Вы находитесь на странице: 1из 215

Graduate School ETD Form 9

(Revised 12/07)

PURDUE UNIVERSITY
GRADUATE SCHOOL
Thesis/Dissertation Acceptance
This is to certify that the thesis/dissertation prepared
By Aleksandra Zofia Radlinska
Entitled
Reliability-based analysis of early-age cracking in concrete

For the degree of Doctor of Philosophy

Is approved by the final examining committee:


W. Jason Weiss
Chair

Jan Olek

John E. Haddock

Surendra Shah

To the best of my knowledge and as understood by the student in the Research Integrity and
Copyright Disclaimer (Graduate School Form 20), this thesis/dissertation adheres to the provisions of
Purdue Universitys Policy on Integrity in Research and the use of copyrighted material.

W. Jason Weiss
Approved by Major Professor(s): ____________________________________

____________________________________
Approved by: Rao S. Govindaraju
Head of the Graduate Program

July 2, 2008
Date

Graduate School Form 20


(Revised 10/07)

PURDUE UNIVERSITY
GRADUATE SCHOOL
Research Integrity and Copyright Disclaimer

Title of Thesis/Dissertation:
Reliability-based analysis of early-age cracking in concrete

Doctor of Philosophy
For the degree of ________________________________________________________________

I certify that in the preparation of this thesis, I have observed the provisions of Purdue University
Executive Memorandum No. C-22, September 6, 1991, Policy on Integrity in Research.*
Further, I certify that this work is free of plagiarism and all materials appearing in this
thesis/dissertation have been properly quoted and attributed.
I certify that all copyrighted material incorporated into this thesis/dissertation is in compliance with
the United States copyright law and that I have received written permission from the copyright
owners for my use of their work, which is beyond the scope of the law. I agree to indemnify and save
harmless Purdue University from any and all claims that may be asserted or that may arise from any
copyright violation.

Aleksandra Zofia Radlinska


________________________________
Signature of Candidate

July 2, 2008
________________________________
Date

*Located at http://www.purdue.edu/policies/pages/teach_res_outreach/c_22.html

RELIABILITY-BASED ANALYSIS OF EARLY-AGE CRACKING


IN CONCRETE

A Dissertation
Submitted to the Faculty
of
Purdue University
by
Aleksandra Radlinska

In Partial Fulfillment of the


Requirements for the Degree
of
Doctor of Philosophy

August 2008
Purdue University
West Lafayette, Indiana

3344112




3344112


ii

PREFACE

There exists a world. In terms of probability, this borders on the impossible


Jostein Gaarder

iii

ACKNOWLEDGMENTS

I would like to begin with expressing sincere appreciation and gratitude to my


advisor, Professor Jason Weiss. Thank you for your guidance, invaluable advice
and support you provided me with during the course of this research work. I am
thankful for your constant motivation and a great example of research and
teaching excellence I received from you.

I sincerely thank Professor Jan Olek, who served on my defense committee, but
has also provided me with constant support and encouragement throughout my
time at Purdue. I thank you and your wife Anna for all your time and patience,
leading me through the maize of graduate student life. I would also like thank
Professor John Haddock and Surendra P. Shah for serving as my committee
members and providing me with critical review of my research work and
dissertation.

I greatly acknowledge financial support provided by Purdue University through


Bilsland Fellowship and by Portland Cement Association for honoring me with
PCA Education Foundation Fellowship (Project F07-05).

The research

presented here was conducted in the Materials Characterization and Sensing


Laboratory at Purdue University and I would like to acknowledge the support that
had made this laboratory and its operation possible.

I specially thank Janet Lovell and Mark Baker for their valuable assistance in
preparation and performance of the experimental work.

I also would like to

iv

appreciate the much needed help and assistance of Maeve Drummond and
Cathy Ralston.

I have very special thanks to my great friends: Farshad Rajabipour, Ryan


Henkensiefken, Gaurav Sant, and Brooks Bucher. I greatly value our friendship
and also the rewarding research we conducted together Chapter 8 of this
thesis is based on the work we performed jointly. Additionally, I would like to
extend my acknowledgments to my friend Pietro Lura. Our discussions have
always been extremely stimulating and have broadened my horizons.

I would also like to thank Mateusz Radlinski, Guy Mazzotta, Thomas Schmit,
Jae-Heum Moon, Vinit Barde, Francois Paradis, Amir Poursaee, Kambiz Raoufi,
Mohammad Pour-Ghaz, Mukul Dehadrai, Javier Castro, Brad Pease, Jon Couch,
Kevin Coates, Mike Norfleet, Adam Rudy, Karol Kowalski, and Chadi El Mohtar.
Thanks to you all, my time at Purdue was really enjoyable and memorable.

Finally, I express my most sincere gratitude to my Mom. Whatever I would like to


say here, you understand without the words.

TABLE OF CONTENTS

Page
LIST OF TABLES .................................................................................................ix
LIST OF FIGURES ...............................................................................................xi
ABSTRACT ....................................................................................................... xvii
CHAPTER 1. INTRODUCTION ............................................................................ 1
1.1. Research Motivation................................................................................ 1
1.2. Concrete in Sustainable Design .............................................................. 2
1.3. Objectives................................................................................................ 3
1.4. Organization ............................................................................................ 5
CHAPTER 2. QUANTIFYING THE VARIABILITY IN THE SHRINKAGE
MEASUREMENTS IN CEMENT PASTE AND CONCRETE EXPOSED TO
DRYING AND REWETTING................................................................................. 8
2.1. Introduction.............................................................................................. 8
2.2. Experimental Approach in Paste Shrinkage Studies ............................. 10
2.2.1. Materials ........................................................................................ 10
2.2.2. Mixture Proportions, Specimen Preparation, and Specimen
Geometry ................................................................................................. 10
2.2.3. Experimental Procedures............................................................... 12
2.3. Experimental Results............................................................................. 14
2.3.1. Shrinkage and Rewetting Cycle for w/c=0.30 and
w/c=0.30+5%SRA Pastes........................................................................ 14
2.3.2. Variability in Shrinkage and Weight Measurements (w/c=0.30
and w/c=0.30+5%SRA)............................................................................ 18
2.3.3. Weight Change versus Shrinkage (w/c=0.30 and
w/c=0.30+5%SRA) .................................................................................. 21
2.3.4. Shrinkage and Rewetting Cycle for w/c=0.40 and
w/c=0.40+5%SRA Paste.......................................................................... 25
2.3.5. Initial Shrinkage and Weight Loss of Paste (w/c=0.40 and
w/c=0.40+5%SRA) .................................................................................. 27
2.3.6. Rewetting of Paste and Secondary Shrinkage (w/c=0.40 and
w/c=0.40+5%SRA) .................................................................................. 30
2.3.7. Weight Change versus Shrinkage (w/c=0.4, w/c=0.4+5%SRA)..... 35
2.4. Discussion of Experimental Results ...................................................... 38

vi

Page
2.4.1. Comparison of the Shrinkage of Cement Paste with w/c=0.30
and w/c=0.40 (with and without SRA) ...................................................... 38
2.4.2. Effectiveness of SRA in Shrinkage Reduction of Cement Paste.... 39
2.4.3. Effect of SRA on Recoverable Shrinkage ...................................... 40
2.4.4. Variability in Shrinkage Measurements.......................................... 42
2.5. Experimental Approach in Concrete Shrinkage Studies ........................ 43
2.5.1. Materials ........................................................................................ 43
2.5.2. Experimental Methods for Shrinkage and Weight Loss
Measurements of Concrete..................................................................... 45
2.5.3. Experimental Methods for Concrete Aggregate Content Study ..... 47
2.5.4. Shrinkage and Weight Change of Concrete................................... 48
2.5.5. Aggregate Content of Concrete Specimens................................... 48
2.6. Summary ............................................................................................... 51
CHAPTER 3. MONTE CARLO SIMULATION .................................................... 53
3.1. Introduction............................................................................................ 53
3.2. Background on the Computational Method............................................ 54
3.3. Time-Dependent Material Property Model Inputs .................................. 58
3.4. Model Simulations Deterministic Predictions ...................................... 64
3.5. Model Simulations Considering Variability in Material Properties ....... 66
3.6. Predicting Cracking Performance .......................................................... 72
3.7. Summary ............................................................................................... 74
CHAPTER 4. RELIABILITY BASED APPROACH FOR ASSESSING
CRACKING RISK USING LOAD AND RESISTANCE FACTOR DESIGN ......... 75
4.1. Introduction............................................................................................ 75
4.2. Estimating Variability ............................................................................. 76
4.3. Reliability Approach............................................................................... 77
4.4. Comparison of MC and LRFD ............................................................... 82
4.5. An Example of Application of the LRFD Procedure ............................... 85
4.6. Reducing the Probability of Cracking Using Shrinkage Reducing
Admixtures.................................................................................................... 89
4.7. An Example of Using the LRFD Procedure to Determine the
Required Concentration of SRA in a Concrete Mixture ................................ 92
4.8. Summary ............................................................................................... 94
CHAPTER 5. THE RING TEST: A REVIEW OF RECENT
DEVELOPMENTS .............................................................................................. 96
5.1. Introduction............................................................................................ 96
5.2. Qualitative Restrained Ring Tests ......................................................... 97
5.3. Using the Ring to Quantify Stress Development.................................... 98
5.4. Additional Calculations to Quantify Ring Results................................. 101

vii

Page
5.4.1. Quantifying the Effect of Stress Relaxation (Creep) using the
Ring ....................................................................................................... 101
5.4.2. Influence of Fiber Reinforcement ................................................. 102
5.4.3. Influence of Specimen Geometry and Bond between the
Concrete and Steel ................................................................................ 103
5.4.4. Influence of Moisture Gradients ................................................... 104
5.4.5. Degree of Restraint...................................................................... 105
5.5. Recent Developments in the Restrained Ring Analysis....................... 106
5.5.1. Dual Ring ..................................................................................... 106
5.5.2. Effect of Variable Concrete Wall Thickness (i.e., Eccentricity)..... 108
5.6. Summary ............................................................................................. 108
CHAPTER 6. ASSESING THE REPEATIBILITY OF RESTRAINED
SHRINKAGE TEST USING ASTM C-1581 ...................................................... 110
6.1. Introduction.......................................................................................... 110
6.2. Research Significance......................................................................... 110
6.3. Experimental Equipment...................................................................... 111
6.3.1. Restrained Ring Geometry .......................................................... 111
6.3.2. Ring Instrumentation.................................................................... 111
6.3.3. Data Acquisition ........................................................................... 112
6.3.4. Environmental Chamber .............................................................. 113
6.4. Experimental Program ......................................................................... 114
6.4.1. Mixing Proportions ....................................................................... 114
6.4.2. Sample Preparation Procedure.................................................... 115
6.4.3. Testing Program .......................................................................... 116
6.4.4. Temperature Compensation ........................................................ 117
6.5. Experimental Results........................................................................... 118
6.5.1. Temperature ................................................................................ 118
6.5.2. Variability within a Single Restrained Shrinkage Test and
Repeatability of the Test ........................................................................ 119
6.5.3. Influence of Shrinkage Reducing Admixture (SRA) ..................... 120
6.5.4. Influence of Relative Humidity Conditions.................................... 121
6.6. Discussion of the Results .................................................................... 121
6.6.1. Accuracy of the Test .................................................................... 121
6.6.2. Assessing Probability of Cracking................................................ 122
6.7. Explaining Variability in the Cracking Behavior of Restrained
Concrete..................................................................................................... 123
6.7.1. Model Description ........................................................................ 123
6.7.2. Results of Simulations ................................................................. 126
6.8. Summary ............................................................................................. 129
CHAPTER 7. ANOVA ANALYSIS OF THE RESULTS OF THE
RESTRAINED RING TEST .............................................................................. 130
7.1. Introduction.......................................................................................... 130

viii

Page
7.2. Testing Methodology ........................................................................... 130
7.3. Single ANOVA Model .......................................................................... 131
7.3.1. Confirmation of Applicability of ANOVA ....................................... 133
7.3.2. GLM Procedure............................................................................ 138
7.4. Random-Effect Model.......................................................................... 143
7.5. Conclusions ......................................................................................... 143
CHAPTER 8. IMPORTANCE OF BOUNDARY AND DRYING CONDITIONS
IN CEMENTITIOUS SYSTEMS........................................................................ 145
8.1. Introduction.......................................................................................... 145
8.2. Experimental Measurements for Sealed and Unsealed Systems ........ 146
8.2.1. Materials ...................................................................................... 146
8.2.2. Mixing .......................................................................................... 147
8.2.3. Experimental Procedures............................................................. 147
8.3. Experimental Results........................................................................... 149
8.4. Theoretical Considerations for Shrinkage in Concrete ........................ 153
8.4.1. Internal Drying (Self-Desiccation) ................................................ 154
8.4.2. External Drying ............................................................................ 155
8.4.3. Relationships between Capillary Pressure, Pore Size, and
Relative Humidity................................................................................... 156
8.5. Interpretation of Experimental Results Based on the Theory .............. 160
8.5.1. Sealed (Autogenous) Shrinkage .................................................. 160
8.5.2. Unsealed (Autogenous + Drying) Shrinkage................................ 162
8.6. Conclusions ......................................................................................... 167
CHAPTER 9. SUMMARY AND CONCLUSIONS ............................................. 168
9.1. General Summary Related to the Assessment of Magnitude and
Variability in Shrinkage Measurements ...................................................... 170
9.2. General Conclusions Related to Reliability-Based Assessment of
Probability of Cracking in Concrete ............................................................ 170
9.3. General Conclusions Related to the Accuracy of the Restrained Ring
Test ............................................................................................................ 171
9.4. Summary of the Studies related to the Importance of Boundary
Conditions in Cementitious Systems .......................................................... 172
LIST OF REFERENCES .................................................................................. 173
APPENDIX A: SAS CODE ............................................................................... 186
VITA ................................................................................................................. 188
PUBLICATIONS ............................................................................................... 189

ix

LIST OF TABLES

Table

Page

Table 2-1 Paste samples investigated in the course of study............................. 14


Table 2-2 Mean shrinkage values and standard deviations for paste at
equilibrium (w/c=0.3, w/c=0.3+5%SRA)................................................... 18
Table 2-3 Mean shrinkage values and standard deviation for paste at
equilibrium (w/c=0.4, w/c=0.4+5%SRA)................................................... 28
Table 2-4 Coarse aggregate gradation............................................................... 44
Table 2-5 Fine aggregate gradation ................................................................... 44
Table 2-6 Constituents for 0.047 m3 (1.65 ft3) batch ......................................... 45
Table 2-7 Wash-out test results (mass based) ................................................... 49
Table 2-8 Wash-out test results (volume based) ................................................ 49
Table 2-9 Statistical parameters for aggregate content ...................................... 51
Table 3-1 Baseline model inputs used for simulations........................................ 64
Table 3-2 Baseline log-logistic simulation parameters considering variability
due to: a) shrinkage only, b) shrinkage, tensile strength and elastic
modulus assuming they are uncorrelated, and c) shrinkage, tensile
strength and elastic modulus assuming they are correlated .................... 71
Table 4-1 Simulations used in the preliminary comparison of the direct
computation and Monte Carlo simulation (7 days of drying) .................... 83
Table 4-2 Modeling inputs .................................................................................. 85
Table 6-1 Sieve analysis of fine aggregate....................................................... 115
Table 6-2 Mixtures performed in the experimental process (for all mixtures
w/c=0.40) ............................................................................................... 117
Table 7-1 ANOVA model .................................................................................. 132
Table 7-2 Mean values and standard deviations for each test ......................... 138
Table 7-3 95% confidence intervals.................................................................. 138
Table 7-4 One-way ANOVA results sum of squares...................................... 140

Table

Page

Table 7-5 One-way ANOVA results R-square ............................................... 140


Table 7-6 One-way ANOVA results Type I sum of squares........................... 140
Table 7-7 One-way ANOVA results Type III sum of squares......................... 140
Table 7-8 Bonferroni confidence intervals for strain at the time of cracking ..... 141
Table 7-9 Tukey grouping................................................................................. 142
Table 7-10 Test Comparison using CLDIFF command .................................... 142
Table 7-11 Results of contrast statement ......................................................... 143
Table 8-1: Change in the shrinkage parameters in SRA and LWA concretes
in comparison with plain concrete.......................................................... 166

xi

LIST OF FIGURES

Figure

Page

Figure 1.1 Concept of sustainability...................................................................... 3


Figure 2.1 View of the cylindrical paste specimen .............................................. 11
Figure 2.2 a) Reusable plastic molds for paste samples and b) Modified
comparator to measure the length change .............................................. 12
Figure 2.3 Test specimens in VENA environmental chamber ............................ 13
Figure 2.4 Shrinkage and rewetting of plain cement paste (w/c = 0.30) ............. 16
Figure 2.5 Weight change of plain cement paste (w/c = 0.30)............................ 16
Figure 2.6 Shrinkage and rewetting of plain cement paste
(w/c = 0.30+5%SRA) ............................................................................... 17
Figure 2.7 Weight change of plain cement paste (w/c = 0.30+5%SRA) ............. 17
Figure 2.8 Variability in shrinkage measurements of plain cement paste
(w/c = 0.30).............................................................................................. 19
Figure 2.9 Variability in weight measurements of plain cement paste
(w/c = 0.30).............................................................................................. 20
Figure 2.10 Variability in shrinkage measurements of plain cement paste
(w/c = 0.30+5%SRA) ............................................................................... 20
Figure 2.11 Variability in weight measurements of plain cement paste
(w/c = 0.30+5%SRA) ............................................................................... 21
Figure 2.12 Schematic illustration of drying in a porous material a) saturated
microstructure b) larger pores dry out first while holding an adsorbed
layer of moisture (Rajabipour 2007)......................................................... 22
Figure 2.13 Weight change vs. shrinkage for different cement pastes
a) w/c=0.30 at 30% RH b) w/c=0.30+5%SRA at 30% RH
c) w/c=0.30 at 50% RH d) w/c=0.30+5%SRA at 50% RH........................ 23
Figure 2.14 Shrinkage and rewetting of plain cement paste (w/c = 0.40) ........... 26
Figure 2.15 Weigh change of plain cement paste (w/c = 0.40)........................... 26

xii

Figure

Page

Figure 2.16 Shrinkage and rewetting of cement paste with SRA


(w/c = 0.40+5%SRA) ............................................................................... 27
Figure 2.17 Weigh change of cement paste with SRA
(w/c = 0.40+5%SRA) ............................................................................... 27
Figure 2.18 Initial drying shrinkage of a) plain paste with w/c = 0.4 and
b) w/c = 0.4+5%SRA, and weight change of a) plain paste with
w/c = 0.4 and b) w/c = 0.4+5%SRA (results for the first 365 days) ......... 29
Figure 2.19 Results of rewetting for different pastes a) w/c=0.40 at 50%
RH b) w/c=0.40+5%SRA at 50% RH c) w/c=0.40 at 70% RH d)
w/c=0.40+5%SRA at 70% RH ................................................................. 31
Figure 2.20 Comparison of initial (dash line) and secondary (solid line)
shrinkage for plain cement pastes (w/c=0.40) ......................................... 33
Figure 2.21 Comparison of initial (dash line) and secondary (solid line)
drying for plain cement pastes (w/c=0.40+5%SRA)................................. 34
Figure 2.22 Comparison of initial (dash line) and secondary (solid line)
shrinkage for SRA cement pastes (w/c=0.40+5%SRA) ........................... 34
Figure 2.23 Comparison of initial (dash line) and secondary (solid line)
drying for SRA cement pastes (w/c=0.40+5%SRA)................................. 35
Figure 2.24 Weight change vs. shrinkage for different pastes a) w/c=0.40 at
50%RH b) w/c=0.40+5%SRA at 50%RH c) w/c=0.40 at 70%RH d)
w/c=0.40+5%SRA at 70%RH .................................................................. 36
Figure 2.25 Comparison of shrinkage for cement paste with a) w/c=0.30
and w/c=0.30+5%SRA and b) w/c=0.40 and w/c=0.40+5%SRA and
weight change of cement paste with c) w/c=0.30
and w/c=0.30+5%SRA and d) w/c=0.40 and w/c=0.40+5%SRA ............. 39
Figure 2.26 Shrinkage reduction for cement paste after the first cycle of
drying. Difference in shrinkage of cement paste with and without
SRA a) w/c=0.30 and b) w/c=0.40 ........................................................... 40
Figure 2.27 Ratio of recoverable shrinkage for cement paste a) w/c=0.30
and b) w/c=0.40 (Arrows indicate much higher rate description in
the text).................................................................................................... 41
Figure 2.28 Histogram and probability density function describing standard
deviation observed in paste shrinkage measurements ............................ 43
Figure 2.29 Shrinkage measurement of a concrete sample ............................... 47
Figure 2.30 View of the end cap: outer side (left) and inner, sand-treated
side (right)................................................................................................ 47

xiii

Figure

Page

Figure 2.31 a) Drying shrinkage and b) Weight change of concrete cylinders


150300 mm (w/c = 0.40) ........................................................................ 49
Figure 2.32 Sieve analysis results for all four specimens ................................... 50
Figure 3.1 Time-dependent residual stress and tensile strength development
in concrete: a) deterministic and b) stochastic approach ......................... 54
Figure 3.2 Stress development in restrained shrinkage specimens
(Weiss 1997)............................................................................................ 55
Figure 3.3 Development of time-dependent cracking potential for different
shrinkage values a) Drying initiated at 1 day and b) Drying initiated
at 7 days .................................................................................................. 56
Figure 3.4 a) The shrinkage of neat cement pastes at various relative
humidities with a water-to-cement ratio of 0.50 and varying amounts
of SRA (Tetraguard) and b) the corresponding shrinkage coefficient
for pastes with a water-to-cement ratio (w/c) of 0.30 and 0.50
(Pease 2005) ........................................................................................... 62
Figure 3.5 a) The relative shrinkage of mortars containing varying amounts
of SRA (Tetraguard) and 3 day sealed curing with a) water-to-cement
ratio of 0.50 and b) water-to-cement ratio of 0.30 (Pease 2005).............. 63
Figure 3.6 a) Results of simulations using the data in Table 3-1 to illustrate
typical results: a) autogenous and total shrinkage, b) elastic and
residual stresses, c) residual stress and tensile strength to illustrate
the deterministic age of cracking, and d) the influence of variation in
the ultimate shrinkage coefficient on the predicted time of cracking ........ 65
Figure 3.7 a) Results from simulations: a) assuming only variability in
shrinkage, illustrating the cumulative age of cracking distribution and
the age of cracking from the deterministic equations and b) 5%, 50%,
and 95% probability of cracking limits for specimens with different
shrinkage coefficients .............................................................................. 67
Figure 3.8 a) Cracking probability density for different ages of the specimen
and b) cumulative distribution of cracking probability for different ages
of the specimen ....................................................................................... 69
Figure 3.9 A comparison of the model simulation predictions for the age of
the specimen at the time of cracking assuming a) uncorrelated
variability in material properties and b) correlated variability in
material properties ................................................................................... 70

xiv

Figure

Page

Figure 3.10 a) A comparison of the influence of the shrinkage coefficient on


the time of cracking a) varying shrinkage only; varying shrinkage,
elastic modulus, and tensile strength (uncorrelated variability); and
varying shrinkage, elastic modulus, and tensile strength (correlated
variability) and b) An illustration of the influence of a delay in the time
when drying is initiated............................................................................. 72
Figure 3.11 Development of time-dependent stress and tensile strength for
a) time of drying equal to 7 days and b) corresponding probability of
cracking ................................................................................................... 73
Figure 4.1 Load (Q) and resistance (R) as the normally distributed variables .... 78
Figure 4.2 The reliability index ......................................................................... 79
Figure 4.3 a) Probability of cracking for concrete with different coefficients of
variation and magnitude of shrinkage and b) reliability index values
for those cases ........................................................................................ 82
Figure 4.4 Comparison of direct computation of COV for multiple sources of
variability and Monte Carlo simulation ..................................................... 83
Figure 4.5 Time-dependant stress and strength development for
N = 3000 ............................................................................................ 86
Figure 4.6 a) Density function and b) Cumulative function plot for
shrinkage coefficient N = 3000............................................................ 87
Figure 4.7 Time-dependant stress and strength development for
N = 3500 ............................................................................................ 88
Figure 4.8 a) Density function and b) Cumulative function plot for
shrinkage coefficient N = 3500............................................................ 88
Figure 4.9 Probability of cracking for coefficient of variation COV=0.0866
with respect to a) Shrinkage coefficient and b) Cracking potential........... 89
Figure 4.10 The relationship between shrinkage coefficient for paste with
a w/c of 0.50 and surface tension of the pore solution (Pease 2005,
Pease et al. 2005).................................................................................... 91
Figure 4.11 Surface tension measurements for deionized water-SRA
mixtures for Tetraguard, Eclipse Plus and Eclipse Floor (Pease 2005,
Pease et al. 2005).................................................................................... 91
Figure 4.12 Steps in the shrinkage based design approach............................... 93
Figure 4.13 The results of simulations showing how addition of SRA can
decrease shrinkage of concrete and reduce the probability
of cracking ............................................................................................... 94

xv

Figure

Page

Figure 5.1 Ring geometry: a) standard ring, b) dual ring, and c) eccentric
ring geometry........................................................................................... 98
Figure 5.2 A comparison of equations 5-1,5-2 and 5-3: a) varying concrete
wall thickness, and b) varying steel wall thickness ................................ 100
Figure 5.3 a) The elastic stress and residual stress that develops in a
concrete ring and b) an illustration of results from restrained ring
with concrete containing fibers (Hossain and Weiss 2004).................... 102
Figure 5.4 a) Influence of moisture gradients on residual stress distribution
and b) the effect of eccentricity on maximum stress development........ 105
Figure 6.1 Geometry of the ring test (description of the symbols in the
text)........................................................................................................ 111
Figure 6.2 Detail of the strain gage: a) Gage being placed and b) The
completed gage ..................................................................................... 112
Figure 6.3 Photo of the rings setup in a chamber and data acquisition
(Note: doors closed during the test)....................................................... 114
Figure 6.4 a) Average strain readings and temperature change and b)
Temperature change and average strain readings ................................ 118
Figure 6.5 Comparison of the strain development as a function of time for
the plain mixture and the mixture with SRA (strain results for each
cast ring) ................................................................................................ 119
Figure 6.6 a) Variability between different tests performed on the same
plain mixture (only the average strain values for each mixture
presented) and b) The influence of humidity condition the samples
were stored in on strain development and the time of cracking ............. 120
Figure 6.7 Standard deviation of the measured strains (plain mortar
mixture B) .............................................................................................. 122
Figure 6.8 Cumulative plot of cracking events for the plain mortar mixture ...... 123
Figure 6.9 Results of simulations: a) Probability of cracking described with
cumulative lognormal curve for plain mortar and mortar with SRA
and b) Comparison of experimental and simulation results ................... 127
Figure 6.10 a) Relationship between shrinkage coefficient and probability of
cracking b) Cracking probability in time for mixtures with different
shrinkage coefficients (COV=12%, DOR=100%)................................... 128
Figure 7.1 Six molds prepared for the ring test................................................. 131
Figure 7.2 Illustration of significance level and confidence level....................... 133
Figure 7.3 Strains at the time of cracking in each test ...................................... 134

xvi

Figure

Page

Figure 7.4 Average strains at the time of cracking in each test ........................ 134
Figure 7.5 A quantile-quantile plot .................................................................... 136
Figure 7.6 Plot of residuals versus sequence number...................................... 137
Figure 7.7 Plot of residuals versus predicted value of strain ............................ 137
Figure 7.8 95% confidence intervals................................................................. 139
Figure 8.1 Comparison of sealed specimens behavior a) RH
measurements, b) Free shrinkage measurements (zeroed at the
time of set) c) Restrained shrinkage measurements.............................. 151
Figure 8.2 Comparison of unsealed specimens behavior (at 50% RH)
a) Weight loss, b) Free shrinkage measurements (zeroed at the
time of set) c) Restrained shrinkage measurements.............................. 153
Figure 8.3 Illustration of drying mechanisms in sealed and unsealed
systems: (a) Sealed only internal drying, (b) Only external drying,
(c) Unsealed internal plus external drying........................................... 155
Figure 8.4 Illustration of drying mechanisms in sealed and unsealed
systems: (a) Sealed only internal drying, (b) Only external
drying, (c) Unsealed internal plus external drying ............................... 158
Figure 8.5 a) Surface tension as a function of SRA concentration
(Pease 2005) b) Desorption characteristic of lightweight aggregate
(Henkensiefken et al. 2008) ................................................................... 160

xvii

ABSTRACT

Radlinska, Aleksandra. Ph.D., Purdue University, August, 2008. Reliability-based


analysis of early-age cracking in concrete. Major Professor: Jason Weiss.

With recent concern regarding the environmental impact of the construction and
urban development, there has been an increased emphasis on understanding
how the concrete industry can become more sustainable. Sustainability relates
to the application of energy efficient materials with low impact on environment
and ensured durability.

By improving the long-term durability of concrete

elements, the life of the infrastructure can be extended, saving resources and
environment.

One problem that leads to premature deterioration in concrete

structures is the development of cracks.

As a result, there is an interest in

developing procedures to produce crack free concrete elements. This research


describes how experiments and computer simulations can be used to relate
fundamental material properties and variability to the cracking performance of
cement and concrete materials.

When thermal, hygral or chemical volume changes in concrete are restrained,


residual stresses arise. If the residual stresses exceed the tensile strength of
concrete, cracking occurs. Previous research has focused on the development
of test methods and computer models to predict cracking in concrete. While
these models are a great step forward, they are generally deterministic and do
not consider inherent variability in material properties, construction processes,
and environmental conditions.

As a result, these models do not accurately

capture the true risk of cracking in concrete elements.

xviii

In this research, Monte Carlo method and Load and Resistance Factor Design
(LRFD) approach have been applied to incorporate different sources of variability
in investigating the probability of cracking in restrained concrete members.
Simulations are performed to determine the extent of free shrinkage reduction
that is required to minimize the probability of cracking to an acceptable level. An
approach is presented that allows engineers to select and incorporate the
probability of cracking during the material design process. With this information,
concrete can be designed using new materials, like shrinkage reducing
admixtures (SRA) or by internal curing using for example lightweight aggregates
(LWA), to meet the specified shrinkage performance.

CHAPTER 1. INTRODUCTION

1.1. Research Motivation


Early-age cracking occurs all too often in concrete structures. When thermal,
hygral or chemical volume changes in concrete are restrained, residual stresses
arise.

If the residual stresses exceed the tensile strength of the material,

cracking occurs. Over the recent years substantial research efforts have been
devoted to developing new materials and construction procedures to minimize
the potential for cracking. In addition, several predictive models have been
developed to better estimate the combination of conditions that can cause
cracking to occur.

The majority of these predictive models rely on the

comparison of time-dependent residual stress development and the timedependent strength development (or fracture resistance) since the time of
cracking can be expressed as the age at which the residual stresses exceed the
materials tensile strength. However, inherent variability exists in the factors that
contribute to both the stress and strength development. If this variability is not
taken into consideration, the results obtained in the deterministic calculations can
be significantly different from those encountered in the actual structure.

This research describes how computer simulations can be used to relate


fundamental material properties and both material and construction variability to
the cracking performance of cement and concrete materials. Two methods have
been used to quantify variability in material properties: Monte Carlo simulations
and the Load and Resistance Factor Design (LRFD). Results determine the
extent of free shrinkage reduction that is required to minimize the probability of
cracking to an acceptable level. An approach is presented that allows engineers

to select a level of an acceptable probability of cracking during the material


design process.

With this information, concrete can be designed using new

materials, like shrinkage reducing admixtures (SRA), or by internal curing, e.g.,


using saturated lightweight aggregates (LWA), and meet the specified shrinkage
performance.

These shrinkage-based prediction models and new design

procedures can improve the long-term durability of concrete.

Further, these

models can be used to assess the impact of changes to cement binder


compositions.

This has applications for using new supplementary cementing

materials and fillers which can substantially reduce the amount of CO2 produced
per ton of cement and contribute to sustainability of concrete construction.

1.2. Concrete in Sustainable Design


With recent concerns regarding the environmental impact of the construction
industry, there has been an increased emphasis placed on understanding how
the concrete industry can become more sustainable. Since a prevalent problem
in the concrete industry is cracking, there is an interest in making crack free
concrete elements.

Sustainability requires the application of energy efficient

materials with low impact on environment and ensured durability. By improving


the long-term durability of concrete elements, the life of the infrastructure can be
extended, saving resources, and environment.

The concept of sustainable design has been schematically presented in Figure


1.1.

It should be noted here, that the term environmentally friendly and

sustainable should not be used interchangeably, as sustainability is much


broader concept and aims at balancing and optimizing all three aspects:
economical, environmental, and societal impact, i.e., Triple Bottom Line (ACI
Strategic Development Council, 2007).

Figure 1.1 Concept of sustainability

The economical impact relates to the initial and long-term cost (includes repairs
and maintenance), durability of concrete structures, safety of the buildings, and
possibility for recycling and reuse of materials. The impact on the environment
translates among others to efficient energy usage, alternative fuel use in cement
manufacturing process, use of supplementary cementitious materials, and
management of the waste products. The impact on society can be expressed
through safety and durability of concrete structures, indoor air quality, noise
reduction and thermal comfort.

1.3. Objectives
The primary goal of this research is to develop a modeling approach that allows
the probability of concrete cracking to be assessed taking into account variability
in the material properties. It should be noted that while cracking can be initiated
due to several causes including structural (e.g., improper design, early age
loading beyond capacity, etc.) and non-structural (e.g., volume instability,

shrinkage, thermal contraction, etc.), the focus of this research is on nonstructural sources of cracking, especially drying and autogenous shrinkage. The
modeling approach developed in this work has a particular value for the concrete
industry as it allows cracking to be predicted based on the amount of shrinkage
expected to develop in the system. If the probability of cracking is higher than the
acceptable level, the proposed approach provides a tool that can be used to
redesign concrete mixture proportioning and lower shrinkage to a sufficient level
where cracking potential can be minimized.

The specific objectives of this work were to:


1. Analyze the different sources of variability and quantify the effect of variability
on stress and strength development.
2. Develop a stochastic modeling approach that allows calculating the
probability of shrinkage cracking in concrete elements.
3. Assess the variability and repeatability of the restrained ting test.
4. Examine different shrinkage mitigation techniques and their effectiveness in
minimizing shrinkage cracking.

The work starts with the description of an experimental program conducted on


cement paste and concrete samples to quantify the magnitude and variability in
shrinkage measurements of cementitious systems with different water to cement
ratio. Next, information obtained in the experimental work is used as an input for
material properties in a computer model.

The modeling process involves

numerical simulations with application of the Monte Carlo method and the Load
and Resistance Factor Design approach.

In the following chapters of this

dissertation, the restrained ring test is analyzed as a method to experimentally


determine the time of cracking in concrete.

The accuracy of the test and

variability within one batch and between different batches are verified and
experimental results are compared with the outcome of computer simulations. In
the last chapter of this work, the importance of boundary and drying conditions

are discussed and the differences in the mechanism of sealed and unsealed
drying are analyzed. The detailed content of each chapter is presented in the
next section.

1.4. Organization
This thesis is primarily focused on the assessment of the probability of shrinkage
cracking in concrete members. It has 9 chapters. Chapters 2, 3, 4, 5, and 8 are
written based on the previous publications. The content of the chapters is as
follows:

Chapter 1 provides research motivation and introduces the reader to the concept
of sustainability. This chapter presents the content of the dissertation and briefly
discuses topics covered in other chapters.
Chapter 2 describes the experimental program conducted on cement paste and
concrete to evaluate the magnitude and variability in shrinkage measurements.
The chapter presents detailed description of experimental approach, including
a series of drying and rewetting length change measurements on cement paste
specimens. The results are provided and the findings are discussed in terms of
parameters affecting the magnitude, rate, and variability of shrinkage, including
the w/c, the ambient relative humidity (RH), and the use of shrinkage reducing
admixtures (SRA). The impact of SRA is quantified at various ambient relative
humidities and the interconnection between the surface tension, RH, moisture
content, and shrinkage are discussed. In addition, shrinkage measurements are
performed on concrete specimens and the potential impact of variability in the
specimens aggregate content is analyzed.
Chapter 3 provides background on Monte Carlo simulation technique and shows
how variability in the system can be assessed and incorporated into the
computations. This chapter presents in detail the modeling approach used to

assess the probability of cracking in restrained concrete elements. The timedependent equations used in the model are presented and strength and residual
stress development discussed. Typical results of simulations are shown and the
effect of material variability on the time of cracking demonstrated.
Chapter 4 introduces the concept of Load and Resistance Factor Design (LRFD)
approach and presents how LRFD can be effectively used to assess the
probability of cracking in concrete, reducing the amount of time required for
computation as compared to Monte Carlo simulations. It provides background on
LRFD and presents how load and materials resistance can be treated as random
variables described by probability distribution. The results are compared with
Monte Carlo method and favorable agreement is observed.

Additionally, an

example is included to present how LRFD can be used to develop a new mixture
design procedure, in which the amount of shrinkage reducing admixture needed
to lower the shrinkage to an acceptable level can be determined.
Chapter 5 provides a review of recent developments in the restrained ring test.
The chapter presents how the ring test became a qualitative technique to assess
stress development and cracking potential in concrete element. It also discusses
the influence of fiber reinforcement, specimen geometry and bond between the
concrete and steel, moisture gradients, and degree of restraint. This chapter
also covers recent developments in the analysis of the restrained ring test,
including a dual ring capable of capturing shrinkage and expansion, and the
effect of eccentricity i.e., variability in the concrete wall thickness.
Chapter 6 comments on the interpretation of results from the restrained ring test.
Six rings were cast at the same time from the same batch and stored in carefully
controlled environmental conditions. The effect of shrinkage reducing admixtures
and environmental conditions on strain development are evaluated.

The

accuracy of the test is determined and variability between the test results
discussed. An approach to assess probability of cracking is presented and the

results of simulations relating probability of cracking with shrinkage and time are
described.
Chapter 7 examines the results from the restrained ring test using analysis of
variance (ANOVA). Six rings were cast from the same batch of mortar and this
procedure was repeated four times. The time of cracking for each ring was
recorded and the obtained data set was subject to ANOVA analysis. It was
verified that data can be assumed to have a normal distribution and that ANOVA
is an applicable modeling tool to evaluate statistical difference between the times
of cracking in the four tests. The results obtained from single ANOVA analysis
gave inconsistent results: one test concluded that there is enough statistical
evidence to believe that the mean times of cracking are different while the other
test did not detect statistically significant difference. The random-effect model on
the other hand showed that the variance within the test is relatively small and the
ring test can be considered a test with high repeatability and accuracy.
Chapter 8 discusses the importance of boundary and drying conditions in
cementitious systems and underlines the difference between sealed (internal)
drying and unsealed (external and internal) drying where samples are exposed to
50% relative humidity. Three systems are analyzed: plain mortar, mortar where
5% of water was replaced with a shrinkage reducing admixture (SRA), and
mortar where approximately 40% of the sand was replaced by saturated
lightweight aggregate (LWA). The chapter provides experimental measurements
of drying shrinkage, weight change, autogenous shrinkage, restrained shrinkage,
and internal relative humidity.

Next, it discusses the implication of boundary

conditions on shrinkage mitigation strategies.

It is presented that SRA reduces

shrinkage by reducing capillary pressure through surface tension reduction while


the LWA reduces shrinkage through providing additional water reservoirs and
increasing the size of the menisci radius.
Chapter 9 presents the summary and conclusions.

CHAPTER 2. QUANTIFYING THE VARIABILITY IN THE SHRINKAGE


MEASUREMENTS IN CEMENT PASTE AND CONCRETE EXPOSED TO
DRYING AND REWETTING

2.1. Introduction
It is known that concrete shrinks in response to drying, chemical reaction, or
temperature reduction.

Shrinkage that occurs in a sealed system is due to

internal drying (i.e., self-desiccation) and is referred to as autogenous shrinkage


(Jensen and Hansen 1995). Shrinkage that is caused by the loss of water to the
surrounding environment (i.e., external drying) is considered as drying shrinkage.
It should be noted, however, that the total shrinkage in an unsealed system
consists of both: drying shrinkage and autogenous shrinkage.

If the specimens subject to drying are subsequently immersed in water (or pore
solution with similar composition to that encountered in concrete), some
shrinkage can be recovered (Helmuth and Turk 1967). The portion of shrinkage
that can be recovered is referred to as recoverable shrinkage. The part of strain
that is not recovered is permanent and is referred to as irreversible shrinkage.
Studies on cementitious system subject to drying and rewetting have been
conducted for decades.

For example Feldman (1972) described hysteresis

between the adsorption and desorption shrinkage curves for d-dried samples,
and concluded that the solid exposed to more than 50% RH has potential for
irreversible shrinkage. In his following studies, Feldman (1974) demonstrated
that major changes occur to the layered structure of dried hydrated portland
cement and that new alignment of layers occurs as a result of water penetration.
This was claimed to cause aggregation of C-S-H layers leading to an increase in
the solid volume and decrease in the surface area.

Parott et al. (1980)

9
investigated differences in shrinkage behavior of pastes with different water/alite
(C3S) ratios. Hwang and Young (1984) examined the influence of specimen
thickness on drying shrinkage of cement pastes, and concluded that the
thickness of specimens in the range of 1 to 3 mm did not affect equilibrium
shrinkage, while it did affect drying and rewetting rates.

In recent studies

Juenger and Jennings (2002) quantitatively compared shrinkage values with


microstructural properties of cement pastes and showed that high internal
surface areas and pore volumes correspond with high values of total and
irreversible drying shrinkage.

In the same work the authors showed that

reversible drying shrinkage is independent of surface area and pore volume (as
measured by nitrogen adsorption). Jennings et al. (2008) presented that
subsequent cycles of drying and rewetting do not generally result in increased
portion of irreversible shrinkage.

Despite a significant amount of research devoted to shrinkage, shrinkage


cracking remains a prevalent problem in concrete structures. Cracking can be
predicted, as it will be discussed in Chapter 4 and Chapter 5, but in order for the
models to adequately predict future concrete behavior, accurate information
about shrinkage and its variability needs to be obtained. This chapter provides
information about an experimental program conducted to investigate the
magnitude of shrinkage and the influence of material variability on shrinkage of
concrete and cement paste. The main scope of this work was to:
1) Analyze shrinkage and shrinkage recovery of plain cement paste
(w/c=0.30 and w/c=0.40) and cement paste containing shrinkage reducing
admixture (w/c=0.30+5%SRA and w/c=0.40+5%SRA).
2) Determine the amount of variability in paste shrinkage measurements.
3) Analyze variability in shrinkage and variability in aggregate content in
concrete specimens (w/c=0.40).

10
2.2. Experimental Approach in Paste Shrinkage Studies

2.2.1. Materials
All laboratory specimens were prepared using Type I portland cement that
complied with ASTM C 150-04. The cement was produced at the Lonestar plant
in Greencastle, Indiana. It should be noted that all mixtures were prepared using
the cement from the same production batch. This cement had a blaine fineness
of 3670 cm2/g and an estimated Bogue composition as follows: 61% C3S, 13%
C2S, 10% C3A, loss on ignition 1.75%, insoluble residue 0.5%, and Na2O
equivalent of 0.61%.

A mid-range water reducing admixture (MRWRA) WRDA-

82 was added at the rate of 0.5% by weight of cementitious material. The


MRWRA (manufactured by W.R. Grace) complied with ASTM C 494-05 and had
specific gravity of 1.15.

2.2.2. Mixture Proportions, Specimen Preparation, and Specimen Geometry


Free shrinkage of cement paste was assessed using a series of cylindrical paste
specimens (Figure 2.1) with a diameter of 12.5 mm (1/2 in.) and the length of 100
mm (4 in.). The gage length was approximately 87 mm (3.4 in.). The small
cross-sectional geometry was chosen to minimize temperature and moisture
gradients.

Four

series

of

cement

pastes

were

prepared:

w/c=0.30,

w/c=0.30+5%SRA, w/c=0.40, and w/c=0.40+5%SRA. These were plain cement


pastes with water-to-cement ratio of 0.30 and 0.40 and similar cement pastes
with 5% of the water replaced with a shrinkage reducing admixture (Tetraguard
AS20), so that constant liquid-to-cement ratio was maintained. A water reducer
was added to the w/c=0.30 specimens at the rate of 0.5% by weight of cement.
The paste was mixed using a standard Hobart mixer in accordance with ASTM C
305-99. To minimize the influence of entrapped air, the specimens were
1

Any mention of commercial products within the dissertation is for information only; it does not
imply recommendation or endorsement by the thesis author or Purdue University

11
prepared using a de-aired procedure. The water was de-aired prior to mixing by
boiling de-ionized water (to remove the dissolved air) and then cooling it down to
room temperature prior mixing. Details on the de-aired procedure are provided in
Sant et al. (2006).

87 mm (3.4 in.) Gage Length

Figure 2.1 View of the cylindrical paste specimen

Plastic, reusable molds were used (Pease 2005) to prepare the paste samples
(Figure 2.2a).

Paste was placed in the molds under continuous vibration to

minimize air entrapment. The specimens were cast with embedded stainless
steel screws (Figure 2.1) with ground ends to ensure a sharp point and to obtain
repeatable length measurements using a modified comparator (Figure 2.2b).
After 24 hours in sealed condition and 231C, specimens were demolded and
initial weight and length measurements were taken.

Specimens were then

wrapped in plastic (sealed) and immediately placed for 90 days in moist curing
room at 231C.

12

Detachable
Base of the

sample molds

mold

(a)

(b)
Top of the mold

Figure 2.2 a) Reusable plastic molds for paste samples and b) modified
comparator to measure the length change

2.2.3. Experimental Procedures


Following the 90 days of conditioning, specimens were subjected to drying in five
different environmental conditions.

A series of samples were sealed with

aluminum tape and placed in plastic bags. The remaining samples were stored
in a computer controlled Vena VC-10 Environmental Chambers (Figure 2.3) at
30%, 50%, 70%, and 87% (1%) RH at 23C (1C) for pastes with w/c=0.30 and
w/c=0.30+5%SRA and at 50%, 70%, 87%, and 95% (1%) RH at 23C (1C) for
pastes with w/c=0.40 and w/c=0.40+5%SRA. Chambers were nitrogen purged to
reduce the potential for carbonation shrinkage. The 95% RH was maintained
using potassium nitrate solution, as that high RH was above the upper limit of the
chamber.

While being tested, specimens were covered with plastic to minimize

drying and the influence of changing environment. After samples were one year
old and had reached equilibrium they were submerged in a synthetic pore
solution for rewetting (0.35 molar KOH + 0.05 molar NaOH, saturated with
Ca(OH)2) (Rajabipour 2008) and kept in that environment for an additional year.
When equilibrium was reached after a re-wetting period, samples were subject to

13
a second drying cycle (w/c=0.4 and w/c=0.40+5%SRA series only). To provide
information about variability in shrinkage measurements, at least 8 samples were
prepared per each condition. Table 2-1 provides a detailed list of the shrinkage
specimens used in this study.

Figure 2.3 Test specimens in VENA environmental chamber

14
Table 2-1 Paste samples investigated in the course of study
Shrinkage Study (Time in Days)
st

nd

Mixture

RH

Sealed
(wrapped
in plastic)

1
Drying

Rewetting

2
Drying

Age at
Completion

Number of
Samples

w/c=0.30

30%
50%
70%
87%
Sealed
30%
50%
70%
87%
Sealed
50%
70%
87%
95%
Sealed
50%
70%
87%
95%
Sealed

90
90
90
90
90
90
90
90
90
90
90
90
90
90
90
90
90
90
90
90

320
320
320
320
320
300
300
300
300
300
275
275
275
275
90
275
275
275
275
90

120
120
120
120
n/a
120
120
120
120
n/a
395
395
395
395
n/a
395
395
395
395
n/a

120
120
120
n/a
n/a
120
120
120
120
n/a

530
530
530
530
n/a
510
510
510
510
n/a
880
880
880
760
n/a
880
880
880
760
n/a

9
9
10
10
8
12
11
12
10
9
8
12
11
10
11
10
10
8
9
10

w/c=0.30
+5%SRA

w/c=0.40

w/c=0.40
+5%SRA

2.3. Experimental Results

2.3.1. Shrinkage and Rewetting Cycle for w/c=0.30


and w/c=0.30+5%SRA Pastes
It should be noted here that while samples were kept sealed and stored in the
moist room, they were undergoing autogenous deformation. To fully understand
volume changes in cementitious system (especially for systems with low w/c)
autogenous deformation has to be captured from the time of solidification which
usually takes place few hours after mixing (Jensen and Hansen 1995, Jensen
and Hansen 2001, Lura et al. 2001, Lura 2003, Lura et al. 2003). This volume
change can later be added to the conventional length change measurements,

15
like the method described by ASTM C 157-04 (Sant et al. 2006, Sant et al. 2007,
Sant 2007, Radlinska et al. 2008). For measurements described in this chapter,
it is assumed that the autogenous deformation was essentially completed by 90
days when samples were unsealed and as such, this work will focus primarily on
the drying component of shrinkage. Consequently, data will be normalized to the
time when drying was initiated, i.e., shrinkage is 0 at 90 days when samples
were exposed to drying.

Figure 2.4 to 2.7 present the average volume and weight change of plain cement
paste and cement paste with 5% SRA. It can be seen that for both systems (with
and without SRA), rapid change occurs when drying or rewetting is initiated. As
observed before (Pease 2005), the lower the relative humidity at which
specimens are stored, the higher the observed shrinkage and weight loss. We
will begin the analysis by comparing Figure 2.4 and Figure 2.5. It can be noticed
that for the paste stored at 50% RH and 70% RH, the weight change
corresponds to the shrinkage curves.

However, during rewetting not all the

shrinkage was recovered, even though weight gain was observed.

It is

interesting to notice that more shrinkage is recovered for the plain and SRA
paste stored at 30% than for samples stored at 50% RH. Additionally, some
swelling for these pastes is observed, as compared to initial state at 90 days.
Also, more shrinkage is recovered for paste stored at 50% RH than for the one
stored at 70% RH. Plain paste stored at 87% RH shows relatively small weight
gain while shrinkage is being observed. It should also be noticed here that testing
procedure for the sealed samples was interrupted and will not be discussed.

16
500
0% SRA

Shrinkage []

0
-500

-1000
-1500
-2000

30% RH
50% RH
70% RH
87% RH
sealed

-2500
-3000
-3500
-4000
0

50

100

150

200

250

300

350

400

450

500

550

Time [Days]

Figure 2.4 Shrinkage and rewetting of plain cement paste (w/c = 0.30)

Weight Change [%]

0% SRA
2
0
-2
-4

30% RH
50% RH
70% RH
87% RH
sealed

-6
-8

-10
0

50

100

150

200

250

300

350

400

450

500

550

Time [Days]

Figure 2.5 Weight change of plain cement paste (w/c = 0.30)

Figure 2.6 and Figure 2.7 show, respectively, the average values for volume and
weight changes of cement paste with 5% SRA. Lower shrinkage is observed as
compared to plain paste when a shrinkage reducing admixture is added. As in
the case of plain cement paste, more shrinkage is recovered for samples stored
at 30% RH. All drying components of shrinkage are recovered for samples stored
at 30% RH and 50% RH. However, only part of the shrinkage for samples stored

17
at 70% is recovered. After rewetting, additional weight gain as compared to 90
days was observed for samples stored at each RH condition.

500
5% SRA

Shrinkage []

0
-500

-1000
-1500
-2000

30% RH
50% RH
70% RH
87% RH
sealed

-2500
-3000
-3500
-4000
0

50

100

150

200

250

300

350

400

450

500

550

Time [Days]

Figure 2.6 Shrinkage and rewetting of plain cement paste (w/c = 0.30+5%SRA)

Weight Change [%]

5% SRA
2
0
-2
-4

30% RH
50% RH
70% RH
87% RH
sealed

-6
-8

-10
0

50

100

150

200

250

300

350

400

450

500

550

Time [Days]

Figure 2.7 Weight change of plain cement paste (w/c = 0.30+5%SRA)

18
2.3.2. Variability in Shrinkage and Weight Measurements (w/c=0.30 and
w/c=0.30+5%SRA)
After the shrinkage and weight data was collected, their mean values and
standard deviations along with the coefficient of variations were determined.
Data presented on the graphs (Figure 2.8 to Figure 2.11) are mean values
together with one standard deviation above and below the mean2. Data lying
further than three standard deviations from the mean were treated as outliers and
removed from the data set.

The maximum shrinkage and weight change

observed during initial drying and maximum standard deviation observed


between measurements has been summarized in Table 2-2.

Table 2-2 Mean shrinkage values and standard deviations for paste at
equilibrium (w/c=0.3, w/c=0.3+5%SRA)
Mixture

RH

Max
Shrinkage
[]

w/c=0.30

30%
50%
70%
87%
sealed
30%
50%
70%
87%
sealed

-3062
-2118
-1529
-90
-57
-1930
-1191
-1070
-170
-50

w/c=0.30+
5%SRA

Max.
SD
(Shr.)
[]
397
149
192
84
148
201
208
161
130
39

Max.
Max
Weight
SD
Change (Weight)
[%]
[%}
-7.41
0.28
-5.33
0.40
-1.07
0.12
0.78
0.07
0.38
0.02
-6.31
0.10
-4.47
0.09
-1.48
0.09
0.76
0.11
0.33
0.06

Comparing Figure 2.8 to Figure 2.10 the effect of a shrinkage reducing admixture
can be evaluated.

As already mentioned section 2.3.1, it can be seen that

reduction in the shrinkage is observed for samples stored at 30%, 50%, and 70%
RH. However the highest shrinkage reduction is observed for lower values of

The testing cycle for sealed samples was interrupted and only the initial part of data for
shrinkage and weight change is available.

19
RH: shrinkage is reduced by -1132 for paste stored at 30%RH (40%
reduction), by 927 for paste stored at 50%RH (40% reduction), and by 459
for paste stored at 70%RH (30% reduction). The higher effectiveness of SRA at
low humidities has been previously reported by other researchers (Weiss et al.
2008).

Specimens stored at higher RH (87% and sealed) show very little

shrinkage.

The results of weight change show that rapid weight loss occurs at early age and
then specimens come to equilibrium. It can be also noticed that more water is
lost for plain cement paste at w/c=0.30 than for paste with SRA. Both plain
system and paste with SRA show similar variability in the weight measurements.

500
0% SRA

Shrinkage []

0
-500

-1000
-1500
-2000

30% RH
50% RH
70% RH
87% RH
sealed

-2500
-3000
-3500
-4000
0

50

100

150

200

250

300

350

400

450

500

550

Time [Days]

Figure 2.8 Variability in shrinkage measurements of plain cement paste


(w/c = 0.30)

20
4

Weight Change [%]

0% SRA
2
0
-2
-4

30% RH
50% RH
70% RH
87% RH
sealed

-6
-8

-10
0

50

100

150

200

250

300

350

400

450

500

550

Time [Days]

Figure 2.9 Variability in weight measurements of plain cement paste


(w/c = 0.30)

500
5% SRA

Shrinkage []

0
-500

-1000
-1500
-2000

30% RH
50% RH
70% RH
87% RH
sealed

-2500
-3000
-3500
-4000
0

50

100

150

200

250

300

350

400

450

500

550

Time [Days]

Figure 2.10 Variability in shrinkage measurements of plain cement paste


(w/c = 0.30+5%SRA)

21
4

Weight Change [%]

5% SRA
2
0
-2
-4

30% RH
50% RH
70% RH
87% RH
sealed

-6
-8

-10
0

50

100

150

200

250

300

350

400

450

500

550

Time [Days]

Figure 2.11 Variability in weight measurements of plain cement paste


(w/c = 0.30+5%SRA)

2.3.3. Weight Change versus Shrinkage (w/c=0.30 and w/c=0.30+5%SRA)


In the next step of analysis, weight change was plotted against shrinkage, for
each relative condition separately (Figure 2.13). It should be remembered that
the initial shrinkage that occurred during the first 90 days when samples were
sealed was not taken into consideration. The weight loss as a function of relative
humidity provides information about the size distribution of pores (Dullien 1979)
and can be described using Kelvins law:

r=

2 Vm
ln( RH ) RgT

Eq. 2-1

where r is the maximum pore radius that remains saturated (i.e., due to capillary
condensation), is the surface tension of the pore fluid, Vm is the molar volume
of the pore fluid, Rg is the universal gas constant, T is the absolute temperature
(in K), and RH is the ambient relative humidity. Depending on the ambient RH,
large pores dry out first (except a layer of the adsorbed moisture on the pore

22
walls) while small pores remain saturated.

This has been schematically

presented in Figure 2.12 (Rajabipour 2007).

(a)

(b)

Figure 2.12 Schematic illustration of drying in a porous material a) saturated


microstructure b) larger pores dry out first while holding an adsorbed layer of
moisture (Rajabipour 2007)

The analysis of the initial drying and rewetting curves allows assessment of the
amount of shrinkage that can be recovered. When analyzing plots presented in
Figure 2.13a-j, it can be noticed that more shrinkage can be recovered for
systems with SRA. It can also be seen (Figure 2.13a) that the rewetting curve
does not follow the initial drying path. Rather there is a noticeable difference
between these two curves (vertical separation).

That difference corresponds

partially to the irreversible shrinkage, as the materials visco-elastic (i.e., timedependent) response can also contribute to the drying-rewetting hysteresis
observed.

23
1000

1000
0% SRA, 30%RH

0
-500

-1000
-1500

al
iti
In

-2000

5% SRA, 30%RH

500

Shrinkage []

Shrinkage []

500

0
-500

-1000

ng
yi
r
D

-1500
-2000

-2500
R

-3000

ing
ett
w
e

-2500
-3000
-3500

-3500
-8

-6

-4

-2

-8

-6

(a)

-2

(b)
1000

1000
0% SRA, 50%RH

500

5% SRA, 50%RH

500

Shrinkage []

Shrinkage []

-4

Weight Change [%]

Weight Change [%]

-500

0
-500

-1000

-1000

-1500

-1500

-2000

-2000

-2500

-2500
-3000

-3000

-3500

-3500

-8

-6

-4

-2

Weight Change [%]

(c)

-8

-6

-4

-2

Weight Change [%]

(d)

Figure 2.13 Weight change vs. shrinkage for different cement pastes a) w/c=0.30
at 30% RH b) w/c=0.30+5%SRA at 30% RH c) w/c=0.30 at 50% RH d)
w/c=0.30+5%SRA at 50% RH (continued)

24
1000

1000
0% SRA, 70%RH

5% SRA, 70%RH

500

Shrinkage []

Shrinkage []

500
-500

-1000

0
-500

-1000

-1500

-1500

-2000

-2000

-2500

-2500

-3000

-3000

-3500

-3500
-8

-6

-4

-2

-8

-6

Weight Change [%]

-4

(e)

(f)
1000

1000
0% SRA, 87%RH

500

5% SRA, 87%RH

500

Shrinkage []

Shrinkage []

-2

Weight Change [%]

-500

-1000

0
-500

-1000

-1500

-1500

-2000

-2000

-2500

-2500

-3000

-3000

-3500

-3500

-8

-6

-4

-2

Weight Change [%]

(g)

-8

-6

-4

-2

Weight Change [%]

(h)

Figure 2.13 Weight change vs. shrinkage for different cement pastes e) w/c=0.30
at 70% RH f) w/c=0.30+5%SRA at 70% RH g) w/c=0.30 at 87% RH h)
w/c=0.30+5%SRA at 87% RH (continued)

25
1000

1000

0% SRA, sealed

5% SRA, sealed

500

Shrinkage []

Shrinkage []

500
-500

-1000

0
-500

-1000

-1500

-1500

-2000

-2000

-2500

-2500

-3000

-3000

-3500

-3500

-8

-6

-4

-2

Weight Change [%]

(i)

-8

-6

-4

-2

Weight Change [%]

(j)

Figure 2.13 Weight change vs. shrinkage for different cement pastes i) w/c=0.30
sealed j) w/c=0.30+5%SRA sealed

2.3.4. Shrinkage and Rewetting Cycle for w/c=0.40 and


w/c=0.40+5%SRA Paste
Figure 2.14 to 2.17 show the overall shrinkage and weight change response of
w/c=0.40 and w/c=0.40+5%SRA pastes, including information about the
measurement variability. The shrinkage behavior can be separated into three
components: initial drying, rewetting and secondary drying.
components will be discussed separately in the following sections.

Each of these

26
1000
0% SRA

Shrinkage []

500
0
-500

-1000
50% RH
70% RH
87% RH
95% RH
sealed

-1500
-2000
-2500
-3000
0

100

200

300

400

500

600

700

800

900

Time [Days]

Figure 2.14 Shrinkage and rewetting of plain cement paste (w/c = 0.40)

Weight Change [%]

0% SRA

4
2
0
-2
50% RH
70% RH
87% RH
95% RH
sealed

-4
-6
-8

-10
0

100

200

300

400

500

600

700

800

900

Time [Days]

Figure 2.15 Weigh change of plain cement paste (w/c = 0.40)

27
1000

5% SRA

Shrinkage []

500
0
-500

-1000
50% RH
70% RH
87% RH
95% RH
sealed

-1500
-2000
-2500
-3000
0

100

200

300

400

500

600

700

800

900

Time [Days]

Figure 2.16 Shrinkage and rewetting of cement paste with SRA


(w/c = 0.40+5%SRA)

Weight Change [%]

5% SRA

4
2
0
-2
50% RH
70% RH
87% RH
95% RH
sealed

-4
-6
-8

-10
0

100

200

300

400

500

600

700

800

900

Time [Days]

Figure 2.17 Weigh change of cement paste with SRA


(w/c = 0.40+5%SRA)

2.3.5. Initial Shrinkage and Weight Loss of Paste (w/c=0.40 and


w/c=0.40+5%SRA)
Figure 2.18 presents initial drying shrinkage and weight change. It can be seen
that the initial rate of shrinkage is rapid and then reduces over time. Specimens
reach equilibrium at approximately 120 days. The addition of SRA decreases the

28
shrinkage. This shrinkage reduction is mostly observed at lower RH: 1066 for
paste stored at 50%RH (40% reduction), but only 36 for paste stored at
70%RH (2% reduction). Also at 87% and 95%, both the specimens with and
without SRA demonstrate similar behavior. The maximum shrinkage observed
during initial drying and the maximum standard deviation between measurements
has been summarized in Table 2-3.

The variability in the shrinkage

measurements is similar between plain and SRA paste.

Table 2-3 Mean shrinkage values and standard deviation for paste at equilibrium
(w/c=0.4, w/c=0.4+5%SRA)
Mixture

w/c=0.40

w/c=0.40+
5%SRA

RH
Max
condition Shrinkage
[]
50%
70%
87%
95%
sealed
50%
70%
87%
95%
sealed

-2562
-1464
-429
148
200
-1496
-1428
-492
118
152

Max.
SD
(Shr.)
[]
158
203
153
132
241
162
162
147
142
132

Max.
Max
SD
Weight
Change (Weight)
[%}
[%]
-8.08
0.27
-3.18
0.31
0.07
0.09
0.82
0.09
0.85
0.05
-7.92
0.21
-3.67
0.42
0.06
0.07
1.10
0.13
1.00
0.13

1000

1000

500

500

Shrinkage []

Shrinkage []

29

-500

5% SRA

-500

-1000

-1000

-1500

-1500

-2000

-2000
-2500

-2500

-3000

-3000
0

50 100 150 200 250 300 350

50 100 150 200 250 300 350

Time [Days]

Time [Days]

(b)

Weight Change [%]

Weight Change [%]

(a)

2
0
-2
-4
-6

2
0
-2
-4
-6

-8

-8

-10

-10

50 100 150 200 250 300 350

5% SRA

50 100 150 200 250 300 350

Time [Days]

Time [Days]

(c)

(d)
50% RH
70% RH
87% RH

95% RH
sealed

Figure 2.18 Initial drying shrinkage of a) plain paste with w/c=0.4 and b)
w/c=0.4+5%SRA, and weight change of c) plain paste with w/c=0.4 and d)
w/c=0.4+5%SRA (results for the first 365 days)

As shown in Table 2.3, very similar weight change has been observed for plain
paste and paste with 5% SRA. This is consistent with earlier work of Weiss et al.
(1998), Weiss et al. (1999), Weiss and Shah (2001), Weiss and Shah (2002),

30
Bentz (2005), Pease (2005), Pease et al. (2005) and Weiss et al. (2008). For
samples stored at 50% and 70% RH there is a rapid water loss at early age and
then samples reach equilibrium. Specimens stored at 87% RH do not exhibit
such a drastic change in weight. Samples stored at 95% and 98% RH, on the
other hand, show weight gain and expansion. This suggests that internal relative
humidity in these samples is lower than the condition they are stored at and
water is reabsorbed over time (Weiss et al. 2008). It should be noticed that
variability in weight measurements is lower than that for shrinkage.

2.3.6. Rewetting of Paste and Secondary Shrinkage (w/c=0.40 and


w/c=0.40+5%SRA)
The results of the second component of shrinkage studies, i.e., rewetting, has
been presented in Figure 2.19. It can be seen that for each case the drying
shrinkage was recovered by 760 days and at higher relative humidity conditions
(87% and 95%) there is also an additional expansion. However, it is important to
note that this relates to drying shrinkage only, as autogenous component was not
included in the analysis, as mentioned earlier in section 2.3.1.

Weight measurements suggest that by 14 days after resaturation started,


samples were saturated nearly to the same degree as when the measurements
were started (at 90 days).

Beyond 14 days an expansion is being noticed

(comparing to 90 days). This may correspond to penetration of water into the


interlayer space of CSH sheets causing them to separate.

31
1000
50% RH

0
-500

-1000

Recovered
Shrinkage

-1500

500

Shrinkage []

Shrinkage []

500

1000
0% SRA

5% SRA

0
-500

Recovered
Shrinkage

-1000
-1500
-2000

-2000
-2500

-2500

-3000

-3000
0 100 200 300 400 500 600 700

0 100 200 300 400 500 600 700

Time [Days]

Time [Days]

(a)

(b)

1000

1000

0% SRA

70% RH

500

0
-500
Recovered
Shrinkage

-1000
-1500

Shrinkage []

Shrinkage []

500

50% RH

5% SRA

70% RH

0
-500
Recovered
Shrinkage

-1000
-1500

-2000

-2000

-2500

-2500

-3000

-3000

0 100 200 300 400 500 600 700

0 100 200 300 400 500 600 700

Time [Days]

Time [Days]

(c)

(d)

Figure 2.19 Results of rewetting for different pastes a) w/c=0.40 at 50% RH b)


w/c=0.40+5%SRA at 50% RH c) w/c=0.40 at 70% RH d) w/c=0.40+5%SRA
at 70% RH (continued)

32
1000

1000
0% SRA

87% RH

0
-500

5% SRA

0
-500

-1000

-1000

-1500

-1500

-2000

-2000
-2500

-2500

-3000

-3000
0 100 200 300 400 500 600 700

0 100 200 300 400 500 600 700

Time [Days]

Time [Days]

(e)

(f)
1000

1000
0% SRA

95% RH

500

Shrinkage []

Shrinkage []

500

87% RH

500

Shrinkage []

Shrinkage []

500

-500

5% SRA

95% RH

0
-500

-1000

-1000

-1500

-1500

-2000

-2000
-2500

-2500

-3000

-3000
0 100 200 300 400 500 600 700

0 100 200 300 400 500 600 700

Time [Days]

Time [Days]

(g)

(h)

Figure 2.19 Results of rewetting for different pastes e) w/c=0.40 at 87%RH f)


w/c=0.40+5%SRA at 87%RH g) w/c=0.40 at 95%RH h) w/c=0.40+5%SRA
at 95%RH

At the age of 760 days, specimens were taken out of the solution and subjected
to the second cycle of drying by exposing them to the same RH conditions as
initial drying. Samples originally stored at 95% RH were not included in this part
of the experiment due to space limitation in the VENA chambers. Figure 2.20
presents the secondary shrinkage in comparison with the initial shrinkage. To
enable a direct comparison, the secondary shrinkage curves have been zeroed

33
to the time when the second drying cycle was initiated (i.e., 760 days). It can be
noticed that although secondary shrinkage for cement paste stored at 50% RH is
lower than that the initial shrinkage, the difference is negligible for cement paste
stored at 70% RH.

Additionally, paste stored at 87% RH shows secondary

shrinkage curve being higher than the initial one.

Weight measurements

presented in Figure 2.21 appear consistent with shrinkage data. The effect of
SRA on secondary shrinkage has been presented in Figure 2.22. It can be seen
that for SRA specimens, higher secondary shrinkage has been observed for
samples stored at 50% RH and 87% RH. The samples stored at 70% show
similar initial and secondary shrinkage. Weight measurements for SRA paste
show slightly higher weight change for samples with SRA (Figure 2.23).
Additionally, it can be realized that the variability in shrinkage measurements for
both, plain and SRA systems is higher for secondary shrinkage than for the
initial one.

Secondary Shrinkage, Time [Days]


700

750

800

850

900

950

1000

1000
0% SRA

Shrinkage []

500
0
-500

-1000
50% RH
70% RH
87% RH
95% RH
sealed

-1500
-2000
-2500
-3000
0

50

100

150

200

250

300

350

Initial Shrinkage, Time [Days]

Figure 2.20 Comparison of initial (dash line) and secondary (solid line)
shrinkage for plain cement pastes (w/c=0.40)

34
Secondary Drying, Time [Days]
700

750

800

850

900

950

1000

Weight Change [%]

0% SRA

4
2
0
-2
50% RH
70% RH
87% RH
95% RH
sealed

-4
-6
-8

-10
0

50

100

150

200

250

300

350

Initial Drying, Time [Days]

Figure 2.21 Comparison of initial (dash line) and secondary (solid line)
drying for plain cement pastes (w/c=0.40+5%SRA)

Secondary Shrinkage, Time [Days]


700

750

800

850

900

950

1000

1000
5% SRA

Shrinkage []

500
0
-500

-1000
50% RH
70% RH
87% RH
95% RH
sealed

-1500
-2000
-2500
-3000
0

50

100

150

200

250

300

350

Initial Shrinkage, Time [Days]

Figure 2.22 Comparison of initial (dash line) and secondary (solid line)
shrinkage for SRA cement pastes (w/c=0.40+5%SRA)

35
Secondary Drying, Time [Days]
700

750

800

850

900

950

1000

Weight Change [%]

5% SRA

4
2
0
-2
50% RH
70% RH
87% RH
95% RH
sealed

-4
-6
-8

-10
0

50

100

150

200

250

300

350

Initial Drying, Time [Days]

Figure 2.23 Comparison of initial (dash line) and secondary (solid line)
drying for SRA cement pastes (w/c=0.40+5%SRA)

2.3.7. Weight Change versus Shrinkage (w/c=0.4, w/c=0.4+5%SRA)


The amount of recoverable shrinkage can be determined from weight change
versus shrinkage plots.

Figure 2.24 presents the weight change versus

shrinkage curves for initial drying, rewetting, and secondary drying. It can be
seen that as in the case of w/c=0.30 paste, the rewetting curve does not follow
the initial drying path and some irreversible shrinkage occurs. For cement paste
stored at 50%RH it can be clearly seen that addition of SRA decreases the
amount of irreversible shrinkage. However, for cement paste stored at 70%RH
addition of SRA does not alter the shrinkage and rewetting response significantly.
Specimens stored at 87% RH and in a sealed condition do not present very
significant hysteretic properties.

36
1000

1000
0% SRA, 50%RH
D
ry
in
g
Se

Shrinkage []

co
nd

0
-500
al
D
In
iti

-1500

0
-500

-1000

ry

-1000

5% SRA, 50%RH

500

in
g

Shrinkage []

500

-1500

-2000

n
etti
Rew

-2500

-2000

-2500

-3000

-3000

-8

-6

-4

-2

-8

-6

(a)

-2

(b)

1000

1000
0% SRA, 70%RH

500

5% SRA, 70%RH

500

Shrinkage []

Shrinkage []

-4

Weight Change [%]

Weight Change [%]

-500

-1000

0
-500

-1000

-1500

-1500

-2000

-2000

-2500

-2500

-3000

-3000
-8

-6

-4

-2

Weight Change [%]

(c)

-8

-6

-4

-2

Weight Change [%]

(d)

Figure 2.24 Weight change vs. shrinkage for different pastes a) w/c=0.40 at
50%RH b) w/c=0.40+5%SRA at 50%RH c) w/c=0.40 at 70%RH d)
w/c=0.40+5%SRA at 70%RH (continued)

37
1000

1000
0% SRA, 87%RH

5% SRA, 87%RH

500

Shrinkage []

Shrinkage []

500

-500

-1000

0
-500

-1000

-1500

-1500

-2000

-2000

-2500

-2500

-3000

-3000
-8

-6

-4

-2

-8

Weight Change [%]

-6

-4

(e)

(f)
1000

1000
0% SRA, 95%RH

500

5% SRA, 95%RH

500

Shrinkage []

Shrinkage []

-2

Weight Change [%]

-500

-1000

0
-500

-1000

-1500

-1500

-2000

-2000

-2500

-2500

-3000

-3000

-8

-6

-4

-2

Weight Change [%]

(g)

-8

-6

-4

-2

Weight Change [%]

(h)

Figure 2.24 Weight change vs. shrinkage for different pastes e) w/c=0.40 at
87%RH f) w/c=0.40+5%SRA at 87%RH g) w/c=0.40 at 95%RH h)
w/c=0.40+5%SRA at 95%RH

38
2.4. Discussion of Experimental Results

2.4.1. Comparison of the Shrinkage of Cement Paste with w/c=0.30 and


w/c=0.40 (with and without SRA)
Figure 2.25 presents a summary of results for shrinkage and weight
measurements for cement pastes with w/c=0.30 and w/c=0.40. It can be noticed
by comparing Figure 2.25a and Figure 2.25b that while significant shrinkage
reduction has been noticed for the SRA cement paste with w/c=0.30 at the
relative humidity of 30%, 50%, and 70%, the effectiveness of SRA for w/c=0.40
has been observed primarily at 50% RH.

This shows that SRAs are more

effective in low water-to-cement systems. Additionally, SRAs are more effective


at lower relative humidities. The effectiveness of SRAs at lower RH is consistent
with theoretical considerations, as capillary tension can cause significant
shrinkage in plain mixtures in the relative humidity range 60%~100%, while in the
SRA mixtures curved liquid-vapor menisci cannot form below approximately 80%
RH (Weiss et al. 2007, Weiss et al. 2008). This will be discussed in greater
details in Chapter 8, section 8.5.

The weight change plots presented in Figure 2.25c and Figure 2.25d illustrate
similar behavior between plain and SRA paste, although paste with w/c=0.40 has
slightly lower mass loss at lower humidities. This was observed before by Bentz
(2001) and Pease (2005).

39
500

500

-500

Shrinakge []

Shrinakge []

w/c=0.30
w/c=0.30+5%SRA

-1000

-500

-1000

-1500

-1500

-2000

-2000

-2500

-2500

-3000

-3000

-3500

-3500

20 30 40 50 60 70 80 90 100

20 30 40 50 60 70 80 90 100

Relative Humidity [%]

Relative Humidity [%]

(a)

(b)
6

w/c=0.30
w/c=0.30+5%SRA

2
0
-2
-4
-6
-8

-10

Weight Change [%]

Weight Change [%]

6
4

w/c=0.40
w/c=0.40+5%SRA

w/c=0.40
w/c=0.40+5%SRA

2
0
-2
-4
-6
-8

-10

20 30 40 50 60 70 80 90 100

Relative Humidity [%]

(c)

20 30 40 50 60 70 80 90 100

Relative Humidity [%]

(d)

Figure 2.25 Comparison of shrinkage for cement paste with a) w/c=0.30 and
w/c=0.30+5%SRA and b) w/c=0.40 and w/c=0.40+5%SRA and weight change of
cement paste with c) w/c=0.30 and w/c=0.30+5%SRA and d) w/c=0.40 and
w/c=0.40+5%SRA

2.4.2. Effectiveness of SRA in Shrinkage Reduction of Cement Paste


In addition to the analysis of variability in shrinkage measurements, the effect of
SRA on shrinkage mitigation of cement paste was investigated. It was found that
the addition of SRA can substantially reduce shrinkage, especially at lower
relative humidities.

Figure 2.26a and Figure 2.26b present the effectiveness of

40
SRA at reducing the shrinkage during the initial drying period for cement paste
with 0.30 w/c and 0.40 w/c, respectively. The shrinkage difference on y-axis
corresponds to the difference in shrinkage after the first drying. This value was
obtained after subtracting shrinkage of paste with 5%SRA from the shrinkage of
plain cement paste (0%SRA 5%SRA). It is shown that while significant shrinkage
reduction can be observed for samples stored at 30% and 50% RH, there is not a

750
w/c=0.30

500
250
0

30% RH
50% RH
70% RH
87% RH

-250
-500
-750

-1000

Shrinkage Difference []

Shrinkage Difference []

substantial effect of SRA for samples stored at higher RH.

750
w/c=0.40

500
250
0
-250

50% RH
70% RH
87% RH
95% RH

-500
-750

-1000

-1250

-1250

-1500

-1500

50 100 150 200 250 300 350

Time [Days]

50 100 150 200 250 300 350

Time [Days]

(a)

(b)

Figure 2.26 Shrinkage reduction for cement paste after the first cycle of drying.
Difference in shrinkage of cement paste with and without SRA a) w/c=0.30 and b)
w/c=0.40

2.4.3. Effect of SRA on Recoverable Shrinkage


Analyzing hysteretic behavior after the shrinkage and rewetting cycle, the amount
of irreversible shrinkage can be determined.

It was observed that more

shrinkage can be recovered in the case of SRA mixtures and more shrinkage can
be recovered at lower RH.

To quantify the difference, the magnitude of

shrinkage after the first cycle of drying was compared with the length of samples
after the period of rewetting. This allowed the amount shrinkage that can be

41
recovered for the 0.30 w/c and 0.40 w/c cement pastes to be determined. Figure
2.27a presents the ratio of the maximum shrinkage that occurred after rewetting
and the maximum shrinkage that occurred during the first drying cycle
(Rewetting/Drying). The ratio higher than 1.0 indicates that swelling was observed
as compared to the initial state at 90 days when the samples were exposed to
drying. It can be noticed that in the case of the w/c=0.30 the addition of SRA
increased the amount of shrinkage that can be recovered (for samples stored at
30%, 50%, and 70% RH). The recovery rate for samples stored at 87% RH was
much higher, reaching the ratio of 5.2 and 2.1 for plain and SRA paste,
respectively. Figure 2.27b presents the results obtained for w/c=0.40 paste. It
can be seen that at higher water-to-cement ratio, the increase in the shrinkage
recovery due to SRA presence can be seen only at 50% RH. At higher relative
humidities either the same recovery is observed (70% RH) or recovery rate for
plain paste is higher (87%, and 95% RH). The highest recovery rate for 0.40 w/c
was noticed for samples stored at 95%, at the rate of 3.0 and 2.3 for plain and

2.0
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0

Ratio of Shrinkage Recovered

Ratio of Shrinkage Recovered

SRA paste, respectively.

0.3
0.30+5% SRA

30%

50%

70%

87%

Relative Humidity [%]

(a)

2.0
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0

0.4
0.40+5% SRA

50%

70%

87%

95%

Relative Humidity [%]

(b)

Figure 2.27 Ratio of recoverable shrinkage for cement paste a) w/c=0.30 and b)
w/c=0.40 (Arrows indicate much higher rate description in the text)

42
2.4.4. Variability in Shrinkage Measurements
It was found that even though the magnitude of shrinkage varies significantly
depending on the humidity conditions samples were stored at, the overall
standard deviation for these measurements remains within similar range.
Additionally, similar variability has been observed between mixtures with and
without SRA.

The information about standard deviation for shrinkage

measurements of plain and SRA cement paste (w/c=0.30 and w/c=0.40) was
plotted together as presented in Figure 2.28. In addition to the histogram, Figure
2.28 presents a log-normal distribution function (continuous, solid line), described
with a mean equal to 165 . This means that in the average experiment of
shrinkage measurements, a standard deviation of 165 could be expected.
Additionally, it can be seen in Figure 2.28 that among many experiments
performed, 5% of the data would be expected to have standard deviation less
than 49 and 95% of the data would be expected to have standard deviation
less than 320 .

43

X <= 49
5.0%

X <= 320
95.0%

Frequency x 10^-3

6
5
4
3
2
1
0
0

100 200 300 400 500 600 700


Standard Deviation [
]

Figure 2.28 Histogram and probability density function describing standard


deviation observed in paste shrinkage measurements

2.5. Experimental Approach in Concrete Shrinkage Studies

2.5.1. Materials
All laboratory specimens were prepared using the same cement as paste
samples described in section 2.2.1. The coarse aggregate (CA) used in all the
mixtures was a dolomite produced by Vulcan Materials (Monon, Indiana).
Coarse aggregate had a specific gravity 2.75 and absorption 1.5%. To remove
the finest particles, the aggregate was washed through a set of sieves with the
smallest sieve having a mesh of #100. For washing the aggregate, a procedure
from ASTM C 117-04 was adopted. After washing out, aggregate was oven
dried, sieved (following ASTM C 136-05), and recombined to ensure desired
gradation (Table 2-4).

44
Table 2-4 Coarse aggregate gradation
Sieve Size
[in]

Sieve Size
[mm]

Target %
Retained

Target %
Passing

1"
3/4"
1/2"
3/8"
#4
#8

25
19
12.5
9.5
4.75
2.36

100
90
49
29
7
1
Total

0
10
41
20
22
6
100

The fine aggregate (FA) used in all the mixtures was local natural sand (Vulcan
Materials Co. "Swisher Pit") with specific gravity 2.65, absorption 1.5%, and
fineness modulus 3.17. Fine aggregate was not prepared specifically for each
mixture, but was taken directly from a storage.

Sieve analysis for FA is

presented in Table 2-5.

Table 2-5 Fine aggregate gradation


Sieve Number

Sieve Size

% Passing

3/8 in.
#4
#6
#8
#16
#30
#50
#100
#200

9.50 mm
4.75 mm
3.35 mm
2.36 mm
1.18 mm
600 m
300 m
150 m
75 m

100
99
97
84
60
35
6
2
1

A mid range water reducing admixture (MRWRA) WRDA-82 that complies with
ASTM C 494-05 has been used.

The MRWRA was manufactured by W.R.

Grace and had a specific gravity of 1.15.

45
2.5.2. Experimental Methods for Shrinkage and Weight Loss Measurements
of Concrete
Cylindrical specimens 150300 mm (612 in.) have been prepared to measure
shrinkage, weight change, and additionally aggregate volume (Figure 2.29). All
specimens have been made following ASTM C 192 Standard Practice, using the
same mixture proportions. Mixtures proportions have been determined assuming
68% of aggregate by volume and setting volumes of coarse and fine aggregate
to be 38% and 30%, respectively. Batch weights for a volume of 0.047 m3 (1.65
ft3) are shown in Table 2.6.

The Over-Mortaring recommendation from

ASTM C 192-05 has been followed to compensate for mortar retained on the
walls of mixing pan. This resulted in an additional 0.002 m3 (0.07 ft3) of mortar of
the same proportions as the mixture in Table 2.4 added to each batch of
concrete (Thier 2005).

Table 2-6 Constituents for 0.047 m3 (1.65 ft3) batch


Material

Batch Weight for


0.047 m3 (1.65 ft3)
[kg (lb)]

Batch Weight for 1.0 m3


(1.0 yd3)
[kg (lb)]

cement
Water

20.83 (45.92)
9.31 (20.52)

443 (751)
198 (336)

FA

36.87 (81.29)

784 (1330)

CA
WR

48.08 (106.00)
0.078 (0.017)

1023 (1735)
1.660 (0.281)

Total:

115 (254)

2450 (4152)

Mixing was performed in accordance with all recommendations stated in ASTM C


192-05. Initially coarse and then fine aggregate were loaded into the pan mixer.
Once aggregate was placed, the mixer was started and 40% of the total water
was added into the mixer. The mixer ran for 30 seconds and then the cement
was added. After cement, subsequent 40% of water was added and all materials
were mixed together for 20 seconds. Finally, the last 20% of water mixed with

46
water reducer was added. Concrete was mixed for 3 minutes, than mixer was
stopped for 3 minutes and again concrete was mixed for 2 minutes fulfilling
ASTM C 192-05 3 minutes on, 3 minutes off, 2 minutes on procedure. The
mixture was placed in the cylindrical forms in three layers; each layer was rodded
25 times. After all three layers were placed, each cylinder was tapped 15 times
with a rubber mallet.

Specimens were sealed and allowed to cure for

approximately 23 hours at 23C under plastic, wet burlap, and additional plastic
cover. Each batch resulted in eight specimens, six of which were designed for
shrinkage measurements and the remaining two for the measurements of the
aggregate content.

After 23 hours of curing, specimens were demolded and the top and bottom 20
mm (0.79 in.) was removed using a wet-saw, resulting in a concrete specimen
with a length of 260 mm (10.25 in.). While a cylinder was being cut, remaining
samples were covered with plastic and wet burlap to prevent moisture loss. To
enable length change measurements pre-made PVC end-pieces were attached
to each cylinder using 5 minutes rapid set epoxy. The end-pieces were in the
form of round plates, each 9.8 mm (3/8 in.) thick and with diameter of 150 mm (6
in.). Each plate had a hole in the center were a gage stud was placed (Figure
2.30). The inner side of the plate was sand treated to assure better bond
between with the concrete surface (Figure 2.30). During the process of attaching
end-pieces, specimens were covered with plastic and wet burlap whenever
possible. After the epoxy set, the initial length and weight of specimen was
measured (Figure 2.29) which was approximately at the time when samples were
24 hours old. Following the initial measurements specimens were wrapped with
multiple layer of a thin plastic sheet to prevent moisture loss and put into
moisture curing room (at 23C) for 90 days.

47

Figure 2.29 Shrinkage measurement of a concrete sample

Figure 2.30 View of the end cap: outer side (left) and inner, sand-treated side
(right)

2.5.3. Experimental Methods for Concrete Aggregate Content Study


All the concrete specimens were cast at the same time and in the same manner.
Two specimens of each batch have been taken aside and washed out through
the #100 sieve to remove all the cement, following indications present in ASTM C
117-04. Total aggregate content was measured and then aggregate was sieved

48
through the set of sieves: 25, 19, 12.5, 9.5, 4.75, 2.36 mm, to check the
gradation of aggregate in a random sample.

2.5.4. Shrinkage and Weight Change of Concrete


After 90 days of moist curing, specimens were unwrapped, and their initial weight
and length measurements were taken.

Immediately after measurements

specimens were moved into 50 3% humidity room kept at the temperature of


23 2C. To ensure that the specimens were drying uniformly, they were stored
on racks that enabled free circulation of the air around the specimens. The
shrinkage and weight loss measurements were taken at the same time intervals
as for paste specimens, i.e., after 6 and 12 hours, and then after 1, 2, 3, 7, 14,
28, 56, 90, 120, 180, 240. 270, 300, and 370 days. Together with shrinkage
measurements, weight loss was measured. The results (zeroed to the time when
drying was initiated), together with information about standard deviation are
presented in Figure 2.31.

It can be seen that regardless of the age of the

specimens, the shrinkage measurements have certain amount of variability that


on average could be reported as 17 during the first 7 days after samples were
exposed to drying and 42 afterwards. Meanwhile, weight measurements have
variability 0.03% during the first 7 days and 0.05% afterwards.

2.5.5. Aggregate Content of Concrete Specimens


After casting, two concrete samples from each batch were subjected to
aggregate wash-out procedure, as explained in section 2.4.3.

The results

obtained from the wash-out test are presented in Table 2-7 and Table 2-8.

49
0.2

0.0

Shrinkage []

Weight Change [%]

100

-100
-200
-300
-400
-500

-0.2
-0.4
-0.6
-0.8
-1.0
-1.2
-1.4

-600
0

50 100 150 200 250 300 350

50 100 150 200 250 300 350

Time [Days]

Time [Days]

(a)

(b)

Figure 2.31 a) Drying shrinkage and b) weight change of concrete cylinders


150300 mm (w/c = 0.40)

Table 2-7 Wash-out test results (mass based)

Total Mass of
Aggregate and
Paste:
Total Mass of
Aggregate
Mass % of
Aggregate

MIX-1_#1
kg (lb)
13.045
(28.76)

MIX-1_#2
kg (lb)
13.122
(28.93)

MIX-2_#1
kg (lb)
13.036
(28.74)

MIX-2_#2
kg (lb)
13.122
(28.93)

9584.36
(21.13)
73.47

9534.46
(21.02)
72.66

9502.71
(20.95)
72.89

9602.50
(21.17)
73.18

Table 2-8 Wash-out test results (volume based)

Total Vol. of
Aggregate and
Paste:
Total Vol. of
Aggregate
Vol. of Aggregate

MIX 1-1
cm3 (in3)
5435.5
(331.7)

MIX 1-2
cm3 (in3)
5467.6
(333.7)

MIX 2-1
cm3 (in3)
5431.7
(331.5)

MIX 2-2
cm3 (in3)
5467.6
(333.7)

3686.3
(225.0)

3667.1
(233.8)

3654.9
(223.0)

3693.3
(225.3)

67.8%

67.1%

67.3%

67.5%

50
After weighing, the total aggregate content of a single specimen, the aggregate
was sieved. Results of the sieving analysis are presented in Figure 2.32 and
indicate that there was only minimal variability of aggregate content in a random
specimen. The volume of aggregate and the gradation of coarse aggregate are
similar (67.40.3%) and it can be concluded that variability in shrinkage is not
caused by variability in the aggregate content.

The statistical parameters

summarizing the aggregate content study have been presented in Table 2-9. It
should be noted, however, that the study was performed in carefully controlled
laboratory conditions and may not fully correspond with field measurements.
Further studies would be needed to determine the influence of aggregate
variability on the shrinkage of concrete performed in the filed.

Cumulative % Retained

80
Mix 01-1
Mix 01-2
Mix 02-1
Mix 02-2

60

40

20

0
5.0

10.0

15.0

20.0

25.0

Sieve Size [mm]


Figure 2.32 Sieve analysis results for all four specimens

51
Table 2-9 Statistical parameters for aggregate content
Total Mass
(Aggregate + Paste)
Mass of Aggregate
% of Aggregate
(Mass based)

Mean
13081.5 g

SD
47.3

COV [%]
0.4

9556.0 g
73.0%

45.7
0.4

0.5
0.5

19.7

0.4

17.6
0.3

0.5
0.5

Total Volume
5450.6 cm3
(Aggregate + Paste)
Volume of Aggregate 3675.4 cm3
% of Aggregate
67.4%
(Volume based)

2.6. Summary
This chapter presented detailed analysis of the shrinkage measurements of
cement paste and concrete. These general conclusions can be made:

In accordance with previous studies, the ultimate value of the shrinkage


depends on the relative humidity conditions that the specimens are
stored in,

As seen in previous research, the addition of SRA alters the magnitude of


the ultimate shrinkage and the shrinkage rate, however higher
effectiveness of SRA has been observed at lower humidities,

Similar variability has been observed between mixtures with and


without SRA,

The average variability in shrinkage measurements of cement paste was


found to be 165 . There is a 5% probability of obtaining a standard
deviation less than 49 and a 95% probability of a obtaining standard
deviation less than 320 ,

There is a little difference in mass loss between plain paste and paste with
SRA at w/c=0.30, however slightly lower mass loss for the cement paste
with SRA (as compared to the plain paste) has been observed for
w/c=0.40 at lower humidities,

52

The analysis of hysteretic behavior after the shrinkage and rewetting


cycle, allows the amount of irreversible shrinkage to be determined. It
was observed that more shrinkage can be recovered in SRA mixtures and
more shrinkage can be recovered at lower RH. The rate of recovered
shrinkage was found to be between 1.1 to 1.4 times greater for SRA
mixtures than for plain cement paste at 30% to 70%RH, while the rate was
as high as 1.7 to 5.2 for samples stored at 87%RH and 95%RH,

In properly performed shrinkage studies of concrete, variability that


originates from inherent variability in material properties can be expected.
In this study, the maximum standard deviation was observed to be 58 .
However, initial measurements had much lower standard deviation, being
less than 26 during the first 28 days,

Very low variability exists in the aggregate content between different


concrete samples from the same laboratory batch (0.3%).

53

CHAPTER 3. MONTE CARLO SIMULATION

3.1. Introduction
Residual stresses arising from the restraint of drying, chemical, autogenous or
thermal shrinkage can result in the development of tensile residual stresses. If
the stresses that develop are high enough, they may cause cracks to develop in
the concrete. Substantial research has been focused on the development of test
methods to assess stress development and the potential for cracking in concrete.
In addition, these test methods frequently focus on the determination of material
properties that can be used in deterministic computer programs to simulate
stress development and cracking potential. While these models are a great step
forward, variability is inherent in the material properties, the construction
processes, and the environmental conditions (i.e., temperature and relative
humidity).

This chapter presents results from an investigation aimed at

examining the effects of variability on the risk of cracking. A Monte Carlo method
is described for assessing the anticipated material variability and the results of
this approach are discussed.

It is anticipated that this type of modeling approach will be able to be used to


quantify the sensitivity of the model to variability in modeling inputs, thereby
enabling the most sensitive inputs to be determined.

This could enable

researchers to determine which inputs require the most significant scrutiny and
could signal where new and more accurate standardized testing protocols are
needed. In addition, it is also anticipated that this type of approach can be used
by contractors to assess the risk associated with the selection of construction
operations and constituent materials with greater confidence. This approach can

54
also be used by specification developers to indicate how specifications based on
anticipated construction variations can be written more effectively to minimize the
risk of early-age cracking.

3.2. Background on the Computational Method


Shrinkage cracking can be expected to occur when the stress that develops in
response to the prevention of shrinkage (residual stress) exceeds the strength of
the concrete. In deterministic approach, the time of cracking (the intersection of
stress and strain curves) is denoted simply by a single point where the stress and
strength are equal (Figure 3.1a). If variability in material properties is considered
however, a slightly different picture emerges. Figure 3.1b illustrates variability in
the materials resistance to cracking (strength) and the load (residual stress). In
that case no longer is a single time of cracking given, but rather a window where
cracking may occur. This variability in the time when cracking can be expected is

Stress
(Solid)
Strength
(Dashed)

Predicted Age of
Cracking

Specimen Age (Days)

(a)

Stress or Strength (MPa)

Stress or Strength (MPa)

shown as the shaded area in Figure 3.1b.

Stress
(Solid)
Strength
(Dashed)

Predicted Age of
Cracking

Specimen Age (Days)

(b)

Figure 3.1 Time-dependent residual stress and tensile strength development in


concrete: a) deterministic and b) stochastic approach

55
Determining whether cracks may develop in concrete is even more complicated
by the fact that both the residual stress development and strength are timedependent. Tensile strength, like other material properties, develops over time
as it is related to the degree of hydration while residual stresses develop over
time because shrinkage occurs in response to moisture loss, which is a slow,
diffusion-controlled process.

It should be noted that the residual stress that

develops in concrete that is restrained can not be computed directly using the
product of the free shrinkage and elastic modulus (i.e., Hookes Law) since the
concrete is sensitive to stress relaxation creep (Figure 3.2).

12
Calculated Stress (MPa)

Shrinkage Stress
Stress After Creep
Relaxation
8
Stress Relaxation
Due To Creep
Shrinkage
Stress

Stress In
Specimen
0
0

14

21

28

Age Of Specimen (Days)

Figure 3.2 Stress development in restrained shrinkage specimens (Weiss 1997)

The potential for shrinkage cracks to develop in concrete depends on several


time-dependent factors including material stiffness, shrinkage rate, shrinkage
magnitude, stress relaxation, and material toughness.

Time-dependent

development of material properties can be described mathematically using the


cracking potential (CR) which is the ratio of the time dependant stress and
strength of concrete, as shown by Equation 3-1 and Figure 3.3.

56

CR (t ) =

(t )

Eq. 3-1

f 't (t )

In theory, a cracking potential equal to or higher than 1 denotes the time of


cracking (intersection of the curves on Figure 3.1, but laboratory experiments
show that cracking can also be observed for lower values, i.e., beginning as low
as 0.7 (Breugel and Lokhorst 2001, Hossain and Weiss 2004, Schiel and
Rucker 2005) which can be explained by variability present in material properties

1.2

1.2

1.0

1.0

0.8

0.8

/f't

/f't

or in creep.

0.6
0.4

0.4

1 Day Drying
shr = 600

0.2

0.6

7 Days Drying
shr = 600

0.2

shr = 800

0.0

shr = 800

0.0
0

14

28

42

Time [Days]

(a)

56

70

14

28

42

56

70

Time [Days]

(b)

Figure 3.3 Development of time-dependent cracking potential for different


shrinkage values a) drying initiated at 1 day and b) drying initiated at 7 days

A model has been formulated to estimate the potential for restrained shrinkage
cracking in concrete elements (Weiss 1997, Weiss et al. 1998, Weiss 1999).
Although this model makes some basic assumptions such as uniform shrinkage
throughout the element, previous research showed a favorable comparison
between the model predictions and experimental observations (Shah et al. 1998).
Similar models have been developed to predict shrinkage cracking as well as to

57
consider the potential for thermal cracking (Schindler et al. 2004, Schlangen et
al. 2004).

This approach has been used to illustrate stress development in

higher strength concrete, to simulate the behavior of structures with imperfect


restraint, and to illustrate the role of reducing the magnitude and rate of free
shrinkage (Shah et al. 1998, Parameswaran 2004, Pease et al. 2005).

Previous work (Weiss 1997) has shown that Equation 3-2 may be used to
estimate the stress development in a restrained concrete element:

d ( )
1
1
+ (t , )
+ ( ) d
E ( ) Ec
d
0

Permit (t ) =

Eq. 3-2

where, Permit(t) is the total strain that is permitted to develop in the actual
restrained concrete (i.e., for complete restraint Permit(t) = 0 which will be assumed
throughout this work), E() is the time-dependent elastic moduli (at time

(0<<t)), EC is a reference elastic modulus (i.e., a 28 day value), (t,) is the


creep coefficient, and () is the differential shrinkage with respect to time (t),
i.e.,

Tot Shr (t ) = ( ) d

Eq. 3-3

The ability to compute the residual stress allows for comparison of residual
stresses with the tensile strength in an attempt to ascertain whether cracking
may occur in a given concrete.

While this allows complicated interactions

between various material properties to be considered in relatively simple fashion,


it does not account for effects of moisture gradients in its current form, and does
not account for mesostructural scale issues like aggregate distribution or
microcracking.

Additional work is being performed to account for moisture

gradients, aggregates, and microcracking [Weiss and Shah 2001, Moon et al.

58
2004, Moon et al. 2005]. This chapter provides results from an investigation to
demonstrate how variability would influence the modeling predictions of the time
of cracking.

A Monte Carlo approach has been used to include material

variability in the model for the purpose of quantifying the level of uncertainty in
the cracking risk assessment.

3.3. Time-Dependent Material Property Model Inputs


The first step in performing a series of simulations is to describe how the material
properties would develop over time. Rather than specifically considering a single
material, this work will focus on describing general series of equations to
describe material properties development. There will be specific parameters in
these equations that can be varied and, as such, these equations can be used to
simulate a wide range of material behaviors.

The study will begin by describing the time-dependent modulus of elasticity using
the equation described by McIntosh (1956):

E c (t ) = E

C1 (t t s )
1 + C1 (t t s )

Eq. 3-4

where E is the theoretical maximum elastic modulus that would develop at


a very late age, C1 is a material constant that describes the rate of elastic
modulus development, t is the age of the specimen, and ts is time of initial set. It
should be noted that while this type of equation can generally well fit the data
from concrete specimens, current work is investigating whether this is the most
appropriate equation for modeling very early-age behavior and it is anticipated
that an equation that is capable of describing the transition from a liquid to a solid
may be more beneficial for future model developments (Bjontegaard 1999, Weiss
and Shah 2002).

59
Second, a formulation for the time-dependent tensile strength will be considered
that, again, is based on the hyperbolic equation proposed by McIntosh:

f ten ( t ) = f ten

C2 (t t s )
1 + C2 (t t s )

Eq. 3-5

where ften- is the theoretical maximum tensile strength that would develop at
a very late age and C2 is a material constant that describes the rate of tensile
strength development. It should be noted that this formulation should remain
somewhat flexible and, as a result, the constant is intentionally selected to be
independent of the constant used in the equation for elastic modulus. This would
be consistent with the suggestion of Olken and Rostasy (1994) who proposed
that the elastic modulus develops at a much more rapid rate than the tensile
strength.

Recent work by Barde (2004) also demonstrated that different

mechanical properties may have different rate constants due to the fact that in
higher strength materials the tensile strength behavior may need to be modified
to account for the transition in property development that occurs when cracks
that propagated through the matrix and interfaces at early ages begin to
propagate through the aggregate as well.

An expression should be included to account for the shrinkage that would


develop over time. An equation of the form developed by Bazant (1989) is an
attractive option (Equation 3-6):

t tdry
TSH

Shr Dry (t ) = SH 1 RH AMB 3 tanh

Eq. 3-6

where SH- is a parameter that describes an ultimate material shrinkage, RHAMB


is a parameter that describes the relative humidity of the environment, tdry is time
of initial drying and TSH is a constant that accounts for the time it takes for the

60
moisture to diffuse from the specimen (including both parameters for the
materials diffusion and the size of the specimen).

This work chose to rewrite the shrinkage expression as seen in Equation 3-7 by
considering the autogenous and drying shrinkage to be additive. While Equation
3-7 provides an easy to use approach it should be noted that the additive
behavior of drying and autogenous shrinkage is not strictly true:

Shr Tot (t ) = Shr Auto (t ) + Shr Dry (t )

Eq. 3-7

The autogenous (Equation 3-8) and drying (Equation 3-9) shrinkage terms can
be approximated initially using the forms proposed by Bazant (1989) after
introducing an internal humidity term (RHINT) and two constants to account for the
time-dependent nature of the shrinkage (C3 and C4):
t t set
C3

3
Shr Auto (t ) = SH (1 RH INT
)tanh

t t dry

C4

3
3
RH AMB
Shr Dry (t ) = SH (RH INT
)tanh

Eq. 3-8

Eq. 3-9

In this work the constants C3 and C4 were selected based on comparison with
sealed and drying shrinkage specimens. However these constants as well as the
choice of the hyperbolic tangent function may need to be investigated more
thoroughly in subsequent work. Finally, creep will be considered in accordance
with the CEB-FIP model (Muller 1992).

It should be noted that the elastic

modulus used in the creep computations will be based on the time-dependent


equation shown in Equation 3-4.

At this point, it should be noted that the material shrinkage coefficient proposed
by Bazant (1989) can be estimated using neat paste specimens. Figure 3.4a

61
illustrates the length change that was observed in neat paste specimens that
were stored at various relative humidities in a CO2 free environment. Further
details on the neat paste specimen preparation and measurement techniques
were presented in Chapter 3 and are also provided elsewhere (Pease 2005),
however to illustrate this approach Equation 3-8 was fitted to the data from those
tests as shown in Figure 3.4a.
3
Shr = N (1 RH AMB
)

Eq. 3-10

where N is the shrinkage coefficient that can be thought of as a coefficient of


hygral length change. Pickett (1956) and LHermite (1960) illustrated that the
shrinkage of the paste can be linked to that of mortar and concrete using an
equation of the form:
3
Shr = N (1 VF )n (1 RH AMB
)

Eq. 3-11

where VF is the volume fraction of the aggregate and n is a coefficient that


describes the stiffness of the aggregate and paste which is typically between 1.2
and 1.7 for normal strength concretes.

At this point, it is interesting to note that shrinkage-reducing admixtures (SRAs)


have been proposed to reduce the shrinkage in cementitious materials by
reducing the surface tension of the pore solution. Pease et al. (2005) measured
the surface tension of SRA-water solutions. Figure 3.4b illustrates that when the
surface tension of the SRA-water solutions used in making the pastes is plotted
against the shrinkage coefficient (N) a linear relationship is obtained (Pease et
al. 2005).

This suggests that since the surface tension of the SRA-water

solution, and corresponding surface tension of the pore fluid in a cementitious


material can be controlled through the dosage of SRA, this reduction in surface

62
tension may be able to be directly used to predict the shrinkage reduction that
may be expected for a concrete.

2000

0.50 W/C - Tetraguard

Shrinkage Coefficient, N

Shrinkage Strain ()

4000

3000

2000

Data Points

Fits

0% SRA
1% SRA
2.5% SRA
5% SRA

1000

0
100 90

80

70

60

50

0.5 W/C
0.3 W/C

2500
3000
3500
4000
4500

40

30

Relative Humidity (%)

(a)

30

40

50

60

70

80

Surface Tension (Dyne/cm)

(b)

Figure 3.4 a) The shrinkage of neat cement pastes at various relative humidities
with a water-to-cement ratio of 0.50 and varying amounts of SRA (Tetraguard)
and b) the corresponding shrinkage coefficient for pastes with a water-to-cement
ratio (w/c) of 0.30 and 0.50 (Pease 2005)

The shrinkage behavior of mortar containing SRA was assessed using a mixture
with 55% aggregate.

The specimens consisted of three 50 mm diameter

cylinders that were exposed to a 50% relative humidity environment at an age of


3 days. Shrinkage measurements were started after the time of setting and more
details on the experimental procedure are available (Pease 2005).

Figure 3.5 shows the relative shrinkage of these specimens by dividing the
shrinkage of a specimen containing SRA with the shrinkage of a plain specimen.
It can be observed that the influence of the SRA has two main effects. First, the
SRA reduces the overall magnitude of the shrinkage and second, the SRA
substantially delays the rate of shrinkage.

This is consistent with previous

observations (Weiss et al. 1999) and illustrates that the use of SRA may
substantially reduce the potential for shrinkage cracking, especially at early ages.

1.2
1
0.8
0.6
0.4
0.2
SRA Content

0.0% SRA
0.5% SRA
1.0% SRA
2.5% SRA
5.0% SRA
7.5% SRA

-0.2
-0.4
-0.6
0

50

100

150

200

Drying Time (Days)

(a)

250

300

Relative Shrinkage (X%SRA/0%SRA)

Relative Shrinkage (X%SRA/0%SRA)

63
1.2
1
0.8
0.6
0.4
0.2
SRA Content

0.0% SRA
0.5% SRA
1.0% SRA
2.5% SRA
5.0% SRA
7.5% SRA

-0.2
-0.4
-0.6
0

50

100

150

200

250

300

Drying Time (Days)

(b)

Figure 3.5 The relative shrinkage of mortars containing varying amounts of SRA
(Tetraguard) and 3 day sealed curing with a) water-to-cement ratio of 0.50 and b)
water-to-cement ratio of 0.30 (Pease 2005)

The aforementioned discussions on mixtures containing SRA demonstrate that it


may be possible to control the shrinkage that may be experienced in a concrete
mixture. This work will illustrate that the deterministic model described in section
4.2 may be used to determine the threshold for shrinkage (assuming, creep,
strength, and elastic modulus remain constant) that is necessary to cause
residual stress levels to develop that are sufficient to cause shrinkage cracking.
It will be shown however in the following section that variations in the material
properties can result in a substantial risk for cracking, even if the deterministic
approach outlined in section 4.2 would not predict cracking. It is also possible to
imagine that variations in the environmental conditions (i.e., temperature or
relative humidity) or variations in construction practices (i.e., the time the
concrete is exposed to drying or the specimen geometry) can also substantially
alter the potential for cracking (Schindler et al. 2004). This work provides an
initial illustration of how variability can be incorporated in this model; however
additional study is needed to quantify the extent of correlation between variability
in these properties and to improve the fundamental model inputs.

64
3.4. Model Simulations Deterministic Predictions
The model described in section 4.2 was combined with the material property
development equations described in section 3.3 and used to simulate the
restrained shrinkage response of concrete under various conditions. The timesteps used in the model were 0.1 day increments for the first 8 days, 0.2 day
increments to 14 days, 0.5 day increments to 28 days, 1 day increments to 56
days, 2 day increments to 90 days, and 4 day increments up to 350 days.
A baseline condition was simulated that could be representative of a fairly typical
concrete with a compressive strength of 33 MPa. Table 3-1 provides a listing of
the material inputs that were used in the baseline simulations.

Table 3-1 Baseline model inputs used for simulations


fcomp
MPa
33.0

E28
GPa
27.5

ften
MPa
5.0

C1
C2
C3
1/day 1/day 1/day
2.0
0.3
15.0

RHAMB RHINT
%
%
50
95
C4
1/day
30.0

tset
day
0.3

SH-Ult

600
tdry
day
1.0

Figure 3.6 illustrates the typical results from the model for the base concrete with
Figure 3.6a illustrating the shrinkage, Figure 3.6 illustrating the elastic stress
development and residual stress development, and Figure 3.6c illustrating the
residual stress and tensile strength. The information from Figure 3.6c was used
to determine the time of cracking which was simply determined to be the first age
in the simulation where the residual strength exceeded the tensile strength of the
concrete. The model inputs were then kept constant except for the magnitude of
the ultimate shrinkage strain which was varied to illustrate how variations in the
magnitude of shrinkage can influence the predicted time of cracking. Figure 3.6d
illustrates that as the magnitude of shrinkage is reduced, the time of cracking

65
increases until a point at which cracking would no longer be expected to occur
(i.e., SH ~ 550 ).

16

600

Stress (MPa)

Shrinkage ()

Elastic

Total Shrinkage

500
400
300
200

Autogenous Shrinkage

100

12

8
Residual
4

0
0

14

28

42

56

70

14

(a)

42

56

70

(b)
1000

6
Residual
Residual Stress
Strength

Strength

2
Age of
Cracking

Shrinkage ()

Stress or Strength (MPa)

28

Age of the Specimen (Days)

Age of the Specimen (Days)

800
600
400
time of drying 1 day
time of drying 3 days
time of drying 7 days

200
0

0
0

14

28

42

56

70

Age of the Specimen (Days)

(a)

14

28

42

56

70

Time of Cracking (Days)

(b)

Figure 3.6 a) Results of simulations using the data in Table 3-1 to illustrate
typical results: a) autogenous and total shrinkage, b) elastic and residual
stresses, c) residual stress and tensile strength to illustrate the deterministic age
of cracking, and d) the influence of variation in the ultimate shrinkage coefficient
on the predicted time of cracking

66
3.5. Model Simulations Considering Variability in Material Properties
The model described in section 3.4 was modified to consider variability in
material properties using a Monte Carlo approach. To begin, only the magnitude
of free shrinkage (Shr-) was varied assuming a normal distribution with
a coefficient of variation of 10%. As a result, it would be expected that only 5%
of the shrinkage values would be below shrinkage of 500 while 95% of the
shrinkage values would fall below shrinkage of 700 .

If the deterministic

model calculations are performed with shrinkage of approximately 550 ,


cracking would not be predicted. However, cracking would be predicted to occur
at an age of 8.7 days for a specimen with shrinkage of 700 . It can therefore
be expected that the model that incorporates variability would predict the majority
of cracks to develop between 8.7 days and some very late age with some
specimens (i.e., the specimens with a shrinkage of less than 550 ) no cracking
at all.

A series of iterative simulations were performed with each simulation using


a randomly selected shrinkage value from the normal distribution. The sampling
was performed using the Latin Hypercube sampling procedure [Guide to Using
@Risk, 2005]. In the simulations described in this work 10,000 iterations were
performed and the time of cracking that was determined from each iteration was
saved and used to develop a cumulative probability distribution, as shown in
Figure 3.7a.

It can be seen that the age of cracking predicted using the

deterministic approach described in section 4 occurs at 27.5 days while the


predictions of this model occur over a wider range and capture the fact that the
age of cracking may not be predicted in some specimens. This type of approach
can be used to better quantify the risk of cracking and to develop a plot like that
shown in Figure 3.7b where the time of cracking is plotted along with a probability
that cracking would be experienced for different values for the average
magnitude of shrinkage.

67

Shrinkage Coefficient ()

Specimens Cracked (%)

100
Age of
Cracking

80
60
40
20
0
0

14

28

42

56

70

Age of the Specimen (Days)

(a)

1000
800
600
400
200

5% Probability of Cracking
50% Probability of Cracking
95% Probability of Cracking

0
0

14 21 28 35 42 49 56

Time of Cracking (Days)

(b)

Figure 3.7 Results from simulations: a) assuming only variability in shrinkage,


illustrating the cumulative age of cracking distribution and the age of cracking
from the deterministic equations and b) 5%, 50%, and 95% probability
of cracking limits for specimens with different shrinkage coefficients

While information on variations of a single property is useful, the variability could


be assessed directly in this case. The benefit of the Monte Carlo approach will
be demonstrated later in this work where more than one material property is
varied. First however, we should discuss the fact that simply having the results
of the simulation from a probability distribution can make the results a bit more
cumbersome to interpret.

As a result, several different expressions were

investigated to describe the probability of cracking using only a few parameters.

The expressions investigated to describe the probability of cracking included loglogistic, lognormal, gamma, and exponential distributions; however several
common trends were observed from the functions that should be mentioned.
The lognormal and exponential curves appeared to slightly underestimate low
cracking probabilities (i.e., < ~25%) while the log-logistic curve tended to
underestimate the higher cracking probabilities (i.e., > ~90%). In addition, the
lognormal curves require the use of an error function that may make their
application more difficult for some users. As a result, the cracking cumulative

68
probability curves were fit using a log-logistic distribution function which was
modified to account for the fact that cracking may not always be observed in
some mixtures. This modification came in the form of a factor (PCrack-) which
describes the probability of cracking after a long time period. In this cases it is
assumed as the amount of cracking observed after 1 year when the simulation
was stopped. Therefore, the probability of cracking is described using Equation
3-12:

PCrack cdf = PCrack

Eq. 3-12

1

1 +
t

where is the shape parameter, is the scale parameter, and is a timelocation factor. The derivative of Equation 3-12 provides an expression for the
cracking probability density which is described in Equation 3-13:

PCrack pdf = PCrack

t 1
t
1 +

Eq. 3-13

which can be helpful for fitting and comparisons with probability histograms.

Figure 3.8a shows a simulation for the probability distribution function obtained
from the simulation that was fitted using Equation 3-13 and a histogram of the
results of the simulation. It can be noticed that the distribution function contains
the most frequent ages of cracking before the time of cracking predicted using
a deterministic model (i.e., 27.5 days). Integrating the probability density function
over time yields the cumulative distribution shown in Figure 3.8b.

Not

surprisingly, it can be seen that after a certain period of time the potential for
additional cracking events decreases dramatically and the number of specimens

69
that can be expected to crack begins to plateau (i.e., this plateau is the ultimate

100

3
Simulation Data
LogLogistic Function
2

Cracking Probability (%)

Cracking Probability Density(%)

probability for a crack to occur, Pcrack-).

Simulation Data
Log-Logistic Function

80
60
40
20
0

0
0

14

28

42

56

70

84

98

Age of the Specimen (Days)

(a)

14

28

42

56

70

84

98

Age of the Specimen (Days)

(b)

Figure 3.8 a) Cracking probability density for different ages of the specimen and
b) cumulative distribution of cracking probability for different ages of the
specimen

While the predictions from the model show reasonable agreement with what one
may expect, as discussed earlier, the real power of this approach will be to
consider simultaneous variations in more than one property.

Three material

properties were considered including the magnitude of free shrinkage (SH-),


elastic modulus (E), and tensile strength (ften-). The material properties were
assumed initially to have a normal distribution with a coefficient of variation of
10%.

Further, the material properties were assumed for some trials to be

uncorrelated with each other.

Other trials, however, will assume that some

correlation exists between the properties. It was assumed in this thesis that an
increase in shrinkage would likely be correlated to a decrease in elastic modulus
and tensile strength. As such, the same coefficient of variation (10% COV) was
used for these in the correlated cases. For the remainder of this chapter, the
data where the three variables were allowed to vary independently will be

70
referred to as uncorrelated and the data in which the variations in material are
related to one another (assuming elastic modulus and strength are positively
correlated and elastic modulus and shrinkage are negatively correlated) will be
referred to as correlated.

Results from one simulation are shown on Figure 3.9 while Table 3-2 provides
the coefficients for Equation 3-12 that were determined from a series of
simulations. It can be seen that the correlated outcomes tend to result in a later
age of cracking. It should however be noted that there is a greater probability of
observing a crack when the results are correlated. At this point, it should also be
noted that while the variability in each material property is assumed to have
a similar coefficient of variation, it is highly likely that each property will exhibit

Uncorrelated Variability

Probability Density of
Cracking Predictions (%)

Probability Density of
Cracking Predictions (%)

a different level of variation with modern quality control practices.

Correlated Variability

0
14

21

28

35

Age of Specimen (Days)

(a)

42

14

21

28

35

42

Age of Specimen (Days)

(b)

Figure 3.9 A comparison of the model simulation predictions for the age of the
specimen at the time of cracking assuming a) uncorrelated variability in material
properties and b) correlated variability in material properties

71
Table 3-2 Baseline log-logistic simulation parameters considering variability due
to: a) shrinkage only, b) shrinkage, tensile strength and elastic modulus
assuming they are uncorrelated, and c) shrinkage, tensile strength and elastic
modulus assuming they are correlated
Only Variation in Shrinkage [Sh]
tdry
Pcrack

1
60.9
2.98 22.54
0.32
3
58.9
2.58 17.33
4.80
7
57.2
2.59 14.94
10.94
Variation in Three Parameters [Sh, ften, E]
Assumed Uncorrelated
tdry
Pcrack

1
67.4
2.05 17.70
-1.29
3
71.2
1.88 13.71
3.41
7
73.7
2.18 13.83
7.48
Variation in Three Parameters [Sh(-),
ften(+), E(+)] Assumed Correlated
tdry
1
3
7

Pcrack
78.3
83.9
85.9

3.08
2.65
2.45

23.37
17.78
14.42

-0.14
4.26
10.98

The difference in the results obtained from varying a) only shrinkage, b)


shrinkage, elastic modulus, and tensile splitting strength assuming they are
uncorrelated, and c) shrinkage, elastic modulus and tensile splitting strength
assuming they are correlated has been shown in the Figure 3.10a. The 5%
potential for cracking curve shows that when we assume all three components
are correlated, the spread of results occupies a smaller range than for
uncorrelated components. This is consistent with what one may expect as the
data in correlated simulations results may tend to offset one another (i.e., higher
strength and higher stiffness) while the data in the uncorrelated results may tend
to combine to create a condition that is either more likely or less likely to result in
cracking.

72

Cummulative Predictions
of Cases of Cracking (%)

800

Shrinkage ()

700
600
500
400

Probabilityof
ofCracking
Cracking
Probability
OnlyShrinkage
Shrinkage
5%5%
Only
Uncorrelated
5%5%
Uncorrelated
Correlated
5%5%
Correlated

300

14

21

Time of Cracking (Days)

Time of Drying - 1 Day


Time of Drying - 14 Days

80
60
40
20
0

200
0

100

28

14

28

42

56

70

Age of Specimen (Days)

(a)

(b)

Figure 3.10 A comparison of the influence of the shrinkage coefficient on the time
of cracking a) varying shrinkage only; varying shrinkage, elastic modulus, and
tensile strength (uncorrelated variability); and varying shrinkage, elastic modulus,
and tensile strength (correlated variability) and b) an illustration of the influence
of a delay in the time when drying is initiated

3.6. Predicting Cracking Performance


To illustrate how the time-dependent development of stress and strength may be
affected by different factors, including: time of set, time when drying was initiated,
ambient relative humidity, shrinkage of concrete and elastic modulus of the
concrete, and example has been shown in Figure 3.11 Figure 3.11a presents an
example of the computed cracking potential for three concrete elements having
volume of aggregate 68%, elastic modulus E = 27.5 GPa, and splitting tensile
strength ft = 5 MPa, but with different magnitude of the shrinkage coefficient:

N = 3000 , N = 3500 , and N = 4000 . This corresponds to long-term


shrinkage values of: Shr = 515 , Shr = 600 ,Shr = 685 , respectively.
Figure 3.11b presents the results of the predicted time of cracking for the same
three types of concrete when 5% variability was assumed for the values of elastic
modulus (E), splitting tensile strength (ft), and shrinkage coefficient (N).

It

should be noted, however, that experimental and field study (Pellinen et al.

73
2005), as well as specifications (ASTM C 157, ASTM C 596) suggest expected
variability in shrinkage as high as 25% or 35%.

To obtain probability density and cumulative distribution curves for the probability
of cracking, Monte Carlo simulations were performed with Latin Hypercube
sampling for 10,000 iterations (Radlinska et al. 2007). Concrete was assumed to
be subjected to drying after 7 days, internal relative humidity was assumed to be
95%, an external humidity was assumed to be 50%. Time of set was taken as 6
hours and other modeling constants has been described earlier in the text. For
the concrete with the highest shrinkage coefficient value, a high probability of
cracking can be expected to occur between 8 and 42 days. The smoother and
flatter curve for the lowest shrinkage coefficient highlights that in this case the
cracks would be expected to occur later (around 14 days) or would not be
expected to occur at all (curve approaches zero at very late ages).

0.12

Probability of cracking

1.4
1.2

/f't

1.0
0.8
0.6
7 days drying
N= 3000
N= 3500
N= 4000

0.4
0.2

14

28

42

Time (Days)

(a)

56

0.10
0.08
0.06
0.04
0.02
0.00

0.0
0

7 days drying
N = 3000
N = 3500
N = 4000

70

14

28

42

56

70

Time (Days)

(b)

Figure 3.11 Development of time-dependent stress and tensile strength for a)


time of drying equal to 7 days and b) corresponding probability of cracking

74
3.7. Summary
This chapter has outlined a procedure for implementing variability in a predictive
model for shrinkage cracking. The work presented here has shown that:

Even small variability results in substantial scatter in the predicted time of


cracking.

It is therefore suggested that a deterministic prediction of

cracking may not provide sufficient information for the selection of material
properties for specifications. Inherent variability in material property should
be considered in specifying a limit on the magnitude of permissible free
shrinkage;

While a normal distribution of variability in the material properties may be


assumed, the time-dependent nature of material property development
and the multiplicative nature of the variability results in a distribution of the
time of cracking that is not normally distributed. A log-logistic curve was
initially used to describe this response since it appeared to describe low
probability of cracking better than other models, however this equation
needed to be modified to account for the probability that cracking may not
be observed;

Three different approaches were considered in implementing this


methodology. The first approach considered only variability in shrinkage.
The results from this approach appeared consistent with what one may
expect from hand calculations. When uncorrelated multiple sources of
variation were introduced, a much higher variability in the results was
observed.

75

CHAPTER 4. RELIABILITY BASED APPROACH FOR ASSESSING CRACKING


RISK USING LOAD AND RESISTANCE FACTOR DESIGN

4.1. Introduction
As stated in Chapter 4, deterministic analysis of the cracking risk results in the
single time (e.g., 15 days) when cracking can be expected to occur. However,
this approach does not provide information about the probability of such an
events occurrence. On the contrary, the Monte Carlo method gives precise and
detailed information about the cracking risk at any age, but requires timeconsuming computations. This chapter presents how these limitations can be
overcome using LRFD (Load and Resistance Factor Design) approach which is
more time-efficient method to predict probability of cracking in restrained
concrete element (Radlinska and Weiss 2006a, Radlinska and Weiss 2006b).
The LRFD method allows calculating probability of load events exceeding
materials resistance in a simplified approach and thus enables to overcome
certain limitation encountered when risk analysis is conducted using Monte Carlo
simulation.

Since even small variability in material properties can result in substantial scatter
of results of shrinkage, stresses, modulus of elasticity, etc., it is important to
quantify the influence of variability of a single input that can reflect the risk of
unwanted event. In this chapter, a reliability approach is proposed that can be
used to establish this requested value. It will be also presented how variability
from many sources can be assessed using direct computation instead of
performing simulations, and how information about the risk of cracking can be
presented by a single term: reliability index.

76
Additionally, this work can be used to develop cracking risk diagrams and to
establish a threshold for the risk of cracking. Further, this work can be used to
quantify correlation in the material variability and practical applications and
incorporating variability in environmental conditions.

4.2. Estimating Variability


One of the possible modeling solutions is to estimate total variability in the
system based on the information about variability in the separate components.
Toward this end the following section will describe how the resistance and
loading may be used to approximate shrinkage cracking in concrete.

This assessment can be done by simplifying time-dependant equations


introduced in Chapter 3.

Section 3.3 introduced equation (equation 3-2)

describing the total strain that is permitted in a restrained concrete element. It


can be rearranged and written in a form (Radlinska and Weiss, 2006a):

1
(t )
+

E (t ) Ec28

Permit Tot Shr = (t )

Eq. 4-1

Throughout this investigation, a complete restraint is assumed, i.e., Permit(t) = 0.


However, these approximations would be similar if variable restraint was
permitted. Recalling Equations 3-4 and 3-5, Equation 4-1 can be rewritten:
1 + C1 (t t s )
C1 (t t s )
1 + C1 (t t s )
1+
(t )
C1 (t t s )

Shr E c 28

Eq. 4-2

After dividing both sides of Equation 4-2 by the term ft, Equation 4-3 can be
obtained:

77

1 + C1 (t ts )
Shr Ec 28

C1 (t ts )

=
f 't
1 + C2 (t ts ) 1 + C1 (t ts )

f 't 28
(t )
1 +
C1 (t ts )
C2 (t ts )

Eq. 4-3

If we focus on the magnitude of the material properties (temporarily omitting


variability in rate dependent constants), an expression shown by equation 4-4 is
obtained:

f 't

Shr Ec28
f 't

Eq. 4-4

In this equation, the numerator (i.e., shrinkage and elastic modulus) can be
thought of as the load imposed on a concrete member (Q), while the
denominator (i.e., strength) can be thought of as the materials resistance (R).
As a result, this work attempts to determine whether the variability predicted in
the age of cracking can be estimated by using a simple ratio which is
approximately related to the risk of cracking such as that given in equation 4-1.

4.3. Reliability Approach


In any reliability analysis, the concept of risk should be described clearly. The
probability of failure in structural design may be understood as structural
collapse; however failure may also be considered in a more general sense as
a loss of serviceability. Generally, the term limit states (Melchers 1987) may be
used to describe material distress.

Material distress could include cracking,

deformation, excessive deflection, local damage or corrosion. While limit state


defines a general concept, this work will use the term cracking, failure, or limit
state interchangeably.

78
The variability that is encountered in loading effects as well as a materials
resistance should be accounted for, as it may alter significantly parameters of
interest and an area of safe design. Detailed analysis of all uncertainties would
not provide practical solution, but simplified methods exist for obtaining
a probability-based assessment of the risk of failure: first-order second-moment
reliability methods. This approach has been used in steel design (Stewart 1997,
Salomon and Johnson 1996) and allows the reliability of a system to be
estimated by treating the load (Q) and resistance (R) as random variables with
assigned distribution, as presented in the Figure 4.1 (Stewart 1997, Stewart
2001, Ellingwood 1993). The area of failure, i.e., the area when the load (Q)
exceeds the resistance (R) may be represented by (however is not exactly equal
to) the amount of overlap of two probability density functions: fQ(x) and fR(x)

Frequency

(Melchers 1987), as shown in Figure 4.1 by the cross-hatched area.

Resistance (R)
Load (Q)

Area
of
failure

Figure 4.1 Load (Q) and resistance (R) as the normally distributed variables

If analyzed R and Q have normal distribution with mean R and Q and standard
deviations SDR and SDQ, the safety margin Z = R S has mean and variance
expressed by:

79

Z = R S

Eq. 4-5

SD 2 Z = SD 2 R SD 2 S

Eq. 4-6

Instead of integrating the area under the two overlapping curves, it is more
convenient to deal with a single curve that consist of the logarithm of R divided
by Q (or the inverse of the cracking potential, as mentioned in chapter 3,
equation 3-1), as shown in Figure 4.2. In this case, the area of failure is the
region under the curve on the negative side of x-axis, marked as the crosshatched region.

The probability of cracking can be expressed as either the

probability of resistance being lower or equal to load, as the difference between


resistance and load being less than or equal to zero, or as the probability of the
resistance to load ratio being less than or equal to one:
pf = P(R Q) = P(R Q 0) = P(R/Q 1)

Eq. 4-7

ln(R/Q)

2.50
ln(R/Q) - standard
deviation of ln(R/Q)

Frequency

2.00
1.50
1.00
0.50

Area of failure

0.00
-0.8

-0.4

0.0

0.4

0.8

[ln(R/Q)]

Figure 4.2 The reliability index

1.2

80
The LRFD approach enables the total variability of the system to be estimated
through the calculation of the reliability index (see Figure 4.2):

ln( R / Q )

Eq. 4-8

SDln( R / Q )

where ln(R/Q) is the mean value of the natural logarithm of R divided by Q and
SDln(R/Q) corresponds to the standard deviation of the natural logarithm of the R
divided by Q. For example, if the maximum stress reaches a long term value of
5.50 MPa and maximum tensile strength is 4.95 MPa, then ln(R/Q) is equal
to -0.11.

Rather than describing the variability in each material property, a simplified


approach can be used to approximate the total coefficient of variation (COV)
using equation 4-10 (Salomon and Johnson 1996):

COV = COV2 + COV E2 + COV f2't

Eq. 4-9

where the single coefficients of variation under the square root are those of the
shrinkage (), elastic modulus (E), and splitting tensile strength (ft):

COV =

SD

Eq. 4-10

COVE =

SDE

Eq. 4-11

COV f 't =

E
SD f 't

f 't

Eq. 4-12

81
It can be assumed that the SDln(R/Q) is equal to the total coefficient of variation
(Salomon and Johnson, 1996).

If all three parameters have a coefficient of

variations equal to 5%, a total coefficient of variation is equal to 0.0866. In that


case, the reliability index can be calculated as:

0.11
= 1.22
0.0866

Eq. 4-13

The probability of failure can be written as:


ln(R / Q ) ln ( R / Q ) ln ( R / Q )
= p( Z < ) = ( )
p f = p(ln(R / Q) < 0) = p
<

SD
SD
ln
ln
(
R
/
Q
)
(
R
/
Q
)

Eq. 4-14

where Z is the standard normal variable and denotes the cumulative density
function, i.e., probability that a variable has a value less than or equal to ln(R/Q)
(Melchers 1987). The use of this approach to calculate the reliability index allows
a rapid estimation of the cracking probability using commonly available standard
normal distribution tables. For the reliability index calculated in equation 4-14,
the probability of cracking can be found as:
pf = () = (1.22) = 0.8887

Eq. 4-15

The probability of cracking in concrete depends on the coefficient of variation in


the system.

Figure 4.3a presents how the probability of cracking changes

depending on the magnitude of the shrinkage coefficient, while the corresponding


figure (Figure 4.3) shows the reliability index values for the same example. The
horizontal dashed lines in the Figure 4.3 denote the maximum and minimum
values usually reported in the mathematical tables for standard normal
distribution; for reliability index equal to 3.4, probability of failure is minimal:

82
pf = () = (-3.4) = 0.0003

Eq. 4-16

while for reliability index equal to -3.4, probability of failure is very high:
pf = () = (3.4) = 0.9998

Eq. 4-17

6.00
COV = 0.0866
COV = 0.1658
COV = 0.2449

0.80

Reliability Index

Probability of Cracking

1.00

0.60
0.40
0.20
0.00
2000

2500

3000

3500

4000

Shrinkage Coefficient ()

COV = 0.0866
COV = 0.1658
COV = 0.2449

4.00
2.00
0.00
-2.00
-4.00
2000

2500

3000

3500

4000

Shrinkage Coefficient ()

(a)

(b)

Figure 4.3 a) Probability of cracking for concrete with different coefficients of


variation and magnitude of shrinkage and b) reliability index values for those
cases

4.4. Comparison of MC and LRFD


To evaluate correctness of the approach presented in section 4.3, a numerical
model was used. Values of the material properties described by means and
standard deviations were obtained from the deterministic information about the
system. Using approach expressed by Equations 4-8 to 4-10, variability of the
system can be approximated without necessity of performing time consuming
Monte Carlo simulations. The total variability expected in the system [SDln(R/Q)]
has been compared with Monte Carlo simulations as presented in Table 4-1 and
in Figure 4.4. The analysis was conducted by introducing variability in the elastic
modulus, splitting tensile strength, and shrinkage. The magnitude of shrinkage

83
was varied from 400 to 700 (at 100 increments) and results obtained from
two different approaches showed good agreement.

Table 4-1 Simulations used in the preliminary comparison of the direct


computation and Monte Carlo simulation (7 days of drying)
Simulation
Number
1
2
3
4
5
6
7
8
9
10
11
12

%variability
E
f't
sh

sh

10
6
8
10
6
8
10
6
8
10
6
8

700

10
4
6
10
4
6
10
4
6
10
4
6

10
15
15
10
15
15
10
15
15
10
15
15

600

500

400

COV
(from
M.C.)
0.178
0.176
0.182
0.176
0.167
0.167
0.177
0.166
0.166
0.175
0.165
0.167

(DC) Direct
Computation
Using LRFD
0.173
0.166
0.180
0.173
0.166
0.180
0.173
0.166
0.180
0.173
0.166
0.180

Coefficient of Variation

Direct Computation

(COVDC)/COV [%]
2.40
5.66
1.09
1.69
0.51
7.68
2.28
0.35
8.91
1.08
0.78
7.93

MC Simulation

0.20
0.15
0.10
0.05
0.00
1

9 10 11 12

Simulation Number

Figure 4.4 Comparison of direct computation of COV for multiple sources of


variability and Monte Carlo simulation

84
After agreement between Monte Carlo and Load and Resistance Factor Design
has been confirmed, additional analysis was performed to present comparison
between results obtained using deterministic approach, MC and LRFD.

The

differences between the results obtained using the three mentioned methods are
presented in Table 4-2. For a mixture with a shrinkage coefficient of 3000 and
COV = 0.0866, the deterministic analysis would indicate that no cracking is
expected to occur. However, if the Monte Carlo simulations were performed for
the same concrete, the output would indicate that 5% of samples will crack by the
age of 18.6 days and by an age of 300 days, 31.5% of samples will be cracked.
For a shrinkage coefficient of 3500 and a COV of 0.0866, the deterministic
analysis would result in the predicted time of cracking after 26.5 days. If the
Monte Carlo simulations were performed for the same concrete, the output would
indicate that 5% of samples will crack by 12.2 days, 50% will crack by 27.1 days,
and some of them will never crack.

If the LRFD analysis is conducted, the

reliability index () would indicate that the overall probability of cracking for the
concrete with shrinkage N = 3500 is 88.8%.

85

Table 4-2 Modeling inputs


Shrinkage
coefficient
[]

Ultimat.
shr. of
concr.
sh []

3000

515

3500
4000

600
400

Time of
cracking
from
determini
stic
analysis
[days]
no
cracking
26.55
16.55

Time of cracking
(MC) [days]

Probabi
lity of
crack.
at 300
days
(MC)

Reliability
index

Overall
probab.
of
cracking
(LRFD)

5%

50%

95%

18.6

500

500

31.5 %

0.56

28.6 %

12.2
9.5

27.1
17.1

500
120

90.4 %
99.8 %

-1.22
-2.76

88.8 %
99.7 %

4.5. An Example of Application of the LRFD Procedure


The approach developed in this work can be illustrated by a numerical model.
Using the strength and elastic modulus calculations described in Chapter 4,
section 4.3, a series of computations were performed to considered the effect of
changing the magnitude of shrinkage (i.e., N = 3000, 3500, 4000 ).

For

shrinkage coefficient N = 3000 , the ultimate deterministic value (i.e.,


assuming no variability) of the tensile strength and residual stress were 4.95 MPa
and 4.72 MPa, respectively.

This would imply that cracking would not be

predicted for deterministic approach as shown in Figure 4.5:

86

Stress, Strength [MPa]

6.0
5.0
4.0
3.0
2.0

7 days drying
Stress
Strength

1.0
0.0
0

14

28

42

56

70

84

98

112

126

140

Time (Days)

Figure 4.5 Time-dependant stress and strength development for N = 3000

However, if we consider the effect of material variability, the probability of


cracking can be calculated. Figure 4.6 presents plots of density and cumulative
functions for log-normal distribution obtained for a material with a shrinkage
coefficient of 3000 (this would correspond to a concrete with a long-term
shrinkage of 515 ).

Coefficient of variation was calculated as given by

Equation 4-10, assuming 5% variability in: elastic modulus, splitting tensile


strength, and shrinkage.

This results in total coefficient of variation COV =

0.0866. The reliability index was computed to be = 0.058 (Equation 5-8) and
probability of cracking was calculated to be pf = 28% (Equation 5-7). The dashed
line in the Figure 4.6a separates the safe and failure zone, and from
corresponding cumulative function plot in the Figure 4.6b, probability of cracking
can be directly obtained. It can be observed that any negative values of ln(R/Q)
correspond to cracking while positive values denote the non-cracking region.

Cumulative Density Function

Probability Density Function

87
5.00
4.00
3.00
2.00

Failure

Safe

zone

zone

1.00
0.00
-0.6 -0.4 -0.2 0.0

ln(R/Q)

(a)

0.2

0.4

0.6

1.0
0.8
0.6
0.4
0.2
0.0
-0.6 -0.4 -0.2 0.0

0.2

0.4

0.6

ln(R/Q)

(b)

Figure 4.6 a) Density function and b) cumulative function plot for shrinkage
coefficient N = 3000

If deterministic analysis was conducted for higher shrinkage coefficient, i.e.,


3500 (long-term shrinkage of 600 ), ultimate strength and stress would be
equal to 5.50 MPa and 4.95 MPa, respectively. This analysis would result in
predicted time of cracking equal to 26.5 days, as shown in the Figure 4.7.
However, if variability was taken into consideration, the plot given by the Figure
4.8 would be obtained. In this case the majority of the density function would fall
into failure region and probability of cracking would be very high (pf = 89%).

88

Stress, Strength [MPa]

6.0
5.0
4.0

Time of cracking: 26.5 days

3.0
2.0

7 days drying
Stress
Strength

1.0
0.0
0

14

28

42

56

70

84

98

112

126

140

Time (Days)

Cumulative Density Function

Probability Density Function

Figure 4.7 Time-dependant stress and strength development for N = 3500

5.00
4.00
3.00

Failure

Safe

zone

zone

2.00
1.00
0.00
-0.6 -0.4 -0.2 0.0

0.2

0.4

0.6

1.0
0.8
0.6
0.4
0.2
0.0
-0.6 -0.4 -0.2 0.0

ln(R/Q)

(a)

0.2

0.4

0.6

ln(R/Q)

(b)

Figure 4.8 a) Density function and b) cumulative function plot for shrinkage
coefficient N = 3500

To illustrate the value of the aforementioned concept, Figure 4.9a presents how
probability of cracking changes with the shrinkage coefficient, while Figure 4.9b
illustrates the relationship between probability of cracking and cracking potential
for concrete having coefficient of variation COV = 0.0866.

89
1.40

Cracking Potential /f't

Probability of Cracking

1.00
COV = 0.0866

0.80
0.60
0.40
0.20
0.00
2000

COV = 0.0866

1.20

1.00

0.80

0.60
2500

3000

3500

4000

Shrinkage Coefficient ()

0.0

0.2

0.4

0.6

0.8

1.0

Probability of Cracking

(a)

(b)

Figure 4.9 Probability of cracking for coefficient of variation COV=0.0866 with


respect to a) shrinkage coefficient and b) cracking potential

It can be seen from Figure 4.9 that the LRFD approach can be utilized in two
primary ways. First, a plot can be developed to relate the shrinkage properties of
the mixture to the cracking potential. By using this type of an approach, it may
be possible to determine the maximum allowable shrinkage (or the amount of
shrinkage reduction that would be needed) in order to minimize potential for the
cracking. The second way this approach can be used is shown in the Figure
4.9b where the cracking potential is related to the probability of failure. Using this
information, deterministic analysis can be performed to determine the steps that
need to be taken to limit the ratio of /ft to a desired number. This approach
enables the required /ft ratio to be fundamentally correlated to the probability of
cracking rather than relying only on engineering intuition.

4.6. Reducing the Probability of Cracking Using Shrinkage


Reducing Admixtures
The use of shrinkage reducing admixtures (SRAs) decreases the magnitude of
shrinkage in a concrete mixture (Shah at al. 1998, Weiss and Shah 2002, Weiss

90
et al. 1998, Weiss et al. 1999, Bentz et al. 2001,Bentz and Jensen 2004,Bentz
2005). This reduction in shrinkage has been claimed to be due to a reduction in
the surface tension of the pore solution (Pease 2005, Ai and Young 1997). As
mentioned before in Chapter 3. Figure 4.10 shows the results from the
experiments where the shrinkage coefficient was measured for a series of pastes
containing different concentrations of SRA (Pease 2005, Pease et al. 2005). It
can be seen that when the shrinkage coefficient is plotted versus the surface
tension of water-SRA solution, a linear trend is observed.

A relationship between the surface tension of water-SRA solution and the


percentage of SRA was previously measured (Pease 2005) and the results are
shown in Figure 4.11. The surface tension decreases with addition of SRA (up to
15% SRA) and can be described using Equation 5-24 (Pease et al. 2005):

CSRA

Eq. 4-18

SOLN = Water 1 ln

if CSRA < 15

SOLN = 33.3

if CSRA 15

Eq. 4-19

where SOLN is surface tension of the solution, Water is the surface tension of
water, 72 dyne/cm (Weast et al. 1983), and and are fitted constants (0.0795
and 0.0164, respectively) for one of the commercial shrinkage reducing
admixtures, Tetraguard (Pease 2005).

91

Shrinkage Coefficient, N

2000

Tetraguard
2500

0.50 W/C

3000

3500

4000

4500
30

40

50

60

70

80

Surface Tension (Dyne/cm)

Figure 4.10 The relationship between shrinkage coefficient for paste with a w/c of
0.50 and surface tension of the pore solution (Pease 2005, Pease et al. 2005)

Surface Tension (dyne/cm)

80

Water
Fits

60

Experimental Data
Tetraguard
Eclipse Plus
Eclipse Floor

40

20

0
0

10

15

20

25

90

95

100

SRA Concentration (%)

Figure 4.11 Surface tension measurements for deionized water-SRA mixtures for
Tetraguard, Eclipse Plus and Eclipse Floor (Pease 2005, Pease et al. 2005)

Since the addition of SRA decreases the shrinkage of concrete, one may want to
compute the value of long term shrinkage that is required to reduce the
probability of cracking. The following section will describe how the concentration
of SRA can be determined to correspond with a specific level of performance.

92
4.7. An Example of Using the LRFD Procedure to Determine the Required
Concentration of SRA in a Concrete Mixture
Consider an example of a plain concrete mixture with a w/c = 0.5, volume of
aggregate 68%, a total coefficient of variation COV = 0.0086, and a plain paste
with a shrinkage coefficient of 3900 (Figure 4.12). It is assumed that elastic
modulus and splitting tensile strength for this material are 27.5 GPa and 5 MPa,
respectively, while the rate parameters are the same as those used in the earlier
example.

The reliability index for this example can be computed using Equation 4-8
( = -1.29) and probability of cracking is equal to 99%. This denotes that the
concrete will almost certainly crack and therefore the design of the concrete
mixture requires some modification. In order to reduce the probability of cracking
in concrete, a shrinkage reducing admixture could be used, the volume of
aggregate in the mixture could be increased, or different cementitious materials
could be incorporated to reduce the shrinkage coefficient. Here, a procedure is
presented to estimate the amount of SRA that would be needed to decrease the
potential for cracking to an acceptable level.

If variability in the system can be estimated, the LRFD approach allows for the
development of a graphical presentation of a relationship between shrinkage
coefficient and probability of cracking (Figure 4.12). If the designer was willing to
accept a 10% risk of cracking, the shrinkage coefficient would have to be
reduced from 3900 to 2820 , which is equivalent to the reduction of long term
shrinkage value (Sh) from 670 to 480 . As shown in Figure 4.12, lowering
shrinkage to that value corresponds to 3% SRA concentration (for Tetraguard).
This means, that instead of plain system, a mixture with 3% SRA should be used
in order to lower the shrinkage from 3900 to 2820 .

93
The procedure presented in Figure 4.12 can be described in the following steps:

Step 1: determine the shrinkage and variability in the system together with
corresponding probability of cracking

Step 2: determine the acceptable probability of cracking (i.e., 5%, 10%)

Step 3: determine the amount of SRA (%) needed to lower the shrinkage

4500

4500
COV = 0.0866

Tetraguard
0.50 w/c

4000

4000

700

Step 1

3500

3500

3000

3000
Step 2

2500

2500

600

500
Step 3

400

Shrinkage Sh ()

Shrinkage Coefficient, N ()

coefficient and the probability of cracking to the desired level

2000

2000
0.0

0.2

0.4

0.6

0.8

Probability of Cracking

1.0

0 1 2 3 4 5 6 7 8 9 10

SRA Concentration (%)

Figure 4.12 Steps in the shrinkage based design approach

The present example shows that in order to reduce the probability of cracking
from 99% to 10%, a concentration of 3% SRA should be used in the mixture. To
verify this approach using a laboratory experiment, a large number of samples
would have to be prepared and tested. Alternatively, the Monte Carlo method
can be used to perform simulations for a concrete with shrinkage value of 670
and 480 (Figure 4.13 ). It can be seen that cumulative function for Sh = 670

approaches value of one at the age of 50 days, which means that by the time
of 50 days almost all the samples (i.e., 99%) will crack. The response of the
specimen with 3% SRA (Sh = 480 ) reaches a value of 10% and flattens after
that, which indicates that only 10% of all the tested samples will crack.

Probability of Cracking (%)

94
100
80

Sh = 670
Sh = 480

60
40
20
0
0

50

100

150

200

250

300

Age of Cracking (Days)

Figure 4.13 The results of simulations showing how addition of SRA can
decrease shrinkage of concrete and reduce the probability of cracking

4.8. Summary
The work presented in this chapter leads to the following concluding remarks:

The probability of early age shrinkage cracking in concrete depends on


the variability in the material properties;

The ratio of time dependant stress and strength (i.e., the cracking
potential) can be used in statistical analysis (i.e., the Monte Carlo method
or Load and Resistance Factor Design approach) to incorporate variability
into computations;

While the Monte Carlo method allows an accurate estimation of the time
and probability of cracking in concrete, it is a time consuming process.
The Load and Resistance Factor Design approach permits more rapid
assessment of cracking, however some of time-dependant phenomena
may be difficult to apply,

The addition of shrinkage reducing admixture decreases the shrinkage


and cracking potential in concrete. A method is proposed to estimate the

95
amount of SRA that is needed to reduce the cracking risk of concrete to
the accepted level;

The approach developed in this work can be used for optimizing


construction procedures or for design purposes.

96

CHAPTER 5. THE RING TEST: A REVIEW OF RECENT DEVELOPMENTS

5.1. Introduction
Several test methods have been proposed to assess the shrinkage cracking
potential of concrete mixtures (Grzybowski and Shah 1990, Weiss and Shah
1990) including linear specimens (Springenschmidt 1985, Kovler 1994, Altoubat
and Lange 2002, Toma et al. 1999, Parilee 1982) and ring specimens (Hah et al.
1992, Bentur 2002, AASHTO PP 34-99, ASTM C 1581-04, Carlson 1942). The
linear specimens have the advantage of the relatively straight-forward data
interpretation, however, these test methods are generally not used for quality
control procedures, partially due to difficulties associated with providing sufficient
end restraint (Weiss and Shah 1997, Altoubat and Lange 2002). Historically, ring
tests have been performed to evaluate the performance of concrete or mortar
mixtures when they are restrained from shrinking freely.

In the ring test,

a concrete (or cement paste) annulus is cast around a steel ring. If unrestrained
(i.e., no steel ring), the concrete would shrink freely; however, the steel ring
prevents (restrains) this movement, resulting in the development of tensile
stresses. Due to its simplicity and economy, the ring test has been developed
into both AASHTO PP 34-99 and ASTM C 1581-04 standards.

The main

difference between these standards is the relative ratio of the concrete to steel
ring thickness which influences the degree of restraint provided to the concrete
(the AASHTO method provides less restraint which results in a lower potential for
cracking or longer times until cracking is observed).

97
This chapter provides an overview of the qualitative use of the restrained ring
test, the use of the ring for quantitative analysis of stress development, and
emerging developments that can account for materials that expand at early ages.

5.2. Qualitative Restrained Ring Tests


The ring test has been used by many researchers over the last century. The
most common use of the ring test is where the inner (steel) ring provides passive
restraint (through its stiffness) to the concrete ring specimen.

For example,

Carlson (1942, 1988) and Coutinho (1959) used the ring test to assess the
cracking of restrained concrete elements. Douglas and McHenry (1947) and
Brewer and Burrows (1946) used a smaller ring geometry to assess cracking in
the cement paste. Swamy and Stavrides (1970) and Krenchel and Shah (1987)
used the ring test for fiber reinforced materials. The restrained ring test has also
been used to examine the influence of new materials (e.g., shrinkage reducing
admixtures) (Shah et al. 1992, Folliard and Berke 1997, Shah et al. 1998) and
mixture proportions on the cracking potential of concrete (Krauss et al 1995).

Grzybowski and Shah (1989, 1990) used a model to calculate the stresses that
could be expected to develop in a ring specimen (assuming the tangential stress
is uniform along the radius) and predicted the cracking potential using a damagebased model. Weiss et al. (Weiss 1999, Weiss et al. 2000) computed the nonlinear stress distribution that develops in the ring and used a non-linear fracture
mechanics model to predict cracking for various ring geometries.

Other researchers have used the ring geometry with a more active inner core.
Malhotra (1970) used a pressurized ring to determine tensile strength of
concrete, Kovler et al. (1993) used an inner ring made of prespex that expanded
on heating, while Weiss (1999) used a pressurized inner ring to monitor creep.

98
5.3. Using the Ring to Quantify Stress Development
Over the last few decades, the ring tests have been instrumented with strain
gages and used to quantify the stress development inside concrete. Figure 5.1
illustrates the geometry of the ring specimens that will be discussed in this
chapter and indicates the notation that will be used in the computations
described in this work. The inner radius of the steel ring is given as RIS, the outer
radius of the steel and the inner radius of the concrete ring are given as
ROS = RIC, and the outer radius of the concrete ring is given as ROC. In general,
strain is measured using four strain gages mounted on the inner surface of the
steel ring (i.e., at RIS) at midheight along the z-direction (however, a greater or
lesser number of strain gages can be used). The average strain measured on
the inner surface of the steel ring (ST) is used in the stress calculations.

ST-OUT

ST

R OS-O

UT

R OC

RIC

(ROC-RIC) - e

ROC
-INT

IC

ST

RIS
R

ST

ST
RIS

ROC

ST-INT

RIC

(ROC-RIC)

(a)

(b)

(c)

Figure 5.1 Ring geometry: a) standard ring, b) dual ring, and c) eccentric ring
geometry

Swamy and Stavrides (1970) used the strain gages to obtain an indication of the
magnitude of the stress that develops in the concrete (ESteel is the elastic
modulus of the steel ring):

99

Ave

= ST E Steel

(RIC RIS ) RIC2 + ROC

+
1
2

2(ROC RIC )
2 ROC

Eq. 5-1

Equation 5-1 was derived assuming a linear stress distribution throughout the
steel ring.

The authors, then, applied equilibrium concepts setting the

compressive force in the steel ring equal to the tensile force in the concrete ring.
This enabled determination of the average tensile stress in the concrete ring.

Attiogbe et al. (2001) and Weiss and Ferguson (2001) independently proposed
solutions that could use the strain measured in the steel ring to compute the
stresses in the concrete ring.

Attiogbe et al. (2001) considered a thin wall

approximation to determine the maximum stress in the concrete ring using


equation 5-2:

= ST E Steel

RIC (RIC RIS )


RIS (ROC RIC )

Eq. 5-2

Weiss and co-workers (Weiss and Ferguson 2001, Hossain and Weiss 2004)
performed a similar analysis using a thick-walled solution resulting in equation
5-3 for the maximum tensile stress (i.e., circumferential stress for r = RIC):

r = RIC

= ST E Steel

2
2
2
)
RIC
RIS2 (ROC
+ RIC
2
2
2
2 RIC (ROC RIC )

Eq. 5-3

Weiss and co-workers showed equations 5-4a and 5-4b as a general solution to
calculate the circumferential ( in equation 5a) and radial (r in equation 5b)
stresses in the concrete ring.

100

(r ) = ST E Steel

2
RIS2
RIC
2
2
RIC
2 ROC

r (r ) = ST E Steel

2
RIS2
RIC
2
2
R IC
2 ROC

2
ROC
1 + 2
r

Eq. 5-4a

2
ROC
1 2
r

Eq. 5-4b

Attiogbe et al. (2004) compared the results from equations 5-2 and 5-3 and
observed good agreement for thinner rings which are the geometries that are
commonly used. However, the results differ as ROC/RIC increases. Figure 5.2
shows a comparison of equations 5-1, 5-2, and 5-3, for specimens with
geometries similar to the ASTM test method; RIS = 152 mm, RIC = 165 mm, ROC =
203 mm, (it should be noted that equation 5-1 provides an average stress in the
concrete while equations 5-2 and 5-3 provide the maximum stress). Both the
AASHTO and ASTM rings have a similar RIS/RIC (0.92); however, the ROC/RIC is
greater for the AASHTO ring (1.5) than the ASTM ring (1.23).

It has been

suggested that, for most practical geometries, RIS/RIC should be approximately


0.8 to 0.95 since thicker steel rings have the problem of difficulty in resolving
strain measurements and thinner rings can bend along the z axis (Shah 2004).

10
Eq. 5-1 1
Equation
Eq. 5-1
Eq. 5-1 2
Equation
Equation
Eq. 5-1 3

0.1

RIS = 152 mm, RIC = 165 mm


0.01
1.00 1.25 1.50 1.75 2.00 2.25 2.50

Concrete Wall Thickness Ratio (ROC/RIC)

(a)

Normalized Stress in the


Concrete (- /(ST EST))

Normalized Stress in the


Concrete (- /(ST EST))

10

0.1

0.01
0.00

Eq. 5-1 1
Equation
Equation
Eq. 5-1 2
Equation
Eq. 5-1 3

RIC = 165 mm, ROC = 203 mm


0.25

0.50

0.75

1.00

Steel Wall Thickness Ratio (RIS/RIC)

(b)

Figure 5.2 A comparison of equations 5-1,5-2 and 5-3: a) varying concrete


wall thickness, and b) varying steel wall thickness

101
5.4. Additional Calculations to Quantify Ring Results

5.4.1. Quantifying the Effect of Stress Relaxation (Creep) using the Ring
In addition to computing the actual stress level that develops inside the concrete
ring (as described in section 5.3), procedures were developed to assess the
extent of creep or relaxation that occurs in the concrete ring. Hossain and Weiss
(2004) formulated an equation for the theoretical elastic stress that develops if
creep (or relaxation) is not considered (equation 5-5).

SH (t ) EC (t )
Elastic Max (t ) =

EC (R
E S (R

2
OC
2
OC

[(1 +

)RIS2 + (1 S )RIC2 ]

R )
2
+ R ) (1 C )R + (1 + C )ROC
+
R )
(R RIC2 )
2
IC
2
IC

(R

2
IC

2
IS

2
IC
2
OC

Eq. 5-5

where Elastic-Max is the theoretical maximum elastic stress, SH(t) is the free
shrinkage at time t, EC(t) is the age-dependant elastic modulus of concrete, and

C and S are the Poissons ratio of concrete and steel, respectively. The elastic
stress from equation 5-5 can be compared with the actual stress from equation 52 to determine the magnitude of stress relaxation experienced by the material as
shown in Figure 5.3 (after Hossain and Weiss 2004).

102
5

W/C = 0.5
19 mm Steel Thickness
75 mm Concrete Thickness

Elastic
Stress

Relaxation
Cracking

0.50%
Steel
Fiber
Volume

Stress (MPa)

Stress (MPa)

3
0.06%
Steel Fiber
Volume

2
1

Residual
Stress

Plain

10

12

14

14

21

Age of Specimen (Days)

Age of Specimen (Days)

(a)

(b)

28

Figure 5.3 a) The elastic stress and residual stress that develops in a concrete
ring and b) an illustration of results from restrained ring with concrete containing
fibers (Hossain and Weiss 2004)

5.4.2. Influence of Fiber Reinforcement


Shah and Weiss (Shah and Weiss 2006, Shah 2004) used the ring test to
evaluate fiber reinforced concretes. Figure 5.3b shows typical results from the
ring tests performed using fiber reinforced concrete. These results showed that
in low (standard redi-mix) volumes, fibers do not alter the stresses prior to
cracking; however, they can alter the behavior after the crack has formed,
resulting in a greater amount of stress transfer across the crack and smaller
crack widths. The authors developed equation 5-6 to estimate the width of crack
in the ring specimen (w).

AC
w = ( 2 RIC ) SH 1 ST
BC
ST

Eq. 5-6

where is the degree of restraint (described in section 5.5) and STAC is the
strain in the gages immediately after cracking and STBC is the strain in the gages

103
immediately before cracking. Shah and Weiss (Shah and Weiss 2006, Shah
2004) discussed that the relatively small circumference of the typical ring
specimen may underestimate the crack widths that can be expected in field
applications; thereby, overestimating the magnitude of stress transfer across the
crack in field applications.

5.4.3. Influence of Specimen Geometry and Bond


between the Concrete and Steel
Moon et al. (Moon et al. 2004, Moon 2006) used finite elements analysis to
assess the influence of specimen height (z direction), steel ring thickness, and
the appropriateness of the shrink fit solutions while Hossain (Hossain and Weiss
2006) studied these factors experimentally. Moon (Moon 2006) observed that
very thin steel rings (RIC-RIS = 3 mm) may be susceptible to bending along the
z-direction. However, more realistically sized steel rings (RIC-RIS of 9 mm) do
not show substantial deformations along the z-direction.

Moon (Moon 2006,

Moon et al. 2005) also demonstrated that the bond condition between the steel
and concrete need to be modeled properly. In both experiments and simulations,
only the transfer of compressive stresses is considered (i.e., no tensile or shear
stresses are transferred between the steel and concrete rings). In experiments,
the surface of the steel ring is typically coated with a form release agent or
a plastic separation sheet to provide this unbonded boundary condition.

The

results of the finite element simulation compared well with the closed form
analytical simulations (equations 5-3 and 5-4) (Moon 2006, Moon et al. 2004,
Hossain and Weiss 2006, Moon et al., 2005). Other bond conditions provide
additional restraint which does not represent the actual experimental conditions
properly.

104
5.4.4. Influence of Moisture Gradients
Weiss et al. (Weiss 1999, Hossain and Weiss 2004, Moon and Weiss 2006)
showed that the stress that develops in the ring depends on whether the ring
dries from the top and bottom surfaces or from the outer circumference. While
drying from the top and bottom can be described using equation 5-3, Moon and
Weiss (Moon et al. 2004, Moon and Weiss 2006) showed that in addition to the
restraint from the steel, residual stress can develop due to circumferential drying
since the concrete near the outer radius shrinks more rapidly than the inner
concrete.

The total stress development was expressed as the sum of the

restraint that comes from the inner steel ring and the differential shrinkage in the
concrete using equation 5-7.
2
2
RIS2 ROC
ROS
(r, ) = steel (t ) ES
1+ 2
2
2
ROS
2 ( ROC
)
r

Eq. 5-7

2
2
SH const Econ r + RIC

R r 2
( f ( ROC ) f ( RIC )) + f ( r ) f ( RIC ) erfc OC
2
2
r
ROC RIC

where SH-const is a shrinkage coefficient, f is a geometry function given in (Moon


and Weiss 2006), Econ is the elastic modulus of the concrete, erfc is the
complementary error function, and =2(Dt)0.5 when D is the moisture diffusion
coefficient of concrete.

The impact of external drying can be seen in the simulation shown in Figure 5.4.
When drying is initiated (low values of ) the stresses develop along the outer
circumference of the concrete ring.

Over time (as increases) the stresses

become more well distributed throughout the concrete until the stress distribution
matches the stresses predicted by equation 5-3 (when approaches a 100).

=0.002
=0.004
=0.008
=0.02
=0.04
=0.08
=0.2
=0.5
=100

Stress (MPa)

Inner
Surface
of the Conc. Ring
Steel
Ring

Stress Amplification Factor

105

Econ / Esteel =0.105

0.02

ROC/RIC = 1.25
ROC/RIC = 1.50
ROC/RIC = 2.00

1.6

1.4

1.2

1.0
0.5

-1
0

1.8

0.04

0.06

0.08

Distance from the outer surface (m)

0.6

0.7

0.8

0.9

1.0

(ROC-RIC) - e
(ROC-RIC)

(a)

(b)

Figure 5.4 a) Influence of moisture gradients on residual stress distribution and


b) the effect of eccentricity on maximum stress development

5.4.5. Degree of Restraint


The degree of restraint provided by the steel ring is important to consider in
interpreting test results. Hossain and Weiss (Hossain and Weiss 2004) provided
an expression for the degree of restraint:

U SH U S | ROS
U SH

100%

Eq. 5-8a

where USH is the displacement of the inner radius of concrete if the concrete
shrinks without the steel ring and US|ROS is the displacement of the outer surface
of the steel ring. Moon and Weiss (Moon and Weiss 2006, Moon 2006) rewrote
equation 5-8a in terms of free shrinkage of the concrete (SH) and the measured
steel strain (ST) (Equation 5-8b).

106

= 1

1 ST (t ) RIS2
2 (1 + S ) + (1 S ) 100%
2 SH (t ) ROS

Eq. 5-8b

However, the strain in the steel is only known after the test is performed. To
overcome this limitation Moon (Moon et al. 2005, Moon 2006) developed an
approach to estimate the degree of restraint before the test is performed so that
the geometry of the ring can be properly tailored to match a particular field
application of the concrete(EC is the effective elastic modulus of concrete):

E C'
= 100% 100%
E ST

Eq. 5-9

1
2
RIS2
ROC
1 e (1 + C ) 2
E C'
RIC
RIC

E
2
ROC

R IS2
ST

1 2 (1 + S ) 2
R IC
RIC

+ (1 C )

+ (1 S )

5.5. Recent Developments in the Restrained Ring Analysis

5.5.1. Dual Ring


One difficulty that can develop in using the restrained ring test is if the concrete
expands. This can occur when a shrinkage reducing admixture is used (Pease
et al. 2005, Pease et al. 2006) or when specific repair materials with expansive
agents are used (Barde et al. 2006). When the concrete ring expands, it can
come out of contact with the steel ring and as a result, stresses will not be
generated in the steel ring until the concrete shrinks back to its original position.
This difficulty can be overcome by using a dual ring approach with inner and
outer restraining rings. The inner ring would behave similarly as in the earlier
ring tests; however, the outer ring would be used to restrain and quantify the
expansion.

107
Using the approach advocated by Weiss and co-workers (Hossain and Weiss
2004, Moon and Weiss 2006), the strain that is measured on the inner steel ring
can be used to determine an internal fictitious pressure (pINT) that is exerted on
the inner steel ring by the concrete using equation 5-10a:

R2 R2
p INT = ST INT E ST INT IC 2 IS INT
RIC

Eq. 5-10a

A similar approach can be used to determine the fictitious pressure (pOUT) acting
between the outer steel ring and the concrete using equation 5-10b:

pOUT = ST OUT

2
2
ROS

OUT ROC
E ST OUT

2
ROC

Eq. 5-10b

The pressures obtained from the strain gages on the inner (ST-INT) and outer
(ST-OUT) rings can be inserted into equation 5-11 to determine the circumferential
stresses in the concrete ring.
2
2
2
2
2
2
R IC

R IC

ROC
R IC
ROC
ROC
1
1
+

+
p
OUT
2
2
2
2
2
2
2
2
2
2
ROC R IC
ROC R IC
ROC R IC r
ROC RIC r

(r ) = p INT

Eq. 5-11

Tests performed at Purdue (Pease and Weiss 2004, Sant et al. 2007a, Sant et al.
2007b) have used a dual ring apparatus with rings made of Invar (Invar was
chosen to minimize the effects of temperature change on the level of restraint
that is provided). If the temperature of the Invar and concrete are decreased, the
Invar will have negligible temperature-induced deformation while the concrete will
shrink. If the reduction in temperature that causes cracking is known, the residual
strength of concrete can be computed.

108
5.5.2. Effect of Variable Concrete Wall Thickness (i.e., Eccentricity)
Another potential area where errors can arise in the interpretation of the ring test
is if the wall thickness of the concrete ring is not uniform (Figure 5.1c). Equation
5-3 can be modified to form equation 5-12 which illustrates the amplification that
can be expected to occur due to the eccentricity of the inner ring using an
approach that follows the derivation of Jeffery (Jeffery 1921, Timoshenko and
Goodier 1987).

r = RIC

= ST E Steel

2
2
2
RIC
RIS2 (ROC
+ RIC
) K
2
2
2
2 RIC (ROC RIC )

Eq. 5-12

Where the amplification factor ( K ) accounts for the eccentric placement of the
inner steel ring (e is the eccentricity shown in Figure 5.4b) which can be
computed using equation 5-13. Figure 5.4b shows that smaller walled rings are
more sensitive to eccentricity.
2
2
2 ROC
R 2 RIC
(ROC2 + RIC2 2RIC e e 2 ) 1
K = OC

2
2
2
2
2
2
2
+ RIC
ROC
(ROC + RIC )(ROC RIC 2 RIC e e )

Eq. 5-13

5.6. Summary
This chapter reviewed developments on the restrained ring test. Although the
ring test has been used for over a century, it has only recently been used in a
more quantitative fashion.

Equations 5-1 to 5-3 enable the strain that is measured in the steel ring to
be used to compute the residual stress in the concrete.

A solution for the elastic stress has been developed (equation 5-5).

Subtracting the residual stress (equation 5-3) from the elastic stress
(equation 5-5) provides a measure of creep or relaxation.

109

The effects of fiber reinforcement can be quantified in terms of the


reduction in crack width that occurs and the stress that is transferred
across the crack (equation 5-6).

Solutions have been developed to account for moisture gradients


(equation 5-7).

Equations 5-8 and 5-9 can be used to quantify the degree of restraint
provided by a ring of a specific geometry.

110

CHAPTER 6. ASSESING THE REPEATIBILITY OF RESTRAINED SHRINKAGE


TEST USING ASTM C-1581

6.1. Introduction
The restrained ring test has recently been standardized by ASTM (ASTM C
1581-04) as a test method to assess the restrained shrinkage cracking
susceptibility of a concrete mixture. Unfortunately, there is currently a lack of
published information regarding the repeatability of the test.

This chapter

quantifies the variability of the ASTM C 1581-04 ring test procedure.

To

determine the repeatability of the standard test, twenty four restrained ring
specimens were cast from the same mortar mixture and stored in carefully
controlled environmental conditions. Results from the restrained ring test were
assessed using the same mixture proportions. The results indicate that a
standard deviation of approximately 6 can be expected in the strain
measurements from the same mixture and 8 from different batches of the
same mixture.

A probabilistic approach was used to describe the variability

associated with the age of cracking and to provide a tool that allows cracking
prediction, especially when only a fraction of the ring specimens crack.

6.2. Research Significance


Even though the restrained ring test has become a widely accepted experimental
method to assess cracking susceptibility of the cementitious systems, information
about the variability and repeatability of the test procedure is not widely available.
The objective of this work is to provide information regarding the variability within
a single test as well as the variability between repeated tests. This work will also

111
describe how the results of the restrained ring test can be interpreted, especially
when only some of the samples experienced cracking.

6.3. Experimental Equipment

6.3.1. Restrained Ring Geometry


The restrained shrinkage test was conducted in accordance with ASTM C 158104. The geometry of the ring specimen is presented in Figure 6.1, where RIS is
the inner radius of the steel ring, RIC is the outer radius of the steel ring and the
inner radius of the concrete ring (RIC = ROS), and ROC is the outer radius of the
concrete ring. The dimensions of the ring specimen are: 2RIS = 3053 mm, 2ROS
= 2RIC = 3303 mm, 2ROC = 4063 mm.

ST
ROC RIS

ST

ST

RIC
ST

Figure 6.1 Geometry of the ring test (description of the symbols in the text)

6.3.2. Ring Instrumentation


Each ring was instrumented with four CEA-series strain gages mounted at the
mid-height on the inner surface of the steel ring. The CEA gages were general-

112
purpose constantan gages with resistance equal to 120.00.3 , gage factor of
2.0950.5%, transverse sensitivity +0.3%0.2% at 24C, and a temperature
range: -75 to + 175C.

The steel ring was prepared by grinding and cleaning the inner surface of the
ring. A two-part epoxy (M-Bond AE-10) was used to glue the gages to the steel
surface.

The gages were covered with Teflon tape, butyl rubber sealant,

neoprene rubber, and secured with aluminum tape, following procedures


recommended by Vishay Measurements Group (Vishay Measurements Group
2002, Vishay Measurements Group 2005). A picture of a strain gage on the
inner surface of the steel ring is presented in the Figure 6.2a, while Figure 6.2b
shows the completed gages.

(a)

(b)

Figure 6.2 Detail of the strain gage: a) gage being placed and b) the completed
gage

6.3.3. Data Acquisition


The strain readings were recorded in five minute time intervals using a Vishay
System 5000 with Strain Smart software. A time of zero minutes in this work
refers to the time when water was added to the cement in the mixer.
collection

was

initiated

immediately after rings

were

placed

into

Data
the

113
environmental chamber, which took place approximately 60 minutes after the
mixing process began. The readings were collected for the first 24 hours. At an
age of 24 hours, data collection was stopped for approximately 30 minutes while
the rings were demolded. The readings were then continued, recording data at
five minute intervals until the end of the experiment.

Once the test was

completed, the data was interpreted as described in section 6.5.

It should be noted here that each strain gage was systematically connected to
the same port in the data logger and rings were placed in the same place in the
environmental chamber each time.

A complete statistical analysis was

performed to insure that the results were not skewed in any way by the gage, the
ring selected, the channel setup, or by the position in the chamber.

6.3.4. Environmental Chamber


A Darwin environmental chamber (series KB) was used for this study (Figure
6.3). The chamber was equipped with a microprocessor for temperature and
humidity control that allowed continuous control with an accuracy of 0.1C for
temperature and 1%RH for relative humidity conditions.

During testing the

temperature and RH were monitored through the chamber and were found to
vary less than 1C for the temperature and 1% for the relative humidity
conditions. The environmental chamber was used for sample storage during
testing, as well as for materials and molds conditioning before mixing.

114

Figure 6.3 Photo of the rings setup in a chamber and data acquisition
(Note: doors closed during the test)

6.4. Experimental Program

6.4.1. Mixing Proportions


Two mortar compositions were considered in this work: plain mortar with a waterto-cement ratio of 0.40 (w/c=0.40) and a mortar with a water-to-cement ratio of
0.40 where 5% of the water was replaced by shrinkage reducing admixture
(w/c=0.40+5% SRA). Type I ordinary portland cement (ASTM C 150-05) was
used.

The cement had a fineness of 367 m2/kg and an estimated Bogue

composition of a 61% C3S, 13% C2S, 10% C3A, loss on ignition of 1.75%,
insoluble residue 0.5%, and the total alkali content of 0.61%.

The mortar used in this work had 55% of aggregate by volume. The fine
aggregate used in all the mixtures was a local natural sand with a specific gravity
of 2.62, an absorption capacity of 1.87%, and a fineness modulus of 3.23.

115
6.4.2. Sample Preparation Procedure
For each mixture the aggregate was oven dried, sieved, and recombined to
ensure each mixture had a consistent gradation (as presented in Table 6-1).
Cement and water were weighed 24 hours prior to mixing, they were sealed and
placed in a room with controlled temperature of 232.0C. Mixing was performed
in accordance with ASTM C 192-05. First, the aggregate was loaded into the
pan mixer. Once aggregate was placed in the mixer, the mixer was started and
50% of the total water was added. Mixing was performed for 30 seconds and
then cement was added. After the addition of cement, the remaining 50% of
water was added and the materials were mixed for 20 seconds. For mixtures
containing shrinkage reducing admixture (SRA), this part of the water contained
the SRA. The mortar was mixed for 3 minutes, the mixer was stopped for 3
minutes (while the sides and the bottom of the mixer were scraped), and then the
mortar was mixed again for an additional 2 minutes.

Table 6-1 Sieve analysis of fine aggregate


Sieve
Number
3/8 in.
#4
#8
#16
#30
#50
#100

Sieve Size % Passing ASTM C 33- ASTM C 3303 MIN


03 MAX
9.50 mm
100
100
4.75 mm
99
95
100
2.36 mm
83
80
100
1.18 mm
59
50
85
600 m
33
25
60
300 m
2
5
30
150 m
0
0
10

In order to obtain information about the variability within a testing method, six
rings were prepared and monitored simultaneously from each batch of mortar.
The rings and the materials used to make the mortar were stored for 24 hours at
a constant temperature before the mortar was placed. Directly before casting the
rings were removed from the environmental chamber. Each ring was cast in two
layers and each layer was vibrated using external vibration. Immediately after

116
casting, all six rings were moved back into the environmental chamber. After
placing rings in the environmental chamber, the top surface of the samples was
covered with plastic to prevent evaporation. The strain gages were connected to
the data acquisition system and the strain readings were begun. Twenty four
hours after the water came in contact with the cement, the rings were demolded.
The outer steel ring was removed and the top of the concrete ring was sealed
with aluminum tape to enable moisture loss from the outer circumference only
(Figure 6.3). In order to eliminate variations in strain caused by the demolding
process, the Strain Smart Software was paused for the demolding procedure
and restarted right after the rings were covered with aluminum tape.

6.4.3. Testing Program


To characterize the repeatability within the testing procedure, the first series of
six rings was cast from a mortar with a w/c of 0.40. To obtain information about
variability between different batches, the same mixture was batched four times
and the restrained ring test (six rings in each test) was repeated four times,
resulting in a total of 24 rings of the same mixture. The influence of a shrinkage
reducing admixture (SRA) was considered in one additional series of rings (i.e.,
six rings) from the w/c = 0.40 mortar. The influence of variations in the relative
humidity conditions was analyzed by setting the humidity conditions to the
boundary values of the ASTM C 1581-04 recommendation, i.e. 46% and 54%.
The detailed description of the tests performed is presented in Table 6-2.

117
Table 6-2 Mixtures performed in the experimental process (for all mixtures
w/c=0.40)
Notation
0.4_A
0.4_B
0.4_C
0.4_D
0.4_5%SRA
0.4_46%
0.4_54%

SRA
0%
0%
0%
0%
5%
0%
0%

RH
50%
50%
50%
50%
50%
46%
54%

6.4.4. Temperature Compensation


A series of initial tests were conducted on the steel rings before the concrete was
cast around them.

In these tests, the temperature of the steel ring and

surrounding environment was varied and the strain response was recorded to
assess the effect of temperature on the recorded strains. The temperature of the
chamber was changed in 5.5C steps: 23.0C 28.5C 23.0C 17.5C
23.0C, as shown in Figure 6.4a. This enabled the test results from the mortar
specimens to be corrected for the slight change in temperature that occurs during
the first few hours.

30

10

28

26

24

22

20

-2

18

-4

16

-6

14

Strain
Ring Temperature
Air Temperature

-8

12

-10

10
0

24

48
72
Time [Hours]

96

120

Average Strain []

10

Average Temperature [oC]

Average Strain []

118

6
4
2
0
-2
-4
-6
-8

-10
15 17.5 20 22.5 25 27.5 30
Temperature [oC]

(a)

(b)

Figure 6.4 a) Average strain readings and temperature change and b)


temperature change and average strain readings

6.5. Experimental Results

6.5.1. Temperature
Figure 6.4b presents the relationship between the average strain measurements
on the steel ring and the temperature of the steel ring.

A linear equation

(Eq. 6-1) was fit to the data:

T = 1.75 T

Eq. 6-1

where T is strain [] and T is change in the temperature [C]. Although the


temperature effect had minimal influence on the measurements performed in this
work, it should be noted that the thermal effects could be further reduced if the
ring was manufactured from material with a very low coefficient of thermal
expansion (CTE) like Invar (Sant et al. 2007a, Sant et al. 2007b).

119
6.5.2. Variability within a Single Restrained Shrinkage Test and Repeatability
of the Test
For each test, six rings were prepared. The plain mixture was repeated four
times, keeping the mixture proportions and experimental procedure constant.
Typical results from each test series consisting of six rings can be seen in Figure
6.5 with dashed lines representing plain mixtures and solid lines representing
SRA mixtures. A slight rise in the strain is noticed from approximately 6 to 12
hours due to the heat of hydration resulting in an expansion of the concrete ring.
At 24 hours the rings were demolded by removing the outer steel mold, thereby
enabling the samples to lose moisture from the outer circumference. The test
was continued from that point until cracking was observed in all rings. The time
of cracking for the plain mixture (mixture A) occurred between 3.17 and 4.18
days, corresponding to a maximum strain development in the range of -47 to

Average Strain []

-64 with an average value of -55 (Figure 6.6).

20
10
0
-10

Demolding

-20
-30
-40
-50
-60
-70
-80

w/c = 0.4
w/c = 0.4, 5% SRA

24

48

72
96
120 144 168
Age of Specimen [Hours]

192

216

240

Figure 6.5 Comparison of the strain development as a function of time for the
plain mixture and the mixture with SRA (strain results for each cast ring)

120
20
10

20
Demolding

0
-10
-20
-30
-40
-50
-60

0
-10
-20
-30
-40
-50
w/c = 0.4, mix A
w/c = 0.4, mix B
w/c = 0.4, mix C
w/c = 0.4, mix D

-60
-70
-80
0

24
48
72
96
Age of Specimen [Hours]

(a)

Demolding

Average Strain []

Average Strain []

10

w/c = 0.4, 46%RH


w/c = 0.4, 50%RH
w/c = 0.4, 54%RH

-70
-80
120

24
48
72
96
Age of Specimen [Hours]

120

(b)

Figure 6.6 a) Variability between different tests performed on the same plain
mixture (only the average strain values for each mixture presented) and b) the
influence of humidity condition the samples were stored in on strain development
and the time of cracking

When the average strain for all four plain mixtures was compared (Figure 6.4), it
can be observed that the shape and slope of the curves is similar, but differences
in the time of cracking can be observed (i.e.: 3.60.4, 4.00.4, 3.50.4, and
3.20.4 days for mixtures A, B, C, and D, respectively). The average values of
strain at the time of cracking were: -55, -61, -63, and -56 for mixtures A, B, C,
and D, respectively. It should be noted however that the variability in the time of
cracking may be larger for cases where the cracking occurs at later time.

6.5.3. Influence of Shrinkage Reducing Admixture (SRA)


One of the most common uses of the restrained ring test is to compare different
mixtures. Figure 6.5 shows the comparison of the plain mixture with a similar
mixture containing shrinkage reducing admixture (SRA). It can be seen that for
the plain mixture cracking was observed between 3.2 and 4.2 days (76 and 100
hours), while the addition of 5% SRA delayed the time of cracking to the range
between 8.2 and 9.6 days (198 and 230 hours). The average magnitude of strain

121
at the time of cracking was similar for both types of mixtures: -59 for plain
(average for all four mixtures: A, B, C, and D) and -54 for mixture with SRA.

6.5.4. Influence of Relative Humidity Conditions


Figure 6.6b presents the influence of the relative humidity on the strain
development. The lower and upper range for the testing conditions (i.e., 46%
and 54%) were chosen as the lower and upper margin of the RH required in the
ASTM C 1581-04 (50%4%).

As can be seen, the lowest relative humidity

corresponds to the most rapid strain development and the earliest time of
cracking, however variability due to the RH variation (within ASTM limits) is lower
than the variability between the same mixture cast in different batches.

6.6. Discussion of the Results

6.6.1. Accuracy of the Test


Initially, the standard deviation was determined, as shown in Figure 6.7. The
standard deviation was calculated using the average strain for each of the six
rings in a single test (plain mortar, mixture B). As the rings crack, fewer values
are being compared. Similar analysis to the one presented in Figure 6.7 was
performed for all the plain mixtures (A, B, C, and D) as well for the mixture
containing SRA. For each of the analyzed cases, standard deviation in a single
testing series was less than 6 . For all four plain mortars and mortar with SRA
coefficient of variation was less than 12%.

122

Standard Deviation []

8
7

Demolding

1st ring cracked


2nd rd
3

4th

5th

6th

3
2
1
0
0

12

24

36

48

60

72

84

96

108

120

Age of Specimen [Hours]


Figure 6.7 Standard deviation of the measured strains (plain mortar mixture B)

6.6.2. Assessing Probability of Cracking


Deterministic data obtained in the tests preformed on plain mortars (six rings in
each test resulting in 24 data points) was used to create a cumulative plot of the
number of rings that had cracked by a specific time (Figure 6.8). Each step in
the cumulative curve corresponds to a cracking event in a ring sample.
Arranging data in a cumulative plot allows the probability of cracking to be
assessed at any given time. If characteristic probability of cracking curves can
be developed for different mixtures, susceptibility of cracking could be assessed
based on the deterministic data obtained from the experiments. This approach
will be further discussed in Section 6.7.2.

Cumulative Distribution of the


Percentage of Cracked Rings

123

100
90
80

Demolding

70
60
50
40
30
20
10
0
0

12

24

36

48

60

72

84

96

108

120

Age of Specimen [Hours]


Figure 6.8 Cumulative plot of cracking events for the plain mortar mixture

6.7. Explaining Variability in the Cracking Behavior of Restrained Concrete

6.7.1. Model Description


A previously developed model (Weiss 1997, Radlinska et al. 2007) was used to
simulate behavior of a restrained concrete element. The model compares the
time-dependant stress and strength development.

Whenever the developing

stresses exceed the strength of concrete in the model, cracking is predicted to


occur.

To predict the likelihood of cracking, a parameter called cracking

potential, CR(t), has been defined as the ratio of residual stress ((t)) and tensile
strength (ft(t)), as shown in equation 6-2:

CR (t ) =

(t )
f 't (t )

Eq. 6-2

The time-dependent residual stress is computed based on the total free


shrinkage (autogenous and drying) that is permitted to develop in the restrained

124
concrete element.

The mathematical model enables the time-dependant

modulus and viscoelastic properties of concrete to be considered using equation


6-3 that was earlier introduced in Chapter 3 (equation 3-2):

d ( )
1
1
+ (t , )
+ ( ) d
E ( ) Ec
d
0

Permit (t ) =

Eq. 6-3

where, Permit(t) is the total strain that is permitted to develop in the actual
restrained concrete, E() is the time dependent elastic modulus, EC is
a reference elastic modulus (i.e., a 28 day value), (t,) is the creep coefficient,
and () is the differential shrinkage with respect to time (t). Detailed description
of the modeling process can be found in Chapter 3 and Radlinska et al. (2007).

The magnitude of shrinkage was described using a shrinkage coefficient for the
paste. This coefficient was estimated to be approximately 3400 for the plain
mortar mixture and 2200 for mortar mixture with SRA. These values are
based on previous experimental studies.

The shrinkage coefficient (paste

property) can be related to the shrinkage of mortar or concrete using Picketts


equation (Pickett 1956):

Shr = N (1 VF )n

Eq. 6-4

where N is the shrinkage coefficient of the paste, VF is the volume fraction of the
aggregate, and n is a coefficient that describes the stiffness of the aggregate and
paste. The value of coefficient n is typically between 1.2 and 1.7 for normal
strength and normal weight concretes (in this work n was assumed to be 1.43
(Moon et al. 2005). As a result, a shrinkage coefficient of 3400 corresponds to

125
a long term concrete shrinkage of 580 (similar to a long term measurements
from ASTM C 157) with 68% of aggregate by volume.

The other material parameters used in the modeling process: elastic modulus,
compressive strength and splitting tensile strength were obtained from
experimental results (Shah 2004, Radlinska et al. 2008). The long-term elastic
modulus was 30 GPa, compressive strength was 40 MPa and splitting tensile
strength was 3.7 MPa.

Modeling parameters including the time-dependant

constants described in (Radlinska et al. 2007) were assumed to be the same for
the plain mixture and the mixture with SRA. The influence of SRA on shrinkage
behavior of the concrete element was modeled through the value of ultimate
shrinkage and its time dependant characteristic.

Complete restraint (i.e., 100% restraint) was assumed in this work. It should be
noted, however, that less than perfect restraint can also be considered. For the
case of less than perfect restraint, the Shr term used in equation 6-4 would not
correspond to the free shrinkage value measured in a test like ASTM C157. It
should rather be described by equation 6-5 (Moon et al. 2006) where the free
shrinkage that would be permitted in a concrete could be computed based on the
shrinkage from the perfectly restrained (DOR=100%) model:

free var iableDOR =

mod el 100% restra int

Eq. 6-5

(1 DOR)

where free-variableDOR is a permissible shrinkage with variable degree of restraint,

model-100%restraint is shrinkage in a model assuming 100% restraint, and DOR is


degree of restraint. A ring test has a DOR of 65-80% after 24 hrs for most typical
normal strength concrete.

126
In addition to the time-dependant calculation of stress and strength development,
the aforementioned model enables variation in material properties to be
considered by using Monte Carlo analysis. During Monte Carlo simulations,
material properties (E, SH, ft) are treated as random variables and the value
used in each simulation can be randomly sampled from predefined distributions.
The repeated sampling procedure allows calculation of stress and strength and
determination of the time of cracking together with the probability of cracking
occurrence (Radlinska and Weiss 2006a, Radlinska and Weiss 2006b). The
variability observed in material properties can originate from inherent variation in
the constituents or from the production process (concrete placement, changing
weather conditions, etc.) (Pellinen et al. 2005). In this work, information about
experimental variability was implemented in the model assuming 12% variability
in the shrinkage coefficient.

6.7.2. Results of Simulations


The results of the simulations are presented in Figure 6.9a. Even with the
relatively low variability in shrinkage (12%) there is substantial variation in the
predicted time of cracking. Figure 6.9a can be also useful in describing why all
the ring specimens cast from the same batch may not necessarily crack in
practice. The lognormal cumulative curve for plain mortar reaches the value of
1.0 approximately at 12 days and that means that all rings cast in that experiment
would be expected to crack before the age of 12 days. However, the cumulative
curve for the number of cracked rings for mortar with 5% SRA never reaches the
value of 1.0 (up to 365 days) and a plateau is reached at the value of
approximately 0.9 (90% probability of cracking). This means that it should be
expected that out of the series of rings that were cast not all the rings would
crack (i.e., only 90% would crack).

The number of cracked rings in a test

depends on the type of the mixture and environmental conditions the samples
were stored at.

127
Next, the results from experiments were compared with the results obtained in
the simulations. A lognormal distribution of experiments is plotted together with
the results of simulations for a plain mixture (for two different coefficients of
variation) in Figure 6.9b. The variability in material parameters will affect the
shape of the probability curve. Figure 6.9b shows that higher COV in material
properties results in a wider spread of cumulative density function. It can be
seen that a reasonable agreement was obtained between the shape of the
theoretical and experimental curves.

0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2

0.4_0%SRA_50%RH

0.1
0.0

0.4_5%SRA_50%RH

20 40 60 80 100 120 140


Age of Specimen [Days]

(a)

Probability of Cracking

Probability of Cracking

1.0

1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
experiment
simul. COV=6%
simul. COV=12%

0.2
0.1
0.0

0 1 2 3 4 5 6 7 8 9 10
Age of Specimen [Days]

(b)

Figure 6.9 Results of simulations: a) probability of cracking described with


cumulative lognormal curve for plain mortar and mortar with SRA and b)
comparison of experimental and simulation results

To better illustrate how the magnitude of shrinkage is related to the probability of


cracking, Figure 6.10a was developed.

It can be seen that if the shrinkage

coefficient is lower than 1700 (concrete shrinkage of 290 with 100% DOR),
the probability of cracking is relatively low (i.e., less than 10%). However, for
shrinkage higher than 2400 (concrete shrinkage of 410 , 100% restraint),
the probability of cracking exceeds 90%.

128
In the last step of the analysis, a series of simulations were performed to
investigate how the magnitude of shrinkage affects the probability of cracking.
Figure 6.10b shows that at high magnitudes of shrinkage the cumulative
distribution function (CDF) approaches a value of 1.0 at an early age.

This

implies that all the samples with high shrinkage would crack which appears
reasonable. However, if the magnitude of shrinkage is reduced (e.g., ultimate
shrinkage coefficient equal to 2300 or less; shrinkage of concrete 395 or
less), the ultimate value for CDF never reaches a value of 1.0 at 365 days. This
suggests that not all the samples will crack and as such gives insights into

1.0

1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
1000
1500
2000
2500
3000
Shrinkage Coefficient []

(a)

Probability of Cracking

Probability of Cracking

understanding why rings may or may not crack in a particular experiment.

0.9
0.8
0.7
0.6

Shrinkage
Coefficient

0.5
0.4

3500
2900
2300
2100
1950
1800

0.3
0.2
0.1
0.0
0

40 80 120 160 200 240


Age of Specimen [Hours]

(b)

Figure 6.10 a) Relationship between shrinkage coefficient and probability of


cracking b) cracking probability in time for mixtures with different shrinkage
coefficients (COV=12%, DOR=100%)

This approach can be further used to develop CDF functions for different
concrete mixtures.

This way, a tool can be obtained that will enable the

magnitude of shrinkage of any given mixture to be related to the probability of


a cracking event.

129
6.8. Summary
The results presented in this work can be summarized as follows:

In a properly performed restrained ring test following ASTM C1581-04, the


variability between strain readings within a single test results in
a coefficient of variation of less than 12%.

The standard deviation of

measured strain at any time within a single test was less than 6 and
less than 8 between different tests,

The time of cracking can vary between the rings cast from the same
batch. It should also be noted that for some mixtures with moderately low
shrinkage only a certain percentage of specimens may crack, even in
a properly conducted test,

The probability of early-age shrinkage cracking in concrete relates to both,


the average properties (e.g., the magnitude of shrinkage) and variability in
the material properties. If shrinkage can be lowered to a sufficiently low
level and variability can be controlled, the risk of cracking can be
substantially reduced.

It should be noted, however, that the risk of

cracking is not linearly related to shrinkage,

Modeling tools were presented that can be used to describe probability of


cracking.

The tools show, as one may expect, that when all other

parameters are held constant, increasing the magnitude of shrinkage


decreases the average age of cracking, and increases the probability that
cracking will occur at all.

130

CHAPTER 7. ANOVA ANALYSIS OF THE RESULTS OF THE RESTRAINED


RING TEST

7.1. Introduction
This chapter provides information about the statistical analysis of test results
obtained from the ring test (ASTM C 1581-04). Analysis of variance (ANOVA)
was performed to analyze a series of four tests performed on the same mixture:
mortar with water-to-cement ratio of 0.40 and 55% aggregate content by volume
(Mixtures A, B, C, and D). As mentioned in Chapter 6, all the steps in material
preparation and mixture procedure were kept constant to ensure that all four
tests were performed in the same manner and their results can be treated as
representative for that type of the mixture.

7.2. Testing Methodology


The experiment was performed in accordance with ASTM C 1581-04 and
detailed description of the laboratory procedures has been presented in
Chapter 6. Each test consisted of six identical molds as presented in Figure 7.1.
Rings were always cast in the same order and stored in the same place in the
environmental chamber. The environmental conditions were kept constant for
each of the four tests: temperature of 23.01C and relative humidity of 501%.

131

Figure 7.1 Six molds prepared for the ring test

7.3. Single ANOVA Model


Four series of tests on plain mortar were performed and all the results were
collected. In each test six rings were cast and each ring had four strain gages
connected to the data logger. The average of four gage readings was taken for
each ring and the data set shown in Table 7-1 was obtained. In the analyzed
model there were four tests and each test consisted of six rings. There were
6 levels of ring factor in the model (i = 1 to 6) and 4 observations for each level
(j = 1 to 4).

132
Table 7-1 ANOVA model
i (factor)
Ring No
1
2
3
4
5
6

Test #1
Strain []
-57.99
-47.38
-55.75
-54.68
-64.02
-51.26

j (observations)
Test #2
Test #3
Strain []
Strain []
-62.34
-63.43
-64.12
-59.71
-55.90
-61.57
-64.60
-62.68
-62.66
-67.02
-58.46
-61.86

Test #4
Strain []
-63.54
-58.73
-48.04
-58.68
-57.58
-47.69

The test result show that all the rings cracked when the average strain values on
the steel rings were between -67.02 and -47.38 . In order to analyze if the
difference between the mean values of the strain at the time of cracking for each
test is statistically significant, a one-way ANOVA was performed (Kutner et al.
2004). Analysis of variance (ANOVA) is a collection of statistical models that
allows evaluating whether all the means from more than two populations are
equal. The ANOVA model can be written as:
Yij = i + ij

(Eq. 7-1)

where i is the theoretical mean or expected value of all observations at level i


and the ij is the random error term that is independent and identically distributed
with mean 0 and variance 2, N(0, 2). Yij is approximately normally distributed
with mean i and variance 2, N(i, 2), and independent. The ANOVA model
assumes that the response variable observations are normally distributed, with
a mean that may depend on the level of factor and the variance that does not
depend on the level of factor.

Additionally it assumes that observations are

independent. For all the analysis in this research, the significance level was kept
constant, = 0.05, and confidence level (1-) was equal to 95%, as presented in
Figure 7.2.

133

1
/2

Figure 7.2 Illustration of significance level and confidence level

7.3.1. Confirmation of Applicability of ANOVA


All the computations were conducted using commercial software SAS and all the
codes used for the analysis have been included in Appendix A. Initially, data was
graphically analyzed and a plot of strain versus the test number was obtained.
Figure 7.3 indicates that test 1 and 4 had more scatter than test 2 and 3. The
average strain at the time of cracking has been presented in the Figure 7.4 and it
can be observed from the plot that sample 1 and 4 had lower average strains
when compared to sample 2 and 3.

134

-40

Strain []

-45
-50
-55
-60
-65
-70
1

Test Number
Figure 7.3 Strains at the time of cracking in each test

Average Strain []

-40
-45
-50
-55
-60
-65
-70
1

Test Number
Figure 7.4 Average strains at the time of cracking in each test

135
For the analysis of variance to be valid, the data has to approximately follow
normal distribution.

As such, null hypothesis (the claim about one or more

population characteristic that is initially assumed to be truth, Devore 2001) was


stated that that values are normally distributed. The assumption of normality was
verified performing a Shapiro-Wilk test (Kutner et al. 2004).

The p-value

obtained in the analysis was 0.9719. The p-value is the probability of obtaining
the value of the test statistic at least as extreme as the one that was actually
observed, provided that the null hypothesis is true. Since the p-value of 0.9719
was greater than significance level ( = 0.05), the null hypothesis (that values
are normally distributed) was not rejected and assumption about normality was
assumed to be valid (Devore 2001). Next, homogeneity of variance was verified
conducting a modified Levene test (Kutner et al. 2004). The modified Levene
test examines the null hypothesis of variances being the same (H0: 12 = 22 =
32 = 42) versus the alternate hypothesis that at least one variance is different.
It can be reminded here that the square root of variance is equal to standard
deviation. The test results obtained from the modified Levene test are given in
terms of the F statistic. An F-test is any statistical test in which the test statistic
has an F-distribution if the null hypothesis is true. In this case, an F-statistic of
2.18 for 3 and 20 degrees of freedom was obtained, resulting in a p-value of
0.1218. As such, the null hypothesis that variances are equal cannot be rejected
and assumptions for ANOVA model are valid.
Next, a graphical tool called quantile-quantile3 plot (QQ plot) was used to verify if
the distribution differs from normal.

QQ plot compares ordered values of

a variable with quantiles of a specific theoretical distribution. If the data are from
the theoretical distribution, the points on the QQ plot lie approximately on
a straight line.

As can be seen in the Figure 7.5, QQ plot confirms that the

distribution is approximately normal.

Quantiles are points taken at regular intervals from the cumulative distribution function of
a random variable

136
12

Residuals []

8
4
0
-4
-8
-12
-2

-1

Normal Quantiles
Figure 7.5 A quantile-quantile plot

Additionally, it was verified if the residuals4 of strain readings are independent of


time. In order to obtain this information, residuals versus sequence number were
plotted. Figure 7.6 shows that the plot of residuals versus sequence number has
a random pattern. That confirms that residuals of strain readings do not show
time dependency.

Finally, Figure 7.7 presents the plot of predicted value of strain versus residuals.
No clear pattern can be found and this indicates that there is no problem with
homogeneity of variance.

A residual, or fitting error, is the vertical deviation from the estimated fit (Devore 2001)

137
12

Residuals []

8
4
0
-4
-8
-12
0

10

15

20

25

Sequence Number
Figure 7.6 Plot of residuals versus sequence number

12

Residuals []

8
4
0
-4
-8
-12
-64

-62

-60

-58

-56

-54

Predicted value of strain []


Figure 7.7 Plot of residuals versus predicted value of strain

138
7.3.2. GLM Procedure
In this step of analysis, the differences in the means of four different samples
were analyzed. Qualitative data obtained in the experiment was examined to
obtain means and standard deviations for the four test samples, as presented in
Table 7-2. The 95% confidence interval bounds obtained are presented in Table
7-3 and graphically illustrated in Figure 7.8. The GLM (General Linear Model)
procedure was utilized to statistically test the data. GLM uses the method of
least squares to fit general linear models.

Among the statistical methods

available in GLM procedure are regression, analysis of variance, analysis of


covariance, multivariate analysis of variance, and partial correlation.

Table 7-2 Mean values and standard deviations for each test
Level of Test

Mean []

1
2
3
4

6
6
6
6

-55.18
-61.35
-62.71
-55.71

Std Dev
[]
5.70
3.44
2.46
6.42

Table 7-3 95% confidence intervals


Test

1
4
2
3

6
6
6
6

Mean
[]
-55.18
-55.71
-61.35
-62.71

95% Confidence
Limits []
-59.25 -51.12
-59.78 -51.64
-65.42 -57.27
-66.78 -58.64

139

Test Number

4
3
2
1
0
-40

-45

-50

-55

-60

-65

-70

Strains [

Figure 7.8 95% confidence intervals

Table 7-3 presents that means of strains at the time of cracking for the six
different rings are similar. To find out if the differences in the means between
four different tests are statistically significant, ANOVA test was performed. Null
and alternative hypothesis were stated as follows:
H0: 1 = 2= 3= 4 (all four means are same)

(Eq. 7-2)

HA: at least one of the is is different

(Eq. 7-3)

The test results from performing one-way ANOVA have been summarized in
Table 7-4 to 8-7.

140
Table 7-4 One-way ANOVA results sum of squares
Source

DF

Model
Error
Corrected
Total

3
20
23

Sum of
Squares
266.45
457.69
724.14

Mean
Squares
88.82
22.89

F
Value
3.88

Pr>F
0.0245
-

Table 7-5 One-way ANOVA results R-square


R-Square

Coeff Var

Root MSE

0.38

-8.14

4.78

strain
Mean
-58.74

Table 7-6 One-way ANOVA results Type I sum of squares


Source
Test

DF
3

Type I SS
266.45

Mean Square F Value


88.82
3.88

Pr > F
0.0245

Table 7-7 One-way ANOVA results Type III sum of squares


Source
Test

DF
3

Type III SS
266.45

Mean Square F Value


88.82
3.88

Pr > F
0.0245

By analyzing results from the ANOVA test, it can be seen that the F-statistic of
the model was 3.88 for 3 and 20 degrees of freedom and p-value was equal to
0.0245. This suggests that at alpha level = 0.05 the null hypothesis has to be
rejected and there is statistical evidence that at least one of the means of the
strain at the time of cracking is different from the others.

The analysis presented in this chapter consisted of four tests and Bonferroni
correction was employed to statistically adjust for multiple comparisons. The
Bonferroni correction is a procedure that develops joint confidence intervals
assuming the statistical significance level that should be used for each
hypothesis separately is 1/n times what it would be if only one hypothesis were

141
tested (Kutner et al. 2004).

It means that when two hypotheses are tested,

instead of a p-value of 0.05, more conservative value of 0.025 should be used.


The results obtained after the Bonferroni correction was introduced have been
presented in Table 7-8 and . It can be noticed that new simultaneous confidence
intervals obtained are wider then the one presented in the Table 7-3.

Table 7-8 Bonferroni confidence intervals for strain at the time of cracking
Test

1
4
2
3

6
6
6
6

Mean
[]
-55.18
-55.71
-61.35
-62.71

Simultaneous 95%
Confidence Interval []
-60.54
-49.82
-61.07
-50.35
-66.70
-55.99
-68.07
-57.35

In the next step, means of the strain values were compared using Tukey
procedure (Kutner et al. 2004) to determine if they are significantly different.
Tukey multiple comparison procedure performs pairwise comparisons of factor
level means, which can be written as:
H0: i = i

(Eq. 7-4)

HA: i i

(Eq. 7-5)

If the means are categorized to the same group by SAS software, the same letter
from alphabet is being assigned. As presented in Table 7-9 it can be seen that
all the means belong to the same group. Contrary to ANOVA test results, the
output of the Tukey test suggests that there is no significant difference between
the means.

Additionally, detailed comparison between tests was performed

(Table 7-10) using CLDIFF command. Again, it has been shown that the means
of the four tests are not significantly different, as comparisons significant at the
0.05 level would be indicated by the three stars sign (***). This may suggest that
the analyzed results are near the border of null hypothesis acceptation/rejection.

142
Table 7-9 Tukey grouping
Mean
A
A
A
A
A
A
A

N
Test
-55.181 6 1
-55.710 6

-61.347 6

-62.711 6

Since the Tukey and ANOVA test gave different results, Analysis of combination
of means of two or more group was additionally carried out using a contrast
statement in SAS. This option allows verifying if differences between four test
results are statistically significant.

Table 7-10 Test comparison using CLDIFF command


Test
Difference Between
Comparisons
Means
14
0.530
1-2
6.166
1-3
7.530
4-1
-0.530
4-2
5.636
4-3
7.000
2-1
-6.166
2-4
-5.636
2-3
1.364
3-1
-7.530
3-4
-7.000
3-2
-1.364

95% Confidence Limits


-7.201
-1.564
-0.200
-8.260
-2.094
-0.730
-13.896
-13.367
-6.366
-15.261
-14.731
-9.095

8.260
13.896
15.261
7.201
13.367
14.731
1.564
2.094
9.095
0.200
0.730
6.366

Since the difference between the means of a group is at most 6.6 (difference
between test 1&4 and 2&3), it can be concluded that the means of four samples
are less than 7 and this is an acceptable number in an engineering
understanding.

143
Table 7-11 Results of contrast statement
Contrast
1&2 v 3&4
1&3 v 2&4
1&4 v 2&3
1 v 2&3&4
1&2&3 v 4

Estimate
0.95
-0.42
6.58
4.74
-4.036

F value
0.24
0.05
11.36
4.42
3.20

t Value
0.48
-0.21
3.37
2.10
-1.79

P value
0.6330
0.8330
0.0030
0.0483
0.0887

7.4. Random-Effect Model


The results of four tests presented here can be treated as a sample from a larger
population and as such it is reasonable to apply a random effect model. In the
random model, the four test results can be treated as a randomly selected
representation from the bigger population. A random effect AVOVA model used,
in general can be written as:
Yij = Mi + ij

(Eq. 7-6)

where Yij is a response variable, Mi is the theoretical mean or expected value of


the strain at the time of cracking (random variable) in a given test and the ij are
normally distributed random error terms.
The result of the random model show that variance within the test was 11 2.
This was twice as small as the variance due to the error term, 23 2, which
suggests that restrained ring test can be considered a test with high repeatability
and accuracy.

7.5. Conclusions
The analysis of the four restrained ring tests showed that the restrained ring can
be accepted as a repeatable experimental procedure.

144
In order to statistically analyze the data, an ANOVA I fixed model was used first.
It was confirmed through Shapiro-Wilk test that the average strains at the time of
cracking can be treated as normally distributed. The results from ANOVA I table
suggested that the hypothesis about all four average means (of strains at the
time of cracking) being the same should be rejected, but Tukey test did not
detect any significant differences between the test results. ANOVA II random
model showed that the variance for the test was (11 2) is twice as small as the
variance due to error (23 2). This can be qualified as an acceptable variance in
engineering measurements.

145

CHAPTER 8. IMPORTANCE OF BOUNDARY AND DRYING CONDITIONS IN


CEMENTITIOUS SYSTEMS

8.1. Introduction
Chapters 4 and 5 discussed an approach to assess the probability of cracking in
restrained concrete elements.

It was shown that cracking potential can be

minimized if the magnitude of shrinkage can be lowered. Shrinkage reduction


(ACI 231, 2007) can be achieved through the use of higher aggregate volumes
(Shilstone 1990) use of expansive additives (ACI 223, 1998), or use of modified
cement binders. Recently, two innovative shrinkage mitigation strategies have
been developed:
1) the use of chemical admixtures (e.g., shrinkage reducing admixtures, SRA)
that alter the surface tension of concretes pore fluid (Weiss and Berke
2002, Rajabipour et al. 2008) and
2) the use of water saturated porous inclusions (e.g., lightweight aggregate,
LWA) that supply additional water during hydration and can be used for
internal curing (Bentz et al. 2007, Lura 2006, Mack 2006, Henkensiefken
et al. 2008a, Henkensiefken et al. 2008b).

While both of these methods show great potential, to obtain the complete
anticipated benefit of either SRA or LWA, the boundary conditions of the
concrete element must be carefully considered and understood. This chapter
demonstrates the difference in the shrinkage behavior of a sealed concrete that
undergoes self-desiccation and an unsealed concrete that experiences external
drying in addition to the self-desiccation. The performance of SRA and LWA
depends significantly on concretes boundary conditions (i.e., sealed vs.

146
unsealed) and this must be considered when selecting a shrinkage mitigation
strategy.
The objectives of this chapter are:

To provide experimental measurements of the shrinkage performance of


a low w/c plain mixture, a mixture containing SRA, and a mixture
containing saturated LWA. These experiments include the measurement
of drying shrinkage, weight change, autogenous shrinkage, restrained
shrinkage, and internal relative humidity;

To discuss the implications of boundary conditions on shrinkage mitigation


strategies using experimental measurements and theoretical concepts.
This will enable practitioners to determine the most appropriate strategy
for a particular application.

The chapter starts with presenting the results of experimental measurements.


These are followed by a theoretical discussion of sealed and unsealed shrinkage
behavior and how inclusion of SRA or LWA can influence the shrinkage for each
boundary condition.

8.2. Experimental Measurements for Sealed and Unsealed Systems

8.2.1. Materials
Three mixtures were prepared with a water to cement ratio (w/c) of 0.30: a plain
mortar (w/c=0.30), a mortar with 5% of the mix water replaced by a shrinkage
reducing admixture; i.e., 2.2 gal SRA/yd3 (w/c=0.30+5%SRA), and a mortar with
approximately 40% volume of sand replaced by saturated lightweight aggregate
(w/c=0.30+LWA). The cement content was the same for all the mixtures. In all
three mixtures aggregate content was 55% by volume. Natural river sand with
fineness modulus 3.13 was used. The LWA was rotary kilned expanded shale
(manufactured) with a fineness modulus of 3.10. The 24 hour absorption of LWA
was measured as 10.5% according to ASTM C128-04.

147
For all mixtures ASTM C150-05 Type I ordinary portland cement (OPC) was
used with a Blaine fineness of 370 m2/kg and a Bogue phase composition of 56%
C3S, 16% C2S, 12% C3A, 7% C4AF and a Na2O equivalent of 0.68%. A high
range water reducer (Glenium 3000 NS) was added at 6.3 fl. oz. per 100 lb of
cement for the w/c=0.30 and w/c=0.30+5%SRA mixtures, and at the rate of 9.4 fl.
oz. per 100 lb of cement for mortar with lightweight aggregate (w/c=0.30+LWA).
The shrinkage reducing admixture was a liquid admixture (TetraguardAS 20).

8.2.2. Mixing
A mixing procedure was used that was in accordance with ASTM C192-05. The
normal weight aggregate (natural sand) was incorporated into the mixture after
oven-drying and subsequent cooling. The lightweight aggregate (LWA) was first
oven dried, air cooled, and then submerged in water for 241 hours before
mixing. The volume of water used to submerge the LWA was equal to the sum
of the volumes of mixing water and water that would be absorbed by the
aggregate. The water level remained above the top surface of the LWA during
submersion; however, before mixing the excess water was decanted and used
as the mixing water. As such, all of the mixtures should have a very similar w/c
and paste porosity.

8.2.3. Experimental Procedures

8.2.3.1. Measurements of Volumetric Change during the First 24 Hours


The linear autogenous (i.e., sealed) deformation of all the mortar specimens was
measured during the first 24 hours using the corrugated tube protocol (Jensen
and Hansen 1995, Sant et al. 2006). The corrugated tube protocol involves the
encapsulation of the fresh mortar (approximately 30 minutes after water is added
to the mixture) in a corrugated polyethylene tube. The tube has a length-to-

148
diameter ratio of 400 mm to 30 mm and a higher stiffness in the radial direction
than the longitudinal direction.

This allows the transformation of volumetric

deformations into linear deformations. Two specimens were prepared for each
mixture. The specimens were then placed in a dilatometer which was maintained
at 23.01.0C.

The dilatometer was equipped with LVDT displacement

transducers, with a measuring accuracy of 5 m/m.

The transducers were

interfaced to a PC through Labview to record length measurements every


5 minutes.

8.2.3.2. Measurements of Unrestrained Volumetric Change after 24 Hours


Free shrinkage was monitored using 7575285 mm prisms according to ASTM
C157-04 with a slight modification on the area exposed to drying. Three samples
were prepared from each mixture for each testing condition. The sealed samples
were covered with aluminum tape on all sides. The unsealed samples were
exposed to a relative humidity of 50% and had all sides sealed, except two
diametrically opposed faces (75285 mm). Free shrinkage was measured using
a comparator. Measurements were begun at 24 hours using the strain measured
from corrugated tubes (see section 8.2.3.1) as the initial value.

8.2.3.3. Restrained Shrinkage Measurements


The cracking potential was measured using restrained ring test, following ASTM
C1581-04. Six samples were prepared and monitored simultaneously for each
testing condition.

Before casting, molds were placed in an environmental

chamber and were conditioned for at least 4 hours at a temperature of


23.01.0C. Each sample was cast in two layers with each layer being vibrated.
After casting, all samples were kept sealed for 24 hours at constant temperature
of 23.01.0C before they were demolded. Ring specimens that were exposed
to 50% RH after demolding were sealed on the top surface with aluminum tape to
limit moisture loss from the outer circumference only (bottom face was sealed by

149
the rings support).

Ring specimens that were sealed after demolding were

covered on the top and the outer circumference with two layers of aluminum
tape. Restrained shrinkage was acquired using the Strain Smart software and
data was collected in 5 minute intervals.

8.2.3.4. Internal Relative Humidity Measurements


To measure change in the internal relative humidity over time, fresh cement
paste was cast in air-tight plastic vials. At an age of 12 hours, approximately
10 g of cement paste was crushed and placed in a glass vial (with diameter
25.4 mm and length 50.8 mm) and sealed to prevent moisture loss. The sealed
vial was then kept in an environmental chamber at 23.00.1C to maintain
thermal equilibrium. A Digitron 2080R temperature and relative humidity meter
equipped with a Pt-100 temperature sensor and a capacitive RH sensor was
used to record relative humidity and temperature data inside the sealed vial at 2
hour intervals for a period of 7 days. Before and after the measurement period
(i.e., from 12 hours to 7 days), calibration of the sensor was carried out using
saturated salt solutions (K2SO4, KNO3, KCl, and NaCl) with a known relative
humidity in the range 75-100% (ASTM E104-02).

8.3. Experimental Results


The following section describes the results of the tests for sealed and unsealed
systems. For the sealed system, the measurements included the internal relative
humidity,

unrestrained

(i.e.,

free)

autogenous

deformation,

unrestrained

shrinkage, and cracking potential from restrained ring test. For the unsealed
system the measurements included weight (mass) loss, unrestrained total
shrinkage, and the cracking potential from the restrained ring test.

150
8.3.1.1. Sealed Systems
Figure 8.1 shows the average behavior of two samples from each mixture under
the sealed condition. Figure 8.1a shows that the cement paste containing SRA
maintained an internal relative humidity approximately 1% higher than the plain
cement paste at 7 days. This is consistent with observations of Bentz et al.
(2001) and Sant et al. (2007). The cement paste containing LWA maintained an
internal relative humidity that was approximately 10% higher than the plain paste
at 7 days. This is consistent with observations by Geiker et al. (2004).

Figure 8.1b shows the unrestrained length change of each mixture. The data in
Figure 8.1b was determined by combining the corrugated tube measurements
from the first 24 hours with the ASTM C157-04 test results, so that the total
autogenous deformation measured from the time of set could be presented. The
SRA mixture shows shrinkage that was 150 lower than that of plain mixture at
14 days. The LWA mixtures show lower shrinkage (by more than 300 ) than
a plain system at 14 days. For all mixtures weight loss was less than 0.06%,
which means it is essentially negligible, as one would expect for the sealed
system.

Restrained shrinkage measurements (obtained using the restrained ring test)


shown in Figure 8.1c appear to be consistent with free shrinkage and RH
measurements.

The mixture with LWA maintains a higher internal relative

humidity and does not shrink significantly in a sealed condition. As such, the
rings with the LWA show little strain. This would imply a very low cracking
potential. Similar observations have been made by Jensen and Hansen (2001).
The addition of 5% SRA maintains a slightly higher RH, reduces autogenous
shrinkage (Sant et al. 2007) and significantly delays the time of cracking
compared to the plain mixture.

It should be noted that in case of mortar

containing 5% SRA, the test was terminated at 22 days and cracking was
observed only for 4 out of 6 rings when the test was stopped.

100

100

98

96

-100
Strain ()

Relative Humidity (%)

151

94
92
90

-200
-300
-400

88

Sealed

Sealed
0.3_LWA
0.3_5%SRA
0.3_Plain

86
84
0

0.3_LWA
0.3_5%SRA
0.3_Plain

-500
-600

3
4
5
Time (Days)

(a)

6
8
10
Time (Days)

12

14

(b)

Average Strain []

20
0
-20
-40
Sealed
0.3_LWA
0.3_5%SRA
0.3_Plain

-60
-80
0

* Test terminated before


all the rings cracked

8
10
12
14
16
Age of Specimen [Days]

18

20

22

(c)
Figure 8.1 Comparison of sealed specimens behavior a) RH measurements,
b) free shrinkage measurements (zeroed at the time of set) c) restrained
shrinkage measurements

8.3.1.2. Unsealed Systems


The results of measurements performed on unsealed systems (i.e. total
shrinkage consisting of autogenous and drying component) have been
summarized in Figure 8.2. As presented in Figure 8.2a, weight loss is higher for
mortar with lightweight aggregate than plain and SRA mixtures. This can be

152
explained by the fact that mortar samples with LWA have higher amount of water
that can leave the system. The weight change for SRA mortar is slightly lower at
14 days. It should be noted however, that at higher relative humidities (80% RH)
the SRAs loose more water. Previous research (Weiss and Berke 2002) has
shown that the final mass loss of SRA mixture is comparable to the plain mixture
at 50% RH but not at higher relative humidities.

Shrinkage of mortar prisms exposed to an ambient relative humidity of 50%


presents trend similar to those observed in autogenous deformation: the lowest
shrinkage is observed in case of the LWA mortar, while the highest one can be
noticed for a plain system (Figure 8.2b). It should be noted however that these
mixtures are not in equilibrium.

The restrained shrinkage measurements (Figure 8.2c) show that cracking is


observed at later ages for SRA and LWA mixtures than for plain mortar. It can
be also noticed that not only the times of cracking are different, but the
magnitude of strains developed is different as well. For LWA mortar, cracking is
observed at much lower magnitude of strain than the plain mixture. This is due
to the tensile strain capacity of the LWA mixture. The rate of strain development
is reduced in case of the LWA mixture due to a combination of the reduction in
shrinkage and reduction in elastic stiffness.

0.2

100

0
Average Strain ()

Weight Change (%)

153

-0.2
-0.4
-0.6
-0.8

Unsealed
0.3_LWA
0.3_5%SRA
0.3_Plain

-1

-100
-200
-300
-400

Unsealed
0.3_LWA
0.3_5%SRA
0.3_Plain

-500
-600

-1.2
0

6
8
10
Time (Days)

12

14

6
8
10
Time (Days)

(a)

12

14

(b)

Average Strain []

20
0
-20
-40
Unsealed
0.3_LWA
0.3_5%SRA
0.3_Plain

-60
-80
0

5
6
7
8
9 10
Age of Specimen [Days]

11

12

13

14

(c)

Figure 8.2 Comparison of unsealed specimens behavior (at 50% RH) a) weight
loss, b) free shrinkage measurements (zeroed at the time of set) c) restrained
shrinkage measurements

8.4. Theoretical Considerations for Shrinkage in Concrete


One of the main driving forces behind shrinkage in cement paste is capillary
tension. Drying (internal or external) causes the formation of curved liquid-vapor
interfaces (menisci) inside the pores of cement paste.

The menisci cause

154
generation of a negative pressure inside the pore fluid (the liquid phase) which is
commonly known as capillary tension (Pcap). Shrinkage of cement paste can be
described as a function of this capillary tension. An example of this approach is
illustrated by Mackenzie equation (Mackenzie 1950) which was later modified by
Bentz et al. (1998). Equation 9-1 shows the shrinkage of paste (p) as a function
of capillary tension:

p =

1
S
1
Pcap
3
K Ks

Eq. 8-1

where Pcap (Pa) is the capillary tension in the fluid, S (unitless) is the degree of
saturation of cement paste, K (Pa) is the pastes bulk modulus, and Ks (Pa) is the
modulus of the solid skeleton inside cement paste.

8.4.1. Internal Drying (Self-Desiccation)


The hydration reactions that occur between cement and water result in
a reduction in volume of approximately 8% to 9% (Jensen and Hansen 2001a,
Jensen and Hansen 2001b). This means that the volume of hydration products
is smaller than the volume of the reactants. This volume reduction is generally
referred to as chemical shrinkage. Chemical shrinkage is of little concern with
respect to cracking before concrete sets, as the fluid nature of the system allows
the paste to collapse on itself. After setting however, the structure can no longer
collapse completely, resulting in the generation of vapor-filled voids (Couch
2006). This is shown schematically in Figure 8.3a.

155

Sealed

Drying

(a) Sealed

Drying

(b) External Drying Only


solid

liquid

(c) Unsealed
vapor

Figure 8.3 Illustration of drying mechanisms in sealed and unsealed systems: (a)
sealed only internal drying, (b) only external drying, (c) unsealed internal plus
external drying

The formation of this vapor filled space is referred to as self-desiccation (or


internal drying). Self-desiccation occurs in every cement paste regardless of the
water to cement ratio (w/c). Self-desiccation is particularly problematic in the
case of low w/c mixtures, especially when finer powders that modify porosity
(e.g., silica fume) are added. In low w/c mixtures, the size of these vapor filled
voids (i.e., this will be equal to the radius of curvature of the liquid-vapor
interface) would be small which results in a significant shrinkage-inducing
capillary tension (Pcap).

Section 9.4.3 provides more details on how the

magnitude of capillary tension that develops in the pore fluid is inversely


proportional to the curvature radius of the liquid-vapor menisci.

8.4.2. External Drying


For an unsealed specimen which loses moisture externally, liquid-vapor menisci
also form; however in this case they form at the surface of concrete. This is
schematically shown in Figure 8.3b and Figure 8.3c.

Figure 8.3b shows

156
a hypothetical case in which only external drying occurs (i.e., internal drying is
not considered). Figure 8.3c shows a more realistic picture of a cement paste
that is unsealed at one surface. In this case, while external drying occurs at the
exposed surface, self-desiccation develops in the interior of the material. This
suggests that while a sealed specimen only experiences internal drying (Figure
8.3a), an unsealed system would simultaneously undergo both internal and
external drying (Figure 8.3c).

The formation of liquid-vapor menisci corresponds with a reduction in the internal


relative humidity of cement paste.

In an unsealed specimen, as moisture

evaporates from the surface, the radius of the menisci continues to decrease.
This causes the drying front to penetrate into smaller openings and to gradually
recede from the surface towards the interior of the material. The internal relative
humidity is proportional to the menisci radius (as discussed later in equation 9-5)
and as such, decreases continuously as drying progresses. Drying will continue
until the internal relative humidity of cement paste reaches equilibrium with the
ambient relative humidity.

8.4.3. Relationships between Capillary Pressure, Pore Size, and Relative


Humidity
It is well known that the formation of the liquid-vapor menisci inside cement paste
leads to the generation of a capillary pressure (Pcap) inside the materials liquid
phase (i.e., pore solution).

The menisci curvature radius is related to the

capillary pressure generated according to the Laplace equation (Adamson and


Gast 1997):

Pcap =

2 cos( )
r

Eq. 8-2

157
where Pcap (Pa) is the capillary pressure, (N/m) is the surface tension of pore
fluid, (rad) is the liquid-solid contact angle (assumed to be 0 rad), and r (m) is
the radius of curvature of the meniscus. The negative sign indicates that the
pressure is negative (i.e., the liquid is in tension). This capillary tension pulls the
pore walls (solid surfaces) together and causes the volume change we observe
as shrinkage. It should be noted that, at equilibrium, the liquid pressure must be
the same throughout the system, indicating that the curvature radius of all
menisci would be the same.

The capillary pressure can also be related to the relative humidity of cement
paste using Kelvins equation (Adamson and Gast 1997):

Pcap =

RT ln( RH )
Vm

Eq. 8-3

where R is the universal gas constant (8.314 J/molK), T (K) is the temperature,
RH (unitless) is the internal relative humidity, and Vm ( 1810-6 m3/mol) is the
molar volume of pore solution. Equations 8-2 and 8-3 can be combined yielding
the Kelvin-Laplace equation, which correlates the menisci radius and the internal
relative humidity:

2 cos( ) Vm
RH = exp

r
RT

Eq. 8-4

This equation is graphically illustrated in Figure 8.4a which shows the menisci
radius as a function of the internal relative humidity for a fluid in two different
cement systems (one that would be similar to a plain paste with = 0.072 N/m
and the other similar to a paste containing SRA with = 0.036 N/m). According
to equation 8-4, an internal relative humidity of RH = 100% corresponds with an

158
infinite menisci radius (r = or flat menisci)5. As drying progresses, the menisci
radius becomes smaller and the internal relative humidity decreases from 100%.

Relative Humidity (%)

100

r1

90

r2

r1

80
70

Sealed - Plain

r3

60

Sealed - SRA

r4

Sealed - LWA

r3

= 72 x 10 N/m
(Plain)
(Plain
and LWA)
= 36 x 10-3 N/m
(~5% SRA)
-3

50
40
30
0.1

1.0

10.0

100.0 1000.0

Kelvin Radius (nm)

Unsealed - Plain

Unsealed - SRA

Unsealed - LWA

Figure 8.4 Illustration of drying mechanisms in sealed and unsealed systems: (a)
sealed only internal drying, (b) only external drying, (c) unsealed internal plus
external drying

Another important point that must be noted is that the menisci radius is closely
related to the size of the water-filled capillary pores that empty (i.e., dry out). In
order for a liquid-vapor meniscus to penetrate inside (or empty) a water-filled
pore, the meniscus radius must be smaller than or equal to the pore radius. This
can be seen in Figure 8.4b which illustrates the menisci formation in sealed
versus unsealed systems and in materials containing SRA and LWA. In the
sealed-plain system which is shown in the upper left corner of Figure 8.4b, the
menisci radius is r1 which is equal to the radius of the largest pore. As such, only
the largest pore would empty while the smaller pores remain saturated. Only
after the largest pore has emptied and the menisci radius is reduced, the next
largest pore can start to become evacuated.

Note this does not account for the presence of ions in pore solution. The influence of ions can
be accounted for using Raoults law, but since this is a relatively minor correction, it will not be
done here

159
In the following section, equations 8-1 to 8-4 are used to describe the shrinkage
behavior of plain, SRA, and LWA systems for both sealed and unsealed
boundary conditions. Before presenting such discussions, however, it should be
emphasized that these equations describe the fraction of shrinkage induced by
capillary tension alone. While at high relative humidities (approximately above
60% for plain paste and 80% for pastes containing 5% SRA (Weiss et al. 2007))
capillary tension is the main driving force behind shrinkage (Weiss et al. 2007,
Hansen 1987) at lower humidities, other shrinkage mechanisms such as the
surface energy of the solid phase (Hansen 1987) and the movement of interlayer
water (Feldman 1968) can become significant. The discussions of this chapter,
however, are limited to the influence of capillary tension.

At this point the fundamental features of the SRA and LWA should be mentioned.
Figure 8.5a shows the surface tension () of water-SRA solution as a function of
SRA concentration for four commercially available SRAs (type T1 was used in
this work). It can be seen that the SRA drastically reduces the surface tension of
the solution. Figure 8.5b illustrates the most salient feature of LWA, a desorption
isotherm. To obtain this plot, the LWA was dried at different relative humidities
until equilibrium was reached. It can be seen that a majority of water is lost at
high relative humidity (RH>96%) implying that the pores are larger than those in
the cement paste. The samples are nearly completely dry by 80% RH.

160
100
SRA T1
SRA E1
SRA E2
SRA E3

0.07
0.06

% of 24 Hours Absorbed
Moisture Remaining

Surface Tension (N/m)

0.08

0.05
0.04
0.03
0.02

80
60
40
20
0

3
6
9
12
SRA Concentration (%)

15

50

60
70
80
90
Relative Humidity (%)

(a)

100

(b)

Figure 8.5 a) Surface tension as a function of SRA concentration (Pease 2005)


b) desorption characteristic of lightweight aggregate (Henkensiefken et al. 2008)

8.5. Interpretation of Experimental Results Based on the Theory


The influence of boundary conditions (i.e., sealed or unsealed) is discussed for
plain mixtures, mixtures containing SRA, and mixtures containing saturated LWA.
A summary of the theoretical shrinkage predictions and a comparison on the
shrinkage behavior between the plain, SRA, and LWA concretes are provided at
the end of this section in Table 8-1.

8.5.1. Sealed (Autogenous) Shrinkage


In a sealed specimen, vapor-filled voids form in the interior of cement paste due
to self-desiccation (Figure 8.3a). Initially, these vapor-filled voids occupy the
largest spaces available (i.e., largest water-filled capillary pores) while smaller
pores remain saturated. As hydration continues, water is consumed by chemical
reactions and as a result, smaller and smaller water-filled pores are evacuated.
Since the menisci radius is gradually decreasing (remember that menisci radius
is equal to the radius of the smallest pore evacuated), the capillary tension (Pcap)
and shrinkage strain (p) are increasing with continuous hydration (equations 8-1

161
and 8-2). Meanwhile, the internal relative humidity (RH) of the paste is expected
to decrease as the paste hydrates (equation 9-4).

This was observed

experimentally in Figure 8.1a. Since water is not lost to the surroundings, the
mass loss would be zero.
Shrinkage reducing admixtures are known to reduce the surface tension () of the
pore fluid (Figure 8.5a). In a sealed paste containing SRA, vapor-filled voids
form in the interior of the paste in a similar fashion as they form in a plain cement
paste.

In fact, it can be argued that the menisci radius (r) would not be

influenced by the presence of SRA (Figure 8.4b) since the volume of the
generated vapor-filled voids and their curvature radius are controlled by the
hydration reactions (i.e., they are dependent on the degree of hydration which
has been measured to be similar in plain and SRA pastes (Sant et al. 2007)). As
the paste hydrates, the capillary tension (Pcap) continues to increase (equation
8-2); however, since the surface tension () is reduced by the addition of SRA,
the magnitudes of capillary tension (Pcap) and shrinkage strain (p) would be
smaller in the paste containing SRA than in the plain paste (see Figure 8.1b).
This would also result in a smaller reduction in the internal relative humidity as
the paste containing SRA self-desiccates (equation 8-4). This is confirmed by
experimental results shown in Figure 8.1a.

Saturated lightweight aggregates (LWA) also reduce the autogenous shrinkage


of cement paste; however this occurs for a slightly different reason.

LWA

contains water-filled pores (several m in size) that are much larger than the
capillary pores inside cement paste (mostly smaller than 1 m). When the paste
begins to self-desiccate, the largest pores (which are located inside LWA) empty
first. Only after all the large pores in LWA have emptied, the capillary pores
inside cement paste can start to dry out (Figure 8.4b). As a result, the inclusion
of saturated LWA can effectively increase the radius of the liquid-vapor menisci
(r2 > r1) as the material self-desiccates. Consequently, both the capillary tension

162
and the shrinkage strain would be reduced significantly in comparison to plain
cement paste.

This effect can be seen in Figure 8.1a for internal relative

humidity and Figure 8.1b for shrinkage strain.

8.5.2. Unsealed (Autogenous + Drying) Shrinkage


In an unsealed specimen, in addition to internal drying (i.e., self-desiccation),
moisture is lost to the environment at the exposed surface of the specimen. As
a result, liquid-vapor menisci form both in the materials interior and at its
exposed surface.

This can be seen in Figure 8.3c in which an unsealed

specimen experiences simultaneously both internal drying (Figure 8.3a) and


external drying (Figure 3b). As a result, the total specimen shrinkage can be
considered as the sum of autogenous and external drying shrinkage.

It must be noted, however, that these drying mechanisms (i.e., internal versus
external drying) are not completely independent and in fact, they interact with
one another. This can be realized by comparing the relative humidity at the
surface with the humidity of the internal vapor-filled voids.

As the system

approaches equilibrium, the internal relative humidity in all pores (near the
surface and at the interior) must equilibrate with surrounding environment and
become equal. As such, external drying not only causes the menisci formation at
the surface of the specimen, but also controls the humidity and the curvature of
the internal vapor-filled voids. In Figure 8.3c, exposure to external drying causes
a reduction in the menisci radii of the internal vapor-filled voids and causes these
voids to grow larger by penetrating further into the surrounding water-filled
openings.

The ultimate shrinkage of the unsealed specimen (p) is a function of the


developed capillary tension (Pcap) and the degree of saturation (S) of the cement
paste (equation 8-1). At equilibrium, the capillary tension can be calculated using
equation 8-3 by considering RH to be equal to the ambient relative humidity.

163
Equation 8-3 presents an important argument as it suggests that at equilibrium,
the capillary tension in unsealed specimens is independent of the surface tension
of pore fluid (). This means that addition of SRA would not alter the capillary
tension (Pcap) that develops in an unsealed concrete. Note that this does not
negate equation 8-2; as in the case of an unsealed concrete containing SRA, the
menisci radius (r) would decrease proportionally with decreasing pore fluids
surface tension (). This is further discussed in the following paragraph. As
such, the net effect on the capillary tension (Pcap) is negligible.
Since the ambient relative humidity (RH) is the same for plain and SRA mortars,
a reduction in surface tension () would result in smaller menisci radius (r) based
on the Kelvin-Laplace equation (equation 8-4). By rewriting equation 8-4 in terms
of r, it can be seen that for constant RH, the menisci radius (r) is a linear function
of the pore fluids surface tension ():

r = 2 cos( )

Vm
RT ln( RH )

Eq. 8-5

This is graphically shown by Figure 8.4a in which at an ambient RH = 75%, the


menisci radius decreases from 3.6 nm for the plain cement paste to 1.8 nm for
the paste containing SRA (note that 3.6nm/1.8nm = 7210-3/3610-3 N/m).

A smaller menisci radius suggests that smaller pores would evacuate in the
paste containing SRA (Figure 8.4b; r4 < r3). As such, when dried to equilibrium,
in comparison to the plain paste, the paste containing SRA should exhibit
a slightly higher mass loss when exposed to the same ambient humidity. This
was in fact observed in previous work (Weiss et al. 2007), indicating that a higher
mass loss is measured at equilibrium in paste containing SRA at high ambient
humidities. Note that the mass loss data shown in Figure 8.2b for plain and SRA
mortars does not correspond to an equilibrium condition. Previous experiments

164
(Sato et al. 1983, Weiss et al. 2007) show that the mass loss from mixtures with
SRA surpasses the mass loss of the plain mixture at higher RHs (~80%), though
the weight loss is similar at lower relative humidities (~50%).

The shrinkage reduction for the SRA mortar in comparison with the plain mortar
(Figure 8.2b) can be attributed to two factors.

First, the presence of SRA

reduces the rate of water evaporation from the mortar. This is confirmed by
a slower mass loss of the SRA mortar in Figure 8.2a. As a result of this lower
drying, the rate of shrinkage would be lower. The second factor contributing to
the lower shrinkage of the SRA mortar is that SRA can reduce the ultimate
shrinkage of the material.

According to equation 8-1, although the capillary

tension (Pcap) is expected to be similar in both plain and SRA cement pastes, by
reducing the degree of saturation, SRA can reduce the ultimate shrinkage.

Also, it has been suggested (Weiss et al. 2007) that while capillary tension can
cause significant shrinkage in plain mixtures in the relative humidity range
60%~100%, in the SRA mixtures, curved liquid-vapor menisci cannot form below
approximately 80% RH. As such, while the plain mixture continues to shrink at
relative humidities below 80%, shrinkage in the SRA mixture would be
significantly reduced below 80% RH due to the absence of the curved menisci
(Adamson and Gast 1997).

The presence of LWA in an unsealed specimen provides extra moisture that


would be available for evaporation. As water evaporates from the surface of
specimen, the moisture inside the LWA starts to evaporate keeping the smaller
capillary pores of the cement paste saturated.

As the specimen dries, the

internal relative humidity decreases; however, it is expected that the relative


humidity inside LWA concrete decreases at a slower rate, since the LWA pores
act as water reservoirs and prevent rapid RH reduction. This means that the rate
of shrinkage would be smaller in LWA concrete. In short, it takes longer for LWA

165
concrete to come to equilibrium with its ambient. This is consistent with the
results of Figure 8.2b showing less shrinkage of the LWA mortar compared to the
plain mixture.

While the specimens presented in Figure 8.2b have not come to equilibrium with
their ambient, the theory suggests that when the equilibrium is reached, the
magnitude of capillary pressure (Pcap) which is a function of the ambient humidity
(equation 9-3) is not expected to be influenced by the presence of LWA. At
equilibrium, the ultimate shrinkage strain of LWA concrete (c) should be
comparable with the plain concrete, excluding the effect of LWA on the concrete
stiffness.

In summary, for unsealed specimens, inclusion of SRA or LWA can reduce


shrinkage and potential of cracking through two different mechanisms. While
SRA reduces both the ultimate shrinkage (by limiting menisci formation) and the
rate of shrinkage (by reducing evaporation rate), LWA reduces shrinkage rate (by
providing internal water reservoirs), however, the ultimate shrinkage is expected
to be comparable with plain mixtures.

166

Table 8-1: Change in the shrinkage parameters in SRA and LWA concretes in
comparison with plain concrete
comparable with plain concrete
lower than plain concrete
higher than plain concrete
Autogenous Shrinkage
SRA concrete
Volume of water

consumed by hydration
Menisci radius

Capillary pressure
Internal RH
Free shrinkage of paste
Mass loss (evaporation)

LWA concrete

Drying Shrinkage6
SRA concrete

LWA concrete

Internal RH

RHambient

RHambient

Ultimate free shrinkage


of paste
Ultimate mass loss
(evaporation)
Rate of shrinkage prior to
equilibrium

Saturation of paste
Menisci radius
Capillary pressure

Parameters reported for drying shrinkage correspond to the equilibrium state

167
8.6. Conclusions
This chapter presented detailed discussion of the shrinkage of cementitious
materials in a sealed and an unsealed condition.

Two shrinkage mitigation

strategies, shrinkage reducing admixtures (SRA) and lightweight aggregate


(LWA) were described and the difference between their mechanisms underlined.
A summary of the theoretical predictions of the change in shrinkage parameters
of SRA and LWA concretes are provided in Table 8-1.

It can be seen that in sealed system, the SRA reduces shrinkage by reducing
capillary pressure through surface tension reduction; while the LWA reduces
shrinkage through providing reservoirs which provide additional water and
increase the size of the menisci radius. Both the SRA and LWA mixtures have
higher internal relative humidity (1% and 10%, respectively).

For the unsealed case, at equilibrium, more water is lost from the SRA system
than the plain system at the same environmental conditions. At lower relative
humidity, less shrinkage is observed in SRA systems since the capillary stress
breaks down at approximately 80% RH.

The unsealed LWA system loses

a large volume of water resulting in lower rate of shrinkage, although long-term


shrinkage is similar. Both SRA and LWA mixtures show reduction in cracking
potential.

168

CHAPTER 9. SUMMARY AND CONCLUSIONS

A new approach to predict shrinkage cracking in concrete is presented in this


work. This approach considers the inherent variability associated with material
properties, ambient conditions, and construction operations in the calculations of
the risk of shrinkage cracking in concrete elements.

Lowering the cracking

susceptibility in concrete has significant economical benefits, as it allows the


concrete to have improved durability, longer service life, and lower life-cycle
costs as fewer repairs and maintenance operations are needed. Additionally,
crack-free concrete has a positive environmental impact since improved
durability leads to less consumption of materials for repair and rehabilitation and
reduces the energy and carbon foot print of the infrastructure.

To assess the risk of cracking caused by volume changes, a Monte Carlo


method was adopted.

Through a series of stochastic simulations, material

variability was incorporated into shrinkage and stress development calculations,


thereby enabling the probability of cracking to be assessed at any given age of
concrete, as well as the overall likelihood of cracking. To quantify variability in
material properties, a broad experimental program was conducted to measure
the rate, magnitude, and variability in shrinkage development of cement paste
and concrete at various boundary and ambient conditions. It was found that
inherent material variability exists in the material properties and the ranges of
expected variability in shrinkage measurements have been quantified in this
work.

169
The Monte Carlo simulations were found to be a useful tool to describe cracking
probability; however, the method required time-consuming computations and
could only be applied to relatively simply geometries.

To account for this

limitation, a second approach was used (Load and Resistance Factor Design) as
an effective alternative tool.

A reliability-based model was developed which

allows the development of a new mixture design procedure where concrete is no


longer designed based on the required strength solely, but rather it can consider
how to specify shrinkage to minimize probability of cracking.

This research also provided an assessment of the accuracy and repeatability of


the restrained ring shrinkage test (ASTM C-1581). A brief history and recent
developments in the restrained ring test method were introduced. Then, the
variability and repeatability of the test method were evaluated using the analysis
of variance.

The results confirmed that the restrained ring test is a reliable

technique to access cracking susceptibility of cementitious systems. Next, the


Monte Carlo and LRFD modeling approach was applied to compare the results of
the experimental work with the numerical computations and a good agreement
was observed.

Finally, the restrained ring test together with other testing

procedures (including free shrinkage and relative humidity measurements) was


used to evaluate different shrinkage mitigation strategies in cement-based
materials. The effectiveness of shrinkage reducing admixtures (SRA) and lightweight aggregates (LWA) were compared.

The analysis included drying in

sealed and unsealed conditions and provided understanding of the fundamental


differences in the shrinkage mitigation mechanisms of shrinkage reducing
admixtures and light-weight aggregates.

170
9.1. General Summary Related to the Assessment of Magnitude and Variability in
Shrinkage Measurements
The following observations were made by investigating the magnitude and
variability in shrinkage of cement paste and concrete:

Consistent with the observations by others, shrinkage reducing admixtures


(SRA) have been found to reduce the magnitude and rate of shrinkage in
cementitious systems, however higher effectiveness of SRA has been
observed at lower humidities,

The addition of SRA does not significantly alter variability in shrinkage and
weight change measurements,

The average variability in shrinkage measurements of cement paste was


found to be 165 ,

The maximum standard deviation observed in shrinkage measurement of


concrete in this work was found to be 58 .

Very low variability exists in the aggregate content between different


concrete samples from the same laboratory batch (0.3%).

Hysteretic curves obtained after shrinkage and rewetting cycles allow the
amount of irreversible shrinkage to be determined. More shrinkage can
be recovered for the cement paste with SRA than in case of plain cement
paste.

9.2. General Conclusions Related to Reliability-Based Assessment of Probability


of Cracking in Concrete
The following observations were made during reliability-based analysis of earlyage cracking:

The Monte Carlo method shows that even low variability in material
properties results in a substantial scatter in the predicted time of cracking.

171
As such, information about the variability should be included in cracking
prediction computations to accurately describe materials behavior,

Monte Carlo method is a time-consuming technique, but alternatively Load


and Resistance Factor design can be implemented as a simplified and
computationally more effective procedure to predict cracking in concrete,

The addition of SRA decreases the magnitude of shrinkage and as such,


reduces the cracking potential.

This information was used to develop

a new mixture design approach, which allows the amount of SRA needed
to reduce the risk of cracking in concrete to be determined.

9.3. General Conclusions Related to the Accuracy of the Restrained Ring Test

The repeatability of the restrained ring test (ASTM C 1581-04) is evaluated in


this work and the results show that the test method is an accurate and reliable
testing procedure:

The variability between the strain readings within a single test


has coefficient of variation less than 12%,

The standard deviation of strain measurements within a single test


performed on mortar with w/c=0.40 was found to be less than 6 ,

The standard deviation of strain measurements between different tests


performed on mortars with w/c=0.40 was found to be less than 8 ,

Analysis of variance using random-effect model shows that the variance


for the test is very low (11 2), being twice as small as the error term in
the model,

The time of cracking can vary between the rings cast from the same
batch. For some mixtures with moderately low shrinkage only a certain
percentage of specimens may crack, even in a properly conducted test.

172
9.4. Summary of the Studies related to the Importance of Boundary Conditions in
Cementitious Systems
Shrinkage mechanism depends on the boundary condition and is different in
sealed and unsealed systems. As such, different shrinkage mitigation techniques
(SRA, LWA, etc.) have different effectiveness. In the sealed systems, the SRA
reduces shrinkage by reducing capillary pressure through surface tension
reduction, while LWA reduces shrinkage through providing additional water
reservoirs. In the unsealed case, SRA causes shrinkage reduction due to the
reduction in the rate of water evaporation and due to the reduction of the ultimate
shrinkage of material.

LIST OF REFERENCES

173

LIST OF REFERENCES

AASHTO PP 34-99 Standard practice for estimating the crack tendency of


concrete, 2004.
ACI Strategic Development Council Meeting, San Diego, 2007
Adamson, A.W., and Gast, A.P., (1997) Physical Chemistry of Surfaces, 6th Ed.,
Wiley-Interscience, New York.
Ai, H., and Young, J.F., (1997) Mechanism of shrinkage reduction using
a chemical admixture, Proceedings of the 10th International Congress on
the Chemistry of Cement, Vol.3, ed. Justnes, H., Gothenburg, Sweden.
Almudaiheem, J.A., and Hansen, W., (1987) 'Effect of specimen size and shape
on drying shrinkage of concrete', ACI Materials Journal Vol.84 (2) 130135.
Altoubat, A.S., and Lange, D.A., (2002) Grip-specimen interaction in uniaxial
restrained test, SP-206 Concrete: Material Science to Application, A
Tribute to Surendra P. Shah 580.
Annual Book of ASTM Standards (2005) Concrete and Aggregate, Vol. 4.02.
Attiogbe, E.K., See, H.T., and Miltenberger, M.A., (2001) Tensile creep in
restrained shrinkage, Creep, Shrinkage and Durability Mechanics of
Concrete and other Quasi-Brittle Materials, F.J. Ulm, Z.P. Bazant and F.H.
Wittmann (eds.), Elsevier Science, 651-656.
Attiogbe, E.K., Weiss, W.J., and See, H.T., (2004) A look at the stress rate
versus time of cracking relationship observed in the restrained ring test,
Advances in Concrete through Science and Engineering, Northwestern
University, Evanston, IL.
Barde, V., Radlinska, A., Cohen, M., and Weiss, J., (2007) Relating material
properties to exposure conditions for predicting service life of concrete
bridge decks in Indiana, Joint Transportation Research Program Report,
p.202.

174
Barde, A., (2004) Early age flexural behavior of cementitious systems and
factors affecting maturity based predictions, MS Thesis, Purdue
University, West Lafayette, IN.
Barde, A., Parameswaran, S., Chariton, T., Weiss, J., Cohen. M., and Newbolds,
S., (2006) Evaluation of rapid setting cement-based materials for patching
and repair', Report Submitted to the Joint Transportation Research Board,
April 2006.
Bazant, Z.P., (1989) Mathematical modelling of creep and shrinkage of
concrete, John Wiley & Sons, Hoboken, NJ.
Bentur, A., (2002) Chapter 6.5: Early age cracking in cementitious systems,
RILEM State of the Art Report TC-EAS, Ed. A. Bentur.
Bentz, D.P., (2005) 'Curing with shrinkage-reducing admixtures', Concrete
International, Vol.27, (10) 55-60.
Bentz, D.P., Geiker, M.R. Hansen, K.K., (2001) 'Shrinkage-reducing admixtures
and early-age desiccation in cement pastes and mortars', Cement and
Concrete Research 31(7), 1075-1085.
Bentz, D.P., Jensen, O.M., (2004) 'Mitigation strategies for autogenous shrinkage
cracking', Cement and Concrete Composites 26, 677-685.
Bentz, D.P., Garboczi, E.J., and Quenard, D.A., (1998) 'Modeling drying
shrinkage in reconstructed porous materials: application to porous vycor
glass', Modeling and Simulation in Materials Science and Engineering,
Vol. 6, pp. 211-236.
Bentz, D.P., Koenders, E.A.B., Mnnig, S., Reinhardt, H.-W., van Breugel, K.,
and Ye, G., (2007) 'Materials science-based models in support of internal
water curing', RILEM state-of-the-art report.
Bjontegaard, O., (1999) Thermal dilation and autogenous deformation as driving
forces to self-induced stresses in high performance concrete, Ph.D.
Dissertation, NTNU, Trondheim.
Breugel, van, K. and Lokhorst, S.J., (2001) Stress-based crack criterion as
a basis for prevention of through-cracks in concrete structures at early
ages, RILEM Proceedings 23, International RILEM Conference on Early
Age Cracking in Cementitious Systems - EAC'01, Ed. K. Kovler and A.
Bentur.

175
Brewer, H.W. and Burrows, R.W., (1946) Coarse ground cement makes more
durable concrete, ACI Journal, 353.
Brown, M.D., Smith, C.A. ,Sellers, G., Folliard, K.J., and Breen, J.E., (2007) 'Use
of alternative materials to reduce shrinkage cracking in bridge decks', ACI
Materials Journal Vol.104 (6) 629-637.
Carlson, R.W., (1988) Model of studying shrinkage cracking in concrete building
walls, ACI Structures Journal 85 (4) 395-404.
Carlson, R.W., (1937) Drying shrinkage of large concrete members, Journal of
ACI, Vol. 33, January, 327-336.
Carlson, R.W., (1942) Cracking of concrete, Boston Society of Civil Engineers
29 (2) 98-109.
Couch, J., Lura, P., Jenson, O., and Weiss, W.J., (2006) 'Use of acoustic
emission to detect cavitation and solidification (time zero) in cement
pastes', Volume Changes of Hardening Concrete: Testing and Mitigation,
Lyngby, Denmark, ed. Jensen, O., Lura, P. and Kovler, K., 393-400.
Coutinho, A.S., (1959) The influence of the type of cement on its cracking
tendency, RILEM Bulletin 5, 26-40.
Cusson, D., Hoogeveen, T., (2006) 'An experimental approach for the analysis of
early-age behaviour of high-performance concrete structures under
restrained shrinkage', Cement and Concrete Research 37, 200-209.
Day, K., (1999) Concrete mix design, quality control and specification, 2nd Edn.
E & FN Spon, London.
Devore, J.L., (2001) 'Probability and statistics for engineering and the sciences',
5th Ed, Duxbury Press, Pacific Grove, CA.
Dougals, C.T., and McHenry, D., (1947) Long time study of cement performance
in concrete tests on 28 cements used in the Green Mountain Dam, US
Bureau of Reclamation Report 345.
Dullien, F.A.L., (1979) 'Porous media; fluid transport and pore structure',
Academic Press, New York.
Early-Age Cracking: Causes, Measurement, and Mitigation (2007) ACI
Committee 231 Report, American Concrete Institute, Farmington Hills,
Michigan.

176
Ellingwood, B.R., Mori, Y., (1993) Probabilistic methods for condition
assessment and life prediction of concrete structures in nuclear power
plants, Nucl. Eng. Des. 142, 155-166.
Feldman, R.F., (1972) 'Mechanism of creep of hydrated portland cement paste',
Cement and Concrete Research Vol.2, 521-540.
Feldman, R.F., (1974) 'Changes to structure of hydrated portland cement on
drying and rewetting observed by helium flow techniques', Cement and
Concrete Research Vol.4 (1) 1-11.
Feldman, R.F., (1968) 'Sorption and Length Change Scanning Isotherms of
Methanol and Water on Hydrated Portland Cement', Paper III-23, 5th
International Symposium on the Chemistry of Cement, Tokyo, Japan.
Folliard, K.J., and Berke, N.S., (1997) Properties of high-performance concrete
containing shrinkage-reducing admixture, Cement and Concrete
Research 27 (9) 1357-1364.
Geiker, M.R., Bentz, D.P., and Jensen, O.M., (2004) Mitigating autogenous
shrinkage by internal curing. High-Performance Structural Lightweight
Concrete, American Concrete Institute Special Publication 218, eds J.P.
Ries and T.A. Holm, 143-154.
Grzybowski, M., and Shah, S.P., (1989) Model to predict cracking in fiber
reinforced concrete due to restrained shrinkage', Magazine of Concrete
Research 41 (8) 125-135.
Grzybowski, M., and Shah, S.P., (1990) Shrinkage cracking of fiber reinforced
concrete, ACI Mater. J. 87 (2) 138148.
Guide to Using @Risk. (2005) Palisade Corporation, Ithaca, NY
Hansen, W., (1987) 'Drying shrinkage mechanisms in portland cement paste',
Journal of the American Ceramic Society, Vol. 70, 323-328.
Helmuth, R.A., and Turk, D., (1967) The reversible and irreversible drying
shrinkage of hardened portland cement and tricalcium silicate pastes,
Journal of the Portland Cement Association Research and Development
Laboratories, 9 (2) 8-21.

177
Henkensiefken, R., Nantung, T., and Weiss, J., (2008a) 'Reducing restrained
shrinkage cracking in concrete: examining the behavior of self-curing
concrete made using different volumes of saturated lightweight aggregate'
Conference Proceedings of 2008 Concrete Bridge Conference, 4-7 May
2008, St. Louis, Missouri.
Henkensiefken, R., Sant, G., Nantung, T., and Weiss, J., (2008b) 'Comments on
the shrinkage of paste in mortar containing saturated lightweight
aggregate', Conference Proceedings of CONMOD08, 26-28 May 2008,
Delft, Netherlands.
Hossain, A., and Weiss, J., (2004), Assessing residual stress development and
stress relaxation in restrained concrete ring specimens, Cement and
Concrete Composites 26, 531-540.
Hossain, A.B., and Weiss, W.J., (2006) Role of specimen geometry and
boundary conditions on stress development and cracking in the restrained
ring test, Cement and Concrete Research 36 (1) 189-199.
Hwang, C.-L., and Young, J.F., (1984) 'Drying shrinkage of portland cement
pastes I. Microcracking during drying', Cement and Concrete Research
Vol.14 (4) 585-594.
Jeffery, G.B., (1921) Plane stress and plane strain in bipolar co-ordinates,
Philosophical Transactions of the Royal Society of London: Series A, Vol.
221, 265-293.
Jennings, H.M., Bullard, J.W., Thomas, J.T., Andrade, J.E., Chen, J.J., and
Scherer, G.W., (2008) 'Characterization and modeling of pores and
surfaces in cement paste: Correlations to Processing and Properties',
Journal of Advanced Concrete Technology, Vol.6 (1) 5-29.
Jensen, O., and Hansen, P.F., (2001) Water-entrained cement-based materials
II. Experimental observations, Cement and Concrete Research, Vol. 32,
No. 6, 2001, 647-654.
Jensen, O.M., and Hansen, P.F., (2001a) Water-entrained cement-based
materials: I-Principles and Theoretical Background, Cement and Concrete
Research, Vol. 31, 2001, 647-654.
Jensen, O.M., and Hansen, P.F., (2001b) Autogenous deformation and RHchange in perspective, Cement and Concrete Research, Vol. 31, 18591865.

178
Jensen, O.M., and Hansen, P.F., (1995) A dilatometer for measuring
autogenous deformation in hardening portland cement paste, Materials
and Structures, Vol. 28, No. 7, 406-409.
Juenger, M.C.G., and Jennings, H.M., (2002) 'Examing the relationship between
the microstructure of calcium silicate hydrate and drying shrinkage of
cement pastes', Cement and Concrete Research 32 (2) 289-296.
Kovler, K., (1994) Testing system for determining the mechanical behavior of
early age concrete under restrained and free uniaxial shrinkage, RILEM
Materials and Structures 27 (6) 324-30.
Kovler, K., Sikuler, J., and Bentur, A., (1993) Restrained shrinkage tests of fiber
reinforced concrete ring specimens: effect of core thermal expansion,
RILEM Materials and Structures 26 (4) 231-237.
Krauss, P.D., Rogalla, E.A., Sherman, M.R., McDonald, D.B., Osborn, A.E.N.
and Pfeifer, D.W., (1995) Transverse cracking in newly constructed bridge
decks, NCHRP Project 12-37.
Krenchel, H., and Shah, S.P., (1987) Restrained shrinkage tests with PP-Fiber
Reinforced Concrete, SP-105 ACI Fiber Reinforced Concrete, 173-179.
Kutner, M.H., Neter, J., Nachtsheim, C.J., and Li, W., (2004) 'Applied linear
statistical models', McGraw-Hill Education Singapure, 5th International Ed.
LHermite, R.G., (1960) Volume changes of concrete, Fourth International
Symposium on the Chemistry of Cement, Washington, D.C., 659-694.
Lura, P., (2003) 'Autogenous deformation and internal curing of concrete', PhD
Dissertation, Delft University of Technology, Netherlands.
Lura, P., Bentz, D.P., Lange, D.A., Kovler, K., Bentur, A., and van Breugel, K.,
(2006) 'Measurement of water transport from saturated pumice
aggregates to hardening cement paste' Materials and Structures Vol.39
(9) 861-868.
Lura, P., Jensen, O.M., and van Breugel, K., (2003) 'Autogenous shrinkage in
high-performance cement paste: An evaluation of basic mechanisms',
Cement and Concrete Research Vol.33 (2) 223-232.
Lura, P., van Breugel, K., and Maruyama, I., (2001) Effect of curing temperature
and type of cement on early-age shrinkage of high-performance concrete,
Cement and Concrete Research Vol.31 (12) 1867-1872.

179
Mack, E.C., (2006) Using internal curing to prevent concrete bridge deck
cracking, MSCE Thesis, Cleveland State University.
Mackenzie, J.K., (1950) The elastic constants of a solid containing spherical
holes, Proceedings of the Physical Society, Vol. 683, 2-11.
Malhotra, V.M., Concrete rings for determining tensile strength of concrete, ACI
Journal (1970) 354-357.
McIntosh, J.D., (1956) The effects of low-temperature curing on the compressive
strength of concrete, Proceedings of the RILEM Symposium on Winter
Concreting, Copenhagen.
Melchers, R.E., (1987) Structural reliability. Analysis and prediction, Baffins
Lane, England.
Moon, J.H., Rajabipour, F., and Weiss, J., (2004) Incorporating moisture
diffusion in the analysis of the restrained ring test, Proceedings of the 4th
International Conference on Concrete Under Severe Conditions:
Environment & Loading (CONSEC), Vol. 2, eds. K. Sakai, O.E. Gjorv, and
N. Banthia, Seoul, Korea, June 2004, 1973-1980.
Moon, J.H., and Weiss, W.J., (2006) Estimating residual stress in the restrained
ring test under circumferential drying, Cement and Concrete Composites
Vol.28 (5), 486-496.
Moon, J.H., (2006) Shrinkage, residual stress, and cracking in heterogeneous
materials, PhD Thesis, Purdue University, West Lafayette, Indiana, May
2006, pp. 244.
Moon, J.H., Rajabipour, F., Pease, B., and Weiss, J., (2005) Autogenous
shrinkage, residual stress, and cracking in cementitious composites: The
influence of internal and external restraint, in Fourth International
Seminar on Self-desiccation and Its Importance in Concrete Technology,
National Institute of Standard and Technology, Gaithersburg, Maryland,
June 2005, 1-20.
Moon, J.H., Rajabipour, F., Pease, B.J., and Weiss, W.J., (2006) Quantifying the
influence of specimen geometry on the results of the restrained ring test,
Submitted to the Journal of ASTM International, Vol.3 (8) 1-14.
Muller, H.S., (1992) New prediction models for creep and shrinkage of concrete,
Creep and Shrinkage of Concrete: Effect of Materials and Environment,
eds. M.A. Daye and C. C. Fu, American Concrete Institute, Detroit, MI.

180
Olken, P., and Rostasy, F.S., (1994) A practical planning tool for the simulation
of thermal stresses and for the prediction of early thermal cracks in
massive concrete structures, in Thermal Cracking in Concrete At EarlyAges, Ed. R. Springenschmid, E & FN Spon, London, 289-296.
Osterle, R.G., Refai, T.M., and Weiss, J., (1994) Variations in calculated
movements of jointless bridges using Monte Carlo simulation,
Construction Technologies Laboratories Report, Skokie, IL.
Parameswaran, S., (2004) Investigating the role of material properties and their
variability in the selection of repair materials, MS Dissertation, Purdue
University, West Lafayette, IN.
Parilee, A.M., Buil, M., and Serrano, J.J., (1982) Effect of fiber addition on the
autogenous shrinkage of silica fume concrete, ACI Materials Journal 86
(2) 13944.
Parott, L.J., Hansen, W., and Berger, R.L., (1980) 'Effect of first drying upon the
pore structure of hydrated alite paste' Cement and Concrete Research
Vol.10 (5) 647-655.
Pease, B.J., (2005) The role of shrinkage reducing admixtures on shrinkage,
stress development, and cracking, MS Thesis, Purdue University, West
Lafayette, IN.
Pease, B.J., Shah, H.R., and Weiss, J., (2005) Shrinkage behavior and residual
stress development in mortars containing shrinkage reducing admixtures
(SRAs), in ACI SP 227 - Shrinkage and Creep of Concrete, eds. N. J.
Gardner and W. J. Weiss, American Concrete Institute, Detroit, MI.
Pease, B.J., (2005) The role of shrinkage reducing admixtures on shrinkage,
stress development, and cracking, Masters Thesis, Purdue University,
West Lafayette, IN.
Pease, B.J., Shah, H.R., Hossain, A.B., and Weiss, W.J., (2005) Restrained
shrinkage behavior of mixtures containing shrinkage reducing admixtures
and fibers, International Conference on Advances in Concrete
Composites and Structures (ICACS), Chennai, India.
Pellinen, T., Weiss, J., Kuczek, T., Dauksas, G., (2005) Comparison of various
INDOT testing methods and procedures to quantify variability in measured
bituminous and concrete properties FHWA/IN/JTRP-2005/03, Indiana
Department of Transportation and Federal Highway Administration, West
Lafayette, IN.

181
Pickett, G., (1956) Effect of aggregate on shrinkage of concrete and hypothesis
concerning shrinkage, Journal of ACI, Vol. 52, 581-590.
Radlinska, A., Pease, B., and Weiss, J., (2005) A preliminary numerical
investigation on the influence of material variability in the early-age
cracking behavior of restrained concrete, Knud Hjgaard Conference on
Advanced Cement-Based Materials, Lyngby, Denmark.
Radlinska, A., and Weiss, J., (2006a) Quantifying variability in assessing the risk
of early-age cracking in restrained concrete elements Eight International
Symposium on Brittle Matrix Composites, 23-25 October, Warsaw,
Poland, 331-342.
Radlinska, A., and Weiss, J., (2006b) Determining early-age cracking potential in
restrained concrete elements using a Load and Resistance Factor Design
(LRFD) approach Conference proceedings: Advances in Concrete
through Science and Engineering, 11-13 September, 2006, Quebec City,
Canada.
Radlinska A., Pease B., Weiss J. (2007) A Preliminary Numerical Investigation
on the Influence of Material Variability in the Early-Age Cracking Behavior
of Restrained Concrete, RILEM Materials and Structures Vol.40(4), 375386.
Radlinska, A., Rajabipour, F., Bucher, B., Henkensiefken, R., Sant, G., and
Weiss, J., (2008) Shrinkage mitigation strategies in cementitious systems:
a closer look at differences in sealed and unsealed behavior
Transportation Research Record (accepted).
Rajabipour, F., and Weiss, J., (2007) 'Electrical conductivity of drying cement
paste' Materials and Structures Vol.40 (10) 1143-1160.
Rajabipour, F., Sant, G., and Weiss, J., (2008) 'Interactions between shrinkage
reducing admixtures (SRA) and cement paste's pore solution', Cement
and Concrete Research Vol.38 (5), 2008, 606-615.
Rajabipour, F., Sant, G., and Weiss, J., (2007) Interactions between Shrinkage
Reducing Admixtures and Cement Pastes Pore Solution, Cement and
Concrete Research (submitted).
Raoufi, K., Radlinska, A., Nantung, T., and Weiss, J., (2008) Practical
Considerations for Determining the Time of Sawcutting in Concrete
Pavements, Transportation Research Board (submitted).

182
Salmon, G.C., and Johnson, J.E., (1996) Steel structures design and behavior,
3rd Edn Harper and Row, New York.
Sant, G., Radlinska, A., Bucher, B., and Weiss, J., (2007) Minimizing the risk of
early age cracking in concrete through the use of comprehensive
experimental techniques, computer simulations, and new materials
(Keynote lecture) Concrete 2007 Conference: Design, Materials, and
Construction Concrete for the Future, Adelaide, Australia.
Sant, G., Rajabipour, F., Radlinska, A., Lura, P., and Weiss, J., (2007) Volume
changes in cement pastes containing shrinkage reducing admixtures
under autogenous and drying conditions, 12th International Congress on
the Chemistry of Cement (ICCC 2007), Montreal, Canada.
Sant, G., Lura, P., and Weiss, J., (2006) Measurement of volume change in
cementitious materials at early ages: review of testing protocols and
interpretation of results. Transportation Research Record: Journal of the
Transportation Research Board, No. 1979, TRB, National Research
Council, Washington, D.C., 21-29.
Sato, T., Goto, T., Sakai, K., (1983) Mechanism for reducing drying shrinkage of
hardened cement by organic additives, CAJ Review, Cement Association
of Japan, May, 52-55.
Schiel, P., and Rucker, P., (2005) New results on early-age cracking risk of
special concretes ACBM, Concrete Cracking Workshop, Evanston, IL.
Schindler, A.K., Ruiz, J.M., Rasmussen, R.O., Chang, G.K., and Wathne, L.G.,
(2004) Concrete pavement temperature prediction and case studies with
the FHWA HIPERPAVE models, Cement and Concrete Composites 26
(5) 463-471.
Schlangen, H.E.J.G., Koenders, E.A.B., and van Breugel, K., (2004) Multi-scale
modelling of crack formation in the concrete cover zone, Advances in
Cement and Concrete, Engineering Conferences International, eds.
Lange, D.A., Scrivener, K.L., and Marchand, J., Urbana-Champaign,
Illinios, 281-291.
Shah, H.R., (2004) Quantifying the performance of steel fiber reinforcement in
mitigating restrained shrinkage cracking in concrete, MSCE Thesis,
Purdue University, West Lafayette, Indiana, pp. 93.
Shah, H.R., and Weiss, W.J., (2006) Quantifying shrinkage cracking in fiber
reinforced concrete using the ring test, RILEM Materials and Structures
Vol.39 (9), 887-899.

183
Shah, S.P., Weiss, J., and Yang, W., (1998) Shrinkage cracking - can it be
Prevented?, Concrete International, Vol. 20, No. 4, 51-55.
Shah, S.P., Karaguler, M.E., and Sarigaphuti, M., (1992) Effects of shrinkage
reducing admixture on restrained shrinkage cracking of concrete, ACI
Materials Journal 89 (3) 88-90.
Shilstone, J.S.M., (1990) Concrete mixture optimization, Concrete International,
Vol. 12, No. 6.
Springenschmidt, R., Gierlinger, E., and Kiernozycki, W., (1985) Thermal stress
in mass concrete: a new testing method and the influence of different
cements, Proceedings of the 15th International Congress for Large Dams,
Lausanne, 5772.
Standard Practice for the Use of Shrinkage-Compensating Concrete, ACI 22398, American Concrete Institute, Farmington Hills, Michigan, 1998.
Stewart, M.G., Melchers R.E. Probabilistic risk assessment of engineering
system, Chapman & Hall, London 1997
Stewart, M.G., (2001) Reliability-based assessment of aging bridges using risk
ranking and life cycle cost decision analyses, Reliability Engineering &
System Safety 74, 236-273.
Swamy, R.N., and Stavrides, H., (1970) Influence of fiber reinforcement on
restrained shrinkage cracking, ACI Journal 76 (3) 443-460.
Their, T.A., (2005) 'Examining the time and depth of saw-cutting guidelines for
concrete pavements' MS Thesis, Purdue University, IN.
Timoshenko, S.P., and Goodier, J.N., (1987) Theory of elasticity, McGraw-Hill
Inc.
Toma, G., Pigeon, M., Marchand, J., and Bissonnette, B.L., (1999) Early-age
autogenous restrained shrinkage: stress build up and relaxation, SelfDesiccation and its Importance in Concrete Technology, 6172.
Vishay Measurements Group (2002) Instruction Bulletin B-134-4 M-Coat F
application instructions.
Vishay Micro-Measurements (2005) Instruction Bulletin B-137 Strain gage
applications with M-Bond AE-10, AE-15 and GA-2 adhesive systems.

184
Weast, R.C., Astle, M.J., and Beyer, W.H., (1983) CRC Handbook of Chemistry
and Physics 64th ed., CRC Press, Inc., Boca Raton , FL, F-33.
Weiss, J., Lura, P., Rajabipour, F., and Sant, G., (2008) Performance of
shrinkage reducing admixtures at different humidities and at early ages,
ACI Materials Journal (in press).
Weiss, J., Radlinska, A., Paradis, F., Niemuth, M., Sant, G. (2007) Cracks in
concrete: an overview of an approach to assess their development, Their
Physical Features, and Their Impact on Durability, RILEM Workshop:
Transport Mechanisms In Cracked Concrete, Ghent.
Weiss, J., and Ferguson, S., (2001) Restrained shrinkage testing: the impact of
specimen geometry on quality control testing for material performance
assessment, Creep, Shrinkage and Durability Mechanics of Concrete and
other Quasi-Brittle Materials, , F.J. Ulm, Z.P. Bazant and F.H. Wittmann
(eds.), Elsevier Science, 645-650.
Weiss, J., (1999) Prediction of early-age shrinkage cracking in concrete
elements, Ph.D. Dissertation, Northwestern University, Evanston, IL.
Weiss, J., (1997) Shrinkage cracking in restrained concrete slabs: test methods,
material compositions, shrinkage reducing admixtures and theoretical
modeling, MS Thesis, Northwestern University, Evanston, IL.
Weiss, J., and Berke, N.S., (2002) Shrinkage reducing admixtures Ed. A.
Bentur, Early Age Cracking in Cementitious Systems, RILEM State of the
Art Report, 2002.
Weiss, J., Borischevsky, B.B., and Shah, S.P., (1999) The influence of a
shrinkage reducing admixture on the early-age behaviour of high
performance concrete, Fifth International Symposium on the Utilization of
High Strength/High Performance Concrete, Vol. 2, Sandefjord, Norway,
1418-1428.
Weiss, J., Schiebl, A., Yang, W., and Shah, S.P., (1998) Shrinkage cracking
potential, permeability, and strength for HPC: influence of w/c, silica fume,
Latex, and shrinkage reducing admixtures, International Symposium on
High Performance and Reactive Powder Concrete, Sherbrooke, Canada,
Vol. 1, 349-364.
Weiss, J., and Shah, S. P., (2002) 'Restrained Shrinkage Cracking: The Role of
Shrinkage Reducing Admixtures and Specimen Geometry', Materials and
Structures, Vol. 35 (2) 85-91.

185
Weiss, J., Yang, W., and Shah, S.P. (2000) Influence of specimen size and
geometry on shrinkage cracking, Journal of Engineering Mechanics
ASCE, 126 (1) 93-101.
Weiss, J., and Shah, S.P., (1997) Recent trends to reduce shrinkage cracking in
concrete pavements Aircraft/Pavement Technology: In the Midst of
Change, Seattle, WA, 217-228.
Weiss, W.J., Lura, P., Rajabipour, F., and G., Sant., Shrinkage reducing
admixtures at different humidities and at early ages, ACI Materials
Journal, 2007 (submitted)

APPENDIX

186

APPENDIX A: SAS CODE

data a1; infile 'G:\COURSES\STAT512\Statprj\rings04.txt';


input strain test ring; seq=_n_; options nocenter;
proc print data=a1;
run;
proc glm data=a1; class test; model strain=test;
means test;
run;
symbol1 v=circle i=none;
proc gplot data=a1; plot strain*test;
run;
proc means data=a1; var strain; by test;
output out=a2 mean=avstrain;
proc print data=a2;
symbol1 v=circle i=join;
proc gplot data=a2; plot avstrain*test;
run;
proc glm data=a1; class test; model strain=test/solution;
means test/t bon clm; means test/cldiff tukey;
run;
proc univariate normal;
run;
*Contrasts;
proc glm data = a1;
class test;
model strain = test;
contrast '1&2 v 3&4' test .5 .5 -.5 -.5;
estimate '1&2 v 3&4' test .5 .5 -.5 -.5;run;
proc glm data = a1;
class test;
model strain = test;
contrast '1&3 v 2&4' test .5 -.5 .5 -.5;
estimate '1&3 v 2&4' test .5 -.5 .5 -.5;run;
proc glm data = a1;
class test;
model strain = test;
contrast '1&4 v 2&3' test .5 -.5 -.5 .5;
estimate '1&4 v 2&3' test .5 -.5 -.5 .5;run;
proc glm data = a1;
class test;
model strain = test;
contrast '1 v 2&3&4' test 3 -1 -1 -1;
estimate '1 v 2&3&4' test 3 -1 -1 -1/ divisor=3;run;

187
proc glm data = a1;
class test;
model strain = test;
contrast '1&2&3 v 4' test 1 1 1 -3;
estimate '1&2&3 v 4' test 1 1 1 -3/ divisor=3;run;

*Residuals and QQ Plot;


proc glm data = a1;
class test;
model strain = test;
output out = a3 r= resid;run;
proc print data = a3;
run;
proc rank data = a3 out=a4
normal = blom;
var resid;
ranks zresid;
symbol1 v=circle i=sm70s;
proc gplot data = a4;
plot resid*zresid;
run;
proc gplot data = a4;
plot resid*zresid;
plot resid*seq;
run;
*levene test;
proc glm data=a1; class test; model strain=test;
means test/hovtest=levene(type=abs);
run;
*random desig;
data a1; infile 'G:\COURSES\STAT512\Statprj\rings04.txt';
input strain testt ring; seq=_n_; options nocenter;
proc print data=a1;
run;
proc glm data=a1; class testt; model strain=testt;
means testt;
random testt/test;
run;
proc varcomp data=a1;
class testt;
model strain=testt;
run;

VITA

188

VITA

Aleksandra Radlinska obtained her MS in Civil Engineering from Szczecin


University of Technology in Poland.

She got her degree in 2005 from the

department of Structural Design. Her Master Thesis focused on optimization of


a spatial truss. Immediately after obtaining her Master Degree she joined the
Cement and Concrete Materials Group at Purdue University. During completion
of her PhD degree at Purdue, Aleksandra worked as a research and teaching
assistant and was an officer of ACI Purdue Student Chapter.

Her research

interest were related to early-age cracking in concrete and reliability-based


analysis of the behavior of cement-based materials.

Aleksandra has been

conducting research for four years and the results of this research have been
presented in 4 journal and 8 conference papers. After completing her degree
she will join the faculty group at Villanova University as an assistant professor.

PUBLICATIONS

189

PUBLICATIONS

Radlinska A., Rajabipour F., Bucher B., Henkensiefken R., Sant G., and Weiss J.
(2008) Shrinkage Mitigation Strategies in Cementitious Systems: a Closer Look
at Differences in Sealed and Unsealed Behavior (Bryant Mather Paper Award for
the best paper in concrete materials section), Accepted by TRB
Raoufi K., Radlinska A., Nantung T., and Weiss J. (2008) Practical
Considerations for Determining the Time Of Sawcutting in Concrete Pavements,
Accepted by TRB
Weiss J., Radlinska A., Paradis F., Niemuth M., Sant G. (2007) Cracks in
Concrete: an Overview of an Approach to Assess Their Development, Their
Physical Features, and Their Impact on Durability, RILEM Workshop: Transport
Mechanisms In Cracked Concrete, Ghent, 7th September 2007
Barde V., Radlinska A., Cohen M, and Weiss J (2007) Relating Material
Properties to Exposure Conditions for Predicting Service Life of Concrete Bridge
Decks in Indiana Joint Transportation Research Program Report, p.202
Radlinska A., Pease B., Weiss J. (2007) A Preliminary Numerical Investigation
on the Influence of Material Variability in the Early-Age Cracking Behavior of
Restrained Concrete, RILEM Materials and Structures Vol.40(4), 375-386
Sant G., Rajabipour F., Radlinska A., Lura P., W.J. Weiss (2007) Volume
Changes in Cement Pastes Containing Shrinkage Reducing Admixtures under
Autogenous and Drying Conditions, 12th International Congress on the
Chemistry of Cement (ICCC 2007), Montreal, Canada
Sant G., Radlinska A., Bucher B., and Weiss J., (2007) Minimizing the Risk of
Early Age Cracking in Concrete Through the Use of Comprehensive
Experimental Techniques, Computer Simulations, and New Materials (Keynote
lecture) Concrete 2007 Conference: Design, Materials and Construction
Concrete for the Future, Adelaide, Australia

190
Radlinska A. and Weiss J. (2006) Quantifying Variability in Assessing the Risk of
Early-Age Cracking in Restrained Concrete Elements Eight International
Symposium on Brittle Matrix Composites, 23-25 October, Warsaw, Poland, 331342
Radlinska A. and Weiss J. (2006) Determining Early-Age Cracking Potential in
Restrained Concrete Elements Using a Load and Resistance Factor Design
(LRFD) Approach Conference proceedings: Advances in Concrete through
Science and Engineering, 11-13 September, 2006, Quebec City, Canada
Radlinska A., Moon J.H., Rajabipour F., and Weiss J. (2006) The Ring Test: a
Review of Recent Developments Conference proceedings: Volume Changes of
Hardening Concrete, 20-23 August 2006, Lyngby, Denmark, 205-214
Radlinska A., Pease B., and Weiss J., (2005) A Preliminary Numerical
Investigation on the Influence of Material Variability in the Early-Age Cracking
Behavior of Restrained Concrete, Knud Hjgaard Conference on Advanced
Cement-Based Materials, Lyngby, Denmark, June 2005
Radlinska A., Bucher B., and Weiss J., Comments on the Interpretation of
Results from the Restrained Ring Test Submitted to Journal of ASTM
International

Вам также может понравиться