.

out the entire series and were experienced in this type testing, giving uni-
formity of sampling technique.
The soils all had the same general geological history. Even though they
came from locations that differed horizontally by up to 3000 ft, they were
initially deposited by the waters flowing in the Mississippi River, and any
variation was not considered to be significant for the testing program.
Correlations between densities and some property of the grain size
distribution were generally very poor. It appeared that better correlations
exist when using a coarser portion of the materials, especially if the coef-
ficient of uniformity indicated something other than an extremely well
sorted material. The best correlation was maximum density and Dr0. For
all of the samples considered, the coefficient of correlation (r) for "r m,x based
upon dry testing methods and De0 was r = 0.570 for all samples. If only
those with a coefficient of uniformity greater than 2.0 were used, the corre-
lation was r - 0.654.

Conclusions
For the testing of fill materials placed by hydraulic methods on this con-
tract, correlations of control densities with gradational properties of sam-

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lic
,452 RELATIVEDENSITY INVOLVING COHESIONLESSSOILS

I10' i i I I I I

IO~t--

,oo-:

Q 9

9 o~ 9
9 moo
9 0 - -I
e ~e ~ o

gO O e e

85-- O'O

80
LSO 2.00 Z50 3.00 3.50 4.00

C"= 'D.)
FIG. 6---Minimum density versus coeOicient of uniformity.
9 I I-- I I

o9

|
@. -

9 9 ooe B
eo
9 9 ee eo ~

oe 9 9

20--

IIIlillillllltllllli[llli l
905 .|0 .15 .20 .25 30

Dto (ram.)

FIG. 7~Pazn0eof densities versun effective diameter.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
REITZ ON HYDRAULICALLY PLACED MATERIALS 453

u J T I I I r J

l
:35--
c'

ee
IE

a 25-- 9I l l e O L 0
IlO
i -

jllfllflflJfllllfIllJJltJIllfflllJil[ilT

Deo (ram.)

FIG. 8---Ranoe of densities versus Doo.

.... l I I I 1 I
5--

C
30--q e ~

OlOo o 9
I 25-Z 9 9 ooe 9 9
X
Q
9 aJ o e
E
e

2O--

LSO 2.00 2.50 :3,00 ~ 4.00

Cu = D ~
D0o

FIG. 9---Ranae of densities versus coe~cient of uniformity.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreeme
454 RELATIVE DENSITY I N V O L V I N G COHESIONLESS SOILS

U.S. STANDARD SIEVE SIZES

#200#140~100 #60 #40 ~20 #10 #4


I00 u i I lI I I I I v I I -i I

80--

60- LIMITS

40--

20--

0 n I U
9Oe "(~ O. I 0.2 0.6 1.0 2.0 7.0 I0.0
.01

GRAIN SIZE (mm.)

FIG. lO---Size range of test samples.

ples was generally unsatisfactory. Specifications for acceptance based upon


minimum relative densities encounter more erratic appearing field results
than percent compaction specifications and, therefore, dependable control
densities are greatly needed.
The best correlation obtained for independent variables of D10, De0, and
coefficients of uniformity and dependent variables of maximum dry
density (dry basis) (~d m~z), minimum dry density (~/~rain), and their differ-
ence ( ~ maz -- ~d rain) was for ~ max with D~. When all samples were used
which had coefficients of ,miformity of 1.25 to 3.25, the line of best fit had
r -- 0.570. When the samples considered had coefficients of uniformity
>2.0, r = 0.654.
A comparison of seven maximum densities (~d m~) determined for the
same samples for the wet and dry methods of ASTM D 2049-69 showed no
consistent difference in density due to method. Further investigation of this
comparison is suggested.
Acknowledgments
The author wishes to thank Bernard McMenamy Contractor Inc., the
dredging contractor, for cooperation in the field operations. The con-
tributions of R. T. Healey, J. W: Orr, and F. E. Palmerton all of Reitz &
Jens, Inc. in testing, statistical evaluations, and coordination and develop-
ment are especially acknowledged.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License
R. A. BelP and J. P. Singh ~

Comparison of Relative Densities Estimated


Using Different Approaches

REFERENCE: Bell, R. A. and Singh, J. P., "Comparison of Relative


Densities Estimated Using Different Approaches," Evaluation of Relative
Density and Its Role in Geoteehnical Projects Involving Cohesionless Soils, A S T M
S T P 5~3, American Society for Testing and Materials, 1973, pp. 455-462.
ABSTRACT: This is a case history illustrating data correlation on relative
densities measured by laboratory and field techniques on loose to medium dense,
saturated fine to medium sands. The field techniques include standard pene-
tration tests, Dutch cone soundings, and undisturbed sampling. Densities were
determined in the laboratory. A discussion is presented on how the results
were used to evaluate liquefaction potential of the sands under seismic vibra-
tions. Variability in the relative density results is presented in terms of sta-
tistical parameters. These parameters are compared with the statistical param-
eters obtained from other published results.
KEY WORDS- density (mass/volume), sands, liquefaction, tests, variability
cohesionless soils

One of the m a j o r seismic hazards on saturated eohesionless soil deposits


is liquefaction with accompanying ground m o v e m e n t s and foundation
failures. Similar phenomena caused serious damage to oil storage t a n k s
during the N i g a a t a earthquake of 1964. This paper outlines a case history
of a proposed t a n k f a r m at a site on the north coast of Java, where there was
the potential of a similar liquefaction problem in a s t r a t u m of loose
saturated sand. Since the basic soil p a r a m e t e r used in liquefaction analyses
is relative density, the results of subsurface d a t a obtained b y standard
penetration tests, D u t c h cone soundings, and undisturbed sampling were
interpreted to evaluate this i m p o r t a n t parameter.

Site Conditions
T h e t a n k f a r m site is on a broad, nearly fiat, alluvial or deltaic plain
which extends about 30 miles from the mountains to the sea. The ground

1Senior engineer and project engineer, respectively, Dames & Moore, San Francisco,
Calif. 94111.

455

Copyright by ASTM Int'lby


Copyright9 (all rights reserved);
ASTM Fri Marwww.astm.org
International 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No furth
456 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

elevation is 4 to 6 ft above high tide in the Java sea, which is located


approximately one third of a mile from the site. Ground water is within a
foot of the surface. The top 20 ft of the soil consist of loose fine-grained silty
sand, and the next 40 ft consist of normally consolidated soft clay.

Seismicity and Liquefaction Potential


The area is subject to moderate seismic activity, mostly from deep distant
earthquakes to the south toward the Java Trench. Based on statistical
analysis of the past 68 years of recorded earthquakes in the site vicinity,
the characteristics of the future earthquakes having a return period of 25
and 100 years were estimated to produce maximum ground accelerations of
about 0.1 and 0.15 g. The site is underlain by deep alluvium, and the
fundamental period of strong shaking was estimated to be about 2 s.
The liquefaction potential of the sand during the two levels of the
stipulated seismic activity was evaluated using the stochastic procedures
developed by Donovan [1].2 No laboratory liquefaction tests were performed
for this investigation. However, the results of cyclic compression tests on
Sacramento River sands with the particle size distribution shown on Fig. 1

GRAVEL SAND
SILT
COARSE I FINE COARSEI MEDIUM FiNE

\\
~?
\
-r
r
LU

).-
ca
SAMPLES COMBINED FOR
MIJ~II~/IIII~A ~ [~JIP.VI~,AIIP, A
ILl "
Z DENSITIES.
D50=0.075 to 0.1 mm
I--
Z OTHER SAMPLES
W DSO = 0.075 to 0.15 rnm
0
~2 ~--SACRAMENTO RIVER SAND
Q_
D50 = 0.20 ram(SEED et ol

I00 I0 1.0 0.1 0.01


PARTICLE DIAMETER - m m

FIG. 1--Particle size distributions.

The italic numbers in brackets refer to the list of referencesappended to this paper.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. N
BELL AND SINGH ON COMPARISON OF RELATWE DENSITIES 457

reported by Seed and Lee [2] were used. The cyclic compression test data
were appropriately modified to cyclic simple shear by using the correction
factors proposed by Seed and Peacock [3] and Finn et al [4]. The test data
was further modified for grain size distribution using the published data by
Lee and Fitton [5] and unpublished Dames & Moore data. It was concluded
that to prevent liquefaction, the required relative density would be greater
than 35 percent for a 25 year return period earthquake, whereas a relative
density greater than 50 percent would be required for a 100 year return
period earthquake.
Investigation Data
The subsurface conditions were investigated utilizing wash boring
techniques and Dutch cone soundings. The primary purpose of the sub-
surface explorations was to locate a site where sufficient sand blanketed the
weak underlying clays so that reasonably high tanks could be adequately
supported without overstressing the clay. After successfully selecting a tank
farm site with ample thickness of sand, primary attention was devoted to
evaluating the settlement and stability induced from the underlying weak

RELATIVE DENSITY
IN PERCENT
0 2O4O6O 8O I 0 0
I 1
DENSITY IN POUNDS PER CUBIC FOOT
0 20 40 60 80 I00 120
0 o I c,
I 0
I I
Q
o
0 o
o
I
go 9 oe ~ 0 o
0
I0 9 Pw.,1~
! ? u ~ ~-~)
9 o l
I--
W
1
I
' ke:
W I
~o
I g I
Z I
20 o~O8o
'1"
Q. KEY;
f,
MINIMUM MAXIMUM
W DRY DENSITY OF DRY DENSITY OF
O
o DRY DENSITY COMBINED COMBINED
WET DENSITY SAMPLES SAM PLES
30
=' MOISTURE CONTENT
+#~--~COMBINATION OF SAMPLES USED FOR
MINIMUM- MAXIMUM DENSITY TESTS
(I) SPECIFIC GRAVITY = 2.75 (2) SPECIFIC GRAVITY = 2.70
40
0 20 40 60 86 I00 120
MOISTURE CONTENT IN PERCENT

FIG. 2--Mccsured relative densities t ~ g laboratory techniques.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions
45~ RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

clays. The potential for seismically induced liquefaction was not immedi-
ately recognized, and thus, only a minimal number of specimens was
obtained from the sand stratum. Relatively undisturbed specimens were
obtained from 13 borings using the 2]/~-in. diameter Dames & Moore
underwater-type sampler. The borings were drilled using rotary wash
methods with the 4-in. diameter casing being driven to within a few inches
above sampling depth. Representative specimens were tested in the
laboratory to determine moisture content (ASTM Laboratory Determina-
tion of Moisture Content of Soil (D 2216-71)), dry density, specific gravity
(ASTM Test for Specific Gravity of Soils (D 854-58)), and particle size
distribution (ASTM Particle-Size Analysis of Soils (D 422-63)). The
results of these tests are summarized on Figs. 1 and 2. Because of the limited
amount of material that was available at the laboratory, three specimens
indicated on Figs. 1 and 2 were combined to provide sufficient material to
perform minimum and maximum density determinations (ASTM Test for
Relative Density of Cohesionless Soils (D 2049-69)). To accommodate the
small specimen, a mold 2 ~ in. diameter, 3 in. high, and with a surcharge
weight of 5 lb was substituted for the standard apparatus. The combined
specimen had a minimum dry density of 71 lb/ft 8 and maximum dry density
of 102 lb/ft 3. A relative density scale, based on these minimum and
maximum densities, is included in Fig. 2.
Supplementary to the undisturbed specimens, eleven standard penetra-
tion tests (ASTM Penetration Test and Split-Barrel Sampling of Soils
(D 158667)) were performed at random locations and depths in the same
borings. The results of the standard penetration tests are presented on
Fig. 3, containing the relationships of vertical overburden pressure,
penetration resistance, and relative density developed by Gibbs and

u.
O~
I000
O.
RELATIVE _
2000
ne
r
u~
I,IJ
5000
n-

:\>(
O. \
..J 4 0 0 0
u
i-
a: 5 0 0 0

60000 10 20 30 40 50 60 70
PENETRATION RESISTANCE, BLOWS/FT.

FIG. 3--Relative densities from sfandard penetration data.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
BELL AND SINGH ON COMPARISON OF RELATIVE DENSrrlES 459

CONE BEARING Kg/cm 2


o I0 2o 30 40
0-

0
(/)
(/)

z_
W
F- I1~
I0
~ W

I,,i,I

W
=,20 ,~

~t

2"~0= ! 2 3 4 5 6 7 8 9 I0
PENETRATION RESISTANCE IN BLOWS PER FOOT

FIG. 4--Comparison of relative densities using standard penetration and Dutch cone
sounding data.

Holtz [6] and plotted in the manner suggested by Coffman [7]. It should be
noted that the low overburden pressures and low penetration resistance
values were obtained at the proposed tank farm site plot in the extrapolated
pdrtions of the Gibbs and Holtz relationships.
Eighteen Dutch cone soundings were taken throughout the site, mostly
at locations between the borings. The cone bearing data are shown on Fig. 4.
C o m p a r i s o n o f Test Data
A comparison of standard penetration tests and Dutch cone soundings is
shown on Fig. 4. To compare, the standard penetration test data have been
plotted using the scale factor, q~/N = 4, where qc = cone bearing in k g / c m =,
and N = blows per foot. This correlation factor was proposed by Meyerhof
[8]. Gibbs and Holtz relationships axe also shown, in order to readily see the
variation in relative density indicated by the Dutch cone data.
A comparison of relative densities measured on the undisturbed specimens
and those determined from standard penetration resistance using Gibbs and

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lic
460 RELATIVE DENSITY I N V O L V I N G COHESIONLESS SOILS

Holtz relationships with depth is shown in Fig. 5. Though the usual variance
in data exists, there is still a reasonable correlation between the two
procedures.
In order to determine variations in the data, the relative densities of the
undisturbed specimens and the standard penetration resistances were
statistically analyzed. The results are presented on Table 1 which contains
the statistical parameters in terms of mean value and standard deviation.
The results indicate a mean realtive density of 25 and 32 percent for
standard penetration tests and undisturbed specimen data, respectively.
The standard deviation above and below the mean is 16.6 and 19.8 percent,
respectively. For comparison, the standard penetration data and measured
relative densities from other published case histories were also statistically
analyzed. The results are also presented on Table 1 and suggest a standard
deviation range of 9 to 19 percent. These ranges appear consistent with
those at the Java site. Such variations are to be expected and are attributed
to the many factors which have been described in great detail by deMello [9]
for the standard penetration test, by Tiedmann [10] for the laboratory
minimum and maximum densities, and by Tavenas [11] for the relative
density.

RELATIVE DENSITY IN PERC ENT


25 5O 7,.5 I00
I
IUNDISTURBED
SAMPLES
APENETRATION
RESISTANCE
(GIBBS 8= HOLTZ
1

I,-
hi IO
u.I
1,1..
Z
"1"
9 $
Q S 9 9

ESTIMATED RELATIVE
2G 9 9 DENSITY TO PREVENT
9 LIQUEFACTION FOR
tl THE STIPULATED
I00 YEAR EARTHQUAKE

25

FIG. 5---Comparison of relative densities (laboratory and field methods).

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further
BELL AND SINGH ON COMPARISON OF RELATIVE DENSITIES ,46]

TABLE 1--~tatistical analysis of relative density data.

Data Analysed Number Relative Density, Percent


of Data
Points Range of Mean Standard Reference
Values Value Deviation

Java site,
Penetration resist- 11 5-57 25 16.6 Fig. 3
ance
J a v a site,
undisturbed speci- 41 0-60 32 19.8 Fig. 5
mens
Sand Pile at Grand
Coulee Dam, Wash-
ington,
penetration resist- 65 29-60 43 8.8 Gibbs [12]
ante
Nuclear power plant
60 miles south of
Omaha, Neb.
penetration resist- 38 39-98 62 13.1
ante
undisturbed samples 19 15-98 60 18.8 Gibbs [12]
MississippiRiver 10
miles northwest of
Baton Rouge, La.,
penetration resist- 90 0-95 66 15.5
ance
undisturbed samples 29 25-88 65 15.1 Gibbs [1~]
Dames & Moore pro- 36 17-78 47 17.6
ject

Conclusions

Acknowledging the limitations in obtaining undisturbed specimens of


cohesionless soils, the limited ability to repeat laboratory m i n i m u m and
m a x i m u m density determinations, the inherent crudeness of the standard
penetration tests, the lacking precedent of evaluating relative densities
from D u t c h cone sounding, and the expected variations in density and
particle size distribution within natural alluvial deposits, a foundation
investigation was conducted utilizing each of these subsurface exploration
methods. The results were interpretated to evaluate the relative density of
the top 20 ft of saturated soils t h a t would provide support for several large
tanks. Reasonable correlation in the relative density of the s t r a t u m was
found within and between each method of evaluation.
The sand s t r a t u m was judged to have relative densities from 25 to 45
percent. This entire range of relative density was low enough to expect t h a t
liquefaction would probably occur during earthquakes t h a t were predicted
within the economic life of the proposed tanks. Based on the results of this
investigation, a program was undertaken to densify the sands beneath the
perimeter of the planned tanks.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant t
462 RELATIVEDENSITY INVOLVING COHESIONLESSSOILS

References
[1] Donovan, N. C., "A Stochastic Approach to Seismic Liquefaction Problem," paper
presented at the Conference on Applications of Statistics and Probability to Soil
and Structural Engineering, Hong Kong, Sept. 1971.
[5] Seed, H. B. and Lee, K. L., Journal of the Soil Mechanics and Foundations Division,
American Society of Civil Engineers, Vol. 92, No. SM6, Nov. 1966, pp. 105-134.
[3] Seed, H. B. and Peacock, W. H., Journal of the Soil Mechanics and Foundation Di-
vision, American Society of Civil Engineers, Vol. 97, No. SM8, Aug. 1971, pp. 1099-
1119.
[$] Finn, W. D. L., Pickering, D. J., and Bransby, P. L., Journal of Soil Mechanics and
Foundation Division, American Society of Civil Engineers, Vol. 97, No. SM4,
April 1971, pp. 639-659.
[5] Lee, K. L. and Fitton, J. A. in Vibration Effects of Earthquakes on Soils and Founda-
tions, A S T M S T P 650, American Society of Testing and Materials, 1969, pp. 71-95.
[6] Gibbs, H. J. and Holtz, W. G., Proceedings, Fourth International Conference on
Soil Mechanics and Foundation Engineering, Vol. 1, 1957, pp. 35-39.
[7] Coffman, B. S., Civil Engineering, Vol. 30, No. 10, Oct. 1960, pp. 78-79.
[8] Meyerhof, G. G., Journal of ~ail Mechanics and Foundation Division, American
Society of Civl Engineers, Vol. 82, No. 1, Jan. 1956.
[9] deMello, V. F. B., "The Standard Penetration Test," Fourth Pan American Con-
ference on Soil Mechanics and Foundation Engineering, Vol. 1, State-of-the-art
Papers, 1971, pp. 1-86.
[10] Tiedmann, D. A., "Validity of Laboratory Relative Density and Gradation Tests,"
Report No. REC-ERC-71-17, Engineering and Research Center, U.S. Bureau
of Reclamation, Denver, Colo., 1971.
[11] Tavenas, F. A., Journal of the Soil Mechanics and Foundation Division, American
Society of Civil Engineers, Vol. 98, SM4, April 1972, pp. 433-436.
[15] Gibbs, H. J., "Standard Penetration Test for Sand Denseness," Fourth Pan Ameri-
can Conference on Soil Mechanics and Foundation Engineering, Vol. 11, 1971, pp.
27-44.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
K . - J . Melzer 1

Relative Density---Three Examples of Its


Use in Research and Practice

REFERENCE: Melzer, K.-J., "Relative D e n s i t y m T h r e e Examples o f Its


Use in Research and P r a c t i c e , " Evaluation of Relative Density and Its Role
in Geotechnical Projects Involving Cohesionless Soils, A S T M S T P 523, American
Society for Testing and Materials, 1973, pp. 463-477.
ABSTRACT: The paper demonstrates not only the usefulness but also the
limitations of relative density by three examples from research and practice.
The first example deals with the problem of relating the angle of internal
friction to relative density. When sands do not differ significantly in their
grain-size distributions, the parameters can be related. However, for a large
variety of sands, compactibility has to be taken into account as a third param-
eter. The second example describes the use of relative density in subsoil explo-
ration. For a specific case, the number of blows of the standard penetration
test, relative density, friction angle, and density axe empirically related to
each other, enabling the designer of a harbor project to reduce soil investigation
costs. In the third example, performance of pneumatic tires tested in two
different sands is different when related to cone penetration resistance. How-
ever, when cone penetration resistance is replaced by relative density, per-
formance is comparable for tests conducted in the two sands at the same relative
density.
KEY WORDS: cohesionless soils, penetration tests, internal friction, tires,
density (mass/volume), sands, subsurface investigations

R e l a t i v e d e n s i t y is one of t h e oldest soil p a r a m e t e r s u s e d in t h e field of


soil m e c h a n i c s a n d f o u n d a t i o n e n g i n e e r i n g [I],~ b u t i t s usefulness h a s b e e n
d e b a t e d for m a n y years. S t a n d a r d i z a t i o n of p r o c e d u r e s for e v a l u a t i n g t h e
v o i d r a t i o s of t h e loosest a n d d e n s e s t s t a t e s of a cohesionless m a t e r i a l ( t w o
of t h e i n p u t p a r a m e t e r s t o t h e r e l a t i v e d e n s i t y c a l c u l a t i o n ) h a s c o n t r i b u t e d
t o i t s usefulness; however, d e t e r m i n a t i o n of t h e v o i d r a t i o of a soil in i t s
n a t u r a l s t a t e (the t h i r d p a r a m e t e r n e c e s s a r y t o c a l c u l a t e r e l a t i v e d e n s i t y ) is
still a p r o b l e m in m o s t cases. T h e i n t e n t of t h i s p a p e r is t o d e m o n s t r a t e b y

i Research civil engineer, Mobility Research Branch, Mobility and Environmental


Systems Laboratory, U. S. Army Engineer Waterways Experiment Station, Vicksburg,
Miss. 39180.
2 The italic numbers in brackets refer to the list of references appended to this paper.

463

Copyright by ASTM Int'l


Copyright9 by(all
ASTMrightsInternational
reserved); Fri Mar 11 16:13:06 EST 2016
www.astm.org
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
464 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

three examples from research and practice not only the usefulness but also
the limitations of relative density. The first example deals with the problem
of relating the angle of internal friction to a pertinent soil parameter, for
example, void ratio or relative density or both; the second describes the use
of relative density in a specific study of subsoil exploration; and the third
deals with soft-soil mobility.

Relative Density and Angle of Internal Friction


Within the last 15 years a variety of relations has been published [2] in
attempts to describe the angle of internal friction as a function of the void
ratio of cohesionless soils, particularly of sands. In these relations the
influence of normal stress on the friction angle (see Ref 3) is generally
neglected. Most of these relations contain either one or two coefficients that
must be determined by shear tests, such as triaxial tests. Probably two of
the better known equations are
cot 4, = ae + b [6] (1)
and
1
cot ~ = - e [5] (2)
C

SiLT SAND GRAVE L


-r I 0 0
I-,.
dA iVO d--- =~,~
0
b,I
Y: 8O t"
In
rlb
w SAND A - ~ u ~ , r
Z
c /J
I- 4O
Z SAND 8- " ~
U
Iz 2 0 il
W //
oi
0.002 0.006 0.02 0.06 0.2 0.6 I 2 8 I0 20 83
GRAIN DIAMETER, MM
e e . e -e . C
Sand max mln max mln u

A 0.799 0.454 0.345 2.43


B 0.755 0.425 0.330 2.46
C 0.763 0.435 0.328 2.37

FIG. 1---Grain-size distribution and soil properties of sands A through C [7].

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reprod
MELZER O N RESEARCH A N D PRACTICE 465

J
1.8
COT r :0.086 ~:2 0 8 1 ,
-*0 0 5 6
r = 0.88
1.4

t9 1.2

SAND

[
lO Q A
~ B
O C

08 I
03 0.4 0.5 0.6 07 0.8 0.9
VOiD RATIO 9

FIG. 2 - - R e l a t i o n between cot r and void ratio e for three sands [7].

where
~b = friction angle measured at failure in a triaxial test,
e = initial void ratio in the corresponding test, and
a, b, c = coefficients.
Generally, both Eqs 1 and 2 describe the relation between ~ and e
adequately as long as they are applied to the results of tests with a given
sand [2]. However, difficulties might occur if an attempt is made to describe
the results of tests with various sands by means of these relations, as the
following examples show.
During investigations with penetrometers for subsoil exploration (Refs 6
and 7), the shear strength characteristics of the materials used in the
penetration tests had to be established. These materials were three sands
with only a small variation in grain-size distribution. Coefficient of
uniformity (C~) and minimum and maximum void ratios (emi~ and em~x,
respectively) varied slightly, but noticeably (Fig. 1). Drained triaxial tests
were conducted on fully saturated specimens having medium dense to very
dense relative densities at confining pressures ranging from about 50 to
300 kPa (7.3 to 43.5 psi). The cotangents of the friction angles measured at
failure were plotted versus the initial void ratio (e), and a linear regression
analysis was performed. This led to the general form of Eq 1 (Fig. 2). An
attempt was made to reduce the scatter of the data by substituting relative
density3 for void ratio (e) in Eq 1 ; this led to the following relation between
and relative density (Dr) in general terms:
cot ch = cl - - c 2 D r (3)
A linear regression analysis, now performed with Dr instead of e (Fig. 3),
increased the correlation coefficient (r) from 0.88 for Eq 1 to 0.97 for Eq 3,
and reduced the standard error of estimate from -~0.056 to ~0.028. Thus,

8 Defined as Dr = (emaz - e)/(e~z - emln)-

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agree
466 RELATIVE DENSITY INVOLVING COHESIONLESS SOltS

in this case where the basic soil properties (C,, emi., e~,~) of the three sands
varied only slightly, the use of relative density led to better results t h a n the
use of the void ratio. Furthermore, relative density is, in many respects,
a more correct description of cohesionless soils than is void ratio, because
the possible overall void ratio range is taken into account.
Using Eq 3 instead of Eq 1 has its limitations, however. In a comprehen-
sive study concerning the angle of internal friction of cohesioniess soils,
Schultze [~] collected results from triaxial tests with 37 different sands, from
his and other research. He analyzed the data statistically b y means of Eqs 1
and 2 and found that both equations described reasonably accurately the
relation between ~ and e for a given cohesioniess soil. He recommended the
use of Eq 2 for all practical purposes, because only one coefficient must be
determined. In addition, he discovered a slightly pronounced, but notice-
able, relation between coefficients a and b of E q 1, a fact t h a t decreases the
value of this equation. Schultze tried to collapse the 37 Eqs 1 and 37 Eqs 2,
respectively, by correlating each of the coefficients (a, b, and c) with such
pertinent soil properties as grain size and coefficient of uniformity to derive
one general equation for the relation between ~ and e, based either on E q 1
or on E q 2, but he did not succeed.
For this paper, a collapse of the relations was attempted b y correlating
a, b, or c or all three with "compactibility ''4 instead of with the coefficient
of uniformity. (Compactibility had replaced the coefficient of uniformity
successfully in a previous study, Ref 8.) However, only 22 of the 37 relations
evaluated b y Schultze could be used because only in these 22 cases (Table 1)
were values of e~=~ and emi, given. Just as when C= was used, no statistically

16
COT ~= ~ 7hS'-OOOT,,D,-
~0D28 i
r=097 ~35 Q
1.4

v I 2
O~ 40 ~
z

SANO
1.0 _ _ D A _ _
45 ~
8
0 C L.

o8
[ - 50

20 40 60 80 I00
RELATIVE DENSITY Dr. , ~

FIG. 3--Relation between cot r and relative density Dr for three sands [7].

4 Defined by Terzaghi [1]: D' = ( e ~ -- emi.)/emi,. Introducing compactibility at this


time serves mainly to demonstrate that other parameters must be taken into account in
complex cases of relating the angle of internal friction to other pertinent soil parameters,
such as relative density, where the use of relative density alone is not adequate.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
T A B L E 1--Soil properties and coe$~cients for Eqs I, ~, and ~ for various sands (adapted from Ref ~).

No. Reference Soil~ dso, m m C~ emax e m i n em~x D' Eq 1 Eq 2 Eq 4


-- emin ....
a b c ca c4

1 Hansen and Odgaard [9] mS 0.23 1.6 0.85 0.50 0.35 0.70 2.46 --0.10 0.42 2.02 0.0083
2 Bjerrum and Kummeneje [10] fS 0.15 2.8 0.86 0.64 0.22 0.34 2.22 --0.27 0.55 1.56 0.0040
3 Broms and Jamal [11] mS 0.30 1.7 1.00 0.56 0.44 0.79 1.72 0.16 0.51 1.96 0.0086
4 Cornforth [12] mS 0.26 2.1 0.79 0.48 0,31 0,65 2.07 0.09 0.47 1.68 0.0066
5 De Beer and Vesid [13] fmS 0.20 1.9 0.84 0.58 0.26 0.45 2.97 --0.74 0.53 1.87 0.0049
6 L'Herminier [14] cS 0.80 2.0 0.82 0.61 0.21 0.34 1.94 0.09 0.49 1.67 0.0043
7 Leussink et al. [15] mS 0.49 2.2 0.82 0.43 0.39 0.91 1.98 0.10 0.46 1.78 0.0085
8 Moussa [16] cS 1.65 1.6 0.88 0.57 0.31 0.54 2.20 --0.16 0.52 1.69 0.0060
9 cS 0.83 1.5 0.80 0.53 0.27 0.51 2.20 --0.10 0.49 1.63 0.0055
10 mcS 0.58 2.7 0.79 0.46 0.33 0.72 1.60 0.41 0.43 1.84 0.0055 m
11 mcS 0.57 2.9 0.82 0.49 0.33 0.67 2.75 --0.36 0.46 1.78 0.0072
12 mcS 0.56 2.5 0.78 0.46 0.32 0.70 1.43 0.74 0.47 1.66 0.0068
13 mS 0.43 3.1 0.80 0.47 0.33 0.70 1.20 0.54 0.46 1.74 0.0072 0
0.86 0.54 0.32 0.59 1.00 0.55 0.53 z
14 mS 0.37 2.3 1.62 0.0060
15 mS 0.30 1.5 0.88 0.59 0.29 0.49 0.60 0.72 0.58 1.52 0.0050
16 mS 0.24 1.8 0.90 0.59 0.31 0.53 0.78 0.60 0.55 1.64 0.0056
17 mS 0.22 1.6 0.94 0.57 0.37 0.65 2.17 --0.20 0.53 1.77 0.0070
18 fS 0.20 1.8 0.94 0.61 0.33 0.54 0.91 0.50 0.61 1.54 0.0054
19 Nash [17] fmS 0.18 1.9 0.88 0.61 0.27 0.44 2.08 --0.17 0.53 1.66 0.0051 zc~
20 Vesi5 [18] mS 0.37 2.5 1.10 0.62 0.48 0.78 1.56 --0.08 0.68 1.62 0.0071
21 mS 0.26 1.5 0.82 0.54 0.28 0.52 2.78 --0.45 0.50 1.64 0.0056
22 Schultze [2] mS 0.2-0.5 1.8-2.0 0.96 0.56 0.40 0.71 1.98 0.04 0.50 1.89 0.0080

a fS = fine sand; mS = medium sand; finS -~ fine to medium sand; cS = coarse sand; mcS = medium to coarse sand.

O,
~4

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
468 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

C 4 = 0 . 0 0 1 3 "1" 0 . 0 0 8 2 01 "1" 0 . 0 0 0 6
r = 0.901
1.0
rt
)-
Ira-
/

~
m

0
.<
0.5

f
0
0
0
0 0.002 0.004 0.006 0.008 0.010 0.01 ?.
COEFFICIENT C4

FIG. 4---Relation between compactibility D' and coe~icient c4 of Eq. 4.

significant relations could be developed between one of the coefficients of


Eqs 1 and 2 and compactibility D'.
As the next step, relative density (D,) was substituted for void ratio (e)
in Eq 2:
cot r c3 - c,Dr = (4)
where
c = coefficient of Eq 2 evaluated by regression analysis,
1
C3 = -- emax, and
c

emax -- emin
C4 = "
100c
Equations 3 and 4 have the same general form, the basic difference being
the method of determining the coefficients. Coefficients cl and c2 in Eq 3
were determined by statistical means directly from the test results; whereas
coefficients c3 and c~ in Eq 4 were calculated from the statistically deter-
mined coefficient c, and the known soil properties em=xand emln (Table 1).
A sound relation was then found between compactibility and c4, with a
relatively high correlation coefficient of 0.901 (Fig. 4). Coefficient c3 can be
represented reasonably accurately by a constant of 1.720, with a standard
deviation of • Equation 4 then becomes
cot ~ = 1.720 - ( 0 . 0 0 1 3 + 0.0082 D')Dr (5)
Thus, for practical purposes, a general relation can be established for
determining the angle of internal friction without costly triaxial tests by
using D, (instead of e) as one independent variable and by introducing D' as

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licen
MELZER O N RESEARCH AND PRACTICE 469

a second independent variable. However, Eq 5 is of limited value. It is valid


only for sands with ds0 (diameter of sand particle having a size greater than
that of 50 percent, by weight, of the particles) ranging from 0.15 to 1.65 mm
(0.006 to 0.065 in.) and with D ' ranging from 0.34 to 0.91 (Table 1). The
influence of grain size and grain roughness still must be investigated.

Relative Density and Its Use in Subsoil Exploration


In one of the larger European river harbors, the paved slope of one of the
piers was to be replaced by a sheet pile wall. In addition, the bottom of the
harbor basin was to be lowered about 2.5 m (Fig. 5). The subsoil in which
the wall was to be embedded was primarily an alluvial river deposit of
sandy gravel (Fig. 5). A consulting bureau 5 was asked to conduct the subsoil
exploration of the project site and to determine the angle of internal
friction of the subsoil for use in designing the wall. General relations also
had to be established for possible estimation of the density and the friction

BORING B lit
PENETRATION PENETRATION
S4 S3
(+z5 5) (+z5 9)
NUMBER OF BLOWS N20 NUMBER OF BLOWS NZO, N30

4 M 20 M "~q

;'5
FILL

SILT
vM•w +2,9 / r (CL)

3 ]
ao

-~ +15 3/..--EXISTING BAS/N I


, s - -
ELEVATION
1 ,s ~
Z- - V.~_ ~PLANNEDBASIN
+128
ELEVATION
It
F" %
w

,oL - t

~s +83V
JLT It CLAY (CL)
~o
+5.5 +5 4
_ V
5~

FIG. 5--Representative soil profile for harbor area.

s See Acknowledgmentat the end of the paper.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
470 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

angle of the subsoil of the entire harbor area, which was assumed to be the
same sandy gravel as that found at the project site.
Penetration tests with a heavy dynamic penetrometer (hammer weight,
500 N (112 lb); height of fall, 50 cm (19.7 in.); number of blows, N20,
that is, the number counted per 20-cm (7.9-in.) penetration) and borings
were conducted at the project site. During the borings, standard penetration
tests were conducted and disturbed specimens were taken. A representative
soil profile determined from the results of the penetration tests and the
borings is shown in Fig. 5.
Grain-size distribution, specific gravity, and void ratios of the sandy
gravel in its loosest and densest states were determined by laboratory tests.
Also, drained triaxial tests were performed on fully saturated specimens of
the material [diameter of a specimen, 30 cm (1 ft) ; height, 60 cm (2 ft)] at
various initial relative densities (12 to 82 percent) and various confining
pressures ( ~ 5 0 to 400 kPa; 7.3 to 72.5 psi).
Results of the laboratory and field tests were used to develop the
following relations. Cotangents of the friction angles at failure from the
triaxial tests were plotted versus the corresponding initial void ratios, and
a relation was established between cot ~b and e having the general form of
Eq 1 (Fig. 6). In addition, the corresponding relative densities and densities
were calculated and plotted as horizontal axes parallel to the e-axis. The
void ratios for the loosest and densest states of the cohesionless material
were assumed to vary only slightly over the entire harbor area; thus, slight
variation in emax and emin would be compensated for by substituting D, for
e, and the relation between cot ~ and D, could be assumed to be generally
valid for this alluvial deposit.
Existing relations among relative density, number of blows (N3o) of the
standard penetration test, and overburden pressure (see Refs 6, 8, 19, and
P0) for various cohesionless materials were combined into one average
relation among Dr, Na0, and penetration depth (D) (Fig. 7). e D had been
calculated from the overburden pressure (~D) by assuming an average unit
weight (~) for the soil layers under consideration (Fig. 5). The relation in
Fig. 7 was established for air-dry cohesionless material; thus, before this
relation could be used to determine Dr, the number of blows measured in
penetrations below the groundwater level had to be corrected for the effect
of groundwater on the penetration resistance [8]. Because the relation had
been established only for N30, values for N2o and Na0 were correlated by
means of regression analysis of the results of the penetration tests conducted
at the project site. This correlation made possible the plotting of N2o, in
addition to Na0, on the horizontM axis of Fig. 7. This was an important
factor, since the standard penetration test (N30 blows) is conducted only in
e The description of the relations in this paragraph appears to be very brief; however,
detailed descriptions of their development would have been too lengthy. Therefore, the
reader is referred to the literature cited for more detailed information.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No fu
MELZER ON RESEARCH AND PRACTICE 471

32.5 COT m'= O.~47 + Z . / 4 4 e


~_ aoez /

35.0

w
r

w
37.2 .,/o
Z
<
Z
0
U
40.0
/ I
I
i
I
f
I

/
J I
I
I
J I

I
I
I
[ 0.20 0.25 0.3C 0.35 0.40 0.45
0.19 VOID RATIO e
I I I I I ~L i I I I I I
IO0 90 80 70 60 ]'50 40 30 20 I0 0
RELATIVE DENSITY Dr~ =/o
I l I ~ I I I
2.39 2.35 2.30 I 2.25 2.20 2.15
2.26
DENSITY P FOR SATURATED S O I L , G/CM 3

F I G . 6--Relations among density p, void ratio e, relative density D,, and friction angle ~h
for sandy gravel.

connection with a boring; whereas the heavy dynamic penetration test


(N2o blows) is conducted as a continuous penetration from the soil surface
and thus is less expensive. Therefore, further subsoil exploration in the
harbor area was expected to be conducted mainly by using the heavy
dynamic penetrometer, with only a few exploratory-type borings with
standard penetration tests. This procedure has become common practice
for subsoil exploration in Europe in recent years.
By means of the relations in Figs. 6 and 7, the relative density of the
sandy gravel (Fig. 5) could be estimated for either N~0 or N30 (Fig. 7),
either for an average number of blows for the entire soil layer or, if

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License A
472 RELATIVE DENSITY I N V O L V I N G COHESIONLESS SOILS

IOC D=2.SM 5.0M ZSM

9C
//, ,/,,5.0. / /
M

//
80

o"~
<
Jul

40 i
30

2e I., ~ /

I0

0 20 J 40 60 80 N3o
C I I , I J l , I ,
0 20 40 60 80 I00 120 NEO
N U M B E R OF B L O W S

FIG. 7--Relations among relative density Dr, penetration depth D, and number of blows
(N~ and NB0).

necessary, according to the changes in the number of blows for cases of


considerable changes in penetration resistance within the main layer. With
these estimated relative densities, the corresponding angles of internal
friction and the corresponding densities were evaluated by the relation
given in Fig. 6. Average friction angle for the sandy gravel of the project
site was determined to be 37.2 deg, average density was 2.15 g/cm s
(133 lb/ft3), and density in the submerged state was 1.26 g/cm 8 (78 lb/ft3),
with a corresponding relative density of 54 percent.
Relative density was thus demonstrated to be a useful parameter for

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licen
MELZER O N RESEARCH AND PRACTICE 473

establishing certain relations between soil properties or parameters or both


for the designer of earth structures in foundation engineering. By utilizing
relative density, not only could the necessary design parameters of the given
project site be estimated, but a tool was devised for use in further design
work in the entire area under consideration with a minimum of additional
field exploration.
Relative Density and Soft-Soil Mobility
In mobility research at the U. S. Army Engineer Waterways Experiment
Station (WES), the dimensional analysis approach was and still is used to
establish analytical systems for the prediction of the performance of
pneumatic tires. Performance parameters (dependent variables) are related
to so-called "mobility numbers." One such dependent variable is the ratio
of pull-to-load developed by a tire at 20 percent slip (P2o/W), which
corresponds approximately to the maximum pull-to-load ratio. The
mobility numbers contain pertinent independent variables, such as tire
geometry parameters, loading condition, and a soil parameter, the latter
determined from tests with the WES standard cone penetrometer within
the top 15-cm (6-in.) layer of the soil under consideration. In the following
example [21], relative density was used in explaining divergent test results
for parameter P~o/W.
Five tires of different sizes under loads ranging from 900 to 6000 N
(200 to 1350 lb) were tested in a single-wheel dynamometer system on a
dune sand (mean grain diameter -- 0.12 mm or 0.005 in.) from Yuma,
Ariz., and in a "mortar" sand (mean grain diameter = 0.27 mm or 0.011 in.)
from the Big Black River near Vicksburg, Miss. The average strength
profile of these sands ranged from 0.86 to 6.08 M P a / m (3.1 to 22.5 psi/in.),
with corresponding relative densities between loose and very dense.
0.5
YUMA SAND---------..,~_ ......, " " ~ ~ -
/

0.4 - II / / ~

o_- i1".,~/- o
0.2 LEGEND

s Id"
/ /
-

A 16X11.50--6~ 2-PR
O. I t [ 0 16XI5.00-6~, 2-PR
IV/O ~ 26X16.00-10, 4-PR
-~ I 0 31XlS.50-13j, 4-PR
oo I1~,o ] I I "1
zo :so
40 so eo
SAND MOBILITY NUMBER NS

FIG. 8---Relations between pull coe~cient P~/W and sand mobility number N~for Yuma
and mortar sands [21] (data paints for Yuma sand are given in Fig. 10).

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
474 RELATIVEDENSITY INVOLVING COHESIONLESSSOILS

The P~o/W values from the tire tests were plotted (Fig. 8) versus the
dimensionless sand mobility number N,, defined as
G(bd) 3/~
Ns-
W "h
where
G = cone penetration resistance gradient (average strength profile),
b = maximum outside width of the cross section of the inflated, but
unloaded, treadless tire,
d = outside diameter of the inflated, but unloaded, treadless tire,
W - vertical force applied to the tire through the axle,
- deflection, namely, the difference between the unloaded and loaded
section heights, and
h = section height of the unloaded, but inflated, tire.

IOO
""•OO VERY
DENSE
80

~0

~0~ DENSE
40

LOOSE
~0
VERY
LOOSE
~- o
....'; 2 3 ,4 S 6 7
z
o. YUMA SAND
~OO
..t~ ~ . . . . . . . - vERY
O~ DENSE
a: 80
/~o~~ ~'~~I DEN.~

6O .~ ~ ~ STANDARD

DENSE
40

/~// LOOSE
~0
VERY
LOOSE
% I 2 3 4. S e 7
GRADIENT G, MRa/r~
b. MORTAR SAND

FIG. 9--Relation between relative density and penetration resistance gradient [22].

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further re
MELZER O N RESEARCH A N D PRACTICE 475

0.6

rUMA SAIVOCURVg 0
0.5 FRO~~/G. a ~
]r
0
0 ~ V
~'~ 0.4
I.- 9 V
Z
bJ
U 0.31
~. i
b.

u 7^ OI A I S X i t , 5 0 - S t 2-OR
+j 0.2
.J 4)m~ V 28XI8.00-10~ 4-PR
& 0 31X1~50-13) 4-PR
o.
NOTE" OPEN SYMBOLS OENOTE YUMA
I O SAND TESTS.
ed-'~O CLOSED SYMBOL DENOTES MOR-
~I~ TAR SAND TESTS WITH YUMA
IA SAND EQUIVALENT G.
I I i I I
10 20 30 40 50 60
SAND MOBILITY NUMBER N S

FIG. lO---Relation betweenpull coe:~cientP20/W and sand mobility number N~ [21].

Unexpectedly, the plot showed two different relations between P2o/W and
N~, separated by sand type (Fig. 8). In other words, for a given tire under
constant load and deflection, P~o/W in Yuma sand was different from
P~o/W in mortar sand.
Because different curves were needed for predicting pull on the two
sands, it seemed likely that the cone penetration resistance gradient (G)
alone might not be an adequate common denominator for representing the
strength of all sands. In an attempt to arrive at a common denominator, at
least for the two sands under consideration, relations recently developed [22]
between G and relative density (Dr) of the same two sands were examined.
These relations showed that at the same relative density, G was higher in
mortar sand than in Yuma sand (Fig. 9).
To test a hypothesis that the two sands would provide similar perform-
ance if their relative densities were the same, the G values for the various
tests conducted in mortar sand were converted into "equivalent" G values
for tests in Yuma sand at the same relative density. For example, in mortar
sand a value of G -- 2.0 M P a / m (7.4 psi/in.) corresponds to Dr = 62
percent (Fig. 9b); the same relative density in Yuma sand corresponds to
G = 1.4 M P a / m (5.2 psi/in.) (Fig. 9a). Thus, G = 2.0 in mortar sand is
equivalent to G -- 1.4 in Yuma sand. This value of G = 1.4 M P a / m
(5.2 psi/in.) was then used in the sand mobility number for plotting mortar
sand test results.
The data from the tire tests in Yuma sand are plotted in Fig. 10, together
with the Yuma sand equivalent values for the tests in mortar sand
developed by using the described procedures. Results show that the curve

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to L
476 RELATIVEDENSITY INVOLVING COHESIONLESSSOILS

for Y u m a sand fits b o t h Y u m a sand data and m o r t a r sand data, thus


demonstrating the feasibility of explaining the divergence of the two sets of
d a t a b y introducing relative density, at least in this specific case.

Conclusions
Three examples from broad fields of research and practice demonstrated
t h a t the p a r a m e t e r relative density is useful as: (a) a correlator of other
pertinent soil properties, such as the angle of internal friction; (b) a con-
necting p a r a m e t e r between two other soil parameters or properties, such as
between angle of internal friction and penetration resistance; and (c) an
explanatory aid, as in the case of divergent test results. On the other hand,
relative density cannot always be used to solve all problems, as demon-
strated in the development of a relation of the friction angle generally valid
for all sands. I n addition, the uncertainties involved in the determination
of this p a r a m e t e r m u s t be kept in mind. Thus, the use of relative density
will still require common sense and engineering judgment in each case before
it is applied.

Acknowledgment
Acknowledgment is made to Professor Dr. E. Schultze, director of the
Institute for Waterways, Foundation Engineering and Soil Mechanics,
Technical University, Aachen, of whose staff the author was a m e m b e r at
the time of the subsoil explorations for the European river harbor, for
permission to publish certain results in this paper.

References
[1] Terzaghi, K., Erdbaumechanik auf bodenphysikalischer Grund/age, F. Deuticke,
Leipzig, 1925.
[2] Schultze, E., "Der Reibungswinkel niehtbindiger BSden," Der Bauingenieur, Vol.
43, No. 9, Sept. 1968, pp. 313-320.
[3] Brinch Hansen, J., Proceedings, The Geotechnical Conference, Oslo, Vol. 1, 1967,
pp. 175-177.
[~] Schultze, E., "Lockere und diehte BSden," Mitteilungen Institut fiir Baumaschinen
und Baubetrieb, Technical University of Aachen, Germany, No. 9, 1966, p. 107.
[5] Caquot, A. and K~risel, J., Trait~ de M~chanique des Sols, 3d ed, Gauthier-Villars,
Paris, 1956.
[6] Schultze, E. and Melzer, K.-J., Proceedings, Sixth International Conference on Soil
Mechanics and Foundation Engineering, Montreal, Vol. 1, 1965, p. 354.
[7] Melzer, K.-J., "Sondenuntersuchungen in Sand," Mitteilungen Institutfi~r Verkehr-
swasserbau, Grundbau und Bodenmechanik, Aachen, Germany, No. 43, 1968.
[8] Melzer, K.-J., Proceedings, Fourth Panamerican Conference on Soil Mechanics and
Foundation Engineering, San Juan, Vol. 2, June 1971, p. 37.
[9] Hansen, B. and Odgaard, D., "Bearing Capacity Tests on Circular Plates on Sand,"
Bulletin No. 8, Danish*~Geotechnical Institute, Copenhagen, 1960.
[10] Bjerrum, L. and Kummeneje, O., "Shearing Resistance of Sand Samples with
Circular and Rectangular Cross Sections," Publication No. 44, Norwegian Geo-
technical Institute, Oslo, 1961.
[11] Broms, B. B. and Jamal, A. K., "Analysis of the Triaxial Test--Cohesionless Soils,"
Publication No. 10, Swedish Geotechnical Institute, Stockholm, 1965.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No fu
MEI.ZER ON RESEARCH AND PRACTICE 477

[12] Cornforth, D. H., G$otechnique, Institution of Civil Engineers, London, Vol. 14,
1964, pp. 143-167.
[13] De Beer, E. E. and Vesid, A. S., "Etude expdrimentale de la capaeit~ portante du
sable sous des fondations directes dtablies en surface," Annales de L'Institut Tech-
nique du Batiment et des Travaux Publics, Vol. 3, 1958, p. 5.
[14] L'Herminier, R. L., Proceedings, Sixth International Conference on Soil Mechanics
and Foundation Engineering, Montreal, Vol. 3, 1965, p. 410.
[15] Leussink, H. et al, "Unterschiede im Scherverhalten rolliger Erdstoffe und Kugel-
schtittungen im Dreiaxial- und Biaxialversuch," Ver6ffentlichungen des Instituts
fiir Bodenmechanik und Felsmechanik, Karlsruhe, No. 21, 1966.
[16] Moussa, A., "Untersuchungen iiber die Scherfestigkeit und die Durchl~issigkeit yon
Sanden," M itteilungen Institut fi~r V erkehrswasserbau, Grundbau und Bodenmechanik,
Aachen, Germany, No. 39, 1967.
[17] Nash, K. L., Proceedings, Third International Conference on Soil Mechanics and
Foundation Engineering, Zurich, Vol. 1, 1953, p. 160.
[18] Vesid, A. S., "A Study of Bearing Capacity of Deep Foundations," Final Report,
Project B-189, Soil Mechanics Laboratory, Georgia Institute of Technology,
Atlanta, Ga., 1967.
[19] "Second Progress Report of the Penetration Resistance Method of Subsurface Ex-
ploration," Report No. EM-356, U. S. Bureau of Reclamation, Design and Con-
struction Division, Earth Materials Laboratory, Denver, Colo., 1953.
[:~0] Schubert, K., "Untersuchungen des sandigen Baugrundes durch Sonden," W/s-
senschaftliche Zeitschrift, Technical University Dresden, Vol. 5, 1955, p. 49.
[21] Patin, T. R., "Prediction of Performance of Rectangular Cross-Section Tires in
Sand," Paper No. 71-603 presented at Winter Meeting of the American Society of
Agricultural Engineers, Chicago, Ill., 1971.
[2~] Melzer, K.-J., "Measuring Soil Properties in Vehicle Mobility Research; Relative
Density and Cone Penetration Resistance," Technical Report No. 3-652, Report 4,
U. S. Army Engineer Waterways Experiment Station, Vicksburg, Miss., July 1971.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
F. A . Tavenas ~

Difficulties in the Use of Relative Density as


a Soil Parameter

R E F E R E N C E : Tavenas, F. A., " D i f f i c u l t i e s i n t h e U s e o f R e l a t i v e D e n s i t y


as a Soil P a r a m e t e r s " Evaluation of Relative Density and Its Role in Geo-
technical Projects Involvino Cohesionless Soils, A S T M S T P 5~3, American So-
ciety for Testing and Materials, 1973, pp. 478-483.
/~BSTRACT: The results of a comparative test program reported by Tavenas
et al at this symposium have shown that the relative density determination
is highly variable and nonreproducible. In this paper, the consequences of
this fact are drawn.
Because of the high variability of relative density measurements, the com-
parison of the properties of different soils as investigated by different engineers
will be impossible. For the same reason an engineer cannot make an efficient
use of any published correlation between the relative density and the me-
chanical properties of sands since he will not be able to reproduce the same
relative density. Finally, the relative density can be used as a compaction
criterion only if the same engineer who has written the specification is also in
charge of the compaction control.
K E Y W O R D S : eohesionless soils, density (mass/volume), tests, compaction

I n a paper to this symposium, Tavenas et al have presented the results


of a comparative test program on the measurement of the minimum and
maximum densities of cohesionless soils. The conclusions of this investiga-
tion were rather dramatic, since the results of the so-called standard tests
performed according to A S T M Test for Relative Density of Cohesionless
Soils (D 2049-69) were shown to be very much dependant on the operator
or the laboratory.
Due to the variability of the minimum and maximum densities and
essentially to the formulation of the relative density (ratio of two small
differences between large numbers), the resulting error on the relative
density was shown to be larger t h a n 4-15 percent in the most favorable case
and usually of the order of 4-30 to 4-47 percent. The most important point,
however, was t h a t a given soil at a given i n situ density would be affected a

1Associate professor, Civil Engineering Department, Laval University, Quebec,


Canada.

478

Copyright by ASTM Int'l


Copyright9 by(all rights International
ASTM reserved); Fri Marwww.astm.org
11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No fur
TAVENAS O N RELATIVE DENSITY AS A SOIL PARAMETER 479

completely different relative density and would therefore be assumed to


have completely different mechanical properties depending on the individual
considering it.
Major consequences on the use of the relative density follow from these
observations.
relative density,%
0 IO 20 30 40 50 60 TO 80 90 IO0
IGO 150

145

140

135 135

130 U30

IZ5

IZO

115 tl5

105 105 =~
i

E
tOO
t~ Z
i
95 95

90 90

85 85

80 80
0 tO 2O 30 40 50 60 70 8O 9O I00
mlQti~ density, %

FIG. 1--Reproducibility of relative density measurements.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to
480 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

Validity of the Concept of Relative Density


As stated by Burmister [i], 2 the concept of relative density was introduced
to "bring the behavior characteristics of soils together on a common basis
in consistent and practically useful relationships," and to provide a tool for
communications between engineers.
The results of the comparative test program show t h a t this basic purpose
cannot be satisfied, since the relative density of a soil at a given density is
a function of the operator who determines it. In other words, two engineers
considering the same deposit will affect to it a different relative density, the
range of possible differences being as large as 60 percent relative density for
a sand deposit and 100 percent relative density for gravelly materials. In the
same way two engineers speaking of a given soil at, say 70 percent relative
density, will correlate this value with entirely different mechanical
behaviors of this soil. Under such circumstances no communication between
engineers is possible by means of the relative density. At the most, and
considering t h a t the variations of the relative density within a test series
are much smaller, as shown by Tavenas et al or by Tiedemann [2], the" use
of the relative density by a given engineer will allow him to communicate
not with others, but with himself, that is, to make use of his own past
experience. Even in this very restrictive condition, the relative density
should be used with much care, particularly on field works, since, as shown
on Fig. 1, the minimum probable error is of the order of -4-15.5 percent on
any single measurement of Dr.

Validity of Published Data Involving Dr


Since 1950 many papers have been published which present correlations
between Dr and most of the mechanical characteristics of cohesionless soils
such as the internal friction angle, the modulus of compressibility, the
compression or shear wave velocity, the dynamic penetration resistance,
and the susceptibility to liquefaction.
Such correlations may certainly be valid for those who have established
them, provided due consideration be given to the fact t h a t they are neces-
sarily affected by a certain inaccuracy directly related to the lack of
reproducibility of the relative density as shown in Fig. 1.
However, a major problem arises when an engineer tries to make use of
these published correlations in practice. As shown by Tavenas et al, the
probability that this engineer can reproduce minimum and maximum
densities, identical to those on which the correlation is based, is rather
limited if at all existent. He will thus use this correlation on the basis of a
"wrong" relative density and will therefore necessarily come to an erroneous
result.
An example to illustrate this point would be a deposit of the sand tested

The italic numbers in brackets refer to the list of references appended to this paper.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
124 , , I , r i i l , , l I ~ ] p ] , , I ] ' i I I , I 1 1 I (

i20 -
Jo ) -- ~ 9 m~n+S

0--
9 9 e ill
9 I e; , o o o 9
tl6 9 8 ; e
- I ~ '~ 0.8 i ~ 8 o ,
meon "Yd mox.
9 b~ 9 9 o
9 D 9
112
o o e .<
o
b
@.. meon -S
108 , $
0
~W 9 W : wet melhod Z

L04 0 7"mox. port.


-a 9 Ymox. ASTM
0 Xmin, port.
<~
9 )train. ASTM
IO0

-- - - - - - . P. - I - - - ~--
. ~
- I F _.--, "q:~-' 9 - :nil -] 1 meon+S
96 9 9 Ip C, 9 m R--I 9 9 [] ~ In ~ meon ~'dmin. >.
-'111 ~ m- ,,m I~l tn
0
-- t --- ~ ~ ~ .I---o-- r - mean-S r'..
94
-~ ! 8 o U
[]
! I

0 4 8 12 L6 20 24 28 52 36 40 44 48 52 56 P~
no. of port icipont

FIG. 2--Minimum and maximum densities of fine sand.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authorized.
482 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

in the comparative test program reported by Tavenas et al, in which


standard penetration tests have been performed and an average value
N = 25 determined at a depth corresponding to a, = 20 psi. An engineer
has to decide on the liquefaction potential of that deposit. He will probably
procede by first determining the relative density on the basis of Gibbs and
Holtz [3] correlation and then perform laboratory test to determine the
parameters necessary to apply the method for assessing the liquefaction
potential proposed by Seed and Idriss [$]. By so doing, important errors
will be introduced at two different steps of the analysis:
1. While no evaluation of the quality of their correlation was made by
Gibbs & Holtz [$], Schultze and Melzer [5] have shown that the standard
deviation on such N-Dr-av correlation is of the order of • percent
relative density. Thus, and assuming that N is not affected by any un-
certainty, the relative density corresponding to N = 25, a~ = 20 psi will be
73 • 7 percent. This is a first inaccuracy.
2. If the engineer works for participant 35, of the comparative test
program (Fig. 2), he will conclude that the in situ density of the sand is
103 • 1.5 lb/ft 3, and he will perform the laboratory tests on samples at that
density. On the contrary, if he works for participant 32, he will assume
~,~ = 110 • 1.5 l b / f P and will obviously get completely different results
from his test program on the very same sand.
It appears, therefore, that the result of the analysis will be essentially a
function of the consultant this engineer is working for, and, consequently,
that the reliability of the results will be very poor and certainly not
sufficient to provide a basis for sound engineering.

Use of the Relative Density in Compaction Works


There is a definite trend toward a generalized use of the relative density
in compaction specifications. Values of the relative density of 85 percent or
so are now a common compaction criterion.
From the results reported by Tavenas et al, it is evident that the quality
of fills compacted according to this common criterion will vary widely
depending on the laboratory in charge of the control tests. Thus, the
relative density requirement fails to provide what it was supposed to do,
that is, a uniform and acceptable quality of all fills compacted on that basis.
It is the author's opinion t h a t the control of compaction should be
performed on those parameters on which the design of the fill is based,
namely, on shear strength, compressibility, modulus of subgrade reaction,
etc., and not on any expression of the density since this parameter has no
real influence on the performance of the fill. Tools are now available to
measure these parameters in situ, before, during, and after compaction. The
static penetrometer and the pressuremeter, are tests which are well suited

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lic
TAVENAS ON RELATIVE DENSITY AS A SOIL PARAMETER 483

for the control of compaction which are certainly more reliable and, in most
cases, more economical in their application tha~l any density measurement.

Conclusion
The comparative test programs performed b y Tavenas et al and b y
Tiedemann [2] have shown that the minimum and maximum densities of
cohesionless soils cannot be measured accurately and that the resulting
relative density is highly dependent on the operator performing the tests.
The main consequences of this observation are:
(a) The relative density cannot be used as a communication tool between
engineers dealing with cohesionless soils. Thus, the basic reason for which
the relative density was introduced is proved wrong.
(b) All published correlations between the relative density and the
mechanical properties of cohesionless soils are useless to anyone but those
who have established them. Therefore, any engineer interested in the use
of the relative density should imperatively develope his own correlations
between Dr and the standard penetration index, the shear strength, the
compressibility, the liquefaction potential, etc. of cohesionless soils. Since it
is obvious that only very few organizations can afford to develope such
personal correlations, the potential of relative density as a valuable
engineering tool is more than limited.
(c) The relative density is not a good criterion for the control of compac-
tion since the quality of a fill at a given relative density will be essentially
a function of the laboratory performing the control test.
(d) The relative density being proved useless in all three major cases
where it was supposed to be a valuable concept; the question of its validity
as soil parameter should urgently be considered.

References
[I] Burmister, D. M., Proceedings, American Society for Testing and Materials, Vol. 48,
1948, pp. 1249-1268.
[2] Tiedemann, D. A., "Variability of Laboratory Relative Density and Gradation
Tests," Report REC-ERC-71-17, Bureau of Reclamation, Denver, 1971.
[3] Gibbs, H. J. and Holtz, W. G., Proceedings, 4th International Conference on Soil
Mechanics and Foundation Engineering, London, Vol. 1, 1957, pp. 35-39.
[4] Seed, H. B. and Idriss, I. M., Journal of the Soil Mechanics and Foundation Engineer-
ing Division, Vol. 93, SM3, 1967, pp. 83-108.
[5] Schultze, E. and Melzer, K. J., Proceedings, 6th International Conference on Soil
Mechanics and Foundation Engineering, Montreal, Vol. 1, 1965, pp. 354-358.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
Summary

CopyrightbyASTMIntl'(alrightsreserved);FriMar1116:13:06EST2016
Downloaded/printedby
(UFPE)UniversidadeFederaldePernambuco((UFPE)UniversidadeFederaldePernambuco)pursuanttoLicenseAgreement.Nofurtherreproductionsauthorized.
E. T. Selig ~ and R. S. Ladd ~

Evaluation of Relative Density


Measurements and Applications

REFERENCE: Selig, E. T. and Ladd, R. S., " E v a l u a t i o n o f Relative


Density Measurements and Applications," Evaluation of Relative Density
and Its Role in Geoteehnieal Projects Involving Cohesionless Soils, ASTM S T P
523, American Society for Testing and Materials, 1973, pp. 487-504.
A B S T R A C T : The souree and type of errors in relative density are assessed.
Based on reported values of maximum, minimum, and in situ density errors,
expected errors in relative density are determined. These values may be used
as a basis for setting confidence limits on the results of studies involving relative
density. Engineering applications of the relative density parameter are dis-
cussed, and alternatives to relative density are given. Factors influencing the
mazdmum and minimum densities used in calculating relative density are
summarized. Experience in correlating relative density to blow count and
strength of cohesionless materials is reviewed. Finally, based on all of the in-
formation gathered in the symposium, a series of recommendations are given
for modifications to the ASTM test procedures and for needed new procedures.

K E Y W O R D S : cohesionless softs, density (mass/volume),

The collection of papers in these proceedings provides an extensive


amount of experience ~ith relative density. The content of these papers as
well as the symposium discussion indicates that the concept is widely used
in engineering practice. However, in many cases the results can be mis-
leading because the relative density (Da) values are not what would be
expected for the given situation. For example, Durham and Townsend 8
report Da < 15 percent for a compacted sand, although most experience
shows that it is difficult to prepare soil that loosely. In another case, re-
ported by Osterberg and Varaksin, sand deposited through water without
compaction gave values of Dd as high as 90 percent. In this case it is ex-

z Associate professor, Department of Civil Engineering, State University of New York


at Buffalo, Buffalo, N. Y. 14214.
2 Laboratory director, Woodward-Moorhouse & Associates, Clifton, N. J. 07012.
a Reference to authors without any reference given indicates a paper included in this
symposium.

487

Copyright by ASTM Int'l


Copyright9 by(all
ASTMrightsInternational
reserved); Fri Mar 11 16:13:06 EST 2016
www.astm.org
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
488 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

pected that considerable compaction should have been required in order to


achieve such a high value. Finally, Stephenson reports that most of his
in situ relative densities in a compacted boulder-gravel fill ranged from
Dd = 100 to 130 percent. This is apparently a result of difficulty in deter-
mining correct values for the maximum density for this material which
had too large a particle size for measurement in the manner prescribed by
ASTM Test for Relative Density of Cohesionless Soils (D 2049-69).
This variety of experience points out the uncertainty in applying to any
job the correlations between Da and performance obtained in another job.
An example of such a correlation is the Gibbs-Holtz relationship between
blow count from the standard penetration test and relative density of
sands. As Holtz points out in his Session II discussion, considerable care
must be taken in the quantitative use of this correlation. Another example
is the relationship between relative density and liquefaction potential which
is usually based on curves developed in the laboratory for a particular set
of conditions. The setting of quantitative limits for field control, such as
described by Bell and Singh, must recognize the inherent errors involved.
This problem is compounded by information from Emery, Finn, and Lee
that the uniformity of density within the test sample is a significant factor
controlling liquefaction in addition to the average D~.
This paper summarizes the results of the two symposium sessions, the
first dealing with the measurement of relative density and the second deal-
ing with applications. Because most of the criticism of relative density
centered around the errors in its determination rather than the concept
itself, the first half of the paper discusses the types and magnitude of error
involved. The second half of the paper evaluates the test procedures and
applications of relative density. Finally, recommendations are provided
for future action.
The opinions given in this paper are those of the authors. They do not
necessarily represent those of the other symposium participants or the
officers and members of the American Society for Testing and Materials
(ASTM) Committee D-18 on Soil and Rock for Engineering Purposes.
Indeed, even at the end of the symposium, opinion was still divided on
many of the issues. It is the hope of the writers that this paper will provide
the basis for understanding these various viewpoints, however.
Types of Error in Relative Density
Errors are generally classified into three categories: (1) systematic, (2)
random, and (3) mistakes. Examples of each of these in connection with
relative density measurement are the following:
Systematic
1. Deviation from prescribed values of vibration frequency, amplitude
and surcharge.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License A
SELIG AND LADD ON MEASUREMENTS AND APPLICATIONS 489

2. Weighing scales out of adjustment.


3. Calibration errors in in situ density measuring devices.
4. Incorrect volume for mold used in determining maximum and mini-
mum density.
5. Use of nonstandard test procedures.

Random
1. Sample variability--particle size, shape, and gradation.
2. Variation in measurement techniques from test to test such as in (a)
pouring sand in minimum density test, and (b) preparation of hole for
in situ density measurement.
3. Changes in environmental conditions.
4. Round-off errors in reading weights and volumes.
5. Density variations in fill.

Mistakes
1. Misreading scales.
2. Using wet unit weight instead of dry unit weight.
3. Calculation errors.
4. Taking single amplitude value as double amplitude.

Errors by mistake must be avoided and can only be avoided by careful


work and constant checking.
Systematic and random errors often can be difficult to distinguish. For
example, if scales are out of adjustment by different amounts in different
labs, the variation caused in relative density among the labs will appear
as a random error. However, the difference in average values obtained be-
tween two independent labs will appear as a systematic error from this
cause. Distinguishing whether an error is systematic or random is important
because the method of error analysis is different in these two categories.
Two other commonly used terms which require understanding are ac-
curacy and precision. Although often used interchangeably, their meanings
are quite different. Systematic error is a measure of accuracy, and random
error is a measure of precision of a quantity. The difference between these
two terms is illustrated in Fig. 1. Accuracy is measured by the difference
between the correct value and the measured average of a set of repeated
tests. Because an inaccurate value can be too low or too high, this error
will have either a plus or minus sign. Precision is measured by the scatter
in the results of a group of repeated tests. Hence, if only one test is run, no
informatig~ on precision is indicated and inaccuracy cannot be distin-
guished from imprecision. Common ways of expressing precision or random
errors are (1) standard deviation, (2) range, (3) variance, and (4) coefficient
of variation. These terms are defined in the paper by Tavenas, Ladd, and
LaRochelle.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lic
4~0 RELATIVEDENSITYINVOLVING COHESIONLESSSOILS

C)
~ verage of Data
-,.7
i
-o-
0
- -
0
-
~
- o - - ~ - -
0
-
l ~ Random Error

J Systematic ~ o
Error ~ o o
(D
r~ . . . . . . . -- _ _ _ o _ _ _ _

\, 'True Value" >


~
o o
O O
~p - r- O
nr n--

Test Number Test Number

a.) Precise, Not Accurate b.) Accurate, Not Precise

0
0 0
0
0
. . . . . . . o

o~
r r 0 0
0 0
r~ o- ~ -~ -~-~ ~ -o-j -n- o - E3
0

(U q)
2*

_o
O
(E r

Test Number Test Number

c.) Accurate & Precise d.) Inaccurate & Imprecise

FIG. t--Illustration of accuracy and precision.

To calculate relative density, values are required for m a x i m u m density,


minimum density, and i n situ density, each of which has an error associated
with it. T h e error ir/~relative density resu ring from systematic errors in the
three densities m a y be calculated from the expression [1]4

AD~ - ODd OD~ A'Ymin "~ OD~

in which y is the i n situ density, and A indicates the systematic error.

4The italic numbers in brackets refer to the list of references appended to this paper.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem
SELIG AND LADD ON MEASUREMENTS AND APPLICATIONS 491

The partial derivatives which m a y be obtained from Eq 16 are given b y


ODd (?_eml._-_ ?..)_) (?min~ X 100 (2)
~?max- (?max- ?mi.)2 \ ? /
and
aDd (? _--__~max) (?max)
O?min -- (?max -- ?rain) 2 \ T / X 100 (3)
and
ol).= (loo (4)
a? (?max -- ?mi.) \ 72 !
After the derivatives have been substituted into Eq 1, the result is
~/max~min(1O0) [ (?rain _--_?) (h?max~
ADd
?(?max -- ?min) lL(?max - - ?min) " ?max
' \ /

-]- (?max -- ?rain) -~- (5)


or after dividing both sides b y Dd,
ADd ?mi: [ (?rain__--_?) (h?max~
D, (? - ?mi.) L(?m~x -- ?mid \ ?max /

+ (?max -- ?rain) ~- (6)


Equation 6 is equivalent to Eq 10 of Yoshimi and Tohno, who point out
that this equation is accurate only for density errors smaller t h a n about 2
percent. Because the expression for Da is a nonlinear function of the three
densities involved, for larger density errors the systematic errors in Da m a y
be calculated more exactly b y substitution into the following expression
ADa = Dd(? + A?) -- Da(?) (7)
and rearranging terms. The result will be the set of Eqs 0 b y Yoshimi and
Tohno. These authors show the error relationships graphically in their
Fig. 4. For simultaneous errors in ? . . . . ?mln, and ? the resulting values of
ADa/Dd may be added, keeping in mind the correct signs. An example of
the combined effect of errors in both ?max and ?ml, is given in Fig. 5 b y
Yoshimi and Tohno.
Using standard deviation (a)5 to express random errors, the error in

s Standard deviation is a measure of data scatter. About 68 percent of the data are ex-
pected to fall within • of the average; 95 percent of the data should fall within • of
the average,

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
492 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

relative density caused b y random errors in the three densities m a y be


calculated from the expression [1]

( OD~~2 (OD~ ~2 (ODDS2


which is the same as Eq 3 b y Yoshimi and Tohno. Examples of the effects
of the random errors in the densities on the resulting random error in Da
are given by Yoshimi and Tohno.
One significant difference which may be seen between Eq i for systematic
errors and Eq 8 for random errors is that systematic errors add algebrai-
cally, while random errors add as the square root of the sum of the squares.
In this latter case the signs do not affect the result.
Another important difference between random and systematic error
affects the accuracy of soil testing. Random errors can be reduced to any
desired degree by repeating the test and averaging the results. Thus, if
r is the standard deviation error associated with a particular single
value of maximum density, the error in the average of n repeated tests for
maximum density ~vm~, is
1
(~/max = O'~rnax.V/7~ (9)

Hence, as long as the repeated tests include the major sources of error
and the errors are essentially normally distributed, repeating the test will
reduce the variability of D~. In contrast, no amount of repetition will reduce
systematic errors, t h a t is, the error in the average of n repeated tests con-
taining the same systematic error will be the same as the error in a single
test. The only way to reduce systematic errors is to eliminate them b y
controlling the procedures as pointed out by Yoshimi and Tohno.

Magnitudes of Error in Relative Density


The possibility of large errors in Dd has been recognized b y a number of
previous writers. As early as 1953 D'Appolonia [2] demonstrated t h a t a
4-2 lb/ft 3 error in two of the three densities would cause an error of 4-20
percent in Dd. At t h a t time, he expressed the opinion that a maximum
error of 4-10 percent in D~ should be attainable in practice. Considerable
information on errors associated with relative density is also contained in
these proceedings papers. Errors in Dd as large as 4-50 percent have been
indicated which raise serious question about the quantitative value of the
concept.
A reevaluation of these reported errors is needed using information pre-
sented in the previous section because of the possible misinterpretation of
the data. In most cases systematic and random errors have not been dis-
tinguished even though the Dd errors have usually been calculated based

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to L
SELIG AND LADD ON MEASUREMENTS AND APPLICATIONS 493

on systematic error theory and the density errors have usually been random.
T h e extent to which D~ error can be reduced b y averaging repeated tests
has also not been fully analyzed.
Based on the writers' past experience and an examination of the data in
the symposium papers, the values listed in Table 1 were chosen to realis-
tically represent random and systematic errors in the maximum, minimum,
and in situ density determinations. The lowest error expected is indicated
as well as the error under typical conditions; however, larger values than
given in Table 1 are possible in each case.
For nominal values let ~min = i00 Ib/ft 3 and ~m~x --- 120 Ib/ft a. If 104
and 115 Ib/ft 3 are taken as in situ densities, these examples will represent
23 and 79 percent relative density, respectively. Equation 5 for systematic
error will become, after substitution of the three nominal density values:
Dg = 23 percent
AD~ = --1.0 AVm~ -- 4.6 AVml. + 5.5 ~V (10)
Da = 79 percent
ADa = --3.3 i'~max - - 1.3 A~/mi. + 4.5 A~ (11)
And Eq. 5 for random error will become:
Da = 23 percent
a~Dd = 0.9 a ~ + 21.5 a2~ml. + 30.7 a~~ (12)
Da = 79 percent
a:D~ = 10.8 ~ 2 + 1.7 ~mi~ + 20.6 a~~ (13)
The relative effect of each source of error on Da is readily apparent from
the coefficients of Eqs 10 through 13. As expected the error in ~m~x has
the largest effect for high Dd while the error in "y~i, has the greatest effect
for low Da. The in situ density error has the largest effect in all of these
examples. Based on Eqs 10 through 13, the errors in Dd for each error
value in Table 1 are shown in Table 1 assuming t h a t only one source of
error exists at a time, t h a t is, when A3'm~ ~ 0, t h e n A3,~, ----- A~/ = 0,
etc.
Unless the systematic errors are known to be compensating, the com-
bined Da error of all three density measurements will be the sum of the
absolute values of the errors f r o m the individual effects. The combined D~
error from r a n d o m effects is the square root of the sum of the squares.
F r o m the data in Table 1 the combined errors computed in this manner
are as given in Table 2.
These results in Table 2 indicate t h a t the smallest standard deviation
error in Da likely to exist is • percent. I n the field errors less t h a n •
percent are not likely and errors of • percent are quite possible. Sys-

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to
.O

T A B L E 1--Representative errors in density tests. Z

T y p e of T e s t Rando~a error ~Dd Systematic error ADd


s t a n d a r d deviation range Am, l b / f P
for single, Dd = 23% Dd = 79% Dd = 23% Dd = 79% r--
m e a s u r e m e n t a, l b / f t 3
Z

I n situ, field n
lowest 1.0 5.5 4.5 1.0 5.5 4.5 O
typical 2.5 13.7 11.2 3.0 16.5 13.5
O
Lab specimen 0.5 2.7 2.2 0.5 2.8 2.3
M a x i m u m density
lowest 0.8 0.8 2.6 1.0 - 1 -3.3
typical 2.0 1.9 6.6 3.0 - 3 - 9.9
M i n i m u m density
lowest 0.5 2.3 0.7 1.0 -4.6 - 1.3
typical 1.5 6.8 2.0 2.0 -9.2 -2.6

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reprodu
SELIG AND LADD ON MEASUREMENTS AND APPLICATIONS 495

TABLE 2--Combined errors in relative density.

add ADd
Da = 23% D,~ = 79% Dd = 23% Dd = 79%

Lowest-Field 6.1 5.3 11.1 9.1


Lowest-Lab 3.7 3.5 8.4 6.9
Typical 15.6 13.1 28.7 26.0

tematic errors are expected to range from a m i n i m u m of 6 percent to as


high as 29 percent. Note t h a t the standard deviation error represents 68
percent of the values while the systematic error represents the entire (100
percent) range.
Tavenas, Ladd, and LaRochelle report D~ standard deviation errors of
3 to 7 percent within the same lab and 10 to 20 percent when data f r o m all
participating labs are considered. Tiedemann gives values of about 4 and
10 percent of the same two quantities. Because the i n s i t u density was as-
sumed to have zero error in these calculations, the actual errors in Dd will
be higher b y probably 10 percent. Bell and Singh list values of aDd obtained
from the literature ranging from -4-9 to ~ 2 0 with most values varying
from 15 to 20. This is consistent with the " t y p i c a l " values in Table 2.
The papers b y Frost, Stephenson, and Low and Sener concerning rockfill
show t h a t the values indicated as typical in Table 2 are much lower t h a n
should be expected in such coarse materials having m a x i m u m particle
sizes in the range of 6 in. to 2 ft.
The previous analysis has assumed t h a t only one test each of ~%ax, ~'min,
and ~ was performed in order to calculate D~. Suppose t h a t four and nine
repeats of each of these three measurements were made and the values
averaged. The resulting errors in the calculated value of Dd would be those
given in Table 3 based on E q 9. A significant reduction in error is shown.
Thus, a properly designed sampling and testing plan could reduce the error
in Dd to :t:5 to 8 percent under typical field conditions. This, of course,
assumes either t h a t systematic errors have been essentially eliminated, or
t h a t the same lab established the performance correlation curves as con-
ducted the control tests so t h a t the systematic errors do not influence t h e
comparison.

TABLE 3--Standard deviation error in relative density.

aDa for Number of Repeats Indicated, lb/ft 3


1 4 9

Lowest-Field 6.0 3.0 2.0


Lowest-Lab 4.0 2.0 1.3
Typical 15.0 7.5 5.0

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License A
496 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

Applications of Relative Density


Many of the papers in the symposium are based on experience in the
application of Dd to engineering problems. These papers helped to estab-
lish the user needs, which in turn should be the basis for determining
changes and improvements required in the ASTM procedures. Opinions on
the usefulness of Dd ranged from one extreme to the other. The correct
appraisal is somewhere in the middle, but depends upon the application.
The writers' opinion is that Dd has value, but that it has frequently been
overextended with a false sense of reliability, or improperly used. It is not
the purpose of this discussion to arrive at a judgment whether to keep or
eliminate Dd, but rather to examine some of the factors concerning its use
and then to determine what changes in the ASTM procedures are required.
Like any other geotechnical design project, the approach to be taken should
be determined by a competent engineer who has project responsibility.
With full knowledge of the advantages and limitations, he must decide
whether Dd will help him. In addition, he should have control over the de-
sign recommendations, the establishment of specifications, and the quality
control of the compacted material which would include the field and lab
measurements. It is clear from the papers and the discussion that any other
approach will have a high probability of an incorrect conclusion.
Relative density is determined from the well-known equation
emax -- e
Dd = X 100 (15)
emax -- emin

in which e is the void ratio of the soil in situ, and emax and emin are the
reference void ratios. Expressed in terms of corresponding dry densities
this equation becomes

D~ = ~m~x( ~ -- ~/m,, ~ X 100 (16)


~/ \'~max -- ~Ymin/

Because Dd is almost always calculated from Eq 16, Cornforth and others


prefer to use the slightly simpler form

Rdd = ( "y - - ~Ymin ~ X 100 (17)

which results from dropping the ratio (~m~x/~/) from Eq 16. The quantity
Rd~ is called relative dry density. It is unfortunate that Dd is no longer
known as relative void ratio because Dd is sometimes mistakenly computed
using Eq 17 instead of Eq 16. Another suggested alternate to Dd is relative
porosity nr defined as
•max -- ~b
mr - (lS)
nmax -- nmin

However, this appears to offer no advantage over Eq 15.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. N
SELIG AND LADD ON MEASUREMENTS AND APPLICATIONS 497

Another quantity frequently used in compaction control is relative com-


paction or percent compaction given by

R~- X 100 (19)


~nlax

Many of the papers in the symposium discuss the use of Dd to control and
evaluate compacted fill composed of cohesionless materials. Poulos and
Hed favored Re for compaction control because they found a strong linear
correlation between ~/max and '~min for clean uniform sands. Youd's work
also indicates such a correlation. However, Reitz showed very poor corre-
lation between ~'max and ~'mi, for similar materials. He also found poor
correlation between these limiting densities and gradation characteristics.
Lacroix and Horn further examine the relationship between these limiting
densities and how they are affected by particle size distribution and shape.
Percent compaction Rc has several other advantages for compaction
control over Dd. Rc does not require a Vmi, test and it has lower variability
than D~. The project engineer may still prefer to use relative density in the
design stage. In such a case he should consider converting the density re-
quirements to Rc for use in field control. Low and Sener, in fact, believe
that the large particle size in rockfill materials require the use of Rc rather
than Dd because the "~min test is impractical.
The value for ~'m~xused to compute Da is customarily obtained from the
vibration procedure in ASTM Test for Relative Density of Cohesionless
Soils (D 2049-69). To minimize systematic errors, it is therefore necessary
to use ASTM D 2049-69 for control in projects for which specifications are
based on Dd analysis. The use of Re, on the other hand, permits ~'maxto be
obtained however the engineer wishes. The usual alternative is ASTM
D 2049-69 versus the impact tests ASTM Tests for Moisture-Density
Relations of Soils, Using 5.5-1b Rammer and 12-in. Drop (D 698-70) or
ASTM Tests for Moisture-Density Relations of Soils, Using 10-1b Rammer
and 18-in. Drop (D 1557-70). When the latter are used with Re, then the
control procedures are similar for a much broader range of materials than
those for which Dd is applicable. However, for clean granular materials,
ASTM D 2049-69 will probably give the highest ~ , ~ for use in computing
Re.
The correlation between Dd and blow count N from the standard pene-
tration test (SPT) has been the subject of frequent discussion in the sym-
posium. The purpose of this correlation is primarily for estimating Da
from the commonly used penetration test result. Schmertmann and Lacroix
in oral discussion and the paper by Lacroix and Horn have pointed out
some of the factors that effect N but not density of granular materials.
These include particle cementation, vertical and lateral stress level, SPT
procedures, and particle angularity. Furthermore, the SPT test has a large

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Licen
498 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

variability in addition to the De variability. It is not surprising then that


poor correlation generally exists between De and N. After a thorough study
of the literature deMello [8] concludes that the SPT gives only a etude
indication of Dd in general.
Relationships between Dd and N such as developed by Gibbs and Holtz
provide a partial eorreetion for the effect of N variation with depth. Another
relationship is presented by Melzer based on additional field test results.
Laeroix and Horn suggest that the relation between De and N should not
be solely based on the vertieal overburden stress, but should also take into
account the horizontal stress. They suggest a proeedure for doing this
which may help correct for the wide range of lateral eonfining stresses
resulting from the method of fill placement.
Papers by Leafy and Woodward and by Osterberg and Varaksin show
poor correlation between Dd and N while the paper by Bell and Singh
shows good eorrelation. However, in the ease of Bell and Singh, ape was
reported to be a rather large 4-17 to 20. These differences in experience
can be attributed to the particular combination of circumstances. It, is
possible that with careful attention to detail, in local situations, a correla-
tion between N and Dd ean be established. Otherwise the N - D e correlations
should only be used to indicate general trends. As an alternative though,
the statie cone appears to be a better indirect approaeh because it is a more
controllable test. The static eone still is subject to the same influences of
soil conditions as the SPT, however.
Papers by Holubee and D'Appolonia and by Corrdorth, as well as dis-
eussion comments from several persons, indicate that De is not a sufficient
index property of granular materials to represent performance in terms of
the engineering properties. That is, physical behavior such as compressi-
bility, shear strength, and liquefaetion potential are not uniquely related
to Dd. Other factors such as uniformity of size and angularity must also
be involved. Thus for example, twodifferent materials with the same Dd
would probably not have the same value of angle of internal friction, 4-
Cornforth has presented a rational proeedure for estimating ~ of clean
sands using both De and particle size and shape characteristics. In this
approach the drained strength ~ is equal to the sum of two components:
the ultimate strength ~c, which is a function of particle characteristics,
and an additional amount Ode which is a funetion of Dd. The author sug-
gests the substitution of a "minimum intergranular density" for Vr,i~. This
density is defined as that for which, in a triaxial test, the peak stress equals
the ultimate stress, and zero volume change exists at failure. The void
ratio at this density state would be called the "critical void ratio."
A1-Hussaini shows exeellent correlation between De and measured prop-
erties in triaxial and plane strain tests on a uniform, subangular sand. This
is expected, though, because only a small quantity of one material was in-

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lice
SELIG AND LADD ON MEASUREMENTS AND APPLICATIONS 499

volved in the tests. The small variability in D~ resulted from good specimen
reproducibility and because :r and ~m~nwere kept constant for all speci-
mens. Thus, the properties would have correlated equally with the speci-
men density, void ratio, Re, or Rd~.
Durham and Townsend provide data on liquefaction susceptibility of a
given sand as a function of Dd, confining pressure, and loading conditions.
A criteria based on Dd was established. However, these results, too, were
based on one material and, hence, any other measure of density would
have worked as well.
Melzer gives three examples of successful use of Dd. However, the par-
ticular nature of these examples leaves room to question whether Dd really
is as important as it appears. Data representing a wider variety of condi-
tions for each example are needed to substantiate the indication that im-
proved results with Dd were more than coincidence.
Because the particle shape and angularity have an important influence
on performance, suitable procedures are needed for measuring these index
properties. Some examples of procedures are given by Youd, Holubec, and
D'Appolonia, Cornforth, and Dickin. Other methods should also be con-
sidered because most of the methods in these papers are not particularly
suited to design practice.
Experience with a variety of direct methods for measuring in situ density
for Da calculation is discussed by Leary and Woodward, and by Osterberg
and Varaksin. Leafy and Woodward suggest that indirect methods such
as SPT, static cone, and plate load tests should only be used in uniform
fills, and then only after proper calibration. In variable fills a direct method
is required with at least one "Ymin and '~max determination for each in situ
location. Reitz, and Low and Sener also confirm this need for frequent
~min and ~maxtests, that is, at least one each for each in situ measurement.

Comments on Test Procedures


In his keynote address, Holtz briefly reviews the history of the present
ASTM D 2049-69 test procedures. He indicates that although there has
been a continuing receptivity to suggestions for improvement in the pro-
cedures, no better methods have yet been proposed. More strict adherence
to the present standards and the use of a larger number of tests, he sug-
gests, are the best ways so improve the results.
Apparently, from the papers and the symposium discussion, departure
from the specified vibration amplitude of 0.025 in. (double amplitude =
0.050 in.) at 60 Hz frequency for 8 min is common. Values reported in the
papers range from 0.012 to 0.049 in., and the time of vibration was both
much less than and much greater than 8 min. One of the reasons for the
deviation from the amplitude requirement is that 0.025 in. has sometimes
been interpreted as double amplitude. Such an interpretation, however, is

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License
500 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

contrary to the A S T M soil mechanics official terminology. This factor of


two decrease in amplitude may cause either an increase or a decrease in
~max depending on other conditions. Another reason for too low an ampli-
tude is t h a t many of the electromagnetic vibrators apparently cannot drive
the required mass at 60 Hz with this amplitude. A cam driven machine 8 is
now available which can meet the standards and provide an alternate choice
for the user.
Methods of checking the vibration amplitude during a density test are
often inadequate; therefore, the operator is not always aware t h a t he is
departing from specifications. The availability of a suitable method should
reduce the variability of the ~/max determination. However, the selection of
0.025-in. amplitude should also be reevaluated. The relationship between
maximum displacement (amplitude) d, maximum acceleration A, and vi-
bration frequency f for sinusoidal motion is
A = (2~f)2d (14)
The ASTM specifications then result in A = 9 g. The papers b y Dobry
and Whitman, b y Johnston and b y Brand suggest that a much lower maxi-
mum acceleration m a y give larger values of ~ . . . . which is desirable. Other
benefits of a smaller amplitude would be lower power requirements, less
problem in isolating the machine to avoid unacceptable vibration in the
laboratory, and less segregation of particles b y size.
It is generally recognized t h a t the ASTM D 2049-69 procedure does not
produce the real maximum density for a granular material, although it m a y
be difficult to devise a method that does. How important this m a t t e r is,
is a question which needs to be discussed. Certainly, the information on
Dd error indicates t h a t any alternate method which has a larger variability
is not acceptable, even if it gives a larger ~ . . . . A consistently lower ~max
represents a systematic error which m a y be preferable. Because careful
adherence to the ASTM D 2049-69 Procedures results in a small error for
soils, the test m a y be shown to have very good reproducibility which is
unlikely to be improved. However, the systematic error between the ~max
from A S T M D 2049-69 and the true ~m,x is not the same for all granular
materials. The reason is that the optimum combination of frequency and
acceleration (or amplitude) varies with the soil conditions, whereas the
A S T M test holds them constant for all materials. Other factors such as
mold size, surcharge, and whether vibration is vertical or horizontal are
important. The paper b y Cumberledge and Cominsky, and b y Brand have
provided some new data on these effects.
A consideration of the purchase cost and the installation cost of the
equipment for the Ym,~ test in relation to the usefulness of the test sug-

e All American Tool and Manufacturing Co., Inc., Model 100 VP-D.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No
SELIG AND LADD ON MEASUREMENTS AND APPLICATIONS .501

gests to many people that this test has too low a cost/effectiveness ratio.
This is one of the most frequent complaints about the method. However,
in an age of increasing sophistication of testing and analysis, this criticism
seems to be a secondary concern. In fact, some discussers have expressed
the desire for the soil mechanics profession to put emphasis on finding the
most meaningful test with little regard for whether it is a simple, cheap
test. It has been further suggested that ASTM should find a means of
certifying labs to see that they have the proper equipment and use the
proper procedures.
An entirely different approach to determining ~max and ~/min was sug-
gested by Youd based on measurement of index properties. Correlation
curves were presented for clean sands. Additional data is given by Dickin.
It is unlikely that the precision of D~ using this method is as good as with
ASTM D 2049-69, except where careful correlation tests have been per-
formed for a narrow range of soil condition. Although the method described
by Youd circumvents the problems in using the vibrating table, the overall
cost of determining ~'max and ~'min by this method is not necessarily less
than with ASTM D 2049-69. Determination of particle sphericity or angu-
larity, for example, could be time consuming with present techniques.
However, the method has considerable value in estimating ~m,x and ~'minin
the absence of direct measurements and for evaluating variation in Dd
associated with changes in the index properties of the material. Data of
the type obtained by Youd and Dickin is therefore desired for other
gradations.
The applicability of ASTM D 2049-69 procedures depends on the ma-
terial characteristics, particularly the percent and plasticity of the fines.
Townsend provides additional guidance on how to make the choice between
ASTM D 2049-69 and the impact compaction test (ASTM D 698-70)
based on these factors. He also shows the effect of gradation and moisture
content on the difference in ~/.... with the two methods. With data of this
type the possibility may exist to refine the statement indicating limits of
applicability in ASTM D 2049-69 procedures.
Conclusions
In the opinion of the writers, the following items concerning the present
ASTM test procedures should be considered as a basis for modifying or
improving these standards:
1. The title of the ASTM D 2049-69 "Test for Relative Density of
Cohesionless Soils" is misleading. Relative density is not measured, it is
calculated from ~max, ~'m~., and in situ. However, no procedures for in situ
y are included in ASTM D 2049-69. Also, ~/maxand ~'m~ncan be used with
other indices than with D~. Therefore, the title of ASTM D 2049-69 should
be changed to properly represent the contents of this standard. An appro-

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lic
502 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

priate title might be, "Test for Minimum Density of Cohesionless Materials
by Pouring and for Maximum Density by Vibration." The standard should
indicate that ~'min and ~'max are not necessarily the true limiting densities.
Therefore, they could properly be renamed "low index density and high
index density."
2. The scope of ASTM D 2049-69 does briefly indicate the type of
materials for which the procedures are meaningful. However, the allowable
limit on the percent fines, for example, depends upon their plasticity and
the gradation of the granular component. Available data could be used to
refine the guidelines based on these factors to assist the user in determining
the appropriateness of these procedures.
3. Many engineers prefer other density indices such as Rc and Rd~ to
Dd. Therefore, if the calculations for Dd are kept in ASTM D 2049-69,
then the definitions of the other indices should be included as well.
4. Because of frequent misinterpretation, further clarification of the
frequency and amplitude requirements of the vibration test in ASTM
D 2049-69 is needed. A method of checking these parameters during a test
should be recommended. The elimination of significant systematic errors is
essential for valid use of Dd correlations with physical properties. One
source of these errors is the variation in effectiveness of the specified fre-
quency and amplitude in producing ~/maxfor the allowable range of granu-
lar materials. Some recent data also suggest that the specified vibration
conditions result in too high an acceleration. For these reasons further
consideration of the vibration parameters should be given. The possible
benefits of horizontal vibration instead of vertical, and air pressure sur-
charge instead of a weight should also be evaluated.
5. Considerable information on test variability is available. This in-
formation should be incorporated in the standard so that the user can
properly assess the confidence limits on his anMysis. A discussion of the
errors in Dd as a function of the errors in ~min, ~. . . . and in situ "y would
also be useful. The standards covering in situ density procedures should
also have variability statements added to them.
6. The methods of specimen preparation and field sampling procedures
need further elaboration. This discussion should point out the difficulty in
getting meaningful results when the fill material is thinly layered or very
heterogeneous.
The symposium has suggested the following needs which should be con-
sidered for new procedures:
1. The present ASTM D 2049-69 method is unsuitable for materials
having particle sizes above 3 in. Three papers in the symposium involved
maximum particle sizes ranging from 6 to 24 in. Standard procedures are
needed for at least in situ ~/and ~,. . . . Information in the symposium papers
should be considered in developing these standards.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to L
SELIG AND LADD ON MEASUREMENTS AND APPLICATIONS 503

2. It has been demonstrated that the physical properties of cohesionless


materials are a function not only of density, but also of the size, size dis-
tribution, shape, and angularity of the particles. The importance of these
parameters justifies standard test procedures for their determination.
Therefore, procedures should be developed for measuring sphericity and
angularity.
3. Terms such as loose, dense, and compact have been used to describe
the density state of granular materials. Appropriate terms should be se-
lected and their meaning defined in relation to relative density.
4. The large variability involved in D~ determination tends to mitigate
the argument for only one procedure for ")'max and "/min. Alternate pro-
cedures should be considered for these two index densities which are easier
to use, or less costly, or more portable if these objectives can be achieved
without sacrificing accuracy or precision. Systematic errors produced by
these changes may be perfectly acceptable if the user has developed his
own correlation curves between Da or similar index and the physical prop-
erties of the material.
The following conclusions are indicated concerning the application of
D~ to engineering problems:
1. As a concept, relative density has merit and it is useful in expressing
general trends in performance of granular materials.
2. In many applications such as compaction control, other density in-
dices appear to be as suitable or more suitable than Dd.
3. Relative density is not a sufficient index to correlate physical proper-
ties with the density state of cohesionless materials. Other indices like
angularity, sphericity, and uniformity are needed. This situation limits
the application of such correlation curves which are usually based on a
narrow range of material type.
4. In general, relative density has associated with it a random error of
about =t=10 to 15 standard deviation and a systematic error of 25 to 30
range. However, these errors can be reduced by appropriate testing. These
values indicate that D~ is not a precise index. They also indicate that a
large uncertainty exists in using correlation curves produced by another
investigator. Such curves should be used with caution and a considerable
amount of judgment. Replicate measurements are certainly required in
most cases to provide acceptable precision in any quantitative use of D~,
because the error in D~ associated wish a single measure each of "rr~,x,
~min, and in situ "y is too large.
5. Other field measurements such as static cone, bearing plate, and SPT
should be considered as an alternative to density methods for design and
control. Engineering judgment is needed to plan the best approach. Rela-
tive density should not be considered the key to all problems. Its benefits
are also limited when a narrow range of soil conditions are involved.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to Lice
504 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

The decision whether or not to use D~ should be made b y the project


engineer. This is not the purpose of the symposium. No amount of change
in the A S T M standards can substitute for the knowledge and experience
required to make this decision. I t is the desire of everyone participating
in this symposium t h a t the material presented will help the engineer make
the proper decision for his problem.

References
[I] Young, H. D., Statistical Treatment of Experimental Data, McGraw-Hill Book Co.,
Inc., 1962.
[~] D'Appolonia, Elio in Symposium on Dynamic Testing of Soils, A S T M S T P 156,
American Society for Testing and Materials, 1954, pp. 138-162.
[3] deMello, V., Proceedings, 4th Pan American Congress on Soil Mechanics and Founda-
tion Engineering, Puerto Rico, June 1971, Vol. 1, p. 1.

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further r
STP523-EB/Jul. 1973

Index
A Magnitudes (see Reliability and
Reproducibility under In-
Allowable fines, 6, 349, 350, 355, 359 place, Maximum, Minimum,
Angularity (see Roundness criteria) and Relative density)
Types, 75, 77, 488
C
Coarse aggregate correction, 221, F
224-231, 431 Friction angle (see Strength)
Compressibility
Plate load, 277, 395 G
Sensitivity, correlated to
Particle shape, 257, 304, 312 Gelatin solution, 185, 198
Plane-strain consolidation, 338
Relative density, 257, 338, 340 It
Relative compaction, 417
Triaxial consolidation, 338 Humphres method, 221
Stress-strain behavior
Triaxial/Plain strain, 339, 342
Cone penetration (see Dynamic and Impact compaction
Static) Correlations with
Critical void ratio, 329, 342 Kneading, 220
Standard and modified, 45
D Vibratory, 145, 220, 307, 352,
Degree of compaction (see Relative 358
compaction) Dry versus saturation, 411
Density ratio, 252, 492, 496 Large scale tests, 215
Dynamic compaction (see Impact Reliability (variability), 12, 14,
compaction) 43, 44, 153
Dynamic cone penetration, 268, 470 Sensitivity to
Particle distribution, 359
E
Percent fines, 359, 409
Sample heterogeneity, 422
Errors Saturation, 408, 411, 422
Analysis of, 74, 490 In-place density

505

Copyright by ASTM Int'l


Copyright9 (all rights
by ASTM reserved); Fri www.astm.org
International Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreemen
506 RELATIVE DENSITY INVOLVING COHESIONLESS SOILS

Control densities, 197, 385 L


Correlations
Balloon/Cylinders, 388 Liquefaction
Balloon/Nuclear, 274, 293 Critical void ratio, 329
Balloon/Sand cone, 257 General, 183, 319, 455
Balloon/Test pit, 391 Sensitivity to
Equipment calibration, 197 Effective stress, 329
Equipment recommendations Relative density, 329
Nuclear, 274-277 Specimen size, 331
Washington densometer, 257- Specimen uniformity, 183
259, 385 Specimen preparation, 184, 324
Reliability (between organiza- Test procedure, 324
tions), 14, 202, 213, 231, 417, Testing equipment, 321
493
Reproducibility (within organiza- M
tions), 14, 416
Required measurements, 80, 92, Maximum density (laboratory)
206 Controlling factors, 99, 101, 106-
Selection of test method, 385 109, 385, 422, 499
Sensitivity to Correlations with
Freezing, 371 Bagnold's grading parameters,
In-situ water content, 203-205 360
Plastic liners, 213 Co, 94
Sample heterogeneity, 256, 387, C., 94, 108, 135, 255, 388, 414,
407 449
Specimen size, 209 D10, 447
Test depth, 275, 370, 386, 419 Ds0, 106
Test methods D6o, 448
Deep depths Impact compaction, 45, 145,
Freezing, 367, 369 154, 220, 307, 352, 358
Sampling, 368, 382, 393, Particle distribution, etc, 253,
458 360, 388, 431
Shallow depths Wet versus dry, 253, 360, 388,
General, 198, 257, 369, 431, 447
382, 388, 389 Errors
Large scale, 212, 238, 246, Random, 78, 490
427 Systematic, 77, 490
Nuclear, 274, 393 Reliability (between organiza-
Test procedure, 199, 202, 274, tions), 31, 32, 137, 152, 231,
370, 373, 382, 386 493
Time, 279 Reproducibility (within organiza-
Typical results tions), 34, 68, 146, 412
Rock fills, 214, 241 Selection of test method, 349, 385,
Sand fills, 370, 383, 408, 415 5O0

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further reproductions authoriz
INDEX 507

Sensitivity to Particle shape, 11, 99, 106, 108,


Particle distribution, 99, 108, 114, 117, 257, 308
145, 220, 253, 256, 350, 360, Particle size, 11, 36, 99, 108,
388, 422, 431 117
Particle shape, 11, 99, 108, 114, Percent fines, 255, 350
253, 257, 308 Specimen similarity, 10
Particle size, 11, 36, 99, 108, Test procedure, 19, 34, 36, 323
145, 153, 220, 223-231 Testers, 91
Percent fines, 255, 349, 350, 355 Testing errors, 8, 78
Sample heterogeneity (in-situ Testing procedures, 76, 99, 100,
layering), 218, 256, 387, 407, 115, 122, 239, 296, 297, 307,
422 323, 409
Saturation, 293, 350, 355, 411, Testing requirements, 84, 499
422, 447
Specimen similarity, 10
N
Test procedures, 19, 36, 124,
134, 139, 146, 172, 216-220,
Nuclear density testing, 274-277,
307, 323, 369, 385, 408, 497
393
Testers, 91
Testing errors, 8, 78, 489
Type of fines, 350, 359 P
Testing procedures
General, 76, 99, 100, 115, 122, Particle breakdown, 25, 26, 27, 149
134, 172, 293, 323, 354 Particle-size analysis
Large scale, 215, 236, 239, 430 Accuracy, 24-26, 64
Testing requirements, 84, 349, 500 Sensitivity to
Minimum density (laboratory) Specimen selection, 28
Controlling factors, 101, 106-109 Specimen size, 27, 209, 428
Correlations with Weight retained, 28
Co, 94 Specimen weight requirements,
Ca, 94, 108, 114, 135, 452 209, 428
Dlo, 450 Typical results
Ds0, 106 Rock fills, 211, 244, 431
D~o, 451 Sand fills, 365, 384, 406, 446
Errors (type) Penetration resistance (laboratory),
Random, 78, 490 315
Systemic, 77, 490 Percent compaction (see Relative
Reliability (between organiza- compaction)
tions), 28, 31, 91, 137, 493 Plane strain, 334, 340, 342
Reproducibility (within organiza- Plate load test, 277, 395
tions), 34, 68 Porosity, 144-120
Sensitivity to Preparation of test specimen, 122,
Particle distribution, 99, 108, 131, 184, 197, 286, 309, 324,
117, 256, 350 337

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No furt
508 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

R Precision, 489
Reliability (between organiza-
Relative compaction tions), 10, 13, 15, 20, 31, 32,
Applications, 48, 232, 417, 420, 39, 40, 93, 377, 378, 479, 493,
497 5O3
Correlations with Reproducibility (within organiza-
Control curves, 443 tions), 34, 63, 68
Cu, 420 Sensitivity to
De0, 420 Control densities, 197, 422
Maximum density, 421 Controlling factors (see Maxi-
Relative density, 48, 293, 419 mum, Minimum, and In-
Definition, 47, 231, 252, 292, 410, place density)
496 Error analysis, 492
Limitations (shortcomings), 420, Groundwater (see also Stand-
422 ard penetration), 372
Merits (advantages), 232, 417, In-place density, 71
426, 503 In-situ layering, 218, 256, 387,
Sensitivity, to 407, 422
Control curves, 422 Liquefaction, 329
Sample heterogeneity, 422 Particle distribution (see also
Relative density Maximum and Minimum
Accuracy, 489 density), 99, 253, 256, 382,
Allowable fines, 6, 349, 350, 355, 420
359 Particle shape (see also Maxi-
Applications, 18,251,252,279,382, mum and Minimum density),
455,464, 480, 482, 496, 503 11, 99, 114, 253, 257, 382
Correlations with Particle size (see also Maximum
Dynamic cone, 269, 470 and Minimum density), 11,
Plate load test, 277, 395 36, 99
Relative compaction, 48 Sample heterogeneity, 218, 256,
Standard penetration, 262, 375, 387, 407, 422
376, 378, 392, 459, 470 Specimen similarity, 10
Static cone, 398, 459 Test procedures, 19, 34, 36, 139
Definitions (equations), 6, 8, 19, Testing errors, 8, 78, 492
75, 85,231,252, 292, 382, 410, Time, 279
496 Relative dry density (see Density
Errors (type/magnitude), 8, 78, ratio)
490, 503 Roundness (angularity) criteria,
Limitations (shortcomings), 9, 19 101,102, 116, 305, 499
43, 83, 86, 96, 232, 252, 279,
305, 382, 413, 417, 420, 426, S
463, 480, 482, 483, 487, 496,
498, 503 Sample similarity
Limits of acceptance, 71 Dams, 211

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No further repr
~o~ 509

General, 10 Density, 284, 465, 471


Sand fills Density factors, 288
General, 251, 366, 370, 383, 408, Minimum density, 286
415 Minor principal stress, 339
Hydraulic, 403, 419, 445 Particle distribution, 305
Sensitivity to Particle shape, 257, 305, 309,
Groundwater table, 372 466
Unconfined boundaries, 387 Particle size, 305, 466
Water content, 422 Relative density, 257, 281, 344,
Soil strain gage, 188 466, 471, 498
Specimen (test) uniformity Triaxial, 282, 286, 299, 344
Effect of liquefaction, 190 Stress-strain characteristics (see
General, 122, 129, 131, 189, 197 Compressibility)
Tests for, 129, 185
Sphericity (see Roundness) T
Standard penetration test
Controlling factors, 259, 264, 265, Test specimen preparation (see
397, 497 Preparation of test specimen)
Correlation with Testing frequency, 495
Dynamic cone, 268, 269, 470
Relative density, 262, 397 V
Static cone, 270, 398
Sensitivity to Vibratory compaction (laboratory)
Drilling procedure, 260 Controlling factors, 385, 430, 500
Groundwater table, 261 Recommendations, 501
Horizontal stresses, 264, 265, Relationship to
391,398 Acceleration, 126, 137, 147, 157,
Particle size, 261 162, 177, 219, 500
Time, 279 Amplitude, 137, 147, 177
Vertical stresses, 260, 263 Bagnold's grading parameters,
Static cone penetration, 270,'398 360
Statistics, 22, 63, 490 Container size, 128, 147, 162
Strength (sands) Frequency, 126, 151, 157, 162,
Controlling factors, 257, 281, 287, 500
498 Height (mold), 162, 178
Density component, 290 Impact compaction, 45, 145,
Energy correction, 345 154, 220, 307, 352, 358
Liquefaction (see Liquefaction) Mold type, 159
Penetration resistance (labora- Particle shape, 125
tory), 315 Percent fines, 293, 350, 355, 359
Plain strain, 285, 299, 344 Saturation, 147, 293, 355, 447
Prediction, 290, 300 Surcharge, 138, 147, 175
Relationship to Specimen heterogeneity, 122,
Crushing, 290 129, 218

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No furt
510 RELATIVEDENSITY INVOLVING COHESIONLESS SOILS

Test procedure (large scale), Table, 122, 134, 142, 157, 173,
217 215, 354, 369, 500
Test procedure (small scale), Vibrating hammer, 142, 215,
385, 500 293
Type of fines, 350, 359 Theoretical model, 163-167
Vibration time, 124, 175, 219 Volumetric behavior
Test procedures (other than Energy, correction, 345
ASTM) General, 328, 340
Attached vibrators, 173 Triaxial versus plane strain, 340
Hammering, 307, 369
Large scale, 215, 236, 430 W
Providence, 134 Weighing accuracy, 23

Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreement. No furth
Copyright by ASTM Int'l (all rights reserved); Fri Mar 11 16:13:06 EST 2016
Downloaded/printed by
(UFPE) Universidade Federal de Pernambuco ((UFPE) Universidade Federal de Pernambuco) pursuant to License Agreem

Вам также может понравиться