Академический Документы
Профессиональный Документы
Культура Документы
FINITEELEMENTS
"
>.....-
~DE
1 Stv1fX:G'N
TABB
I
~
1.
'- ..../
A SIMPLE SUIDE TO FINITE ELEMENTS
bv D. R. J, OWEN and E. HINTON, Department of Civil Engineering, University
Col/ege, Swensee, U.K.
I~
GEOTECHNICAL ENGINEERING
~ 11
This book provides the absolute beginner with a brief introduction to the finite
':( element rnethod. Steady stats heat flow in a eylinder is eonsidered using linear Z !
'--" one and two dimensional elements. Ali eomputational details are ineluded and a
-l
cr useful FORTRAN program is provided. The further applications of the torsion of
prismatic bars and groundwater f,1?ware also considered. rn
:{ Approx. 150 pages, , 980, Q-906674-<l~2
rn
r-
a AN INTRDDUCTlON TO FINITE ELEMENT COMPUTATIONS
by E. HINTON end D. R. J. OWEN, õepertment of Civil Engineering, University
rn' ~J-i._-;'~
L-LJ -+ .i{).L.L_Ll_~_J~
t~1_LL1_~
Ci) Col/ege,Swensee, UX.
~ k~' P ~/' ,/ .\ , ~I
rn IT
~._-,:)---?- / b i I 1
C! This book oflers a basic introduction to the finite element method. Alter a
detailed Introduction into th~ numerical analysis of diserete systems, such as
Z ~
!3'''o,Z'
O -o., / \ . I
frameworks, consideration is given to the solution of some one-dimensional P"',"< ''O_..J'./'\ 'Q,
'-
.~,,-----
I
..__ .. -<:------_._ ..
.
_--)1I
r{ problems using variational and weighted residual finite element methods. Two
dimensional finite elements are Introdueed and used to solve problems
~ , tf' \ -:-.:Jo • !.
r· IY . 'b G·i·
~ •.•.... J. I ,j
\(
associated with heat and fluid flow. the torsion of prismatie bars and other stress
analysis applieations. Other to pies eonsidered Include : numerieally integrated z I '~'J)), T I )1
r >. isoparametrie elsrnents. the flnite strip method, advaneed equation solvers and
\otvJr~-<)-}--
\(
c
ç
mesh generation schemes. Computer programs, written in FORTRAN, and
worked examples are included for ali applieations.
I,
I
11
Ç'j
c,
professional refaranea. Engineering graduates and under-graduates as well as
Qu.II~.d .ngln •• ,. wlll flnd lhe Itlf·contaln.d IIXt I uaeful inlioductlon to finit.
~
rro·.. t\.,).:Y' }~F/ -~"__
7 -.- ....,,----
I
I
11
elernents espeelally as tha mathematies is kept to a falrly elementary leveI.
?\ .!l,'
.R.':.cr--G'
't,
»<!
I
I I ,)- .. _
.~-"-.
__.__ ,i
(
Approx. 400 pages, 1980, Q-906674-<l6-9
I h.
l}
\.
;-, /
}~
'\ Ç:.! -~--._~
FINITE ELEMENTS IN PLASTICITY: THEORY ANO PRACTICE Z W \ /iJ"à ! I li
( by D. R. J. OWEN and E. HINTON, Department of Civil Engineering, University n' :t":'--<:>---~'
1 \
.
"\.,'
/ I1
~
Col/ege, Swensee, U.K.
This text describes the application of the finite element method to the solution of
»
r-
!
.1
i~
!f
\
\
q
"\~'
/""
/
)i,
-.
-,
I
,
I .,. "~o
~,
elasto-plastic problems, The essentlal steps in the numerical solution of such
nonlinear problerns are introduced by the enalysis of one-dimensional axial bar rn .
11 \,/,j.
\ /''-
I
I I
& systems. The theory Is then generallsed to the case of a continuum and particular
expressions are derived for the two-dimensional situations of plane stres.s/strain
and axial svrnrnetrv, Elasto-plastie plate bending problems are also considered.
Z
C\
~.--..._-o----q
li \
I"
. '..
~
.(
(
'..J
(
',~
(
,( Finite Elements
.
:( zn
(
'~
Geotechnical Engineering
~
\{
i(
(f~ D. J. Naylor and G. N. Pande
\(
~U~-1<v(e- Department of Civil Engineering, University Col/ege 01Swansea, U.K.
v
\ NQ.STP~~moi.t. ~ with chapters contibuted by
B. Simpson and R. Tabb
\
\
''{
\
(
''-'
(
'J
~
.(
~,
( 1981
'-'
1\
'-'
J.
« PINERIDGE
Swansea. U.K.
PRESS
J
J.
.J
\.(
'-..(,
--{
'-\
--r-----'· -----.- ."-~------
(
("-. .:
(
lb
L PREFACE
é
~
,{
v
In January, 1978, a short course, titled as this book, was
~ First Published, 1981 by run under the auspices of the Institute for Numerical Methods
1. Pineridge Press Limited in Engineering in the Civil Engineering oepartment at Swansea.
91, West Cross Lane, West Cross, Swansea, U.K.
It was aimed at practising engineers, and these accounted for
({ most of its participants. It appeared that there was a clear
ISBN 0-906674-11-5
need for such courses to provide information to help engineers
Cê. Copyright C> 1981 by Pineridge Press Limited decide whether or not to use the method in geotechnical appli-
C{ cations, and if 50 how. It also became apparent that the scope
of the course had been too wide and that some of the theory had
({ British Library Cataloguing in Publication Data been presented at too advanced a leveI.
(j Naylor, O. J. A 100se bound book of lecture notes had been prepared for
C; Finite elements in geotechnical engineering
the 1978 course. In planning a repeat of the "FEGE" course in
July, 1981, we decided to up-grade ths notes and 50 produce
1. Mechanics, Applied
ç 2. Engineering mathematics
I. Title 11. Pande, G. N.
this book. The scope has besn narrowed by excluding consolid-
ation and dynamic analyses. The topics that are dealt with,
r= 620.101'51'7 TA350
'\ however, are covered more fully than previously. The notes
ISBN0-906674-11-5
\~ have been completely re-written.
"
( .
\, .J (v í í )
(vil
l, We would lik.e to thank. Mr. Czechowsk.i of The Henderson
Busby Partnership (Consulting Engineers) for permission to
(0 include figure (3-10) which is tak.en from finite element
L: analysis'carried out by orie of us in connection with proposed
roadwork.s in north London. The client for this work. is the
CONTENTS
H.E.C.B. division of the Department of Transport. Separate
~oknowlBdgBmBnt! relating to ohapters 11 and 12 are givsn at Page
c the end of thoss chapters. 1. INTRODUCTION
{ Ws would also lik.e to thank Nancy, Ethel, Elisabeth and 1.1 Objectives 1
Felicity, who managed to do ths final t yp íng between t hern in 1.2 Scope 2
(
their "spare" time; also Mrs. Paul who typed the first draft 1.3 Invariants 4
r
r:
of almost the complete book.. Lastly.we must thank our families
for putt1ng up with the 1ntrus1on this book placed on our time
1.4 Rsfsrsncss 9
with them during the last six months. 2. FINITE ELEMENT BASICS 11
( 2.1 Introduction 11
D. J. NAYLOR 2.2 Elements 12
C' G. N. PANDE 2.3 Shape functions 15
2.4 Co-ordinats transformation 17
~
,-
May, 1981. 42.5 Strain-displacement relations 19
l, ~2.6 Strsss-strain relations 20
\
(
\. .,
, --'" 2.7
-'P 2.8
St1ffnessequatibns
Body forces
20
25
\ Preface to second printing
Cl 2.9 Surface tractions 25
( 2.10 Rsfersncss 27
(1 Inevitably a number of errors relating to po1nts of
detail were revealed after the first printing. These have
(: been corrected in the second, otherwise the text has not been 3. GEDTECHNICAL CONSIDERATIONS 29
( changed.
\....- .• 3.1 Introduction 29
--7> 3.2 Effective stress methods 29
.(
\:...." 3.3 Loadings 40
D. J. NAYLOR 3.4 Init1al stresses 42
Á
'C.;. G. N. PANDE 3.5 Excavation 46
Jan., 1984. 3.6 Fills 51
~ 3.7 References 58
~
,.;
l '* 4.
VARIABLE-ELASTIC STRESS-STRAIN LAWS
4.1 Introduction
59
59
~ 4.2 Si-linear elastic model 61
« 4.3
4.4
K-G model
Hyperbolic model
62
64
C{ 4.5 Comparison of models 65
4.6 Variable elastic or elasto-plast1city?
c{ 4.7 References
66
.,
r:
68
~
ç
\'1 .)
'{
(
\
(
'-.j <, (v í í í )
(ixJ
L Page Page
l ~ 5. ELASTO-PLASTIC ANO ELASTO-VISCO-PLASTIC 8. TECHNIQUES F.DR NON-LINEAR ANALYSIS 149
L STRESS-STRAIN LAWS
5.1 Introduction
69
69
8.1
8.2
Introduction
Classification of techniques
149
149
L 5.2 Basic concepts 71 8.3 Equivalent load methods 150
rl 5.3
5.4
Yield function - specific forms
Oilatancy
75
78
8.4
8.5
Variable st1ffness methods
Assessment of techniques
155
159
t 5.5 Hardening law - specific forms 82 8.6 Referenees 159
[: 6.9
6.10
Kinematically admissiblB strain
rate field
Upper bound theorem
105
105
10.5
jointed rock
Generalized plane strain analysis
in rock mechanics
176
185
6.11 ExamplBs of application of the upper 10.6 Effective stress analysis of undrained
~
c 6.12
6.13
bound theorem
ImprovBd solutions
AppIication of the Iimit thBorems to
105
109 10.7
roek masses
References
189
190
\
(
(
( '<,
(xl
í
\.....' Page
t 12.8
12.9
Acknowledgements
References
240
241
It must be commonplace now for engineers to find them-
selves in the invidious situation of having to choose between
powerful computing techniques which they do not understand and
conventional hand calculation methods. If they choose the
( former they feel at the mercy, if not of the computer, then of
f the distant figure who programmed it. If they choose the
latter they may be concerned that the analysis takes too long
r or else 1s inadequate. With the continuing reduction in
computing costs even the most conservative are being forced to
f rely on computer analyses. They must take advantage of this.
f How can they decide when to use the computer? And if they do
how can they be satisfied that the output is to be trusted?
f There are two ways. First, and perhaps most important,
(' the compute r analysis must be backed up by the proverbial
back-of-envslope calculation. No sxpsrienced engineer needs
f' telling this. Coupled with this will be his assessment of
ç the computer results against his experience (or perhaps his
prejudices!). The second way is to obtain some understanding
c of what the computer is doing, and this is where this book
L
comes in. A complete understanding will not be practicable
in the case of the more sophisticated analyses such as the
f.e.m., nonetheless a working knowledge which can go a long way
\ to bridging the divide between the "long haired" academic who
(2. writes the programs and the engineer, who is preoccupied with
f
(
(
r 2 3
)
\..'
It is not perhaps widely realized that structural programs may
have been referred to above. These two chapters cover ideal-
izat10ns of the geometry (i.e. the finite elements) and the
be quite unsuitable for geotechnical analysis. Thus a struct- various processes (Chapter 3). They say nothing about tne
I ural program, while it might do for a linear elastic analysis material idealization. The formulations and technique3 apply
l.L
of a loaded footing, would be unable to model an excavation, equally [apart from some minor qualifications) whether the
,\(.:. or a layered fill, or predict excess pore pressures in an un- material is idealized as linear elastic ar non-linear.
drained analysis, or compute the effect of a known pore
( pressure change, or take into account initial stresses, to name The material idealization is central to the next four
the more important geotechnical requirements dealt with in this chapters on non-linear laws and behaviour. Variable elastic-
~ book. 1ty is covered in chapter 4, followed by an introduction to
ri. elasto-plasticity in chapter 5. This provides a foundation
Knowledge and understanding of the theories is not enough. for chapter 7 in which the critica 1 state elastic-plastic model
r[. The art lies in judging when it is appropriate to use them, is described in some detail. Chapter B describes techniques
and then in deciding what values should be assigned to para- for applying these non-linear laws. Chapter 6 is in some
t meters, particularly those defining the stress-strain law. respects a "cuckoo" in this family as it describes the clessica
Q. Skill in this art comes from experience of practical applic- upper and lower bound methods of limit analysis. It has been
eticns. in fliledbeok from case studies. 8aok enelysis of in- included because of the value of ~hese powerful and simple
lS. strumented prototypes is needed. The last two chapters by techniques in providing an independent check on f.e. analyses
practising engineers include some such experience. In them which attempt to simulate a failure condition. It follows
~ a number of case histories are described which illustrate the chapter 5 since it draws on ideas introduced in that chapter.
(j type of problem for which f.e. analysis is appropriate. They
also illustrate how parameter values can be assigned, The Chapter 9 is quite separate. It has been included since
final chapter by Dr. Simpson draws on the wealth of experience the neee for a steady seepage analysis 50 often crops up in
\(
of f.e. analysis of excavations in London clay. He shows how geotechnical work. Whereas hand-sketched flow nets are quite
~. feedback of the measured performance of the structures anal- adequate for homogeneous soil conditions the f.e.m. provides a
ysed has been used to enhance future predictions. Such use cheap and versat11e tool for the more usual mult1-1ayered soil
of the "observational method" (Peck 1969) must be central to configurations. Having only one degree of freedom per node
~
successful application of the f.e.m. in geotechnical engin- the f.e. formulation is one of the simplest. This chapter i9
I~ eering. restricted to steady seepage. It was decided that the rather
\- 1.2 Scope
large subject of transient flow shauld be outside the scope of
this book.
"I It is assumed that the reader 15 versed in conventional Chapter 10 likewise stands on its own as the chapter
'( soil mechanics theory (as taught in civil engineering degree dealing exclusively with rock mechanics applications. Unlike
\...
courses) but that he has only a hazy idea about finite elements. chapter 9. hawever. it does draw on thearies developed ~n the
This should be sufficient for him to recognize that a bar in a preceding chapters.
l
\....i pin-jointed frame is a simple finite elemento If, in addition,
.,\ he is familiar with matrix methods of analysis which would Chapters 11 and 12 have already been mentianed. They form
''--" allow him to set up the simultaneous equat10ns relating nodal the practical bias to balance the 1nevitable theoretical mat-
\
\.....
loads to displacement for the aforementioned pin-jointed frame erial which precedes them. Suffice it to say that withoLt
then he is alI set to make light work of chapter 2. Relatively them this book would be badly unbalanced.
~, recent graduates (within the last two decades?) will be
familiar with such methods. Their mone senior colleagues and The Mohr-Coulamb strength parameters c and ~ (or c' and
\
l_, those having different backgrounds should brush up on matrix ~') feature in much of sails analysis. They imply that if
algebra. [Dnly the basics are required.) yielding accurs it is in the plane containing the major ard
~ minor principal stress. The intermediate principal stress
( Chapter 2 deals with the basics of the f.e. methods. The has no effect, nor [if the material is ~igid plasticl is there
,~
approach has been necessarily selective the criterion being to strain in the intermediate stress directian during yielding.
~ provide the reader with the information needed for the later This candition suits the plane strain case since it would be a
chapters and little more. The more mathematical formulations nuisance to have ta calculate the out-of-plane stress sim~ly
~ have been avoided. Chapter 3 deals with the formulations and to provide infarmation for a yield criterion when it is nct
'.( techniques needed specially for geotechnical applications which atherwise needed. With a Mahr-Coulomb criterian alI pIare
strain analyses can be carried out in terms of the in-plane
1
(
\
(
~. 4 5
1.3.1 Plane strain group
L stresses and strains. Furthermore the general three-dimen-
li of the ensuing chapters. The plane-strain stress invariants dE • .'!.. fedE -dE )2 dy 1
in particular occur repeatedly throughout the book. d 2 x y
+
xy
2 '2 (dE1-dE2) (1-4 )
<i
v. 1.3 lnvariants Note that Ed = JdEd is not, in general, the same as
The components of stress and strain (strictly, the com- equation (1-4) with dE replaced by E.
\( ponents of the stress and strain tensor) depend on the chosen
direction of the coordinate axes. The principal stresses and 0d and dEd are always positive. They are respectively
~ strains on the other hand are invariants. Their values are the diameter and rad1us of the Mohr's circle. See figure (1-1)
'{ unaffected by the choice of reference axes. Other invariants a and 1 dE are the abscissae values of the centres of the
may be defined as functions of the principal values. ln general 5 - s
2
() three independent stress or strain invariants may be defined. Mohr circles.
\
\
(
(
6 7
?
\ The invariants have-been chosen 50 that the work increment
o,
C term dW is given as fo110ws:
L,\ dW = S
T
d! = 01dc1 + 02dc2 °sdcs+Oddcd (1-5 )
,/
0.(
0"0.,,
t
~
L
L 1.3.2 General Group JL--~ ~ •.°2
01 ~ 02 ~ 03 and dE1 ~ dE2 ~ dE3
~
Stress invariants OJ
~
~ o
m
= 1
3
(o +0
xyz
+0 ) = 1 (o
3123
+0 +0 ) (1-6)
Figure 1-2 Invariants as cylindrica1 coordinates in
« o 2 .1[(0 -o )2+(0 -a )2+(0 -o )2+6(, 2+, 2+, 2)J (1-7aJ
stress space
(f q 2 x Y Y z z x xy yz zx
Note that for the triaxia1 stress state (01 > 02 = 03)
(j 1 [I.c -a )2+(0 -a )2+(0 -o )2] (1-7b) 0q=01-03 (and becomes the same as 0d) and e = 30°. 0m and Oq
2122331 are the same as the Cambridge p and q respectively. They are
v: 1 .-1 27J 3)
simp1y related~to certain other invariants which appear in the
technical literature. These are the first stress invariant,
(( 8 = 3' Sln [20 3
(1-8à 11, the second deviatoric stress invariant, J2, the octahedral
q q
mean stress. 0oct, and the octahedral shear stress 'oct.
The relations are as follows:
~.
ç-'
(
t-
where
J
3
c -o
sym.
x m T
c -o
y
xy
o -o
z m
T
r
xz
yz
°m = -
°
q
= 13J2
1
3' I1 = °oct
=
12
~ Toct
(1-10)
20
,( 1 2+03]- using a tension positive convention.
'-..,. or 8 Tan -1 [° - (1-8bl
13
cl (°1-03) A useful relationship exists
principal stresses, as follows:
between em' ° , 8 and the
q
~ ° - 3.3 ° Sin(8 + ~ (i-211T] (1-11 )
0m and 0q' suitably scaled, together with the angle 8, ai m q
~ form cylindrical coordinates in principal stress space. This
is i11ustrated in figure(1-2l. The axis is the 1ine equally (i = 1,2,3l
~ inclined to the 01' 02' 03, axes points on which represent
(l states of equal-all-round stress (01=02=03)' 0q measures Strain increment invariants
the deviation from this and 8 - sometimes referred to as the
ri Lode angle - measures the orientation in the normal or "Pi" dE
V
dE
X
+ dE
Y
+ dE
z
dE
1
+ dE
2
+ dE
3
(1 -12)
« plane.
2
de 2 = - [I de -dE ) + ••• + -2
3
(dy 2 + ••• )] (1-13a)
CI q 9 x Y xy
L dW = cr T de
-v '\,
= cr de: +o' de: +o oc
112233
= a oc
mv
+ cr de:
qq
(1-14)
C
The Lode angle i9 not involved. This. however. implies that a
C Lode angle defined for the strain increment is the same as
L that defined by equation (1-8).
l
l
l (i)
This was the main reason for depart1ng from certain pre-
( cedents. Thus 0d is preferred to the "q" of Lambe and
Whitman (1969) which is ~ 0d' 5imilarly 0q (the Cambridge
( "q") 1s preferred to Toct.
r
(
(
r'
(,
r
11
~;
H
G CHAPTER 2
L 2.1 Introduction
L The finite element method now has a very wide range of
L applications. These in addition to structural engineering
include the fields of fluid flow, electricity and magnetismo
l It started with structural applications, and it is ~ith the
([ formulations for relating loads to the displacement of a struc-
ture that we are concerned here. A different formulation is
c required for seepage analysis. This is covered in Chapter 9.
((' The reader familiar with matrix methods for anaIysing
~ structures such as a pin-jDinted frame is half-way to under-
\(
standing the f.e.m. Perhaps more than half-way, because the
~ members Df a pin-jointed frame are simple finite elements.
''('
( j
This chapter sets out to explain these procedures. The
;1. conventional 'stiffness' approach is used. (Alternative deri-
vations based on variational caIculus or, e.g., the Galerkin
'-
,( method, are not given both because of the mathematics involved
L and because they are better suited to formulations where the
starting point is a governing differential equation. This is
~ not the case with load-deformation problems, although it is
~ with seepage.)
Q,
Geometric considerations - elements. shape (or interpola-
tion) functions and certain geometric co-ordinate transforma-
'\'- tions - are treated first. The virtual work method is then
used to set up the stiffness equations, and to determine nodal
~
forces equivalent to body forces and surface tractions.
\L
For a more complete treatment of finite element basics the
([ reader is referred in the first instance to two books from the
([ same stable as the principal authors of this volume: the classic
text by Zienkiewicz (1977) now in its third edition, and the
«
(
\,
(
(
( 12 13
ventional elements (there is a compatibility problem at common
~ text by Hinton and Owen (1979). The notation used herein is nodes). They have, however, a role in soil:structure inter-
( much the same as fn these texts. The second of them devotes a action applications. 3-D elements are used when the cost
( chapter to a review Df the literature. In it there is a use~ (which lies as much or more in data preparation and output
L fuI list Df 31 other books on the subject, with short comments handling as in the actual computing) is justified.
(
x, on each. It is noteworthy that not one in this list is aimed
f primarily at geotechnical engineers. An earlier text by Higher order elements with more than one midside node are
L Hinton and Owen (1977) deals specifically with programming available but although they give better accuracy per element
aspects of the f.e.m. . it is doubtful if they offer any advantage on a 'per nade'
,~,
basis. It seems to be widely accepted that the so-called
2.2 Elements 'parabolic' elements which have one midside node offer the best
c value per nade. The bottom row of elements illustrated in
I( The starting point of an analysis is the division Df the figure (2-1) are therefore widely used, especially the 8 noded
'- structure into elements. Eight basic element types are illus- quadrilateral. Recent work has drawn attention to the
l trated in figure (2-1), Some of these shapes represent a 'Lagrangian' quadrilateral and brick elements which have,
number of different types. Thus a 2 or 3 noded line element respectively, 9 and 27 nades. The ninth node is at the centre
{ can represent a bar having no bending sti~fness in 1, 2, or of the quadrilaterial, and the seven extra nodes in the 3-D
r{ 3 dimensions. For these applications it will have respectively element are at the element centre and the centres of the six
1, 2, or 3 degrees of freedom per node. Alternatively it may faces. These elements may give even better value than the
\f be a bending element in which case an extra rotational degree more commonly used type of parabolic elemento
of freedom per node is added for 2-0 applications and three
ri: extra rotational degrees Df freedom (making a total Df 6) are With the exception Df the beam and plate bending elements,
~'
added for 3-0 applications. Similarly, the triangular and the elements considered here are 'isoparametric'. That is, the
quadrilateral elements may represent membranes or plates in 3-0. equations describing the shape of their boundaries are of the
,-
In the former case their bending stiffness is neglected and
'{ they have 3 degrees of freedom per nade. Plate bending ele-
same arder as those describing the variation of the nodal
ç- ments may have 6 d ,of f , per node.
unknown (e.g. displacement) across the elemento
quadratic in the case of 'parabolic' elements.
Thus both are
1 --
,I,
.J_
---
are rapid changes in stress and strain.
trates this for a footing.
Figure (2-2 a) illus-
The smallest elements are at the
"
~
( corner of the footing. Figure (2-2 b) illustrates a 'spider
web' type of mesh. This efficiently increases the element size
( with distance from the highly stressed region.
'-(,
1,-,
(~
( /60 Remarkably high accuraey can be achieved with a coarse
.mesh when parabolie elements are used. It is good praetiee to
carry out a preliminary analysis using a very coarse mesh,
e.g. 6 soil and one footing element would eonstitute a suitable
coarse mesh for a preliminary analysis of the footing of
( Figure 2-1 Some basie finite elements figure (2-2 a). Substantial agreement of displacements with
~ those obtained from a fine mesh analysis would be expected
,) (say 20% difference in the maximum valuesl. .
'--
For geotechnical work the triangular and quadrilateral
~ elements applied in a plane strain analysis are most eommonly
used. (They then represent a solid bloekof material with an
.~
out of plane thickness which is usually one.) Sometimes the
([ plane elements will represent an axisymmetric geometry. Line
elements can represent ties or props, or flexible linings to
(~ tunnels. Bending elements are seldom used, perhaps beeause
there are not many programs available which mix them with con-
i(
(
('
''-. 14 15
l 2.3 Shape functions
l
•
lhe quantities comprise the nodal unknowns in the first inst-
ance, but include any quantity which is rEquired to vary
l smoothly across the element between fixed (known, or to be
determined) values at the nodes. Let O stand for the value
t of the quantity at some point x,y (~e restrict to 2-D for
.( simplicity), and suffix i indicate the value at node i,
then
n
( (2-1 )
O L -.
0i
« i=1
"--
t simple expression applicable to a particular mapped shape.
lhis can represent a wide range of actual slement shapes.
Figure (2-3) illustrates the mapped and parent elements for
\..~
l
the 6 noded triangle and 8 noded quadrilatsral.
,(
'-' lhe shape functions must have a +cr-n such that when the
< ~.n values corresponding to the same noda, 1.e. node i Tor
'-.. Ni• are substituted the function assumes t1e value 1. Also
~ it must be zero for the ~,n values corresoonding to alI the
other nodes. lhus the three-noded bar having ~ = -1 for node
J 1, ~ = O for node 2, and ~ = +1 for nade 3 (n does not apply
,1 here) has the following shape function for node 1 :
i.l N
1
-~~(1 -~) (2-2 a)
l4
11
versa, are not required o All that is required Ls an expr'ess-
~; 1 ion for the area increment dA = dx.dy in terms of ds and dn,
é 1~6
5
4
_ç
and a means of finding the eartesian shape funetion deri-
vatives, i.e. aNi/ax and aNi/ay, given âNi/as and âNi/ân.
x
~. o
'2 3
1
(
\.::..c
11
~LQ
~ 1
« 7 6 5 / g const
'C -1 8 4 1 ~.d1] //
1]
---.~~-
Q ,
(-
~
((
2
-1
3
x I
Yi
a
~,d1]1
P
"tJ!r:/
r-
d
...J~a
Q _-1]
dt:
const
N = 4~[1 - ~ - n)
2
~., N = ~[2~ - 1)
Figure 2-4 Area transformation
3
(2-3)
N4
(
,( N
=
=
4 ~n Let dA represent the area in the aetual element corr-
s n(2n - 1) esponding to d~.dn in the mapped elemento Let! and ~ be
'r
\..::."
NS = 4n(1 - ~ - 11)
veetors representing the sides PO and PR of theinfinitesimal
area dA as shown in figure (2-4)0
d~,dn as follows:
The veetors are related to
2
=
=
H1-11)(1-~2)
(
\
(" .
~. 18
19
(.) with the columns and rows increased by one termo Zienkiewicz
( dA 1JI dE;d n (2- 5) (1977) and Hinton and Owen (1979) include the 3-D case in their
c) coverage of transformations. Hinton and Owen (1977) provide
f'-..' in which rJ 1 is the determinant of the Jacobian matrix FORTRAN codings.
l
,<
~ aE; ay
élE; 2.5 Strain-displaeement relations
,~; J r (2-6)
l'-..'
{
\~
l l'5.
an
( the shape funetion E;,n derivatives and nodal eo-ordinates by Plane stress and strain
~ differentiating equation (2-1). This gives
I( EX = - a u/a x (2-9 a)
n aN
f
ax I _i
Xi and ete. E = - avia y (2-9 b l
~ i=1 aE; Y
(2-9 e)
Y = - (élu/ay + élv/ax)
C(
Explieit expressions for a~ laE;, élNi/Cln are readily obtained (E
z
= O for plane strain)
f,-, by differentiating the formulae for N
i
(e.g. equations (2-2)
to (2-4)). Axi-symmetric
i(
(~ To obtain the cartesian shape function derivatives we
{ The r~lations are the same as equations (2-9) with x , u
first note that they are functions of x,y so that applying standing for radial and y,v for axial. The hoop tangential
(:(' the chain rule one obtains strain Ez 1s given by :
({ aNi ClNi
u/x
dN
i ãX dx + ay dy (2-7) E
z
= -
l~~lI::'}
I' by a veetor (or column matrix), ~, having 3, 4, or 6 eomponents
aeeordingly as the problem is plane, aXi-symmetrie, or three-
r ax dimensional.
{ :;' }
~,
( , ClE;
c ClNi
=
ClNi
For finite element applications it is neeessary to relate
6, dT)
ax
ãTl
El.
Cln ay
strain to the displaeements at element nodes. Let these be
represented by <3 vector §e whieh in 2-D is (u1' v1' u2' v2 ...•
GJ u , v )T. Using eque t í.on (2-1) to express
n8dalndisplacements one obtains
u , v in terms of
:-~~
of this equation. Invers10n g1ves explieit expressions for the e
,( cartesian derivatives, 1.e. E B 6 (2-10)
\.1-
.~
~,
N Ths matrix B compr1ses a row of n (the number of nodes in
1
-~ I ar- the elementl- submatriees ~i wh1ch for plane problems take the
l
C1 Cly-: { -. }
aélx } ( 2-8) form:
Ci
{
ãY - em
dX
aE; J
i
ar)
B.
-1
= - [N/" O
O
ClN/Cly
I
I (2-11 a)
I{'
Th1s completes the transformat1ons
cations.
needed for 2-D appli-
The matr1x equations der1ved above also apply in 3-D
ClNi/dy ClN/Clx j
r
\-l
\.
(
~ 20 21
(; J and for axi-symmetric problems 2.7.1 Element relations
«
(
Generally
/:0,0
these can be expressed
= 9 /:o,f (2-12)
tual displacements (i , ó', 1s appl1ed to the nodes. Let the
stress at a point in the-element be a and the strain corres-
ponding to the virtual displacements be ~'. Equating the work
done externally (at the nodes) to that done internally gives
«
q
Where a contains the 3, 4, or 6 stress components corr-
esponding to ~ (see preceding section) and '/:o,' indicates
"cheng a 1n' which may or may not be small. is the corres- 'º' (Ó*)T Fe = J (~.)T ~dV (2-13)
r: ponding square modulus matrix. Its components are constant for the integral being taken over the volume of the elemento Now
'{ linear (elastic) applications. For non-linear they are deform-
F
by (2-10)
ation or stress dependent. O is usually symmetric. An excep-
\ tion occurs with non-associative elasto-plastic formulations E* = e s'
as will be illustrated in Chapters 5 and 7. Explicit forms Df
Q are given later, e.g. for elastic formulations in section (E') T (B ó*) T (Ó*)T BT
3.2.1 and for non-linear formulations in Chapters 4, 5. and 7.
~ T he foro Df O is not needed jus t yet.
~- Substituting this in [2-13) and using the fact that the
( coefficients Df ~. are independent Cwhich allows ô' to be
C; 2.7 Stiffness equations effectively cancelledi we obtain
(!
X The global stiffness matrix equation which relates nodal
forces to displacements 1s best considered in two stages: the
E
e
= JêT
Q dV (2-14)
''-'
/{ setting-up of the element stiffness matrices, then their assem-
'-.:.1 Th1s 1s an important relation. It is used whenever a set of
bly to form the global matrix.
nodal forces is required which are to be equivalent to - or in
~
-I ,I .~
(
\~'
~ Noda! forces_
te ) n Stiffness
~e 4 I __ Noda! disp!acements
Qe
[i) If virtual displacements must be interpreted physically,
then the implication is that they must be such as to cause
~ negligible change in the nodal forces. This implies either
C.c' ," _
« /' I 3
<,
that the displacements are very small, or cause rig1d body
motions, or if the body is plastic, cause plastic flow.
({ An essential feature is the independence of the displace-
ment components.
(( Figure 2-5 An elemento
CI
(
(
"--"
(
I
'..•... 22 23
(.
<
y
( 24 25
components at common nodes are simply added, e.g. if a node is a basic treatment which includes computsr codings. Hinton and
\'( at the corner of four slements there will be four force compo- Owen (1977) describe in detail a version of the 'FRONT' equation
( nents to bs added for a particular direction. With displace- solver, a form of Gauss elimination widely used in finite ele-
? ments, of course, no addition is required, the nodal value is ment programs.
the same for each shared elemento The assembly of the elsment
'( stiffnsss matrices i5 not 50 obvious. It is best explained by
~~ an example. 2.8 Body forces
(
'--( 2 3 4 5 6 These forces, which are distribu:sd over the volume of
, elements, must be converted to equivalent nodal forces. They
'- a a ;a a will then form part or alI of the righ:-hand side vector B.
(
2 a i ab , b a ab !
I
Let Bb denote the body force contribu:ion to B. Dsnoting the
body force intensity (force per unit volume) at x,y by the
3 bi b I b vector e= (Px,py)T, and using ths vir:ual work principIe with
~- I I
( 4 a a I I a c a c c an asterisk to denote the virtLal di3placements, ws havs:
'----
5 a ab b a c abc c
t 6 c c c
(cS.)T Rbe J (u· Px + v' Py) d'/
,(
Numbers inside are the local;
.( Using equation (2-10) to expreEs u· and v' in terms of the
those outsids the global.
nodal values, and expanding thE left-hand side:
/(
(
Figure 2-7 Stiffness assembly
Ln
i=1
b b')
[ R, e u ~ + R, e v i
1X 1 ay
= J (P nL N; U
x i=1 - i
•+ P Ln N. v, ") dV
Y i=1 1 1
( For simplicity, suppose that there is only one variable
'r
per node. (This actually is the case with seepage problems,
see Chapter 9.) The example of figure (2-7) thersfore has a
Again, s nce the components of §',
í e, u
independent we can equate their coefficients,
í •
1, vi.s.
1' e t c , are
total of 6 degrees of freedom and the overall stiffness matrix
l(
is 6 x6. The letters represent element stiffness coefficients. R~e
, =JPXNidV (2-19a)
Thus 'a' in row 4,column 5, identifies the local stiffness coef- 1X
ficient Kr3 for element 'a'. Where more than one letter appears
> in a box the stiffness contributions are added. Thus from row R
be
= J PY N. dV (2-19b)
( 1y 1
5, column 2, we obtain the global coefficient
{ with i=1,2, ...n.
Ke(a) Ke(b)
KS2 +
'( 30 21
Equations (2-19) ar:bused to dat errní.na the 2n equivalsnt
-;
Note that only Kss receives contributions from alI three ele- nodal forces comprising R e. These e Larnerit vectors are then
(
ments. Note also that symmetrical element stiffnsss matrices assembled to form Rb. -
\ .. ,-
( result in a symmetrical global matrix.
'-.( If there were more than one degree of freedom per node 3.3.2.
Some applications of thi5 theory are given in section
\.:...
the boxes in figure (2-7) would each represent a submatrix,
( e.g. 2 x2 for 2-0 problems. The assembly procedure is the same.
"-- 2.9 Surface tractions
!t
l
(
( 26 27
S se
( R is obtained by assembling the edge contributions R from
;11 the loaded edges iQ the f.e. mesh.
~
c Some applications of this theory are given in section
l
(
3.3.3.
~,
2.10 References
l'-..'
L FENNER, R.T. (1974), "Computing for Enqineere ", MacMillan.
c HINTON,
Academic Press.
(
f
~ Figure 2-8 Tractions on element side.
~
J1
k [se
Rix ui• + Riy
se vi.] = J '.
[p(v . ·
Cose -u Sine)+ q(vSine ·
+u Cose) J' dA
~
~ Integration is over the side area, i.e. length times out-of-
C. plane thickness. Again using equation (2-10) to relate u·,v·
~i to nodal values, and using the independence of the latter we
obtain
c-> R
se
-f
(p Sine - q Cos e) N
i dA
(2-20)
L
,(
ix
(:, Rse
iy f (p Coa 6- q Sin6) N
i
dA (.2-21)
l
Note that for a horizontal surface with e =1800 (i.e. leveI
L ground, e =0 would represent a ceiling).
L
iL R
se
ix
-J q N dA
i
(2-20 a)
( se
Riy -f P Ni dA (2-21 a)
([
(
(
c'
(
\.....-
29
(
'-"-
l, CHAPTER3
l GEOTECHNICAL CONSIOERATIONS
(
{
3.1 Introduction
f
The principIe of effective stress underlies much of geo-
f technical analysis. Techniques for applying effectiv~ stress
'( methods of analysis using the f.e.m. are therefore dealt with
first in this chapter. They are followed by an explanation of
f how the various forms of loading likely to be encountered (i.e
c
(
discrete, distributed body forces, or surface tractionsl can
be implsmentsd. Ths important qusstion of how to deal with
r· initial stresses (i.e. the stresses and pore pressures which
X
(-
are there at the start of the analysisl 15 next considered, and
finally procedures for using the f.e.m. to model excavations
{
and fills are described.
~"
t
,(
ticated techniques involving the Biot consolidation equations
are required (see e.g. chapter 12 of Oesai and Christian,1977l.
Similarly the time dependency due to creep (ar "Secondary
l..;
Consolidation"l is outside our scope. Some account of this can,
L however, be taken by an appropriats rsduction of stiffness.
~
( ;
30
in which m 1s a column matrix with "1"5 in the direct comp- Substitution of (3-3), (3-4) and [3-7) into the principal
31
.
( onent pos1tions and "0"5 in the shear, e.g, for plane strain of effective stress equation [3-2) 1ntroduces ll~ 1nto each of
L, the three terms. Since the components of ll~ are independent
m = (1,1,0)T. For stress changes (3-1) becomes:
G
'V
of each other: we can effectively "cancelo them to obtain
!l~ !lf + ~!lu (3-2) T
L The soil skeleton and pore fluid constituents can be con-
R n'
'C
+ m m K
'V'V f
(3-8)
sidered as separate elements [not necessarily "finite elements" No restriction has been placed on O, ~' 01' Kf' They can
represent linear elastic properties, in'Vwh~ch case their comp-
L although we shall shortly consider them as such) which share
the same physical space. If conditions are fully undrained onents are constants, 01' virtually any form of non-linear
relation. The derivation of the foregoing in 'terms of
t there will be no movement of pore fluid relative to the skele-
ton, consequently the skeleton and pore fluid elements undergo
I
incremental relationships allows ths important class of
l precisely the same de+ormet.d cns , Their strains (!l~) can be
equated. (The objection that this is a physically unreasonable
I' differential stress-strain laws to be incorporated. These
include the variable elastic laws considersd in chapter 4,
( assumption, since flow w111 start as soon as 10ad is applied, I and the plastic1ty laws (including the critical stats model)
(( qualified by "in the limit as the time for consolidation For elastic isotropic applications and involve only R R'
approaches zero". This becomes a definition of "undrained",) two parameters, These are conventionally Young's modulus and
f Total stress changes are related to strain changes by a
Poisson's ratio, but a variety of alternatives are possibls of
which the bulk modulus, K (01' K'), and the shear modulus, G
C(' modulus matrix R
[this approaches a tangential modulus matrix (01' G'), are particularly useful. Equat10n (3-8) then takes
(( as the changes become small), i.e. the following form for plane strain
ç'
G:
ll~' R'll~
~ f
!lu Kf!lgv (3-5) G = G' (3-10)
G where Kf is the bulk modulus of the pore fluid elemento We This is an important resulto Using it in the four upper
~ shall call it the equivalent pore fluid modulus. It is not left hand elements of the matrices givss
simply the bulk modulus of the pore fluid (K ) as the pore
~ fluid occup1es only the voids, but is a comp~site of the pore K+ ~
3
=K'+~+K
3 f
fluid and the particle (not skeleton) st Ff'nass, Ks.' (5 for
~
í
CT
( ,
c
{
,--,!
32
the undrained case. The only difference as far as the results
.
The advantages of the effective stress method are its
33
( are concerned i9 that the latter gives explicitly the effect- flexibility and its ability to provide explicit predictions of
U ive stress and pore pressure components, [the "prosO and "cons" the "excesso (i.e. load induced) pore pressure, 6u, in un-
l of the two approaches are discussed later in this section). drained enalysis. It is flexible in that since the soil skele-
0'
An example will illustrate the conversion. ton stiffness (R') is defined the method is not restricted to
C Example
undrained analysis. Thus a drained analysis is implemented
simply by putting Kf = O. In this case as far as stress
(
"--' changes are concerned, and provided there is no change in pore
í A clay soil is assumed to have an elastic soil skeleton pressure during loading, there becomes no distinction between
'- heving E' = 30 MPa and v' = 0.25. It is saturated. The effective and total stress analysis = A~' = A~). [R' R,
1 equivalent pore fluid bulk modulus Kf 1s taken as 2000 MPa.
(This approximates the bulk modulus of water, actually about For undrained analysis there is no coupling between direct
~ 2200 MPa. Choice of K is not critical when it is much larger and shear terms as in isotropic elasticity [i.e. where a hydro-
than K'.) The corresp6nding total stress parameters for un- static stress change causes no distortion and pure shear
~ drained analysis, E and v, are required. stress causes no volume changel, and provided the soil is sat-
urated (i.e. virtually incompressible), then Au can be obtained
~( We first obtain G = G' using the standard relation bet- from a total stress analysis. It is given simply by
:;- ween G, E and V (most elasticity texts, e.g. Jaeger, 1962, or
,~, Timoshenko and Goodier, 1951) Au % Acrm [3-12)
~ E' 30
r G G'
2 (1 +v' ) 2 X 1.25
12 MPa where ócr
m
2
3
[Acr
x
+ ócr
Y
+ ócr )
z
'-<.
~ The effective stress procedure described above~could be
'~ then using the stanaard relation between K, E' and V
dispensed with in this case. But where there is coupling
.:
~
K' = 3(1-2v'
E'
l
. 30
3 X 0.5
~ 20 MPa
between spear and direct stresses as in plasticity form-
ulations~ or where the soil has significant compressibility
"(' (Kf not very much greater than K'), the total stress method
r:
Using (3-111 K = K' + Kf = 20 + 2000 = 2020 MPa.
cannot be used directly, and the effective stress method comes
) So far we have corresponding total and effective stress into its own.
parameters in terms of K and G, i.e.
) It is still possible, however, to use 'a total stress
analysis to estimate pore pressures. They are calculated from
Effective: G' 12, K' 20, Kf 2020 }
~ the total stress changes by means of pore pressure parameters.
'--.J All in MPa
Total: G 12, K This approach can be linked to the effective stress method as
,( 2020
there is a relation between Kf and the pore pressure parameter
<: B.
-( To obtain E, v we again use the standard relations which
'-' express them in terms of K and G, i.e.
.( The pore pressure parameters A and B, proposed by
'-' Skempton (1954), are widely used. They relate specifically
9KG 9 x 2020 x 12
( E =
3i<:+G = = 35.9 MPa to triaxial test conditions (Acr1 > Acr2 = Acr3). For general
--..c..' 3 x 2020 + 12
applications the alternative parameters Ah' B are more
"( 3K - 2G
rational. [These have been attributed to HenRel, hence the
'-' 3 x 2020 - 2 x 12
,,{
V = 6K + 2G = 6 x 2020 + 2 x 12 = 0.497 suffix h.) Using these the excess pore pressure is given by
"--
Au Bhócrm + AhAcrq (3-13)
J This example illustrates how an effective stres5 un-
drained analysis of a saturated soil is equivalent to a ,total
J stress analysis with Poisson's ratio close to 0.5. Note that
an approximate value for E (quite good enough for practical (jJ Dilatant 50115 such as dense sands, or negatively dilatant
j purposes) could have been obtained by assuming v = 0.5 and soils such as 50ft clay, exhibit this coupling. The former
using G = G', 50 that by expanding under shear (or developing negative excess pore
J pressures if prevented from expanding), and the latter by
o E 2(1 + v)G ~)E'
(1 +v '
1.5
1.25 x 30 36 MPa. contracting (or developing positive excess pore pressures).
~
l
(
r
1" ••
! , i
34
3.2.2 Known-pore-pressure-change analysis
35
--1 Kf Kf at a point x, y (1n 2-D), then the body force (un1tsl forcei
Bh (3-16) volume) to be applied to the soi1 skeleton is the gradient
r-. K K' + Kf
vector
~ K'Bh
r ar, conversely
'-( Kf = ""1=Ei (3-17J R, - 'í/ 6u (3-17)-
/"
h
'-{ This has components -a(âu)/ax, -a(âu)/<ly.
/"' Note that for the triaxial case 6a (the cell pressure) Px Py
3
'i. equals 60m 50 that Bh can be replaced by Skempton's B. The physica1 validity of (3-17) is 11lustrated by con-
s1dering a 3 m rise in a water t~ble in a coarse sand or
''( Note further that as Bb (or B) varies between O and 1,
r', graveI (figure 3-1). The body force vector p is zero
Kf varies between O and infinity. everywhere except in the 3 m high flooded zo~e
).
Finally, what are the disadvantages of the effective
) stress method applied to undrained situations?
'--) ,',', .\/"
A It is unnecessarily complicated if there is no interest ".
~ y
in the division of stress into effective stress and pore
A Final
pressure. A total stress anal~sis would be perfectly suitable
~I
'-"
,..-{ for the prediction of short term sett1ements of a footing on
'-' clay or, if undrained strengths are avai1ab1e (as they often
/'( are), the prediction of over-stress in the c1ay. . .....
Body force
KN/m3
value to eIement numbers so that the value can, if necessary, (i) The choice of a datum for strain is somewhat arbitrary for
'-( be applied to the eIement on one side of the node and not to
.- that on the other side•
soil. A normally consolidated caly for example, undergoes
very large strains during the process of its deposition.
'i Strains are usually, however, taken as zero at the start
") of the analysis.
'\
~,
3~ 39
~(
, J
stress or strain d~pendent constitutive law (R' = R'(o', ~))
;
\...'
will not be rigorous. The error will depend on the extent of
the non-linearity and may or may not be significant. This
J stress path dependency is similar to that already referred to
u in the known-pore-pressure-change class of analysis. 8713~I- ------,- - .
\
I ------..-
G
_o~o4
,r{ except that the load input section must be altered to read the
\.J
(l,
known-pore-pressure-change and compute the corresponding nodal
forces according to equation (3-18l. .,
\
,
3.2.4 Examples .. .1 _ot ;
Cl ~7\"OJ )~ __ -7'
~--+_._-p
I '
CI Figures (3-2land(3-31illustrate the use of the effective
stress method applied to the construction of a central clay
'3 ,J»
Cl •
1;:
80re damo (Actually hypothetical although representative of a
J
common type of dam section.J Figure(3-21shows computed pore
pressure distribution at the end of undrained construction
z
~-o~_
9/
- - - - - •• r'
1/01'
---..,
!-'l
J. ,J4::t>
4-
ª0"
C7
compared for two different non-linear idealizations (described
in chapters 4 and 7). Figure(3-~ shows displacement vectors
for two sets of analysis. The full lines and solid circles
assume complete drainage as the fill is built up in layers.
J +
(a)
ee
ac a Le
r' (Four were chosen - see section 3.6 below for explanation of
''-( layered analysis.l The broken lines assume no drainage until
r: construction is complete (points "4"). Then a known-pore- <t
ir: pressure-change analysis models the dissipation of the pore
pressure and computes the movement vector "4-0". The analyses La~er4, __\'!..__ ~
'i
r-
),
were repeated for linear and non-linear constitutive laws.
Note that the drained points "4" (full linesl are close to
t.4
\4
<h
I
I 5
97
l'
I.",
1\
"
31 _\4.7
2.i--·-----
f ~~TnJJJ""
I
_>'00
4"--- --- - ---_ .•
•••'7) 7:•• "
.•.
~ 4. n-2,3,4 I a ye r n complete
,~
-...;)
u=ccr e d i s s i pe t i on complete
ri
,---,' ':~.J!I"',:~, G '0-' -
_-
Und r a in ed
Dr e ined
40111)'8.'
ana l y s e s
,--{
000°00
o'~.'..!.. i 1'-\0
~,,~Jr:-
) o'~)04
fll /-·,,-\S
\...-" 47~
,--i
'-.--' u ,,-'300
i f II , o
,1. o
KtJjm'
~
/
.-.
_,I
I
'i
'-
~'o/~
o /0"
I'
I_~"'" O
,~t.~
...':-~"~
"1_ -'-
o 0 ...\.,
.......•.
Yh
~9~'",
, ,•
,
•
n-o
'o
"r o
<---------'
Di sp.
, m.
sC'.11e
'J o , o
",1
/.0 0· .....0
1
o '~7
C1
~G~'~
Total lood on element: 1
,{
"-,,,'
I,-{
"-""
3.3 Loadings
« -'I"
aGI. .
'/6 '/3 -'/9 '13 '/9 '/'8
«
0 "·D"·
In geotechnical applications gravity is the commonest
form of body force loading. With the y direction vertically 2/.
o o 9
({
Surface tractions arise when it is desired to simulate, Lood~ Iram lelt to righl but u.d Irom top to battom
c{ say, an embankment loading by ths direct application of a
normal and shear stress to the ground surface. A probably
~( mors common epplicetion occurs when excev~tion is simulated Figure 3-4 Equivalent nodal forces for body forces
using ths first of the two stratsgies described in section 3.5
)' below. The need to convert both normal and shear surfacs
'-.i tractions into equivalsnt nodal forces is an unlikely require-
/,
<;»
ment outside of gsotechnical applications. Structural f.e.
programs will generally be set up to deal with pressures only.
,A
'--' Another facility sei dom provided in structural programs Total lood on element side:1
A
\--) and commonly needed in geotechnical work is provision for the
~
<;»
'\
variation of distributed quantities across elements or along
their sides. Thus in excavation analysis (by Strategy 11
linearly varying tractions along element sides must be applied.
,/.ITITD'I
2 . 2 '/6
OTIJIII]
2/3 '/6
~.
,'\
'..-' An alternative to automatic calculation is hand calcul-
,i ation using equation [2-191 for body forces, and equations
.~,
(2-201 and (2-211 for surface tractions. This has been done "'~,,, O~'/3
:1 here for some commonly used elements and the rssults presented
in figures(3-~ and 8-5). These results may be used to obtain
2/3
:'(
\..- the equivalent nodal forces when in the case of body forces
the elements are rectangular ar in the case of surface trac-
Figure 3-5 Equivalent nodal forces for
Ci tions the edges are straight. Non-corner nodes must also be
pressures on element sides
central. If these conditions are not met automatic calcul-
\J ation within the program 1s virtually a necessity.
\.j
'J
é 42
43
3.4 Initial st~esses
~ will according to judgement within the extreme limits K and
K where "a" and "pU refer to active and passive respec~ively.
~ 3.4.1 Introduction FBr loosesands and lightly over-consolidated clays K' less
than 1 would be appropriate. (K' = 1-sin.' is a Wid81y used
~ The state Df stress in the ground existing before load is empirical assumption.) For dens~ sands and stiff clays K' is
c applied or excavation carried out is usually significant.
though it will not directly affect deformation calculation for
Al- typically greater than 1 and may approach K' at low stres~
levels. T will be zero. p
an assumed linear e1astic soi1 it wi1l for non-1inear analysis xyo
CL when stiffness is stress leveI dependent. In failure pre- For effective stress analysis the ~nitial pore pressure,
d dictions the collapse load (or unload) will only be unaffected u , must also be specified. This will often be y h where h
w
by the initial stress state when the soil is treated as purely i~ the depth below a water table. For c:ays Uo w~l~ be thus
u cohesive (as in the undrained analysis Df a saturated clay in defined for negative as well as ~ositive hw' The effective
(j terms of total stress). Even in this case the displacements initial stress components in this case w:ll be
prior to collapse will depend on the initial stress state.
({ ,
a = K'a
o yo
For linear analysis it is not necessary, although it may xo
\( be convenient, to incorporate the initial stress in the f.e. ,
a = yh - u
o
, program as they can be added later. The principIe of super- yo (3-21)
"< position applies. In the analysis Df excavations the initia1 T = O
>
stress va1ues are used to determine the unloading forces (see xyo
'-i section 3.5). Consequently in this case the initia1 stresses u = Ywhw
affect the deformations even in the e1astic case. o
\(
How then should 1nitial stresses be incorporated in a The initial stresses may bE read in sets and the sets
\ f.e. analysis? They may be generated by applying a gravity assigned to rows of nodes from wrich Gauss point values can
loading to unstressed ground or may be specified direct:y. be computed. More conveniently y, K and the y values corres-
f The first is unsatisfactory as will now be explained. The ponding to h = O and hw = O can te igpuc and the Gauss point
<,
,-( second leads to two alternative strategies (section 3.4.3). values automatically calculated by equation (3-21). At this
~,
point one of the two strategies ~entioned above is se1ected.
,( 3.4.2 Gravity generation of initial stresses They are as follows:
'-../
-( A vertical body force of intensity Y is applied over the Strategy 1.
'---' whole region. In the case of meshes with a box shaped bound-
( ary having smooth sides, this wi11 produce a vertical stress' The applied loads are read in to form a vector ~ which
',-, becomes the right hand side Df the stiffness equations. These
ayo = yh (h = depth below surface), which is what is required,
and a horizontal stress a = K yh which may, or more likely are solved in the usual way to give displacements from which
may not, be what is requi~~d. °K wi11 be contro11ed by the strains and load induced stresses. õ~, are calculated. Final
~~ material law. If this is isotrop~c elasticity then stresses are then obtained as Z = Zo + õZ·
(
v
'-" K (3-20) This is sufficient if non-linear computing schemes (see
o h chapter B) such as the "tangential" are used. The correct Z
J serves to define the tangential modulus. With the constant
A further drawback is the unwanted downwards deformations
J resulting from the gravity 10ading. They cannot be separated stiffness matrix schemes, however, it is necessary to have a
measure Df "force residuaIs" 50 that fictitious Ioads can be
from the load induced stresses, at least not in one analysis.
'J applied to deform an elastic body (defined by the constant
stiffness matrix) to the shape of the actual body under the
'--! 3.4.3 Specification Df initia1 stresses
actua1 10ad. To determine these f~rce residuaIs it is first
,.
\J
The ana1yst chooses se1f equi1ibrating initia1 necessary to obtain forces ~9uiva18nt to the initia1 stresses.
stresses. This will in practice limit him to an initially Denoting these forces by ~(l and the residuaIs by ~ we have
o ~
''-{ horizontal ground surface. (Otherwise a preliminary analysis
must be carried out - see section 3.4 below). a Will be ~This assumes that there are no initial strains. If there are
'" chosen to equilibrate gravity by making it equalYo yh. then following the notation of section 2.7
K o (or Ko ' in terms of effective stressesl can be selected at - F
''-(
~ ~a
o
~e:o
'-\
1:
'-' 45
o (
44
part of a coarse mesh such as might be used for the problem of
figure (3-6). y here is 18 KN/m3. Note that with Strategy 2
G ~ ~ + ~o - ~ (3-22) each upward force of 72 KN will be cancelled by y acting down-
wards ave r an area of 4 sq. m. AIso note that the net horiz-
~,
when ~ is made up from element contributions ontal ~ forces are zero at inside nades. At the boundaries
they are in overall equilibrium with crxo' i.e. crxo integrated
over the 6 m height equals the sum of the horizontal forces.
'1 (= J ~TR;dv (3-23)
This is 216 KN.
'-,
Q.. (notation as in chapter 2).
and similarly for ~
({ O'yo·lah
']5 Ia 8~8
J6
The forces ~ equivalent to the initial stresses go
C; directly into the ~oad array and the applied loads are added
8
®
to them. With this strategy the applied loads must include ~
~.
the gravity body force which is the source of the vertical
hl 7~ ® 7~ '1)6 16~16
%
18
5L.
lB
51.
ç
(-
cause the mesh to expand upwards. 6m
n® 7~ 36
5'6
3~ ~---Q--tO- 32
;, ©
:
40~40
®
j
"
5·1. 51.
z' 64
#///#////,; 1"'/,'////,'/$ r/////////.t: z,..
o
a
90
\...; 56 56
,''--',
~. .
90
180 180
©
(,
,-\.
\..-,
T
h ~
Assembled lorces I Sol 64
90 90
64
"
slrclogy 1 slralegy 2
'-o For the example of the footing Strategy 1 is clearly
more convenient than Strategy 2. However, for problems
Q Oyo.'Yh in bclh slralogies
involving excavation, particularly if the boundaries are
<l complex, Strategy 2 has advantages.
in section 3.5.
This will be made clear
r: .
~ 46 47
r·
be carried out. The procedure is given in section 3.5. The
\ height of the horizontal surface will be chosen in conjunction
j with'the assumed Ko value to give horizontal stresses approp-
riate to the main analysis.
~.
,
.
(,
3.5 Excavation
G
~
3.5.1 Single stage
ê. U3
au~~ __•
I J o
thê assembly of the nodal forces from the ghost elements. This
probably outweighs the advantages of ghost elements.
s....,. :2
G I II ~p.c~~••..
\~ The procedure for subsequent stages is identical to that
~"'/~j for the second.
(j \l.!9l : I
, ,
2.S'~ 2'$ I
\.(
I
3.5.3 Example
':>'~7:2"
i7e -C.H9'Ct-Cl -0.29791(-01 -, ./ 0.291 -0.140
(.I. eU r..H~<!.H-Ol -O.370r,9('OO
consolidated clayl. The unloading pressures to be applied to
0.011 -0.186
15 ~ C.ll171f ce
-0.139'91' ao--,::=- the excavation boundaries (using Strategy 1) were obtained as
'-" 1" . (I.H0571 CO .•o. U04l[ 00.... " /2&3 0.115 O.OUI
follows. On the side of the wall:
A EUr~fl:t 1(0. 111 /,,/ /'.I26~ 0.1V2 -0.22"
Li HOill ~ .c • J CO~,. Y (OriP. ' "" '285
LS2 i/' -0.037 0.411
6
-O.H'SZE-C1 "G.466!OE-01 ~
~o•• -e ..74Cl H-Cl (I.•l2709. 00 // -p = O = U + O
iH -O.:UUCE-C1 C .• 131021-01 XO O xo
cS
i;'"
Hr
;''''1
:~:~~~~:~.~~
":;:H~~::~:
-ç.eCOZ1I-1C
.;/ -O .•17l0iJi-01,.;7
and ao 011tor oth.r nad••
01\excavated bo\ll\dar,
Introducing 0xo " = Ko 'a yo ' = K
, (O - u 1
U4
;:>H
P.1t2CU CC "v.l;:~:l. DO:; o yo o
C .• 1147U CQ C.1tUj[-01 Lo.d. to b. applhd In
~ tu r.;-;IH :.0. 112 • ta,. 2 (Trpe LD data) and u
o
= Ywhw' a
yo
= yh we obtain
'.Ct'f t.::J. J tO"'.. Y CO"'.
'i z~.; C.•J7UH cr. 1'.4CúH[ 00
v ~~;,-~
:~!i~
~~
-~~~"1~~ ..V
~!~l~- _o' -p = (1-Ko')Ywhw + Ko'yh (3-25)
/\/~o.2.~:"_._--:----
Supl. trOIlStua 1 output On the base of the excavation:
~
G. -p = yH (3-261
\
'--,
i
\':..-i 51
50
lhe soil was treated as linear elastic with Young's Nontheless, data preparation for programs incorporating
~ Strategy 1 is perfectly straight forward for single stage
(( modulus increasing with depth (considerable information 1s now
'.....• available on appropriate E values for London clay: St. John, analysis where the excavation sides are vertical and horizontal
or sloping if Ko = 1. lhe example Df figure (3-10) used a
r 1975, Simpson, 1979).
~ Strategy 1 programo
lhe analysis enables predictions to be made Df the
~ Single stage analysis will be adequate for many purposes.
expected deformations and the bending moments in the wall.
ri Multi-stage will be necessary where distinct construction
'- stages occur such as the incorporation Df struts at different
3.5.4 Sum up
ti levels as an excavation is deepened. When non-linearity is
incorporated in the soil constitutive law a multi-stage
(I In general, programs incorporating Strategy 2, since they
do not require the user to specify unloading forces to de- analysis should give a more rigorous solution which may differ
(j stress the excavated boundaries, lend themselves to excavation significantly from a single stage analysis. (lhe difference
analysis. lhe user must, however, apply gravity to alI between single and multi-stage analysis for non-linear soils
((, remaining elements. lhis is not required with Strategy 1. in the parallel problem Df fills is considered more fully in
the next section.)
~ ;- - - - - - -,- - - - - _.~--- ._--
(\
t' --.-
3.6 Fills
r=:
(i
~.
I
,'
-------.-- ! -,~--;'-+\-~---,-,,---:
- - - - -j -
3.6.1 General
I.:.::
'
" . Most fills are constructed in layers which are thin
\ :::::: ::}; Cr-=::~:'·-I,~''---.l--~·-·~~';-+-,...y"1''J ;
compared with the finished fill depth. It is usually
necessary, therefore, to use much thicker layers in numerical
'1 modelling Df fil1 placement. lhe number needed depends on the
___._ ___ _ .---i.-t--
-·----rr--- - -.- - -- --- information required. If the interest lies in the soil under-
1 ----.-------1'-! lying the fill very few layers are required since both stresses
and displacements are insensitive to the number. Indeed it i5
'{ I
.mm.mm:-"ur-u \ .-u:r···um
i often adequate to assume the fil1 is placed in a single layer.
\..-' However, if the interest lies in the fill itself, more layers
r ,
",,-, are required, relatively few - or even one - if only stresses
i are required. (Clough and Woodward, 1967), more if displace-
~ ments are needed. About 10 layers have typically been used in
I I \ t
analysis of major dams (Clough and Woodward, 1967) Kulhawy
i - - - - - - -.---- - - -' - --- -- -'-
,,
I
- - -.- -- - 't- - - - -- --- - - ---
and Ouncan, 1972; Penman, Burland an~ Charles, 1971; Penman
'\ and Charles, 1973). lhis number could probably be reduced
'-'-' considerably with little effect on the results. Naylor and
~- Jones (1973) have shown that for fills which are wide compared
I with their depth (i,e. the one dimensional case) five layers
(l, ,
,
I will produce an acceptable approximation to an infinity Df
I
layer5 for quite severely non-linear soi15.
C\ ,
I
52 53
)
'-(. analysis of the fill as a whole can be achieved using linear
s
(
'-V(
I
analysis for sach laysr. Ths slastic constants ars simply
altsrsd bstwssn layers to suit ths changed stress (and/or
strainl state.
matically.
This may be done crudely by hand, or auto-
The former has the attraction that the analyst
ia forced into e dBteilBd ewereness of the computed results.
\'0./
I
o 3.6.2 Single lift ~]U
{
v Gravity loading is simply applied to alI the fill elements.
'( The idealization is that of the fill being brought down to
'- sarth from outer space, or more realistically in the case of
r(
\...... a model, being spun up to speed in a centrifugs.
\
\(
-
"-( 54 55
I
) I
\.j
I
\
r'.
I
I ..-/
."........-----
----.
'(
'-'
',---,
(
+
\~~~-.'
,~",-/~,~"C=õ",
\- '.,:.,'
'S"
'e'
/ .• I·O'S'"
'--:~::::::"'/
.J"
,?;] (! ; ) 1
~ ,+
'S"
/ :t \ •
"'-(
(
'-'- ~"
~
t,~
(e J Sfng l e 11ft - oon-Lír-ee r-
I
,I
I:
'
(J
(b) "So+t " f111 - r r-aece
'.I
'~
'-1
,-( /,./", . .,../ / I
/' /. / jt
\' -r .> ...../ /
'\ <,o ".
"1
fb) Single 1ift - lineor Figure 3-13 Overstress ratios
(
ratios have been used to show the predicted development of
~ yielding. In (a) a fill stiffness ajout one quarter of that
'- of the underlying clay (the fill was the same clay type re-
~
I
'~'
compacted) was assumed with Poisson's ratio (v) = 0.35. In
( (b) the fill was given negligible stiffness and v = D. The
'-, analyses were in terms of total stress. The foundation was
,\ idealised as a Tresca material (cu increasing with depth),
1..../
,(
\.!.--
-",
...... .~
-,
the fill as linear elastic. The much greater devslopment
overstress in the foundation in analysis (a) is due to the
of
\
\_j
!
1-..lJ
57
56
--./ A further point arising from the example of figure (3-13l surveyed. Satisfactory interpretation becomss difficult when
concerns the prsdiction of a failure in the foundation. If a datum is not establishsd until the fill has been raised
é failure is to be simulated completely (e.g. by reducing Cu by
a iactor, or increasing the fill unit weightl either the fill
ssveral metres above t,he device.
d
'--' must be assumed non-linsar and capabls of y1elding (or crackingl Displacements within a fill can be obtained bY adding
vectorally the movements at the base of each layer from each
l or if elastic must be assigned negligible stiffness. The
solution (al which indicatss'a low safety factor, as evidenced analysis. This is illustrated for a three layer analysis in,
1 figure(3-14l. The displacement at ths crest is zero. If the
'- by theextensive
istic.
overstress, may thsrefore bs a little optim-
Complete collapse may result if the stiffness were analysis incorporated the effects of consolidation or crsep the
Q rsducsd to that of analysis (bl but with v = 0.35 retained. procedurs would still apply (albeit with some approximationl,
an appropriate time interval being allowsd betwsen layers. The
C1 3.6.3 Layered analysis resulting deformation would relate to the point in time when
construction had just finished. In this case there would, of
C? course, be further movement.
Each new layer is treated as for a single lift analysis.
(( Gravity may bs applisd to the elements of the new layer or,
alternatively, a pressurs distribution of intensity yóH (óH =
\í layer thicknessl applied to its base. The differences referred
CJ to above betwsen theôe two approaches become less as the number
of layers is increased. So also is the effect of the new layer
'\..( stiffness assumption.
Scol es
'-( As in a single lift analysis the displacement within the uo s t r-eem T-----:= Geom:
o .to
l...-.-.I..----
m.
~
l
1
explained in the next section. The selection of the stiffness
of complete layers is clsarly important for settlement pre-
diction, and should relate to the stress leveI as has been ~"+'-'l-
'+-)--
mentioned.
)
l
There is a choiçe with the f.e.mesh. Either the same x • MeesurBd at end of
l· mesh may be retained throughout, or elements can be added as
construct1on
C the new layers are placed. In the former the elements above
ths new layer are assigned neg11g1ble st1ffness to become Leye r ' '<'
-~~>,."..,.f..-->:. ~
~ "ghost" elements. Th1s has ths advantags of simplicity but
G
(
y
ç 58 59
c 3.7 References
Ci
. (
'('
,"
(..1
61
). 60
by strains Ed' Eq' General do = K dE (4-3)
m v
~, 3. A Mohr-Coulomb or similar type of failure criterion 'do = 3 G de (4- 4)
.ri q q
should be satisfied. This implies that the tangential
,--,' shear modulus tends to zero when yielding occurs (see
í( below). These two equations are of general validity (they include
'-...I the plane strain case) except that equation (4-4) is subject to
ri 4. On unloading, there is an abrupt increase in stiffness. the same qualification as equation (4-2).
'-..V
, (
'--' 5. Over-consolidated clay soils experience an abrupt reduc-
tion in stiffness when the pre-consolidation stress is o, Od Bi-lineor
~ reached.
í(
~K,
1 , K-G'
"--' 6. Stiff clays and dense sands tend to be dilatant (i.e.
J expand on shearing), whereas 50ft clays and loose sands
tend to be negatively dilatant. The dilatancy depends on
,
K,+,G,D
1
q
r:
The advantage of \Iorking in invariants has been pointed
out in chapter 1 (section 1.3). The elastic relations in
1
K,f/ ~K
terms of invariants are as follows. The reIations are in 1
'\'
differential formo The moduli are tangential. implying linear-
)1 ity over the small incremento Ev EQ
2~
(~
Plane-strain
do
s
(K +
I
/3 G) dE
S
(4-1)
Figure 4-1
(b) "Conical" yield surfoce
Stress-strain
version in terms of generol invarionts
J
l
(
y
y 62 • 63
,
'-K should this occur. 0m' 0q version
(' (4-9)
Four constants need to be defined: two Rlastic, and two K K1m + c Km o m
j
defining the yield criterion. The former are K ,G o r , if
G prsferred, E
í
0q~MOm+N criterion.
(j
If M and N are made identical to S and T, respectively, the At yisld, G must tend to zero, 50 that equation [4-8)
C( 'compromise cone' yield surface results. They may be made lss5
or greater than this to suit a variety of possible yield sur-
becomes
« G G1s + Cl.GsOs + BGd °d [4-8) The procedure incorporating a conica1 yie1d criterion is
the same. S and T in equations[4-13) and [4-14) are replaced
(j
'-
{ 64 65
.i by M and N, respectiveIy. This imposes the two constraints
The model is unsatisfactory compared with the first two
on the three constants in equation (4-10).
models described here in that it requires up to twice as many
J constants to define it. Its strength ~ies in the extent to
K and CXK in equation (4-7) or (4-9) do not affect the
1 which it has been used. It has evolved over several years with
\~ yieId criterion and must be specified separateIy.
feedback from instrumented prototypes (e.g. Oroville Dam).
li The model can be used for unloading in much the same way
Those experienced in using it can expect to make good predic-
tions. A beginner would be well advised to choose a simpIer
as with the bi-linear modelo Selection of the unIoading stiff-
~ modelo
ness is not, however, 50 obvious. A simple way to do this is
J to set SG to zero on unloading. K is unaffected (as with the
bi-linear model) and G abruptly assumes a higher vaIue. If 4.5 Comparison of models
'J the yieId condition is again approached(i) SG assumes its
<J former value. The bi-linear and K-G models are preferred to the hyper-
bolic on the grounds that they are simpler and involve less
:j 4.4 Hyperbolic model constants in their definition. Which Df these two is to be
preferred will depend on a number Df factors, including the
~ objectives of the analysis. An attempt will now be made to
This model relates accumulated stress to strain, 50 is
clarify the issues.
i,( not a differential modelo It is attributed to Kondner (1963).
It has, however, been extensively developed by Ouncan and his
The bi-linear model requires only four constants (Ke, Ge,
~( associates in the U.S.A., and is sometimes called the Duncan-
c, Ijl) to be defined as against the five (K1' G1, CXK,CXG' BG)
(- Chang model (after Duncan and Chang, 1970). It was originally
-, formulated to fit undrained triaxial test stress-strain curves,
required for the K-G modelo The choice of these five is not
as arbitrary as may seem at first sight since, as has been
and in that role had a simple form with two constants to ne
\1 defined. It subsequently grew in complication as it came to be
shown above, incorporation of the chosen yieId criterion
restricts the choice to three. and of these in many cases only
applied to realistic boundary value problems, both drained and
':r undrained. The two basic parameters ceased to be constants and
one (e.g. SG) will be important. Nonetheless. the task is
\.. more demanding than the choice of the four bi-linear model
were defined by equations which required the definition of up
J constants.
to 9 constants. This formulation is described by Seed et al
l~ (1975). In the analysis of Oroville dam Kulhawy and Ouncan
(~
6y
0"
"o::,
0,>
'--' used in practice). It can be shown that if 02 = 03 the reci- .x:
G
procal of "a"is the initial (1.e. when 01 - 03 = E1 = O) tangen-
tiel Young'9 ModuluB, Eo' which i8 equel to 3GQ [for plene .OL /
3t}' 9-8 .'l:
()
strain condition 1/a = 4 G ). The reciprocal o+ ('b"would be ,0-7 / 0'
J the failure deviator stress~ 0df' approached assymptotically /'\I{I 56 /
J (as with the K-G model) were it not for a refinement. This is / 30 / ./ o
slress po í hs
to select b so that a failure stress in excess of 0df is o~--,,--_~~_-;,!:-_-=z:::==:::::::;~EI!~leeel ive
i-J approached, the stress being cut off when 0df is reached. 500 1000 as
kN/m2
~ Conlours are of be st lil surlaee G, 0-5.270.- 270d MN/m'
Cr' 66 67
the same yield criteria. Since in this case the plasticity is
ç' A pay-off is that the 'K-G' model can better approximate
the abserved behaviour of soil, in particular properties 1 and
'perfect' (see chapter 5 for an explanation of thisl, the sa~e
q 2 Df section 4.1. This is illustrated in Figure (4-1) and
also in Figure (4-2), which shows the results of consolidated
collapse load ror unload in the case of an excavationl
obtained from the two models. 'The only differences
will be
occur in
C
« deviator stress. An arbitrary contour, say 0.9, may be taken
as the boundary of a yielded zone. Ia) Si-linear Ib) Elastic- plaslic
)
~ 4.6 Variable eIesticity or elasto-plasticity? Figure 4-3 Comparison of yieIding.
6 To answer this it is necessary to understand the differ-
ci ences between purportedly equivalent variable elastic and
elastic-plastic models. First, in what way can they be equi-
Figure (4-3l shows a pIane-strain prism of Tresca mater~al
on a smooth table, but anchored at the centre of the base. :t
6
,..(
valent? Consider a bi-linear variable elastic model incorpor-
ating a Tresca yield criter1on. Th1s would be appropriate for
is loaded by apressure p on its sides and p + 2 c on its tup ,
so that it is on the point of collapse. A very small latera:
\......
a total stress plane strain undrained analysis of a saturated load lIq is then applied. The separate deformed shepe s pred~c-
clay. The Tresca 1s the special case of the Mohr-Coulomb ted by the two models are shown by the dotted lines. [ActuaLy
cS criterion w1th ~ =0. Equat10n [4-5) therefore applies and the small Ioading would cause unIimited deformations. It must
/\ takes the form be supposed that 6q is removed after the p r í.srn has deformed
\:.:.,
to the shapes shown.l The point is'that eIasticity models
<l 0d ~ 2 c [4-16) compute incremental strains according to the inorementaL
stresses, whereas with the eIastic-pIastic models it is the
C! The elastic-plastic Tresca model described in the next chapter aooumuLated stress which controls the.pattern of movement.
can have identical pre-yield eIastic constants and incorporate
Cl (ilThis impIies that the frictional material is not
G ·perfectIy plastic·, see chapter 5.
(.
\
y
y'
(,
68 69
Both measurement and intuition suggest that the elastic-
~.
plastic prediction of movement is more realistic. (A mountain-
ser may dislodge a boulder weighing half a tonne. The direction
~ Df its movement will depend on the resuItant forces acting on CHAPTER 5
(, it, not on the force he exerts, which is negligible by compari-
(
c son.) The example given above is an extreme case, and undoubt-
edly exaggerates the difference between the two models. If the
ELASTo-PLASTIC ANO ELASTo-VISCo-PLASTIC STRESS-STRAIN LAWS
G perturbation
between them.
6q were vertical, there would be no difference
G
c.l It is tempting to use variable elastic models since they 5.1 Introduction
can usually be implemented with less trouble and at lower
computing cost than elastic-plastic, yet they must be expected In chapter 4 variable elastic non-linear models were
~ to give a less realistic prediction. In many cases, however, discussed. The simplicity of these models has led to their
~ the differences will not be significant. application in many practical problems. They have, however,
(( 653-665 .. (/)
<:»
c1 SEEo, H.B., oUNCAN, J.M. and IORISS, I.M. (1975). "Criteria Strain, E
and methods for static and dynamic analysis of earth dams",
Cl Cri teria and Assumptions for Numerical Analysis of Dams,
Figure 5-1 Uni-axial stress-strain relationship of an
elastic-ideally plastic material.
cI 564-588. (Proc.of Int.Symp. at Swansea, Sept.1975.
Eds.Naylor, Stagg, Zienkiewicz.) Figure(5-1)shows an idealized stress-strain curve
a obtained in a uni-axial test on a material. Note that the
stress-strain relationship is linear elastic up to the point A.
G
C{
"
'-"
C 70 71
é If the bar is unloaded from a stress lower than Yo. known as
the 'yield stress·. the strains are fully recovered and stress
or functions of them. The strai~ invariants are required
incremental (or ratel form ando again. can be the strain
in
incre-
L and strains are both zero. However. if an attempt is made to ment invariants or functions of them.(il
cL apply a stress even infinltesimally
strains increase indefinitely.
larger than Yo. the
If the uni-axial test was Suitable invariants have been defined in section 1.3. For
strain controlled,in which increments of strains were applied stress we generally use em' 0q.8 for three-dimensional formu-
~ anca the strain Ey is attained, any further increase in strain lations. and 0s.od for plane strain. Much use will be made
\l would not then increase the stress beyond Yo' On unloading of the latter since so many geotechnical situations are ideal-
ized as plane strain. However. in the definitions of yield
« from a strain E. strain Ey is recovered. leaving a permanent
strain (EPl equal to (E -tyl in the material. A material
conforming to the idealizea stress-strain curve of figure (5-1l
functions which follow it is possible to use only two invar-
iants even for three-dimensional applications. provided these
<T is called an elastic-ideally plastic material. . are properly chosen. Thus the Mohr Coulomb yield function can
([ be fully described in terms of 0s.od' Conversely, the Orucker-
Prager criterion requires only 0m and 0q' Two invariant forms
(f will be derived here. For completeness the alternative 0m'o .8
formulation will be given for the Mohr Coulomb function. q
(j' Y'
C( o
5.2 Basic concepts
/"
''""
<11
.!:
~ til 5.2.1 Yield Function
C(
In the uni-axial elastic-ideally plastic case illustrated
~' by figure (5-11 the yield stress Yo indicated the onset of
plastic strains. In the multi-axial situation we cannot talk
~. Ey Ey'
about yield stress as there are more than one components of
stress. Instead, we define a yieZd function (Fl. It is a
scalar function of stress (stress components. principal stress
Stroin,
c. E
ar stress invariantsl which indicates the onset of plastic
C. Plastic (permonent)
strain, EP
strains and can be symbolically written as
Cl ponding strain Ey'. Thus the yield stress has increased from Equation '(5-1) represents a surface in a stress space.
Thus if we express equation (5-1) in terms of principal stress
Yo to y'. A material conforming to this idealized stress-
C1 strain behaviour ís called an elastic-plastic material. the yield function can be plotted as shown in figure(5-3).Such
a plot of yield function in any stress space is called a yield
G Safare we can generalize the uni-axial stress-strain surface. The space enclosed by it is the elastic domain.
« behaviour. we need to define the variabIes equivaIent to uni-
axial stress and uni-axial strains in the muIti-axial. i.e.
(( two- and three-dimensional situations. In the three-dimensional
case there are six independent components of stress and six of
C( strain. Adv'antage can be taken of the assumed isotropy of the
......•."
• "-..•.. Ós-, :~~.
\
'r )
72 73
)
c_ In th~ multi-axial situation, the yield function is a
f \
\....:.
function Df not only stresses but also. plastic strains [plastic
strain components, principal plastic strains, or plastic strain
Ll °3 invariants), It is therefore more generally written as
l) F( 0, e;p) o (5-2)
\~
(
(
(
I' 74 75
"-.J A description of tha flow rule in which the direction of
\J
IP where o (a) i' F (a)
Cf F (Q l = O
Flow rules are of great importance in the theory of
" O" Ê.,
« / r - plasticity. This is particularly the case with an associated
flow rule since the uniqueness of a solution to a boundary
«,"
value problem depends on it. Moreover, the collapse theorems,
which have a useful place in soil mechanics in finding upper
\1 and lower bound solutions, also require the assumption of an
r- associated flow rule [see chapter 6).
I..(
Oz. Éz
rr , Flow rules also govern the dilatancy duriDg plastic
(Al
'<r straining which, in turn, may have great influence on the
strength. We take up this important area of discussion in
~ s8otion 5.7 and olso 1n ohopter 7 where 1t 1s shown to be a
ç °3·Ê.3
central feature of the critical state modelo
r(
\......
5.3 Yis1d function .- specific forms
Q As an alternative to the stress tensor, O, ths compo-
°2.Ê.2
Q (B l nents of which depend on the physical direction of the coordi-
nate axes, the yieId function may be expressed as a function
G of the invariants which, since the material is assumsd iso·
tropic, are not direction dependent. Adopting the ~nvari~nts
Q Figure 5-5 Assoeiated (a), and non-associated (b), we have decided to use we ean write for general applications
flow rule and normality condition.
G
F(o o e) o (5-6)
G m q
-J,
(
..
:
(
7
(
j 76 77
\.' equation (1-2), equation (5-8) becomes
,(
v., 0d) o ( 5-7)
F(OS ' vr:': °Y )2 2
o (5-9)
L F =,tCo"
X
- +4T
xy
- 2c
~,
We shall now discuss various specific forms of yield functions
indicating their applicability in geotechnical analysis. 5.3.2 Mohr-Coulomb yield criterion
<L The form of this yield criterion in terms of shear stress
5.3.1 Tresca yield criterion
~ (T) and effective normal stress(o ) on the failure plane is
In chapter 3 the techniques of geotechnical analysis well known and is given by: n
~ were discussed in terms of effective as well as total stre55.
(I If a total stress analysis is adopted for cohesive materiaIs,
ITI ~On Tan~ + c (5-10)
the angle of internal shearing resistance (~) is obtained as
« zero for undrained situations, as seen from the Mohr diagram
~ In terms Df °
'Od invariants, equation (5-12) becomes parti-
cularly simpl~. It is
( o) oí oj o, o~ o; oi'
'--'
/(
'-.,/ F = 0d - 20s Sin cp - 2 c Cos cp o (5-13)
m q
l l d E P
1
= d ~ aF
aa,;- = d)d1-Sincp) (5-17a)
F = -o Sin $+ ~o cose - Sin6Sin$ J - c Cos$ = o (5-14)
L m 13 13
~ d c P = d À~ = o (5-17b)
5.3.3 Von-Mises yield criterion 2 a02
cl p
((
This yield criterion was proposed by Von-Mises in 1913
d E3 = dÀ~ = - d 1,,(1+ Sin cp) (5-17c)
primarily for metaIs. It is frequently used for total stress ao 3
({ undrained analysis of soil masses and is an alternative to the
Tresca yield criterion(equation (5-8)), discussed earlier. Summing. gives
(( It is given by , =:p = 1
p
dE
+ dE
2
p
+ dE
P
3 = - 2dÀSin cp (5-18)
~- F o -2c
q
o (5-15) Note that dE: also equals dE: in this case.
r:
\{ 5.3.4 Drucker-Prager yield criterion Since dI" is positive, equation (5-18) implies that the incre-
ments of volumetric plastic strains are dilatant (note the
C; Drucker and Prager (1952) presented an approximation to negative sign) and are proportional to Sin cp for the
Coulomb criterion with associated flow rule.
the Mohr-Coulomb yield criterion. They modified Von-Mises
C{ yield criterion to include the influence of mean stress on
\" yielding. The criterion is written as
ç' F o
q
~ 20
m
Sin $ - 2 cCos $ o (5-16)
(
(
81
~ 80
d p associated fIow ruIe. In fact, if W =0. the pIastic potsntial
Ed . From equations (5-17a) and (5-17c) we have
~ function (equation (5-21)) reducss to
(
r
<::
dE P =
s
d E1
P + d EP
3
= - 2 dI. Sin cjJ (5-19a) o = ad - constant o (5-22)
L
,{ and de P =
P
~ (dE1 - de p) = dI. (5-19b)
and conssquently the rats of diIatation of pIastic strains is
zero. This is illustratsd by the vertical arrow in figure (5~B)
d s
<:»
,{ whence the slope of the flow rule vector, dEcF / dE~, is - ~SincjJ.
"--'
It can be. seen immediately from equation (5-13) that the slope
rL of the yield line is 2Sin cjJ. Consequently the fIow ruIe
vector is at right angles to ths yieId surface in accordance
with the assumption of normality.
~ The a ,ad pIot is useful because it shows at a glance
~. whether the fuatsrial is dilatant or noto If the flow rule vector
inclines to the left of vertical dEsP ( = dE p) is negative
({ and thsrefore the material is diIatant. If ~he inclination
fi' is to the right, the material is negatively dilatant.
~ °d
V
It oen be ahowM by e perellel ergument thet the flow ~\
r:
\ rule vector in a 01 ' a3 space is also at right-angles to
the yield surface. This is left as an exerci se for the reader. --~. -211
'( Equotlon (5
r:
The dilatancy predicted by the Mohr-Coulomb yield criter-
\ ion with associated flow rule is unrealistic. Overconsolidated o;
\ materiaIs do dilate on shearing but not so much, nor is the
rate constant. It reduces as shearing progresses. Oilatancy
) controls the collapse load of many geotechnical structures.
Figure 5-B Mohr-Coulomb yield function and non-
If it is prevsnted by, for sxampls, boundary constraints, the
) mean stress increases and this, in turn, increases the strength. associated plastic::FJ:.~~tic.,p(J~Elr:t_~,a_~
function.
\ .' In undrainsd situations ths pore water provides this constraint.
"\
\.:/
The increase in effective stress and strength is then assoc-
iated with a drop in the pore pressure.
One could thus choDse a non-associated
the desired dilatancy.
flow rule to give
An example of this is the treatment of
\ the "super-critical" yield surface in chapter 7, where a non-
<::
A part1al ramedy for th1a a1tuation 11a9 1n edopt1ng e associativs fIow ruIs is dsfined suoh that ~ reduoes as (J
~ non-associated flow rule. Let us consider a plastic potential
\ .•../ increases, becomi2g zero at the critical state. In the eve~t
,~ function of the form of non-availability of dilatancy characteristics of the mater-
v ial it may be prudent to analyse for the extreme conditions,
\ Q = a1 (1 -Sin Wl - a3 (1 +Sin W) constant = O (5.20) viz: an associated fIow rule (O = F) and a non-associated flow
"--'
,'( or O = a d -2a 5 Sin W - constant = O (5.21) rule with W
=0 and then make engineering judgement for the
'-' purpose of designo
where W
is an angle less than~. These equations are the
~ same as equations (5-11) and (5-13), respectively, except that Finally, note that the Tresca yieId criterion is a special
W has been substi tuted in place of ~. The yield function is case of the Mohr-Coulomb obtained by setting ~ =0. The W =0,
~ unchanged. The yield function and the plastic potential func- or zero dilatancy case, therefore results from an associative
tion related to 0s,ad axes are shown in figure (5-8). The
1 normality of plastic strain rates is now with respect to the
fIow rule. This is very useful since it makes this relatively
simple plasticity formulation applicable to the important cIass
Q plastic potential function, as shown in the figure. The rate of analysis of undrained saturated soils in terms of total
(
of dilatancy would be proportional to sin W as shown by equa- stress.
\.> tion (5-18). Since W
< ~, the rate of dilatancy with the non-
assDciated flow rule would be less than that predicted by an
((
C(
(
\
(
82 83
5.5 Hardening law - specific forms position of subsequent yield surfaces after initiaI yieId, on
a 0d' as pIot. It is noted that the yield surfaces simply
5.5.1 Isotropic hardening expand or contract and alI subsequent yield surfaces are para-
11el to the initial yield surface. Hardening or softening of
,-, In the elasto-plastic materiaIs, the hardening law this type is called isotropic hardening ~r softening. The
j
x.. determines the manner in which the yield function (or yield critica 1 state model described in chapter 7 provides another
,( surface in stress space) changes due to plastic strain1ng. As example of an isotropic hardening/softening modelo The criti-
<:: an example. let us consider the Tresca yield criterion given cal state model, incidentally, is unusual in that the paramater
{ by h is the vo~umetric plastic strain.
'\..,-.
,.(
o - 2 c o
"-'
F
°1 3
5.5.2 K1nematic hardening
L ar F 0d - 2c (5-23)
There is another type Df harden1ng
Tresca yield function can be written as
law in which the
rL
If c is assumed as constant, we have ideal plasticity.
~ Suppose c is a function of the hardening parameter h. This F (°1 - 03 - edh)) - 2 c o
will usually be the deviatoric invariant EdP in plane stra1n.
I or E p in three dimensions. Then we can write (Od - cdh)) - 2 c o (5-25)
q or F
.~
,~,." c = c (h) (5-24)
,
I where c ih) represents a hardening function. Equation [5-25)
represents a yield surface in which the elastic domain trans-
lates 1n the stress space (figure 5-10).- Such hardening is
called anisotropic or kinematic hardening. They hav8 been
-( Subsequent yield surfaces recently introduced in soil mechanics te model behaviour of
soils under cyclic and transient loads. Oiscussion Df this
\
==~---~==-
A topic is beyond the scope of this book.
((, hardening
"(
Flod,h) =O
s: o 1 ;' <
) u
B
lnitial yield surface
\,/
( o,
'---- C Subsequent elastic domain
Ia) Ib)
/~
r,\ ~
o
'--./ Figure (5-9) (a) Strain hardening/strain softening functions,
\.
'---" (b) Initial and subsequent yield surfaces after
lnitial elastic domain
h~rdening/softening. C
J
.:z Figure (5-9 a) shows a typical variation of ct'\~f~l(:~ Tresca as
« material with h. In curve A the value of c "incrÉll3seswi th
increase of h, while in curve B, c decreases with h. Curve
J A represents a strain hardening function while curve B repre- Figure 5-10 Representation of kinematic hardening.
sents a strain softening function. Figure (5-9 b) shows the
,J
\
/
~
) 64 85
l.i
5.6 Formulation of elasto-plastic stress-strain law It is a basic assumption that during plastic yield the
L stress remains on the y~eld surface. This implies that
C
/(
We are now in a position to gather the threads together d F (~ r h) o ( 5-30)
x., to obtain a single relationship between increments of stre55
.I
, and increments of strain. We thU5 shall seeK the form of O
We now introduce, for brevity, the symbols êf and a
'-....' in the equation -ep
to repre5ent the vectors dF/da and dQ/aa, respectively. -q
da = O d c (5-26)
~,
_ep
'( Expanding (5-30) by the chain rule
,~ where da and d E are increments of s t re ss and total strain
(elastic-plus plastic) tensors and Qep is the elasto-plastic
rI matrix. Firstly, using the additivity po st ulete , we can write ~f
T d
~ +
dF
ãh .
dh'd
dE P ~
P o (5-31 )
;
'c. ,
or d Ee = O
-e
-1
da (5-28c) and substituting in (5-33) gives
) Substituting (5-27b) in (5-26b) leads to dF
\...,.'
H dh ) T (5-35 )
ãh (
{ dfP êq
da O [dE _ - dEP]
= _e (5-29)
,
'-'
"\
Thirdly, the plastic strain increments are related to the flow Pre-multiplying equation (5-32) by 2fT g and noting that
'.J
rate, i.e. O O' -1 Ls the identity matrix, leads to e
~e ~e
,-j
T T T
"\
d P = d À~ ~f ~e d~ = ~f d ~ + ~f ºe d À ~q (5-36)
E (5-5)bis
,--,' da
'( We shall derive the elasto-plastic matrix for the general T
Eliminating ~f da by (5-34) and rearranging, gives
non-associated case i n which Q" F. The associated case can
11\
•......• always be obtained by setting Q= F. 1 T
d À = - ~f ~e d E
(5-37)
:1 Fourthly, the yield condition is given by T
6
<i"-- in which 6 H + a O a
-f -e -q
F(a,h) = O (5-3)bis
.~.." ,Rearranging (5-32) leads to
.T' in which h= h(fP) is the hardening parameter referred to
...\ previously. Note that if the material is ideally plastic, so da O dE-dÀO a ( 5-38)
-e _ _e -q
.,J that there is neither hardening nor softening, there is no vari-
-tion of F with h, l.e. âF/dh = D.
J
(
.) ,
,\
'.....
\
B7
(( 86
Substitution Df (5-37) into (5-3B) gives the required relation
Y between dç; and d!;;.i.e. equation (5-26). The matrix ~ep is
model has close links with
fact been used as a device
elasto-plastic theory and has in
to obtain results pertaining to
c elasto-p~astic behaviour. The elasto-visco-plastic model is
,T"-',
then :
I
flep = ~e
D - 1:. b b T [5-39) described in the following paragraphs.
S ~q ~f
~, The essential and rather phiIosophical assumption in
in which elasto-visco-plasticity is that alI plastic strains [called
1
'-...' b = D a visco-plastic strains in this theory) in the material are
~q -e -q
developed with 'time'. It may be recalled that elasto-plastic
~ = strains are time-independent or, in other words, are produced
~f ge ~f
'r'-- instantaneously.
elasto-visco-plastic
Figure (5-11 )shows a rheological analogue of
material. It consists of a spring which
In the case of ideal plasticity, the yield function is
({ independent of hardening parameter, i.e.
is in series with a dashpot and a slider system in parallel.
The slider denotes yielding and permanent strains which cannot
~ <lF
take place instantaneously due to the viscous dashpot which
«
r- so that by equation
ãh
[5-35), H
D
D, and
needs time to strain.
behaves purely elastically,
elastic spring.
Thus, instantaneously, the material
alI stress being taken by the
-r
r>,
'-( e aT b (5-40)
r'o ~f -q
'-i
O The elasto-plastic matrix, D ,is in general non-symmetric
'( due to the term _ep
r
Spring
'r<--" b b T
~q _f
)
Only when Q = F, i.e. the flow rule is associated, does this
J term and ths elasto-plastic matrix become symmetric. Recalling
~) that the stiffness matrix of a structure Df elastic material
,,( is given [see chapter 2) by
'-'" Oashpot-- -Slider
,í
'--"'
.-(
K J 6
T
D 6 dv [5-41)
v
~i
~
\..'...I
the tangential stiffness of a structure of an elasto-plastic
material at any stage Df plastic flow would be given by
"
..j
6T
'\
K
~ep J ~O ep _6 dv [5-42)
Figure 5-11 Rheological analogue of
~../ v
elasto-visco-plast~city.
1 If ºep is non-symmetric, the resulting ~ matrix would
also be non-symmetric. Most of the standa~ solution routines The model is intuitively more appealing than plasticity
1 make use of the symmetry property Df the stiffness matrix and theory and it happens that the associated algebra is much
simpler.
:1 such routines will not be applicable if a tangential stiffness
approach is adopted for non-associated flow rules [Q ~ F).
1 This is discussed further in chapter B. Let us consider a uni-axial strain controlled test on an
eIasto-plastic material. At time (t) equal to zero, e uni-
j' 5.7 Elasto-visco-plasticity exiel strain [E) has been applied. Since, instantaneously,
the behaviour is pur-e ly elastic, the stress [a) et t = O is
j A model of elasto-visco-plasticity has been successfully E E where E is the a Ie st c modulus.
í This stress 'mey be less
used in the quasi-static applications of the finite element than or greater than the uni-axial yield strsss [Y). If it i5
J. method in geotechnical engineering in the past decade. This
.;.( D
(
1\
(
(
88 89
A
\.... less than the uni-axial yield stress, no visco-plastic 5.7.1 Assumptions
(
strains are set up. However, if the stress is larger than the
C uni-axial stress, the visco-plastic strain (EVP) is set up at These can be summarised ~s follows:
C a certain finite rate (EVP) which depends on the excess of
O over Y and the elastic strain and stress in the bar reduce (a) It is assumed that instantaneously the material behaves
L as they are related by elast1cally.
« or CJ E (E - E vp) (5-43)
1 (c) Stresses outside the yield surface give rise to a visco-
plastic stra1n rate according to a specified 'flow
:,s. The rate of visco-plastic straining is defined in such a
equation'. Stresses on or inside the yield surface
manner that it is zero only when CJ ~YD' Thus a steady-state
Q- (no further increa~e in visco-plastic strains) after a time require ~vp = O.
r t =T exists only when the stress is on the yield surface in
\í the multi-axial situation. (d) The concepts of yield function, hardening law and flow
rule developed in the context of elasto-plasticity are
~-
Figure(5-12) shows a typical stress-strain curve obtained also applicable to elasto-visco-plastic1ty.
(:. by joining the points at steady-state for the uni-axial strain
V 5.7.2 The flow squatibn
controlled testo The stress path traversed in reaching the
r.
':.{ steady-state is shown by the dotted Iines. If the sample is
unIoaded at any stage, elastic unloading takes place, and There are a number of var1antsl that presented here 1s
r>,
'( visco-plaatic ~ermanent strains remain in the sam~18 in a due to Perzyna [1963) and is the bne wh1ch has been used in
r-; manner similar to the theory of elasto-plasticity. geotechnical problems.
) The rate of visco-plastic strains (~Vp) is exp11c1tly
.,,:)
given by ~
J aQ
~ i
ÊVP = ~ < ~(F) > ãã (5-44)
J ~ ~
"-)
,..-l
where ~ is the 'fluidity parameter' - a material constant
o o
A.
J
Vl
Vl
QJ
<-
-,t =-,O ~(F) 1s the flow function
-..!..
<.Ste~ stote
as normal brackets.
points
Ideally, ~ and ~(F) should be identified and determined from
'\ standard experimental tests. However, the main interest in
'-'"
3. geotechnical problems is in determining steady-state stress
and plast1c stra1ns. The transient stress path 1s not impor-
tanto It is usual 1n such s1tuat1ons to assume
~ • Visco- plostic stroin •
Stroin, E
:1 ~(F) F (5-45)
~ 90 91
J
93
J. CHAPTER 6
J
J LIMIT THEOREMS ANO TrEIR APPLICATION
,J
6.1 Introduction
-J
Although this book is about finite elements, this chapter
V is noto The desirability of incorporating a yield criterion
in non-linear finite element formulations has already been
'< indicated. It is necessary to check the collapse predictions
(\ of these formulations by alternative means. The limit
theorems of plasticity provide a simple and powerful tool for
'~( doing this.
'I
.--- o • •• • -
;~
-(
/
'"( 94 95
~ I 5. At the instant of collapse, the changes in the geometry
I
of the body are smail so that alI variables can be
'-.. referred to the original configuration.
(
Rigid stroin hordening
plostic
,\ 6.3 Corrollaries of the normality principIe
'---
-] Rigid ideolly ptasric
'--
:\
Ê Yo~K~---
Vl
!
Normality principIe was discussed in chapter 5. Accord-
ing to this principIe the rates of the plastic strain (€p)(i)
'-.... are given by (see equation (5-4)) ~
/'(
'-- Rigid stroin softening ~p dÀ ~ (6-2)
plostic ao~
'--( '"
:1 Stroin
where dÀ is a positive proportionality
yield function as previously.
parameter and F is the
We shall now discuss two lemmas
\J relating to the applicability of this principIe which are
necessary for the understanding of the limit theorems.
d Figure 6-1 Uni-axial stress-strain curves for rigid
-:
..J
plastic materiaIs Lemma 1: áT EP = O (6-3)
~ '"
r1g1d-plastic model 1s due to the 'Correspondence Theorem' Proof: If stress situation i9 inside the yield surface, i.e.
'-f
r
which states the following: c
F < O, then P = O (purely elastic behaviour) and hence lemma
~ would be tru~. If the stress situation is on the yield sur-
Correspondence Theorem: The collapse load for a body of an face, à should represent change of stress from one position on
elasto-ideally plastic material is the same as that of a body the yi~ld surface to another on the same yield surface (ideally
~ of rigid-ideaIly plastic material having the same geometr~l plastic behaviour). Thus the vector á is tangential to the
the same fixity conditions and the same yield parameters(~
~ yield surface (figure 6-2) and is ort~ogonal to €p which '
In the following paragraphs the words 'rigid-plastic' are taken proves the lemma. ~
to imply ~igid-ideally plastic' unIess strain hardening/
ç'
(.
softening is expressly specified.
O'3·E3
, \.
v
~ = R,p (6-1)
~gent plane
Ç" 96 97
r
<, Lemma 2: (o-o )T
'V 'Va
EP
'V-
> O (6-4)
On the path BC plastic strains (~p) occur and the com-
where O represents a state of stress on the yield surface and plementary work has to take these into account. Thus comple-
~
(
\....
"
tP' rep'fesents the corresponding rate of plastic strains. R.a' mentary work on th~ path BC is given by
is any other 'allowable' state of stress. The term 'allowable" R:+6R:
() needs some explanation. 1n chapter 5 it was shown that the
theory of plasticity does not admit a situation such that
J {R,e+~p}T di(. (6-5b)
C
,-{ F (o) > O
'V
i?,
Cl JR: {c
e
+6E}'P T
di(. (6-5c)
« •P
°3·E~
i?,+ 6i(.
(1 Yield surface Adding (6-5a). (6-5b) and (6-Sc) we obtain the total com-
plementary work (Wc) done ave r the closed cycle ABCA which is
Q less or equal to zero. Thus
(j
W JR: {~e}T di?, +
r+ 6i?,
{ C e + E p} T do +
Ci c
i(.A i?,
'V 'V '"
~
JR:A e 'P T
di(. ~ O (6-6)
C{ /1 01.t~
{~ +6R,}
ç- R:+tll?,
Combining the terms of elastic strain into one term, we have
2 •P
O2'(2
f {{} T dZ + JZ+
6Z p T di?, +
{ R,} JO'VA {6~p}T
'V
da < O ( 6-7)
C i(. 5?,+65?,
C Figure 6-3 Stress path in loading and unloading from
an arbitrary admissible state of stress The first term on the left hand side of equation (6-7) is
t Proof: The proof of this lemma is based on Orucker's zero as alI elastic strains are recoverable. The second term
t
(l,
stability postuJates. Without going into complicated mathem-
atics it simply implies that no energy can be extracted from
the material in the process of deformation. Let us choose an
is negligible if 6i(. + O.
(6-8)
arbitrary stress path along a closed cycle A B C A, figure [6-3).
ci 1n order to ensure that energy is not extracted from the mat- i?,
+ 6i?,
~ 6~P by kP
T 'p (6-9)
(~ (o
'V
- a)
'Va
E
'V-
> O
C{' °3.
~ield surface /
k;..<:::::::~-------_ •. O,
r-
I..{
C{
r\
'-( °1
Figure 6-5 Concavity of the yield surface
(l let us assume In rigid pIas ti c bodies some parts of the body which have
become plastic may be deforming continuously while other parts
CI ~a = O [6-11 )
may not be deforming at alI. This gives rise to a discon-
Cl [Note that the origin must lie within the yield surface as
tinuity of rate of strain which is in turn related to rate of
change Df displacements o~ in other.words,velocities. In such
(( otherwise it would mean that the material yields without any
situations the components of the velocity normal to the dis-
stresses imposed on it)
continuity surface have to be continuous otherwise a gap will
(( be created or the parts of the body will penetrate into each
other.
C(
r'
'''-{
~
'-I
(
'(
ç
r,
100 101
\", (d) Boundary conditions Df (i) surface loads. (ii) surface
Ci Component of velocity normal to the The theorem which enables us to correlate the SPA stress
discontinuity I Vn ) should be continuous field with the collapse load is the 'Lower Bound Theorem' Df
collapse discussed in the next section.
ç 6.7 Lower bound theorem
C[' _ _ Surface of velocity discontinuity
If aSPA stress field can be found for a body under a
(( (8) given set Df loads. the body will either not collapse or may
ç be just on the verge Df collapse under these loads. In other
words. the load (P ) for which aSPA stress field has been
ç' Figure 6-6 Surface Df admissible stress and
velocity discontinuities
found is equal to 6r less than the true collapse load (P ) i.e.
c
C Pe 2. Pc
Similarly. there can be a discontinuity in the stress
~: f1eld but the components of stres9 normal to the d1scont1nu1ty
This is one of the two lim1t theorems. They are extremely
(, have to be continuous otherwise conditions of equilibrium
would be violated. Figure (6-6) illustrates the conditions
useful as w111 be shown by examples in later sections.
L
by the stress field in each of the three
iL It is required to find the ultimate bearing eapaeity [pc)
of an infinitely long strip footing on purely cohesive soil
and it is satisfied
zones.
having eohesion 'c' and unit weight 'y'.
~ If the load on the footing is 4e per unit width the
I( Solution:
boundary conditions of surface loads will also be satisfied.
'-- Consider the strese field shown in figure [6-7). The Thus with a footing load of 4e the stress field shown in
~ stress field has two lines of diseontinuities A C and B O. figure [6-7) is statieally and plastically admissible. Accord-
AB is the loaded area. We first note that the component of ing to the lower bound theorem the true eollapse load is
~ stress normal to the discontinuities is continuous. The stress greater than or equal to 4c, i.e.
!L is assumed to be constant throughout each zone.
(6-15)
p > 4c
c -
~
a- 4C1unil .••.idlh
Note that the intensity of the collapse load does not depend on
the width of the footing and is also independent of the unit
A ( B X
(J weight of soil.
----11 ------,-,
I{ I Iy )'+4CI I I y.y 6.8.2 Vertical cut
cz
~.
y
[
~.y
I y. y·2CI
ly.y+2CI I
~I
ly.y+2CI IYT2CI
Ilyy+2CI
4Ç' yy
ly.y+2CI
cohesive
Find the height to which a vertical
soil ean stand.
cut in a purely
I I
ç y.y
O I
ly.y+4CI
=O I 1, Y =0
- X
\
1 'Y= 1,y
O
~ ZONE I II ZONE~ ZONE m
\.L ~ y.y ZONE I
(
l- ~ ~ y ~
,.\
I
\~
t
((
Figure 6-7 A statically and plastically admissible stress
field for strip footing on purely cohesive
foundetion
.IY_hl
y.lr-h)
1~.-<f'Y
y.iy-h)
1 y,.ly-h)
discont' . ress
Planes of sI
IOUlty
y,lr-h)
1- The equilibrium equations are ~ yly-h)
I Y·Y
;l dT
ZONE m ZONE n
1.
dO
x xy I
+ay- O
dX
I
(( ao
'-1
h
~
ax
+-'i..
ay y (6-13) h
Since Tx is zero. equation (6-13) is satisfied in each of
\~
the thre~ zones. Figure 6-8 SPA stress field for a vertical cut in
purely cohesive soil
Q Solution: Consider the stress field ehown in figure 6-8.
11
(
7
I
'-./ 104 105
í. It satisfies the equilibrium corditions since (1) there
\
'-, '
6.9 Kinematica1ly admissible strain rate field
l. is no variation Df horizontal stress and the shear stresses are
zero everywhere, and (2) aay!ay = y, where y is the unit weight
of the soil. The stress field also satisfies the boundary A plastic strain rate field (strain rates prescribed at
/ alI the points over a body) is sa~d to be kinematica11y ad-
\ .... conditions. Therefore it is statically admissible. It would
be also plastically admissible if the difference in the prin- missible (K.A.) if it satisfies the requirements (b), (c) and
t cipal stresses was limited to 2c. (d)(ii) listed in section (6.7) and also gives rise to boundary
;{ velocities (vk) such that
'- The difference in the principal stresses in the three
po vk dS ~ O
t zones is f
S
(6-17)
'{ ZONE I yy where po represents the surface lOBes acting over the boundary
«
~.
ZONE II yy - y(y-h) yh
S.
~
"i(
I~
According to the Lower Bound theorem
the vertical cut can stand (h) is given by
the height to which 6.10 Upper bound theorem
\1
We illustrate the use Df the theoren by some examples.
cf
i.( 6.11 Examples of application of the upper bound theorem
C("
= (c. I:2Bl. /2 + (e.Bl. 2 + (e/2 B). /2 6eB (6-21l
ç _
"I ./Pu /unit Io'idth
é
Equating (6-19) to (6-21) we have
r:
r' __ B_
"IB ~ ) O
x Pu B = 6c.8
\
("
E
•
A\)(~So
, , I , ,
1 I [4 //
1 /
or Pu 6c C6 -22)
j <,
" ~1
fi 11 fi /
According to the upper bound theorem, since a kinematically
>f
-, " I /
/ Velocily = O admissible strain rate field has been
Pu = 6c, at this load the collapse is
found for the load
imminent or must have
Velocily = O "" ~ Planes of ~elocily
f. already taken place. Thus, the true collapse load [Pc) is
'--' "" ~ disconlirulty less than Pu i.e.
(
\,.; ioy Pc < Pu = 6e
,(
'-' It may be reealled that we had seen in 6.8 .1 that using lower
(
V bound theorem, the true eollapse should be greater than 4e.
i( Thus,
~
Figure 6-9 A kinematieally admissible veloeity field
4e<Pe<6c (6-23)
"I...:.- for the strip footing problem
Q to a compatibla atrein fiald. Secondly if tha shaaring atresse6 Find the height to which a vertical cut in purely co-
on the boundaries AO, De and OB are l1m1ted to the cohesion hesive soil can stand.
J 'c', the material eonstitutive law would also be satisfied.
We wish to find a solution using the upper bound theorem.
~- Thirdly, sinee the block ABO moves downwards along the direction
A lower bound solution was found in 6.B.2.
Df the force acting on AB,
(~
~
f AB P
u
B dS > O (6-18) eonsider the velocity field shown in figure (6-10). A
wedge making an angle a to the vertical is sliding with a unit
.r:
Thus the proposed veloeity field is k1nematieally admissible velocity along the plane AB. We use identieal arguments as
'{ already used in the solution of the problem in 6.11.1 to show
(-- if the shearing stress on the boundaries AO, De and OB are
'-( restrieted to 'e'.
!
~l
(
I
1
( -,
r'
I,..
108
that the velocity field is kinematically admissible with the
109
6.12 Improved salutions
f shearing stress along AB limited to c. Assuming h as the
L)
L. height of the slope, the rate of work done at the discontin-
.uity AB is '
In the sections 6.8 and 6.11 examples of the application
of lawer and upper bound theorems were given. Using more
elabarate stress and strain fields, the bounds for any problem
C c(_h_).1
COBa
(6-24)
can be found within narrower limits. Theoretically if the
lower bound and the upper bound solutions coincide, the corres-
~ ponding limit load must be the' true collapse load. We shall
~ take up the problem of bearing capacity of a strip footing on
purely cohesive soil and discuss some more elaborate statically
l and kinematically admissible stress and strain rate fields.
;alA I /
1 Velocity = O ically admissible. The dotted lines represent stress dis-
({ continuities and p = y.y where y is the depth af the poin~
r
\{
C'
(
h
cosal\ r
I \J
Plane of velocity
discontinuity
under consideration below the' ground surface. To see if the
components of stress normal to the planes of discontinuity
continuous, one needs to resolve the at r-as aas,
are
The equations
for transformation are particularly simple as shear stress i5
C
'( ai zero everywhere. ~he stress field gives Pc = 5c as the lower
bound solution. Thus with this solution taken into account
~ the collapse load is bounded by 5c and the previous upper
(: A bound, i. e.
1
5c 2. Pc < 6c (6-29)
C
J Figure 6-10 A kinematically admissible velocity field
for a vertical cut in purely cohesive material
~. External work is done by the gravity forces. x
L, Rate of external work weight of wedge ABC x vertical /
I ''
\
I p+3e
pose ,
p+3e
/ I\
~.
L
~
component
Equating
of velocity ~ yh. htana. cosa
/
/P+C
/1
p.3C
P.2~/
p+
I
p.~
47
/ p.2C
,\
p+2C
pose
I
I
I \
I \
\
\
p.)C I \ I \
(( h =~
y sin2a (6-26) --I _
~ _ P+C-;j-
. J'+2 p+2e
\i \
~
/ t.J P.2C\ - --\-
We are interested in finding the smallest value of h for the
/ ~2C /30') ~+2e I\ \
~ upper bound solution.
TI
a = ~ , therefore
Minimising (6-26) we obtain
I ~ / p ,p p' \ \
~ e-zc I \ \
/;r!
p+2C ~P
~ h < 4c
(6-27)
I I \ \
([
- y
Plan
es d stress discontinuity
I
Y
I \\
Using the result of the lower bound theorem obtained in p = y,y
( 6.10.2 we have
<-
\
7
L, 111
I 110
(,
Upper baund: Cansider the velacity field shawn in figure 6-12. Naw,
C The cylindrical zane ABCOA is moving with a canstant velacity
w radius/unit time. The baundary ABC is a velacity discan- Rate of wark dane by tHe external forces; Pu.1. ~ (6-31 )
(J tinuity and the zane autside the cylinder is rigid having zero
~
Pu ; 2rrc ; 6.28c I•
cr Bh
Line of symmetry
ç- L /
di scontinuity
ç \
~ /
/
ç' \
2 Zone of zero
velonty
unçulcr
\
""
<; -
w radius/time
B
_ v/ /
f e of velocity
S"~i:~OOhOOit,
Figure 6-13 Kinematically admissible velacity field
for a strip footing on purely cohesive
foundation
( Rete of work dane by the internal forces ; Wark done over the
c. plane of discontinuity OB + Work done over the plane af dis-
continuity DA + Work done over the surface af discontinuity
cl Figure 6-12 Kinematically admissible velocity field
for a strip load on purely cohesive
AB + Wark done aver the plane of discantinuity
in the region OAB due to shear distortian.
AC + Wark done
Limiting the
~. foundation shearing stress to 'c' on alI the planes of discantinuity, we
{\ can write the expressian for the rate af work done by the
v Let us now consider the velocity field shown in figure internal forces as
f1
'-'
r(
(6-13). o'B is the line af syrnmetry. The zone O'OB is
moving downwards with unit velocity.
with a velocity of 1/12 along AC.
The zone CAO is maving
Along the arc AB of the
= [c.~)
12 12
~ + O + c [rr.~ ~
2 n nJ
II 12
~ + [c ~J ~
12 12
'-
Cl
Cl
circle, the tangential velocity is 1/12.
~ Sc .::.
Pc .::.
5.14c
CHAPTER 7
if Thus the true collapse load has been bounded within very close
C limits. lt can be shown that equation (6-33) is also a p CRITICAL STATE MODEL
,( lower bound so that p = 5.14 is the exact solution. c
v c
,f 7.1 Introduction
'- 6.13 Application of the limit theorems to strain hardening
II materiaIs The critica1 state model is basically a constitutive Zaw
relating strain to effective stress. lt is an e1asto-plastic
((
'-.:- Limit theorems are not strictly valid for strain hardening law and can be classified with the laws cescribed in chapter
plastic materiaIs. They are still, however, useful. In most 5. It has, however, sufficient idiosyncrasies to justify
II materiaIs only a limited amount of strain hardening or strain treatment in a separate chapter. The law is incremental 50
« softening takes place and the final yield parameters become
independent of plastic strains. lf an estimate of the final
that small changes of strain are related to corresponding
stress changes. lt must therefore be applied repetitively
\j parameters can be made, the bounds of collapse load can be and the increments accumulated. These mLst be small.
again worked out assuming the body to be made of a hypothetical
(j' ideally plastic material having the same yield parameters as Superficially the model is a "black box" which can bs
r: the strain hardening/softening material in its final state. "plugged" into a finits elemsnt or other form of computer
\{ program for the analysis of boundary valLs problems. lt has
({ 5.14 References
however, a wider role resulting from its capacity to bring
together in a single model a number of a~parently unconnected
r
(( but generally accepted concepts in soil mechanics. ThDse
CHEN, W.F. (1975], "Limit Analysis and Sail Plasticity",
familiar with conventional soil mechanic~ may obtain deeper
r EIsevier, Amsterdam.
'{ insights by means Df the modelo For others, partieularly if
,," they have a baekground in plastieity thecry, it ean serve as
MARTIN, J .6. (1975). "Plasticity - fundamental and general
'\ results", The MIT Procs.
a soi1 mechanies primer. (The writer i5 dubiDUS, however,
about its suitabi1ity for inclusion in :rtroductory s.m.
\
( SALENCON, J. (1977). "Applicatian af the theory of plasticity
courses) .
(l
lf
(
(
'7
(; 114 115
I.
\.. .. the relationship between the stress and strain invariants.
L. The model was originally developed under Roscoe's leader-
ship at Cambridge in the 1950's and 60's. The text by 7.2.1 State boundary surface
Schofield and Wroth (1968) made the theory.accessible to a
C wider readership and is still the standard reference. This
is now complemented by the book by Atkinson and Bransby (1978)
The first essential feature of the model based on the
l which, since it is.aimed amongst others at undergraduates,
provides a gentle introduction. Most modern texts make at
experimental work of Rendulic (1938) and confirmed late r
(Henkel, 1960) is that a surface can be defined in a stress-
~ voids ratio space inside of which a point representing the
least some reference to the modelo It is nin°. state of a soil must lie. Figure (7-1) illustrates part of
L The model was based largely on triaxial tests on recon-
such a surface in a os' 0d' espace. It is called the state
boundary surface (s.b.s.).
t stituted (i.e. consolidated from a slurry) Kaolin. The stress
state was restricted to that of the triaxial test in which
([ 01 > 02 = 03 with some consideration of ·01 = 02 > 03(i.e. the
Od
r{ triaxial extension test). Roscoe was careful not to claim
validity outside the scope of these tests. Subsequently the
,; model has been extended - rather tentatively perhaps - to
'\.
general stress states. Elostic \<1011
((
ç The model is usefully described in terms of stress and
strain invariants. These have been considered in section 1.3.
For generality three effective stress invariants are needed
(~. State boundary
and, 0m' Oq and e are particularly suitabls. The presentation surface
C( given here is for plane strain and involves Os ano 0d and the
corresponding strain invariants ES and Ed' . Apart from o
1 requiring only two stress and strain measures instead of
three this presentation has the added advantage that, as has
Cá
\ been demonstrated in chapter 5, the Mohr Coulomb yield crit-
erion is very simply incorporated.
) Do o,
( It is unfortunate that the triaxial test is not plane
( strain since the majority of situations requiring analysis
are idealized as such. This is the justification for the use
Figure 7-1 State boundary surface and elastic wall
t. of a plane strain formulation here. If analyses are to be
restricted to the triaxial test a parallel theory using 0m
A soil consolidated from a slurry starts off with a very
and 0q instead of Os and 0d may be used~ This has bsen the
~ conventional approach. For the triaxial tsst Oq becomes
large voids ratio and then traces a path along the s.b.s. If
1<
v squal to 0d' and 0m and 0q (or 0d) equate with the Cambridge
it is consolidated isotropically (01=02' or 0d = O) the path
A B is followed. If consolidation is anisotropic (01/02
p and q respectively.
<S cgn~tant) path A1B1 is followed. Should the soil then be
slowly unloaded the state path (as the path in this space is
ci Since effective stress is used throughout primes will be
omitted for brsvity. only in a few places later in the
called) moves inside the s.b.s. to, e.g. Co or C1. It is
assumed that this path is restricted to what the Cambridge
!l chapter where total and effective stresses need to be dist-
school called an eZastic waZZ. The elastic wall advances with
inguished will it be reintroduced.
the state point during consolidation but on unloading remains
~ fixed. Thus the elastic wall containing Bo B1 Co C1 cannot
7.2 The geometric model
1 move until the state point returns to the s.b.s. If after
So long as the model is limited to relating two com- unloading to Co 0S were again increased to the value corres-
Cc ponents of stress to two of strain it is possible to describe ponding to 00 the corner of the elastic wall previously at
~. alI its fsatures from the geometry of lines in a two dimen- Bo would advance to 00' The wall expands as it advances.
s10nal stress space or surfaces in a three dimensional stress: It is also possible for the elastic wall to retreat and
shrink. This happens, as will be explained later when the
~ voids ratio space. This was the medium of communication used
by Roscoe and his co-workers. It will be used here to develop s.b.s. is intersected by the state path in regions closer to
~ the voids ratio axis.
(\'
(
\
(
C 116 117
/<:»
The state paths referred to above will be familiar when 7.2.2 Hvorslev and Rendulic surfaces and the critical state
L portrayed on an e-Log as plot (figure 7-2). This is obtained line
/. Critical state
This requires that the projection Df the elastic wall is
straight in the e-as plot rather than in the e-Log as.
(, °d
line [K ceases to be a constant).
//~r~'
Hvorslev surface
When a state point moves on the state boundary surface
~ D p~astic yie~ding occurs. The expansion or retraction, il1-
C:.-( • iiil Rendul ic surface ustrated by the paths GH and EF in figure [7-3), are referred
to respectively as hardening or softening. A hardening law,
cl expresses mathematica11y the function of the s.b.s. in con-
ti / /:;'/~( trolling the change of size of the yield surface with stress
/ ?E~:-(.C' change.
D'
- F' I <, (o
~ I ,,/ So far the model provides a relation between os' 0dand e.
« Dó I
/
/'\,\ //
G''4{H'
The voids ratio e, can be determined for a given stress path.
Volumetric strain is related to the voids ratio increase in ~
CI C' ~ \ \ voids ratio from its initial value, ~e, according to
o ~ \
(I \ E
6e
[7-" J
'-J V
1 +e
Cr B~
o,
C( e is strictly the average voids during the change, but may
often be taken as the initial voids ratio, eo' For plane
C( strain EV=ES'
Figure 7-3 The complete surface
C( To complete the relation between the invariants of stress
and strain the deviatoric strain Ed mu~t be included. This is
~
/-
done by means of a flow rule as will be explained in section
7.5. The main implications are that strains must be divided
\ 7.2.3 Relation to e1asto-p1asticity into elastic and plastic parts [the flow ruIe applies to the
( latter) and the stress-strain relation must be incremental.
(\,; .. Certain similarities with the properties of elastic- 80th associative and non-associative flow rules as described
(; p1est1c materiels descr1bed in chapter 5 will by now be
apparent, and it is appropriate to introduce some plasticity
in chapter 5 will be required hers.
a
( E
~
(
(
'-- 120 121
~ and a plastic component Differentiating equation [7-5)
rL 6eP = - [À-K) LogeP [7-3)
I(
da
x; dh x-c [7-6)
°d °c + Ôd
Supercritical regions
~ The denominator ha5 been introduced 50 that F(ii)= F(i) for
(il I (iil
,.(
line ------- I 1~'1
_ln equation (7-9) and (7-10) os' 0d and 0c are variab1es,
\_/ and S is an empirical constant slightly less than one.
L,L_/y
-r:
(lf c' is assumed zero then S = Scs and S = 1). O 1 I \
20,
.
~ O,
(( E O O,
Exactly what happens to the yield surface near ths
'-' orig~n 1s not crucial since the intercept f = 2C'COS~'/SCS on
.(
\-./' the 0d axis is relatively small. The Mohr Coulomb line may
conservativsly be cut off by the "no-tension" line 00 [figure
~ Figure 7-6 Norma1ized yield surface
7-6) which rspresents the limiting condition 02 = O, or
0d = 20s' Tension occurs to the 1eft of this line. Less
~ conservativsly the line CO can be continued to E. An ass-
elliptical shape also applies in the super-critical region.
(i umption between these extremes provides ~ logical compromise.
This has the advantage of simplicity.
discussed in section 7.10.
The pros and cons are
"-, that the principal plastic strain increment directions co- dEP = dÀ ~ (7-11 )
incide with the principal stress directions. This allows the s da
l stress axes to be shared with plastic strain increment axes.
5
-p
é Thus 0s,od axes can be shared with dE~,dE~orOs'~d with dE~ and dE
d
= dÀ ~
aÕd
CL d~~ = Scs dE~. (NB not dE~/Scs' The requirement is that
({ Õd dE~ = 0d dE~ = plastic work). If normality applies, or, dividing the second equation through by Scs gives the altern-
as it is alternatively expressed, the flow rule is associative, ative form
e the plastic potential is the same as the.yield surface, i.e.
O : F. O = O represents the equation Df the plastic potential. dEP = dÀ ~ (7-12)
(( d dOd
Normality is assumed to apply in the sub-critical region. dÀ will be related to the stress increment, the current stress
({
Calladine (1963) in his discussion of the paper by Roscoe, state, and the strain hardening characteristics in the next
ç' Schofield and Thurairajah (1963) must take credit for showing
how this relatively simple rule could be applied to their
section.
( experimental data. Just as normality is unsatisfactory for A plastic potentia1 for the super-critical region which
(
the Mohr Coulomb criterion because it causes excessive ex- satisfies the requirement Df variable dilatancy described
pansion (see chapter 5) so also is it unsat1sfactory in the above is
~
super-critical region. 50 much dilation does not occur in So
\ real soi1s. A non-associative flow ru1e is needed. That
proposed here causes the dilatancy to increase uniform1y from
0(1) °d + 20
c
(a -a
c s
)2 = O (7-13)
r)G
e
f represented by F, rather than just one for which h has a
;J 1:::)
,( de: )
~ (7-15) particular value). Clearly dF also is zero. Applying the
= ~. =
C1 dF
aos
do
5
+ ~
aOd
do
d
+ ~
ah
• dh O (7 -18)
Equation (7-15) is readily derived from the standard elast-
(( icity stress-strain equations, (see equation 3-10, the left
({
hand side of which expresses the matrix in terms of K and G).R Now dh = dEP (by definition), 50 introdueing equation (7-11)
into (7-18)Sto eliminate dh, and rearranging gives
The original Cambridge model assumed that deviatoric elast-
(( ic strains were negligible 50 that G was effectively infinite.
aF
C( The assumption of a straight elastic wall on the semi-log
plot implied that K was proportional to the mean effective
(~dO
ao
5
5
+ 30'
d
do )
d
r stress. Referring to section 7.3 above, and equation (7-2)
dÀ
aF ao (7-19 )
'{
in particular, and noting that Log p beeomes dos/os for a ãh30
C(" stress perturbation in the elastieewall:
5
r
[
K +
An alternative,
2.
3
G = (1+e)
K
°S
aF
ãh
~)Oe
(aO
dOc/dh by equation (7-6)
e X
.(
c. K
1 +e)
= (K °m (7 -171
dÀ -1 (aF
--- do aF)
+ --- do (7-20)
I( H aos 5 aOd d
'--". This results sinee K is also the slope of an eIastie wall on
6, an e:Log
e
°m pIot, and'K = do /de: (by definiiion).
m v
in which °c (~] (~)
~ When ·linear sl!et1clty 1e assumed equation (7-16) or H = -- ao ao
(7-17) may be used to give an idea of a suitable vaIue for K. X e 5
~
~R] tl
K may be estimated or measured in standard consolidation tests,
Substitution of dÀ in equations (7-11) and (7-12) gives
(rt is typieally in the range 0.03 to 0.06 for elays of
Li ° °
rl
medium plastieity). A representative value of or ao
[ will be used. G should not be obtained by eliminatTng K s
between (7-16) and (7-17), =
~aa; ao s aod ao·s s
~F (7-21)
({ H aF ao aF ao d
dEP
f
7.6.2 Plastie eomponents of strain d 30
5
aO
d
aOd ao;; °d
(
y
ç' 128
[ox,oy,T1T as before. Introducing The total strain increment is obtained by adding the
129
( a
-v
~-J
\,
elastic strain increment invariantsfrom equation (7-15) to
J
{
a
",f
· = l-. -,aFaos -aF
aOd
J T and a • =,
"'q
l- -ao
aos' aOd
T
the plastic from equation (7-21). Denoting the 2x2 elastic
compliance matrix in (7-15) by C· we hav~
e
\.' • = • •
L equation (7-21) becomes
de:
'"
C
•
da
",ep 'V
• • • =
(7-22)
G * 1 T'
de: = -H (a ) (afl da (7-21a)
in which de:
-v
c de: + de: and C
"'e "'p "'ep
C
*
"'e
+ C
'Vp
"'p ~q '" '"
~
The inverse of equation (7-22) is
or even more concisely as
~ • • •
cL de:
• = C
• da
• (7-21b)
da
'V
= ",ep
O de:
-v
(7-23)
C( ~ - Scs da
5
S
cs
a
C The elastic modulus matrix D: = (C~)-1 is obtained by inspect-
ion as the inverse of equatíon (7~15), i.e.
C(' dF(i) = _1_ aO(i) = _1_ 1
ç" K + "3 G
ç'
~ Scs ~ Scs
o
'\,e
o
,DG 1 (7-25)
aF(i) (_ 2-.)
'(J1 1
--=-
dOC
1 S
cs In the next section the complete consitutive law will be
derived from equation (7-21). It will be represented by
c) Sub-critical region
equations identical to (7-22), (7-23) and (7-24) but without
the asterisks on the matrix terms.
6 dF (ii)
-a-o-
dQ(ii) 2S
cs
(a -a
5 c
)
i
cS s
= -a-a-
5 °d + Scsoc 7.7 Stress-strain component relation
cS aF (ii) aQ(ii) 20
The assumption that principal plastic strain increment
0. ~=aa;-
d
Scs(od+Scsoc)
directions coincide with the principal stress directions
already been invoked in the definition of the flow rules.
has
It
also allows Mohr's circle to be used to represent both stress
~ and plastic stain incremento (Figure 7-8).
(1.. dF(ii) 2Scsos
-a-o- = - 0d+S a Consider first plastic strain increments. From the geo-
c cs c metry of the Mohr's circle
~
«
3.
Note that in the derivation of the last two expressions use is
mede Df the fact that the numerator of equation (7-10) is zero.
(
(
\... ./
130 131
L dEP = ~ dEP + dEP cos 2a stress that cos2a = (a -a )Ia and sin2a 2. Ia d we obtain
C x 2 s d x y d
~ (dtl! t
~ V2cid
with R as defined
~ above.
q-
C'~UC'
Os I ar (dE:ll ,casa~ . Substituting (7-29) in (7-21bJ and the resulting equation
(1/zdE:f) in (7-26J gives
\{
dEP = ~T C ~ da (7-30)
~ '" "'p '"
L in which ~ =
[0.5 0.5
2:i02J
The inverse is
L cos2a -cos2a
da = O
'" "'ep
dE
'"
(7-32)
~- Differentiating
5 rather than by first deriving the law relating invariants._
This two stage derivation was followed here as it seemed
more suited to the geometric desc~iption of the critical
\Í 20d dOd = 2 ea -a ) (da -da ) + 8. dr state model given earlier in this chapter.
• X Y X Y
\(
Dividing through by 20d and noting from the Mohr circle of
\(
!"
._------------
'\
C, 132 132
" \
7.8 Parameter values If c' is known for a given state Df consolidation measured by
\ 0c' ~~s may be calculated from (7-33). Note that the implied
(
,( The model has now been expressed as a constitutive law c' will vary during the analysis as 0c varies. ~' and ~cs are
\..' in fi form suitable for "plugging" into a finite element or constants. c' is often 5mall and can 50metimes be neglected.
other program suitable for the analysis of boundary value In this case ~' = ~~s.
/'-, problems. The first question asked by the practitioner is:
[ "What parameters must be given vaIues to perform an analysis"?
They should be both few and recognizable. Although it may
~ is the dilatancy angle at zero os' The flow ruIe has
been cRosen 50 that ~ varies from ~o to zero as Os varies from
(
not be immediately apparent from the foregoing the number of zero to 0c' For Iack of better information ~o = 2/3~'
G important parameters are few and they are alI "old friends", is suggested. Note that if yieIding is entirely subcritical
~ even though they may be expressed in unfamiliar forms for ~o is not required.
algebraic convenience. They are in three parts: (1) material
ri
\.~ constants, (2) a preconsolidation measuré, and (3) the initial The hardening constant X may be seIected empirically or
stress state. computed using laboratory measurements (or estimates) of À
( and K from the equations
7.8.1 Material constants
~ À-K
[7-34)
X =-
(l' AlI these relate to the soil skeZeton and are therefore
in terme of effective stress. First are the two elastic
1+8
r- or in terms Df the counterparts Cc 2.3À and Cs 2.3K when
\Í constants. If constant elasticity is used they may be defined a base 10 log plot is used.
by K and G or by E and V. The transformation from one pair
(] to the other has been illustrated by the example in section 7.8.2 Preconsolidation measure
ç, 3.2.1. If variabIe elasticity is preferred~ may be defined
with K, nowa variable, given by equation [7-16) or [7-17). The initial value Df 0c' oco must be supplied. oco is
ç G may be kept constant or if desired made to vary with stress. the radius of the initial yield surface in the ãd, Os space.
K may be obtained from oedometer or triaxial tests. K, f~ Oco i5 half the consolidation pressure in an isotropic con-
C I
not calculated from K, will be selected from experience ~ as solidation test, and a little more than half in an oedometer
will G. If plastic yielding dominates, as in the drained testo
'i loading of soft clays, the choice of the elastic parameters
r-
l.t will not be critical. 7.8.3 Initial stress state
Next come the "5" parameters: 5. 5cs and So. Since The initial effective stress components must be supplied.
these are defined by S = 2 sin~', 5cs = 2sin ~cs' 50= 2sin~0' They must be such that the stress point they define lies on or
~ (
~', ~6s and ~o must be supplied. ~'is the familiar slope within the initial yield surface. The program should check
(~ Df the Mohr circle failure envelope [in terms Df effective this.
'-, ( stress). ~cs is less than or equal to ~', the difference
G between the two being a measure Df the cohesion intercept c'.
In section 7.4 it was noted that T = (Scs-5)oc' Expressing
Since the model is suitable for the effective stress
techniques described in chapter 3 [section 3.2) the initial
(', this in terms of the expansion of T, 5cs and S: pore pressure must also be specified. Introducing now the
"-'
A prime to denote effective stress the components to define the
'\..../ z c: cosé ' 2[Sinq,~s - 5in~' )oc initial stress state are therefore 0xo' 0yo' TO' uO'
ci· or
c'
cosq,' + sinq,' (7-33)
Selection of these has been discussed in section 3.4.
Sinq,~s
~' o
c Df the 5 material constants q,' and X stand out as most
important. The preconsolidation measure, oco is also crucial.
CI The practitioner needs to concentrate his judgement on these
ri (i) three.
~ This is sometimes called "guestimating". It helps to
rJ.. know what answer the analysis is expected to produce.
,.
\.:.
G.
!,{
t- 134 135
'--..-. The effective stress path is ~ully defined in this case
7.9 Examples by 01 increasing and O2 remaining constant at 200 KPa.
\(
Consequently:
( The three examples which follow show how the model can
'- be used to determine stress-strain relations for an idealized
d0 - O
sample in which the stresses and strains are uniform. In dOd 1
~ the second and third examples, the effective stress path is !Cd0 -O)
2
das 1
\l also computed since in these examples, unlike ths first, it
is not controlled dirsctly. It is therefore a line of slope 2:1 in 0d:o space. Since
J ~' = 30° the critical state line is slopedSat 45°. The
The first example also shows how the model parameters iR~ersection of the stress path with the critical state line
(I
are determined from conventional soils data. It complements at C reprssents collapse. C has coordinates as = 0d = 400 KPa.
\:
the Grangemouth oil tank case history described in chapter 11
where similar procedures were used to arrive at the parameters. To obtain the stress-strain curve from A to C the stress
\. change (60d = 400, 60 : 200) must be divided into a large
Example 1 s
number of small steps. A programmable calculator at least
~ 1s required. To illustrate the procedure the calculation of a
A sample of clay has been isotropically consolidated to single incrsment as 0d' increases from 98 to 102 is presented.
'-I an effective stress of 200KPa, at which point the voids ratio To improve the accuracy the accumulated value of the stresses
is 1.0. It is then sheared under fully drained conditions will be taken at the middle of the increment, i.e. as : 250,
\ (zero pore pressure) to failure with held constant. (i.e. 02 0d : 100. The stress increment is das = 2, dOd : 4.
\., as in the triaxial testo The analysis here being plane strain
will not strictly be applicable to the triaxial test although Yielding is entirely sub-critical in this example, con-
\.. the results will not be very differentl. A plot of od
\ sequently ~' and ~o are not required. ~' is used to give
against Ed is required for the shear stage. Scs : 2 sin ~cs : 1. The plastic hardenr~g constant X is
'1 obta1ned assuming
The clay has the following properties:-
'-o
(\ 8 = e , i. e.
À : 0.2, = O. 05, ~ , : 300 o
<;
, K
cs
'--.. Since there is no unloading the soil is normally con- À-K
X =-_
: 0.2-0.05
1+1
= 0.075
( 1+e
solidated at the start of shearing with oco: 100 KPa,oxo:oyo
'-... 200 KPa, and TO : O, (the pore pressure is zero conseque~tly
~ ° represents both total and effective stressJ Constant elasticity is assumed with K obtained from K
(equation 7-16) for the average as value of 300, i.e.
~. Od (K Pol
t K : (1+e) 2x300 = 12000 KPa
'-.. 400 c_ -- a
K 5 D.05
'-.:.
'1
(
300
/ <
Collapse yield
surface
G will be guest1mated at 6000 KPa. (This corresponds to a
Po1sson's ratio Df 0.29. Alternatively v could have been
assumed and G derived.)
él Initicl
.1 200
The parameters required are summarised in Table 7-1 .
\.1
100
"
\{
Intermediate yield surface
1
'(
\' 136
TabIe 7-1 Parameters required for exampIe 1
137
C\
Substituting in equation (7-21)
.'
~..
f dE~
Material properties Stresses (KPa)
= 10-6 1217
207] (7-35:
~
Elastic
K = 12000 KPa
G = 6000 KPa
a
co
= 100
yield surface
(Lní t e I
í
size)
1dE;) l207 197
1::1
S = 1.0 (~' = 30°)
a
xo
= 200
C Plastic X
cs
= 0.075
cs
o
yo
= 200
The elastic strains are obtained
and G in equation (7-15), i.e.
by substituting for K
L
f
idE;
(, (S, So not needed) T
o
= O
dEel = 10-6 71 ol do ) (7-36:
V
calculat10n, however, it i9 eonvenient to obtain oc directly
from ths geomstry of figure (7-9), denoting the 0d • 100,
Os = 250 point by S and noting that 0e is the radius of the
yield surface through S, by Pythagoras:
j"') .
de:d.
10-
6
['"
207
207]
239 1::)
(7-37:
= = =
f (250 - a )2 + (100)2 = a l
c c
Substituting
de: = .00137.
dos 2 and dOd 4 gives de:s .00140 and
Example 2
Ó 2S a
aF cs 9 2 x 250
- =- = The clay sample of example 1 is sheared under undrained
~ ao = - 100+145 -2.041
c Od+SC9OC instead of drained conditions. The initial stresses are the
~ same, and the material properties since they relate to the
From which the parameter soil skeleton are also the same. The clay will be assumed
L saturated so that there is negligible volume change. Since
a
aF aQ _ conditions are plane strain E:s = O.
t H = - ~
X
ãOãO - -
145
0.075 (-2.041) (0.857)
_
- 3382 KPa
We now use a prim~ to
distinguish effective from total stresses.
c s
l
In this problem the effective stress path is not fully
í( defined. We impose a deviator stra1n change dOd and knowing
([ de:s = O must now caleulate das and dOd'
f
'{
\~ 138 13~
(I
ç'
(i
Cld [ Critical state'~
400
(J
) Cld C.S.line
300
lJ 7Final y.s.
G KPa
200 100
'-y-
G 100
KPa
~
(l n V I
10
I
20
I
30 Ed
o V
100
I r
,nn
I
~oo Os
« %
« Es
10
Figure 7-11 Exampls 2 stress paths
(]
(J %
C(
°d
C( 10 20 30 Ed ,-- Critical state
%
C( 100
rr-,
'-( KPa
Figure 7-10 Stress-strain curves for example 1
(~
'1 By equating the sum of dE~ and dE~ from squations [7-15)
)'
\...
and [7-21) rsspsctively to zero an explicit expression for
do~ in terms Df dOd is obtainsd. dos thus calculated, dEa °d
and dE§ are obtained from [7-15) and [7-21) respectively and %
;
,-, added to give dEd' As previously, os' 0d and h [from which
0c is calculatsd) will bs updated after sach incrément. lhe C Critical state
rssulting strsss path is plotted in figure [7-11) and stress-
strain rslation in figure [7-12). lhe pore pressure, which
/o is simply the horizontal distance between the total and
effective stress paths, is plotted against deviator strain 100
C in figure [7-12).
l'~ KPo
V
Exampls 3
Cl
A "simple" shear box, ie of the type which distorts the
(l sample from a rectangle to a parallelogram, contains a clay Ed
which is first consolidated from a slurry under the Ko %
(l condition [ie no lateral strain) to a vertical stress 0y of
e 300 KPa and then unloaded, still under the Ko condition,
until the vertical stress has rsduced to 60 KPa. It is then Figure 7-12 Example 2 stress-strain and pore pressure-
Cl strain rslations
fi
(
c 140
slowly sheared to failure. Conditions are drained through-
141
c out and pore pressures are negligible. (The primes will again
%:~
-Ey ad
1 a
0.236 • no
!] 5
'-l.
Figure 7-14 Shear stress. shear strain and vertical whence a2 2-no
-( strain for example 3 K • - • -- • 0.780
o a1 2+no
r
"I'
l
\
'?
142 • 143
~
i.. This is a little higher than would be expected. The widely °d 2 = (a -o )2 + 4,2
x y
L used empirical formula K = 1 - Sin~' would gíve Ko ~ 0.5.
The theoretical value wi~l however, bs rstainsd sincs the we obtain r = 32.7 KPa,'
L purposs here is to illustrate the working of the modelo
To obtain ths stress path and stress-strain relations
L At the end of consolidation 01 = 0y = 300 KPa (givsn)
and 02 = 0x = 0.789 x 300 = 236.7 KPa. The correspondíng
during super-critical yielding bsyond P the constitutive
is obtained in the form of equation (7-31).
law
Use is then made
tf
'-- invaríants are 0S = 268.3 and 0d = 63.3. These ~efine the
point A in figure (7-13) which lies on the end-of-consolidation
of the fact that during shsar lateral expansion of the shear
box is prevented 50 that d~x = O, and the vertical stress is
~ yield surface. DA represents the consolidation stress path. constant 50 that doy = O. The first of equations (7-31) then
tL 0c is obtained from equation (7-40) or (7-41) as 141.6 KPa.
becomes
(~ Unloading as °
reduces from 300 to.60 KPa now takes o = C do + C d,
place. This will &e elastic (at lsast initially - there is 11 x 13
«
~. the possibility that yielding will re-occur before unloading
C13
is complete). With the assumption of linear elasticity ths
«- oondi tion of no lateral strain imposss the ccns t r-aí.nt or da
x
= --
C11
dr (7-41)
( da x = (~J do y
ç 1-v
ar in terms of K and G
(Suffixes 1 to 3 are used to identify ths slsmsnts Df the
~ep m~trix.) Eliminating dox from the second and third
~( equatlons:
~
=. =
3K-2G)
(3K+4G dJJy
d~ C21 dox + C23 dT [C23 C2~)
C11
dr (7-42)
r: Y
Substituting K = 12000 KPa and G = 6000 KPa and doy = ôOy
\
ç = - 240 KPa (givsn),ôox = 0.4 ôOy = - 96.0 KPa 50 that after
unloading 0x = 236.7 - 96.0 = 140.7 KPa. . dy C31 dox + C33 dT = [C33 -
C31 C13)
C
11
dT (7-43)
:(
•...... -3 [- 5.22 5.75]
The shsaring stage of the test is now carried out. The c 10
( "'p
26.18 -28.84
,-J assumption Df elasticity causes thsre to be no change in 0x
as T is increassd in the·elastic region. Sincs 0y is held
~ constant Os remains constant and 0d increases. Note that
I~ 0d increases whether the sample distorts to the left or to the To use equation (7-30) to compute C the matrix R must
'- right. Ths stress path therefore moves vertically from B. first be evaluated. This requires eva1eation of ths"'prin-
~ Let it reach the yield surface at P where supercritical cipal direction as meas~red by the angle a in figure [7-8).
yielding commences. This can be done at any point on the stress path knowing the
« The dsviator stress at P is obtained from the geometry
stress components, eg from
\i.-
0d = 0e [Scs-S) + 505 At point P T = 32.7 and °x - °y = 80.7 whencs a = 19.60
C
'-l
/-:
'-{
= 0.908 we obtain 0d at P as 103.8 KPa. Solving for T from:
~ -[
0.5
Cos2a
0.5
-Ccs Zc 25i:2 J l =
0.5
0.775
0.5
-0.775 ,:J
,(
,( 144 145
'l and y and T initially increase, thus by equa~ion (7-48) 0x' and
'r [0 6."
23,86 -24.61 therefore 0S, reduces. The stress path starts off downwards
and to the left from P (figure 7-13). Repeated application
C ; ~T ~. ~ ; 10-3 8.04 -31.52 31.89 of the above procedure in a compJter prcgram produced the
'( 'l-p. P
stress path PQ in this figure. Strain softening occurred
~'-.
r -11 .69 44.81 -46.141 (7-45) very slowly 50 that by the time the sheêr strain had reached
10% (at which point the analysis was stcpped) the stress path
The elastic compliance matrix ~e must be added to obtain was still a long way from the critical state.
C This in terms of G and v is
'l-ep
,:y and Y:Ey curves are plotted ir figure (7-14). The
-v O results are interesting and some~hat enigmatic. The soil
'--'
~
C ; ~
'l-e 2G ['"' -v
O
1-v
O
O
21 (7-46)
dilates as it should (ie Ey negative). This would eventually
leveI off as the critical state ~ere ap~roached but has barely
started to do this at 10% shear strain. The reduction in 0x
is not in accordance with physical expectation. One would
,~ expect an expanding soil to push harder ~gainst the sides of
Given G = 6000, K = 12000 we obtain v ; (3K-2G)/(6K+2G) ; 7./7. the box! The peak in T is expected. It is surprising, how-
J Substituting this in (7-46) and adding to (7-45) gives ever that T first increased rather than jecreasing immediately
.( after yielding .
- 6.18 23.84 -24.61
These somewhat quirky resu1ts are probably a consequence
\(
~
ep
; C +
'l-e
~
p
; 10 -3 I 8. 02 -30.92 31,89 of the high constraints in this example. The soi1 cou:d
( strain free1y only in the vertical (y) dire~tion. This con-
-11 .69 44.81 -45.971 (7-47) trasts with examples 1 and 2. The sensitivity of the stress-
strein relat10ns to the compliance (~9p) metr1x has already
Note that the elastic contribution is very small and that the been mentioned. In a similar exampfe to the one presented
matrix is unsymmetric due to the flow rule being non-assoc- here the writer computed a stress path which moved cp and to
iative. Substitution of the terms C11, C1~' etc. of the the right from point P. Such a drastic change could be
.( matrix Cinto equations (7-41), (7-42) anB (7-43) gives
'l-ep caused by, for example, a small change in C33 such as would
( change the sign in equation (7-50). Let the reader beware!
dOx -3.98 dT (7-48)
'(
dCy 10-3 (31.89-31.94) dT ; -0,00005 dT (7-49) 7.10 Conclusions
-3
dy 10 (-45,97 + 46.57) dT ; 0.00060 dT (7-50) A critical state constitutive law suitable for plane
<;
( strain applications has been derived. By expressing i~ in
terms of the plane strain stress and strain increment invariants
It can be seen that equations (7-49) and (7-50) involve it has been possible to do most of the derivation geometrically,
(
the sma1l differences of relatively large numbers. The only in the final stages when the three components of stress
( expressions for dEy and dy should therefore be viewed with and strain were related was it recessary to take a more abs-
suspicion. A ameIl chenge 1n the coefficienta of the ~ep trect epproech.
matrix (equation 7-47) could, for example. change their signo
Nonetheless we shall pursue this example to its conclusion. It can be seen now how thE first ~our conventional soil
(
.- Let us next find the initial slope of the effective
mechanics concepts mentioned at the beginning of the ch3pter
are incorporated in the modelo
~ stress path in os,od space. Since doy = O dos; l dox'
Using equation (7-28) and eliminating dox by (7-48) we obtain 1. Critical state
~ dOd ; -1.82 dTo Consequently
l.. When a stress path reaches the critical state line t,e
dOd model allows distortion to occur without change in eff-
( 0.914 (7-51 )
dos ective stress or volume, (or pore prEssure if undrai,ed).
This is of course the essence of the critica 1 state =oncept.
{
r 147
y 146
2. Voids ratio : e~fective stress relationship
(í surface is encountered. In regions close to fixed boundaries
( there is a "numerical trauma" as major stress adjustments take
(- This is embodied in the state boundary surface which in the place to cause strains which are compatible with the boundary
model ia formalized as a plasticity strain hardening law.
Cl 3. Plastic deformation characteristics
conditions.
straints.
This is a similar problem to that illustrated by
example 3 where the troubles are attributed to excessive con-
There have been some recent developments in non-
\~
,{ linear finite element techniques which make them better able to
Again this is embodied in the plasticity formalization by
'-'
,{
L.
means of the flow rules and hardening law. The main con-
sequence of this is the ability of the model to reproduce
incorporate awkward constitutive laws.
here.
A marriage is needed
the marked difference in stiffness between virgin consol- It was mentioned that an alternative form of the model
~ idation and unloading/reloading of over consolidated soils. continued the sub-critical ellipse into the super-critical
Note that by using plasticity theory to achieve this the
( model differs from conventional [met~ll plasticity in that
region. This has the advantages of simpIifying the model and
allowing an associa tive flow ruIe to be used throughout. It
cI volumetric plastic strains control the strain hardening
[or softeningl.
suffers the disadvantage that there is a large elastic region
above the Hvorslev surface where the Mohr Coulomb criterion is
(f violated. The situation is not quite as bad as might first
4. Mohr-Coulomb strength criterion
(( appear since if the yield surface is intersected in the super-
critical region strain softening will occur eventually bringing
This has been incorporated simply by making the super-
the stress path to the critical state (as in the version des-
~ critical yield surface a straight line in 0s:od space of
cribed herel. Furthermore dilatancy will progressively in-
r slope related to .' and intercept to c' and .'. (A bonus
\~ crease as the stress point moves to the left of the critical
is the ability of the model to vary c' according to the
state. This, as with our version, is in at least qualitative
amount of preconsolidation the soil has experienced.1
~
r:
agreement with experimento
'i In addition to these four points the model is able to There appears to be a future in "kinematic" critical
reproduce, at least qualitatively, the essential features of state models. Instead of there being just one elliptical
~ shear tests. Thus a progressive increase in shear resistance yield surface, one or a family of such surfaces are defined
~n6~de the basic surface.
1 with shear distortion to failure associated with reduction in
voids (or increase in pore pressure if undrainedl is obtained
They move about and change in size
with the stress point, the rules never allowing one surface to
~. for normally consolidated soils. For over-consolidated soils cross another. These surfaces control plastic yield and allow
I~; a peak strength will be reproduced followed by dilatant be- some to occur when the stress point is in the elastic region.
,l haviour (ie increase in voids or negative pore pressure devel- Such models are being developed for cyclic loading. If nothing
L, opmentl which diminishes as the critica 1 state is approached. 91se they ease th9 sharp transition from the elastic to plastic
({ Lastly the model requires material property parameters which states. This should provide some relief to the "numerical
,-,' are readily obtained from conventional constants. (.', À or Cc trauma" referred to above.
;S and the preconsolidation stress are the dominant onesl.
r(
"\,,;.y The model sounds wonderful. What then are its weaknesses?
f( Its credibility in certain applications (Example 31 has already
'-' been questioned. As far as finite element application is
( concerned it is not always easy to implement. Although the
'-' model has been incorporated in finite element programs since
about 1970 there are not many published examples of its app-
J lication to boundary value problems, nor does it appear to
Cl have been used widely in industry. Perhaps one reason for
this is that when used in programs which incorporate some of
(I the earlier non-linear computing schemes - in particular the
constant stiffness methods (see chapter 81 - it does not always
U converge nicely and sometimes not at alI. Solutions can be
r
expensive. The writer suspects that this may be due to the
~ very abrupt change in stiffness which occurs when the yield
,~
\..(
r>.
((
<- '
( , \
"--'
/'-J ,148 149
~
7.11 References
L ATKINSON,J.A. and BRANSBY,P.L. (1978), "The. Me.c.hQYÚC6 06
( soirs", McGraw-Hill. CHAPTER 8
({
CALLAOINE,C.R. (1963), Correspondence, Ge.ote.c.hYÚque., 13, No.3,
TECHNIQUES FOR NON-LINEAR ANALYSIS
CI 250-255.
"-(
( \
(
150 151
(
Figure (8-2) illustrates the procedur~s for the one-dimen-
'i Table 8-1
s10nal case. The ~etterings 1n diagrams (a), (b) and (c)
correspond. The positioning of the points E2, E3, .•.. on
( the 11nes of slope Do requires an explanat10n at the outset.
i Equivalent Load Variable Stiffness
,""' Methods Methods They have been positioned randomly. The actual positions will
(
,- depend on the boundary conditions and the discretization (ie.
( Initial stress Tangent stiffness ~ the number of elements). Thus, if there is a single bar
'- Initial strain Secant stiffness
element with a load applied to the end, E1,E2' ., ..En will alI
be at the same 'leveI' (ie. stress value). In general, the
\,(
'-' Visco-plastic points E may be associated with 1ncreasing ar reducing stress.
(
'~
(
'---
8.3 Eguivalent load methods
,
{
In this group a set of equivalent nodal forces are computed
which, when applied to a linear elastic counterpart of the actual
( body, produce the correct displacements (figure 8-1).
( t
(
t
i
(
A body of
1 non-linear A"
\. material
J (a) Initial stress (b) Initial strain
"
J
~)
1
~
{
~.
(
~ Figure 8-1 Diagrammatic representation of
1:,'
(
'-..-.
r(
equivalent load method.
j~
"- The calculation of stiffness for the elements and assembly
,i. of the stiffness for the entire structure (~o) is performed Strain
only once. Numerous updatin~of the loads tright-hand side of Axes for the three diogroms
,~ equation (2-18D are required to obtain the correct solution.
(l AlI three methods start with a linear elastic analysis (c) Visco - plostic
\ based on an elastic modulus matrix, Qo,with corresponding stiff-
ness matrix ~o' It is in the computation of the corrective
~ forces which form the r.h.s. of the stiffness equations that the
three methods d1ffer.
(
Figure 8-2 Equivalent load methods.
:(
(
(
152 153
(
',. 8.3.1 Initial stress method
r
Let the linear elastic analysis for a particular load Read and check alI except load
increment produce strain 1l€1' Point A1 in figure (8-2 a) data.
represents the stress according to the stress-strain law at
Calculate nodal forces equiva-
this state of strain, and E1 the stress had the material behaved
lent to (actual) initial stre-
elastically during the increment. E1A1 (:11) is therefore a
( sses. put in accumulated load
measure of stress error. It is the 'initial stress' from which
vector~. Put ~o Cusually Ol,
( the corrective nodal forces are calculated. (Initial stress
Zo on a file, say FILE 3.
in this context has nothing to do with initial stresses in the
( ground.) Let 601 represent the correct (as opposed to elastic)
Calculate element stiffnesses,
stress increment corresponding to the strain increment 1l€1'
( Then - ~oe and write them on a separ-
ate file, say FILE 1.
!1 Qo li ~ 1 - llQ 1 (8-1)
.{
IRSOL ~ O
( Element nodal forces corresponding to ~1 are given by W
(see chapter 2, section 2.7) DO 100 IN :1 MAXIN CMaximum number of load
.(
ljJe =
_ 1 --
T
f
8 11 dv (8-2)
increments)
~1 then forms the r.h.s. for a re-solution of the stiff- IF IRSDL = O, assemble stiff-
ness equations. (Note that this is much more economical than ness, eliminate and put diagon-
would be the case if the stiffness matrix, ~o' were changed.) aI matrix on FILE 2.
This produces a new set of incremental displacements from which IRSOL = 1 IF IRSOL : 1, resolve. Return
strains 1l~2 are computed. This defines point E2 on figure
'( (8-2 a). 1 is then computed using equation (8-1) with the I with incrementel dieplecement.
suffixes '172replaced by '2'. ~2 is obtained using the same f, = ó + M Keep updating displacements.
<I:, 'I. 'I.
( procedure as for 1jJ2' The process is continued until ljJis
sufficiently small-to be ignored. When this occurs, the process Compute llE, llo, T (initial
( is said to have converged. stress co~rec~io~, Fig. 8-2bl
or ~ Cinitial strain correc-
( A criterion for convergence is required. A convenient one tion, Fig. 8-2c), whence resi-
is to require that the norm of the vector ~ i.e., dual vector~. Increment Di E
( NO and update FILE 3. 'I. 'I.
:.~
IIjJI :
_
/-:-2'1'1
+ ,/,2
'l'2+.... +IjJN
2 (8-3)
YES
-l (where N is the total number of degrees of freedoml has reduced
below a threshold or tolerance value.
;(
The rate of convergence can often be speeded up. This is 100 ...•
4f------ CONTINUE
\ done by multiplying the corrective forces ~ by an accelerator,
'~
;\
a, which is greater than one. a: 1.8 has been found optimal
for some problems, although numerical instability can occur in
cartain instances when a exceeds one.
é
-..( Figure 8-3 Flow diagram for initial stress/initial
A flow diagram for the initial stress method is shown in strain method
figure (8-3).
(,
,(
(
(
Corrective forces • _
~.I are computed using equation (8-2)
:{ with TI replaced by 11' Re-solution then proceeds exactly as
It has been explained in chapter 5, section 5.7, how the
in the-initial stress method until convergence is achieved.
{ visco-plastic material can either be used to represent a truly
viscous material, in which case t represents real time, or a
Numerical instability leading to non-convergence (with
'-i computed displacements amounting to possibly hundreds of kilo-
time-independent elasto-plastic material, for which t simply
controls the convergence. In either case there is a limit on
\{ meters:) can sometimes occur with the initial strain method.
~ 6t above which instability may occur but below which conver-
Use of a declerator, i.e. a < 1.0, may prevent this.
.{ gence is guer-ent sed b). The limiting value nf t,t is known as
the critica~ time step, t,t • Expr8ssions for it can be der-
The flow diagram for the initial strain method is essen-
,( ived for certain material l~~s when the time-stepping process
tially the same as for the initial stress (figure 8-3).
is exp~icit as it is in the method described here. Cormeau
( (1976) presents the following thre~:
( Y = Y t,t (8-5) In these equations G is the shear modu Ius , v Poisson' s ratio,
_I -I I
where ]J is the fluidity coefficient previously introduced in Note that a high stiffness, represented by a large value
( of G ar v elose to 0.5 (implying large K), in even one
'--" chapter 5.
element can result in a very small critical time-step with the
:r
~ For multi-dimensional elasto-visco-plastic laws (which is result that convergence may be unduly slow.
i yI = <F> ~
In these methods, the stiffness of the ~tructural assem-
]J (8-71 bly is repeatedly updated. This process is mor~-expensive in
,{ (lo
The scalar yield function, F, replaces TI as a measure of the
,--(
stress error, and the gradient of the plastic potential, (lO/(l~ (il i.e., in theorYJ in practice, convergence may sometimes
fixes the ratios of the strain rate components so that the be so slow that it cannot be obtained in a reasonable time
normality condition of plastic flow is satisfied. even on a large computer.
--(
(
156 157
computer time as compared with re-solution (as a rough rule-of-
thumb, one updating of the stiffness is equal to about five re-
solutions).
o
'v
0'0%lo!) See Figure 8-3.
,( ~o~;!)1
-,:",_~,
Incr.
00 100 IN = 1.MAXIN (Maximum number Df load increments)
+
~
'- B = correct end-of- incr. point
The load increment ôR is read.
fi
OOT= crude modulus
"- OT = ideal modulus
"
~
p E + ôf!
'\ This method is based on piecewise linearization of the
stress-strain curve. The tangent modulus matrix in the basic Increment loads and displacements.
......
\ form of the method is the initial tangent matrix (figure 9J s ó + ô§
Reactions added to ôR in FRONT.*
8-4). It is assembled at the beginning of each increment of
'l load. The load increments should, in theory, be infinitesimally
small to be able to edhere to the true stress-strain relation- ComputeÔE,Ôo. Update E,O and write
( ship. With finite increments, at the end of the increment, equi- on FILE 3. -Then calculãte ~
librium conditions will not be satisfied. Thus, (optional but desirable). -
•..•.. '
If required
(~T) 62 - 6B = '1! 'I O
158 J 59
An important point to be noted in respect of this method 8.5 Assessment of techniques
is that in strain softening situations [figure 8-4) the tangent
stiffness is negative. This may lead to numerical instability. The techniqueg in the first group (initial load methods)
are efficient for moderately non-linear problems. In the
0.4.2 Secant stiffness method problems of plasticity and visco-plasticity non-associative
'.~- flow rules and strain-softening can be dealt with quite easily
( by these techniques as they require assemb1y and solution with
\~-'
The method is based on a secant modulus matrix Os. In
soi1s the stress, strain origin for defining the modulus will initial stiffness matrices which are symmetrical and positive
not normally be zero due to the existence of initial stresses, definite. On the other hand. the techniques Df the second
and in some cases initial strains. (Strains are usually taken group Cvariable stiffness) lead to non-symmetric stiffness
s as zero at the start. but sometimes they are non-zero to matrices for non-associative flow rules (chapter 5). Tangent
continue strains calcu1ated in a preceding analysis - see stiffness is nega tive for post-peak behaviour in strain-
~.
chapter 3). The procedure 1s 111ustrated in figure (8-6). softening situations and may lead to numerical instabi1ity.
.{ Most solution routines make use of the symmetry property of the
The f1rst analysis using a stiffness based on an 1nitial
modulus OS produces stresses and strains corresponding to stiffness matrices. 501utions with non-symmetric stiffness
~ point 1. o l' is then identified allowing a secant modulus O~ matrices are about three times more expensive.
to be defined. A second analysis with stiffness based on
J AlI the teehniques within the first group may sometimes
O~ then produces stresses and strains represented by point 2,
,( whence 2' and D~. The process continues until the defined be very slow to converge. Techniques in the seeond group are
points settle down onto the curve. An alternative is to perhaps the only way for solving highly non-linear problems.
identify the points on the curve 1', 2', etc. by a 'sloping', Computing eosts are problem-dependent.
'í
ar even a 'vertical' line instead of the horizontal line from
\{.
the points 1, 2, ete. In the context of elasto-plasticity and elasto-viseo-
plasticity, the visco-plastie method appears to have many
'( This method i5 more ver5atile than the tangential sinee advantages. In addition to being able to deal with non-assoc-
it can ineorporate strain-softening stress-strain laws, but is iated flow rules and strain-softening situations, its ability
t generally less efficient. to incorpora te true time-dependence allows the study of prob-
.; lems like delayed placement of lining and other structural
I
components. Moreover, for a predefined load, even a non-
converged solution provides useful answers, since the correct
)
a solution is being progressively approached.
:(
Stress- strain Ia",
°1 LI --------f,----~--~;?~,
8.6 References
'-o
OWEN, D. R. J. and HINTON, E. (1980), "FJ.nae. Elemmú A..Yl
\
Pla.ó;t,i.ci.;tt/',Pineridge Press, SWl!lnS81!1.
,.(
.
'-
\
ZIENKIEWICZ,O,C. [1977), "FiYL-U:e. Eteme.n.:t Me.:thod J.Yl
EYlg.tYle.eJúYlg SCÜ.Ylc.e.", MeGraw-Hill.
~
i
:i °0Eo E
J
'-Í
Figure 8-6 Seeant stiffness method
J.
161
'{
~
'-...'
\
I
(..
CHAPTER 9
(
~,
u
,~ axa [K
xaxa ] +aya [K
yayau] +ãZa [K
zãZau] -- o (9-1 )
(
where Kx' Ky and Kz are the. coefficients of permeabi li ty in
the x , y, and z directions, 'respectively; x , y, z being the
1 principal directions of permeability.
J. u
u = + z water head or potential
'i, Yw
1.
\'( 163
162
u pore pressure (2) Flow boundary condition; here the intensity of f10w (q)
y is prescribed on some part of the boundary. This can be
unit weight of water
w written as
z elevation head above a datum.
au +K-!/,au +K-!/,au
The derivation of the equation is based on the f0110wing assump- K-!/' +q o (9-4 b l
x ax x y ay y z az z
tions:
(a) Fluid flow is steady. where !/'x,!/'y
and!/'z are the direction cosines of the outward
normal to the boundary.
í (b) Soi1 ske1eton is rigid and is fully saturated.
i (c) The fluid is homogenous and incompressib1e. The special case of an impervious ~oundary is inc1uded in
\..
equation (9-4 b l and is obtained when q = O. As an example
( (d) Inertia effects ere negligible and Darcy's law holds good. we write the equations of seepage and bOJndary conditions for
the situations shown in figure (9-1).
When the soi1 medium is of isotropic permeability (i.e. the
same in alI directions) equation (9-1) rsducss to
'.~
r;;2 U o (9-2)
,~
where r;; 2 is the Laplacian operator, i. e. : t L-Sheet pile
I ""ali
h'
:( a2
r;; 2 ::~ + ~ + - (9-3) A
d
.( dX2 dy2 dZ2
2 -
For two-dimensional problems, the ~z~ term disappears from Assumed
A
~,~~~c~J'--,~~r-T)~!----~.~~~~r.c-r7~w.~&a~
_
(
(
'( 164 165
applies over the entire domain AO EC subject to the boundary Figure (~-21 shows a typical division of a two-dimensional(1)
'(
conditions (see figure 9-1 aI, region into triangular finite elements. The head (ü) at any
\( location within a finite element is assumed a function of the
u h + h' over AS (9-6 a) heads at the nades (ü,) forming the element according to
'-- ~
í u h + d over SC (9-6 b l n
,,~
(lu•
u L Ni Ui (9-81
,{ o over DE (9-6c) i=1
(
ãY
'--
\ where N. is the shape function for nade i, and n is the
Earth dam: The governing differential equation (9-51 is to be
number o~ nades in the element (c+. sec t í.on 2.3).
L applied on the domain AS CE. The location of SC is not known
a prior i and has to be assumed. It is subsequently corrected.
,( Use of equation (9-8) together with standard techniques
The upstream and downstream faces of the dam (figure 9-4b) of variational calculus (see for example, Z1enkiewicz, 1977)
( leads to a set of n algebraic equations which can be written
below the water leveI are equipotential surfaces. At the free
surface (SC) the pore pressures are zero, therefore the poten- in the form
e
'\ tial is simply equal to elevation head (yl. Also there 1s no !Se !! O 19-9)
flow across the free surfaces. The complete boundary condition; e _
can be written as follows: where u is a vector of the n element nodal values of u, and
( Ke represents the element 'stiffness' matrix given by
u ~ h' on AS (9-7 a)
'{ -
u = d on DE (9-7 b I Ke fê T E ê dV (9-101
'( -
u = h on BC, u = O (9-7 c) in which B is a 2 x n matrix made up of a row of n submatrice:
'( u = y on CO, u = O [9-7 d )
I
.... dü (lNi)
\ dy = O on AE (9-7 e I ax'
[9-11 )
!.
B.
-~ aNo
~
-, 9.3 Finite element discretization of seepage equation
( ay'
'--
i( The seepage problem is one of the simplest of problems where i identifies the element nades as before
'- which can be solved by f.e.m. The domain of analysis is discre-
\
,_o
\ tized in a number of finite elements [see chapter 21. Each
nade of the finite element mesh has only a single degree of
freedom, i.e. only one unknown variable [head at the nada 1
pointl.
and p
['~ ': ] (9-12)
-(
,.(
x [il We discuss here only two-dimensional problems. Extension to
three dimensions i5 straight-forward once the reader has
'\ got an understanding Df the ideas.
Figure (9-2) Typical division of two-dimensiDnal region
,( \ into triangular elements
,
{
167
\ 166
9.3.1 Calculation of right-hand side (Q) for seepage equations
'\
"( y We consider here two types of boundary conditions given
by equations (9-4 a) and (9-4 b l.
''\ ~ .'f.~
(a) Prescribed heads: For the prescribed pressure head condi-
I
tion, equation (9-4 a), the procedure adopted is identical to
( that adopted in stress analysis calculations. (see, for example,
'-- Zienkiewicz, 1977.) Let us assume that we have triangular ele-
(
ments leading to the element stiffness equation of the form
I,~~
(
.:
'--
\.~
\.ç
(
K21
K31
K22
K32
K13
K23
K33 ,
-'
{:: ).{
Assuming "e has the prescribed
: )
Io r known) value, 0.,
(9-16 )
l
system of coordinates (figure 9-3). The transformed stiffness i
\( [!Seg ] of the element is given by I !
K
-
e
g
IT !Se T (9-13)
'21
K31
K22 x 108
K32
K23
K33
I
I
{~:} - { '" '~08 ,a} (9-17l
I
'{
where I 15 a transformation matrix defined by
\ ,[ Cos e -sine-: and this procedure would lead to UB being calculated as CL
I (9-14)
( Sin e coseJ (b) Prescribed flow: The flow terms (O) for the condition of
prescribed flow are made up from element-nodal contributions
( whers e is the angle between the global x-axis and the local
(
axes of principal permeabilities. Q,
J.
= f
S
N-:
J.
q dA (9-18 )
i those which may have been prescribed, and Q represents the flow
(which corresponds to force in structural f~rmulations) or 9.4 Computation Df veloc1ties and flows
~ 'forcing function' and 1s assembled from the individual element
contributions. A discussion Df the calculation of the flow The solution Df ~quations (9-15) gives the potential
~ terms follows. heads at the nodes. The engineer is often interested in find-
.~ ing out the velocities Df flow and discharges. These can be
(~
(
168 169
'(
evaluated from the potential heads. Using Oarcy's ~aw. the x For computation of the discharge. a section line ar plane
'7 component of the superficial velocity v 'and discharge q 'are
x x has to be specified as shown in figure (9-4 a), The flow across
given by the part of the sectio" line intercepted by an element can be
v ' = - K 'i' and q' = v ' A (9-19) found by transforming the velocity components normal to the
'- x x x x x section line. Assuming 6 to be the angle which the normal
(,( to the section line makes with the global x-axis (figure 9-4 b).
'-, with similar expressions for the y' components. Here i is discharge through section l' -4' (length L and unit thickness)
( the hydraulic gradient and A is the area of the surface normal
~ is given by
to the flow. Note that the local coordinate system is used
(
'-.:.- here. q (v Cos 6 + v Sin 6)'L (9-21 )
x y
J
'~
!y
~
LlD r 2
Idealizing the problem with two linear elements of unit
cross- sectional area (figure 9-5 b l and treating the total
water head as basic unknown variable. we write
(b)
~ x
i
u
[~ ~l {Ü1 } (9-22)
'\ 2 ., ü2
Figure (9-4) Finite element discretization of sheet
-{ wall problem; (a) section acrDSS which where ~ represents a local system of coordinates. The 8 matrix
flow is to be computed; (b) geometry of
''-(
section line through an elemento
(
(
'( 170
171
9.6.2 Example 2 - Flow through an earthfill dam
r
( Figure (9-6) shóws the details Df the problem Df seepage
'-- through an earth dam (after Taylor and Brown. 1967). The earth
( dam consists Df two zones having different permeabilities. The
\.: Flooded to 6m depth -1 ü\=14m
t
downstream rock toe has very large permeability and 1s assumej
4m origin ~ Element 1 free draining. As the problem involves a free surface. its
•1 2 trial position has to be chosen. Th1s 1s shown in figure (9-6) .
4m
The free surface is assumed to be an impervious boundary. This
~ Element 2
leads to the determination of pressure heads on it. In the
3 üJ = 8m next iteration. the location Df the free surface is modified to
correspond to the computed pressure heads. The final location
10I Physicol problem (b) Finite. element mesh ond
local system of co-ordinates
Df the boundary obtained after 25 iterations is also shown on
,~
figure (9-6). Dotted lines represent equi-potential lines.
,,<
'"~r Figure 9-5 Seepage through two layers of different
-( permeabilit1es (after Naylor. 1978).
15 in th1s case a 1 x 2 matrix and has the form
Equipotential line
( ClN ClN
1 2 , 1 1 '
ê , ãX'" -I
i - -I (9-23) Free surfoce - Initial trial
dX J L ~ j/, I
'i
Applied head = Free surface - After 25
í where ~ i9 the length of the element (depth Df the layer). The 100units
value iterotians
ClN/Clx results from the relationship between local and
( global coord1nate systems. The st1ffness matrix for element 1
can be written using equations (9-10). (9-11) and (9-12) Ko
'( Rock toe
4 r- _ !
K
-Celement
-
1] - o
J ,. + 4
2 [0.02) - ~ • ~ -; dx [9-24)
L 4
-5 -,
-2 , 5
10 :-5 5 (9-25)
( I
:J
J
, -5
O
7.5
-2.5
-2.5
2.5
I
I
f ;;} . E:J (9-26) OESAI, C.S. and ABEL, J.F. (1972). Introduction to the Finite
Element Method for Engineering Analysis, Von Nostrand
Reinhold Company.
-( Solving these. we obtain u 12.00 m and 01 - 03 = 10 NAYLOR, O.J. (1978). "Finite Element Methods 1n Soil Mechanics"
2
---(
litres/year. Chapter 1, Developments in Soil Mechanics, edited
C.R.Scott, Ap~lied Sc1ence Publishers.
~(
.(
\
(
173
( 172
SCOTT, R.F. (1963). PrincipZes of SoiZ Mechanis, Reading,
(
Mass., Addison-Wesley.
TAYLOR, R.L. and BROWN, C.B. (1967). "Oarcy flow equations with
( a free surface", Proc; ASCE J.Hyd.Dn., Voz, 93, SM6. CHAPTER 10
( ZIENKIEWICZ, O.C. (1977). The Finite Element Method in Engin-
eering Science, McGraw-Hil1. ANALYSIS DF JOINTEO ROC< MASSES
(
( 10.1 Introductio n
<-
(
174 175
( tures as the tectonic stresses may be present in the rock maS5. Some rocks, such as chalks and evaporites, exhibit time
The init1al state of stress may not have any resemblance to dependent behaviour. These rocks are examples of extreme be-
the relatively s1mple gravity induced stresses which occur in haviour. Even though intaet rock samples of many roeks may not
50i1s. However, the techniques used in finite element analysis show any pronounced time dependenee under moderate stre5se5,
are the same as already deseribed in chapter 3. the roek mass tempered by excavation often does. What, perhaps,
happens is shown in figure (10.1).
The third aspeet whieh needs to be taken into consider-
ation 15 that of 'seale effects'. In soils, the test samples Excavation produces a zone Df micro-craeking around the
are assumed to represent the soil mass from whieh the sample excavation. The additional load resulting from the exeavation
has been obtained. In other words, the seale effects are is transmitted through the miero-eraeks. The intensity of
ignored. A roek sample in many situations cannot inelude the stresses at the contact points is relatively high thus
influence of the presenee of widely spaeed diseontinuities. eausing ereep and plastie.strains. It may be particularly
It thus represents 'intact roek' and not "roek mass'. In important to incorporate non-linear time dependent behaviour
analysis appropriate allowanee has to be made for this effeet. when other structural components are to be associated with
.;
~4
rocks during construetion, e.g. tunnel linings, rock bolts,
etc., since the load transferred to these eomponents may sign-
Natural rock line
ificantly depend on the time dependent behaviour of the roek.
.(
Finally, the presence of groundwater in the rock massas
~
,(
.~
(
'( 179
178
'no-tension' material is written as
2 2
r Cos s Sin S -2SinSCosS
F -03 - tol F o (10-4 )
'(
2
T = I Sin2s Cos S 2SinSCosS I (10-3)
where 03 represents the minor principal effective stress (we
( again drop the primes as effective stress is implied in what
'--..-'
SinSCosS -SinSCosS cos2S-Sin2s follows) and tol represents a tolerance constant (it is nec-
(
<:: essary from numerical considerations to prescribe a small
( in which S represents the angle defined in figure (10-2). For amount Df tension as allowableJ. Note that when 03 is tensile
\_'
3-D and generalised plane strain problems (see section 10.5) and numerically greater than Tol, F > O and visco-plastic
( the transformation matrix involves direction cosines Df the strain rates (~vPJ would develop givsn by
x', y', z' axes with respect to x, y, z axes and can be set up
J (Jaeger and Cook, 1976). With the R
mat~ix thus defined the
kVP )1 < F > aF
ao (10-5)
rest Df the procedure of the finite element analysis remains
,i. unchanged.
'V
.~ whers Z
represents componsnts of strsss and an 'associated
10.4.2 Rock as a no-tension material flow rule' (ehapter 5~ has been assumsd. < > have their usual
,( meaning, i.e.
If rock is assumed to have a large number Df randomly
,( oriented planes of weakness, and if it is assumed that tensile €vp 3F
F > O
=
)1.F. aR: if
stresses cannot be transmitted across the planes of weakness, 'V
-( [10-6)
we ean model rock as a no-tension material (Zienkiewicz,
Valliappan and King, 1969). Traditionally designers do not .vp
,( C = O if F ~ O
'V
like to rely on geological materiaIs to provide tension. For
-( the analysis of many practical problems such as underground
openings, foundations Df dams, ete., the designer is often The cavern no-tension analysis of chapter 11 used the
'( interested in a stress analysis in which tensile stress has visco-plastie algorithm.
been either eliminated or an estimate Df tensile strsss which
~( would sxist if craeking did not oeeur is obtained. A no- 10.4.3 Time dependent multi-laminate model
tension analysis enables the dssignsr to estimate the extent
'( Df cracking if no reinforcement is provided. He can then do It was emphasied in section 10.2 that an important
a with-tension analysis to determine the amount of reinforce- feature of roeks is the presencs of ths planes of weakness
'~;
ment needed. The cavern example in chapter 11 (section 11.5) forming a fabrico The multi-laminate modal describsd here
'-.. ' illustrates both these approaches. Thus a simple analysis for takes into account this factor.
.(
tension relaxation can be quite useful from the practical
'-- (al Ideelized modal and assumptions
( p01nt of V1iW.
\.J- '
.( In chapter 8 techniques for non-linear analysis were It is assumed that rock material is traversed by n
,-' discussed. Constant stiffness methods, in particular the families of discontinuities, [figure 10-3J. In practice n is
( 'initial strsss' and 'visco-plastic' methods, can be used for limited to 1, 2, or 3. It is assumed that the shear strength
<:»
eliminating tension. It is also possible to use tangential on these discontinuities is considerably les5 than that of the
stiffness methods. 'intact' rock material. Let ~1' T2' •••.•• ~ represent a set
~~
Df transformation matrices which ~ransform th~ global stress
J In the visco-plastic method, the criterion of 'no-tension' (~) to the normal stress and shear stress on the plane of the
dlscontinuities.
has to be expressed as a pseudo 'yield function'. The model
J represents a 'time dependent no-tension material'. The relax-
J, ation of tensile stresses is traced with respect to time [on
an arbitrary scals if an arbitrary value of ths fluidity
J cosfficient, )1, has bsen assumedJ. This scheme has been found
to be quite convenient particularly in problems where the be-
J haviour Df other structural components such as tunnel linings
is to be studied since 'delayed' placement Df the lining can
J be conveniently simulated. The pseudo yield function for
~\
(
\
(
j80 181
(2) The peak shear strength of the discontinuity planes ca~
( be represented by a Mohr-Coulomb type friction law.
Assuming a linear form, we have
(
-{
=
{1 Z envelope of a particular family of discont1nu1ty planes
n 1 represented by equation (10j12) the behaviour of that
set of discontinuities is purely elastic. It 1s indeed
( possible that yielding would take place on some families
{:} = {2 Z (10-7)
of discontinuities while there may be no yielding on
( n 2
others.
,,(
{:n} ~ {n !?, (4) It 1s assumed that d1scontinuities
excursions into the tensile zone.
remember their
By this we mean the
( n
following: once tensile stresses have taken place across
~ The behav10ur of the discontinuities has to be idealized a particular familj of discontinuities, the discontin-
in a manner that can be incorporated in the modelo For this uities open. If subsequently the normal stresses become
( the following assumptions are made: compressive they would not be transmitted unless the gap
clossd. How this 15 achieved 15 d1scussed further in
( section 10.4.3 (d) below.
(1) It is assumed that a discontinuity plane cannot carry any
( tensile stress, i.e.
This idealization' of the behaviour of discontinuities is
(
typical. Slightly different assumptions can be made.
0n > O (10-8 )
-(
(
( 182 183
analogue of a rock mass having n families of discontinuities The visco-plastic strain rate contr~buteá by the kth family of
( or laminates can be represented by (n + 1J dash pot - slider the discontinuity planes can be written as
systems as shown in figure (10-4). Each of the components 30
represents a family of discontinuity planes and the (n + 1) 'vp k
~k < Fk > ~ (10-16)
system represents the intact material. The activation of in- ~k
dividual components is governed by the yield functions some of
which may be dormant at any time. The model thus allows sim- where Fk represents the yield funct~ons of equat10ns (10-8)
ultaneous sliding and separation at several discontinuity and (10-12) and 0k 1s the plastic potential function (chapter
planes. 5) for the kth family. Assuming an assDciated flow rule with
Qk = Fk' equation (10-16) reduces to
8F
'vp k
~k < Fk > ~ (10-17l
~k
(
n+1
'vp aF k
(
2
~ L ~k < Fk > ~ (10-18)
( k=1
(Discontinuily Family 21
, ( With aqu~tion (10-18) defining the reta of visco-plBstic
(
ç n+ 1
(lntact rack I
strains the working of the visco-plastic
multi-Iaminate model can be identified.
algorithm of the
(
(
c
.2
V)
1"
--- ---- ().~
-----
Figure 10-4 Rheological analogue of multi-Iaminate 8 r~\)
t
(
rn-laminates) material "
cl
----
--- --- --
- -::::---- --- .>:
-<
G
(
(
(
( 1B4 1B5
dilatancy in real discontinuities is often assumed to depend (d) Monitoring of joint opening and closing
(
on the physical angle Df inclination Df the asperities to the
( discontinuity plane (0) and is caused by the 'ride up" action Due to the relaxation of tensile stresses normal to plane
(figure 10-6). The dilatancy does not go on unabated and aftér Df discontinuity plastic strains signifying opening of joints
( some sliding it stops. In the model, if assocated behaviour take place. During further loading if the normal stress be-
is assumed, the dilatancy would continue indefinitely. On the comes compressive the compressive stress is transmitted across
other hand, if non-associated behaviour with 0/ = O (figure the open joint implying that the joints become fillRrl. It is
10-7) is assumed no dilatancy would take place at alI. Ob- usually more realistic to assume that the.yclose. Assuming
( viously to model real situations a variable angle 0/ is re- they close perfectly, we write an additional yield function.
quired. Experimental data to define 0/ are meagre and in many If the joint has been under tension, the peak normal total
( situations the extreme conditions represented by 0/ = ~ and
'-.-. tensile strain for each Df the families (Entmax) is monitored.
.( 0/ = O both need to be analysed. If the current leveI of total compressive normal strain on the
planes of discontinuities (E~t) is such that
c max
( Ent < I Ent ( 10-19 )
Ideallzed shape
~( asperities the additional pseudo-yield function to be used is
~(
F (J I O (10-20)
( n (1,2, .... n)
1, .
geometry may be idealized as plane strain. The first category
( of such situations are those in which the orientation of the
.::»
axes of principal initial stresses may not coincide with the
( plane Df analysis.
Do
.~
( Figure 10-7 Non-associated flow rule - non-dilatant
behaviour of discontinuities
J
J
~
(
(
'( 1.86 j87
'7
, \
'>
\
(
~J
,(
Al
. ~.~ General direction of
principal stresses
.~(
AJ (A) GENERAL SITUATION (B I HOMOLOGOUS POI NTS
~
J
./ ....••
I Y
DF A SLICE DF UNIT
THICKNESS
4
Figure 10-9 Families of discontinuities inclined
'-{ to the plane of analysis
!:) . I:)
plane stra1n analys1s.
i
~'
In the second category are the situations where the axis
1 (10-21 )
~ of anisotopy, due to the laminated structure, does not co-
incide with the plane of analysis. In a two dimensional prob-
'~ lem the or1entation of the normal to the discontinuity planes
p p'
for the multi-laminate model discussed in section 10.4 must
li 11e in the x, y plane.
Applying the same argument to a point p on another slice
'" This hardly ever happens in real situations which are adjacent to the slice under consideration, we have
J illustrated in figure (10-9). u
~p
= u,
~p
= u"p ; u
~
= ru,
~
v, w'Jl T (10-221
,\
'(
155 159
Pande (1975) for further details of this type of analysis.
~
r
This implies that the w component of displacement 1s
independent of Z co-ordinate of the points considered on the
body. Mathematically this me~ns
10.6 Effective stress analysis of undrained rock masses
~ = ~ (x, y)
Techniques of effective stress analysis were discussed in
chapter; 3. These are valid for r-ock masses as wel1. However,
In other words the displacements at any point are indep- a few remarks are called for in the context of the effectlve
endent of the z co-ordinate. This implies that stress.analysis of rock masses. In chapter 3, it was shown
(equation 3-6) that the bulk modulus of pore fluid element ~f
dU dV dW is given by
ãZ dZ dZ
o (10-23)
,. 1 n 1-n
'- Substituting in the 3-D strain-displacement relations we
- +-- (10-26)
Kf I< I<
w s
~ obtain
where I<w is the bulk modulus of water, I<s is the bulk modulus
~ dU
-J
E:
x ax of soil particles and n Is the porosity. In problems of soils
the drained bulk modulus of the soil "skeleton" 1<' Is small as
av compared to f(w and I<s' In derivation of equation (3-6) it has
'J. e:
y ãij been tacitly assumed that either the soil grains are incomp-
resslble or the mean stress in the particles equals the pore
'{ E: O pressure.
z
"'1
Yxy
= au
ay
+
dV
-
ax
~ (10-24 ) Consider an element of unit volume of rock skeIeton and
pore fluido Let this element be subjected to a change of
effective stress (óZ'), and pore pressure change of óu.
dV
'( Yyz ay Assume that the element is undrained, i.e. no pore fluid can
-,, escape (ar enter) from the unit volume, which implies that
dW
Yzy ax ó (Volume of the element) O (10-271
\
Turning to the finite element implementation of general-
r' ized plane strain conditions, we describe the topology of the
where ó represents 'change of'.
-' elements and nodal co-ordinates in two dimensions. A sllce Now we work out the various components of volume change.
1.(\,
, I
of unit thlckness Is assumed for analysis. Each nade has
'--' three degrees of freedom (u, v and w displacements). The six (a) Reduction in volume due to change of pore pressure (bu)
',-,' strain components and six stress components are used in the B
matrix in ~ nbu (1-nJóu
r<
w I<s
(10-28)
~
I, -t ~ (10-25)
, The terms correspond to the reduction in the volume of pore
'":,, where ô are nodal displacements. The ~ matrix is 6 x 3n where fluid and rock grain respectively.
n is t~e number of nades in the elemento
(bJ Reduction in t?e volume due to change of effective
stress (óz') (1.
1 This approach allows two dimensional elements to be used
when otherwise 3-D elements would be required. For example
8 noded (24 d. of f.) quadrilateral elements would be used mócr'
(10-29 )
instead of 20 noded (60 d. of f.J "brick" elements with a con- 3f\
J sequent considerable reduction in the cost of computing and where ~ = (1,1,1,0,0,0)
T
for the general 3-D case.
...l data preparation. As a ruIe of thumb, experience indicates
that the cost of computing can be reduced by an order of mag-
(i) This assumes that the mean stress in the solid particles
;,.( nitude. The reader should refer to Zienkiewicz, Taylor and is the same as the mean stress in the rock mass.
'(
'y 191
190
The total reduction is thus equàl to PANOE, G. N. (1979). "Visco-plastic algorithm for mode1l1ng
\ tensile non-linearity in rock and concrete .structures",
mTóo' Mechanics of BimoduIus MateriaIs, ADM Vol. 33.
'I nóu + (1-n)óu '\, '\,
+ ---
~'
kw K'
s
3K
s PANOE, G. N. (1979). "Numerical modelling of rocks - poss-
L, ibilities and problems", Proc. 3rd Int. Conf. Num. Meths.
which must be equal to the change in the volumetric strain of in Geomechanics, A. A. Balkema, Rotterdam.
\..J
the skeleton. Thus
[ ZIENKIEWICZ, O. c.. VALLIAPPAN, S, and KING, I. P. (1968).
mTóo' "Stress analys1s of rock as a 'no-tension material''',
óe: = mTóo = nóu + (1-nlóu + '\, '\,
,I (10-30) Geotechnique, Vol. 18, pp 56-66.
'-, V '\, '\, Kw Ks ~
,
I
ZIENKIEWICZ, O C. and PANOE, G. N. (197:1. "Time dependent
'"'-, Introducing (10-26)
multi-Iaminate model of rocks - a numerical study of
L mTó€ Óu + '\,
mTóo'
'\,
(10-31 )
deformation and failure of rock masses", Int. J. Num. &
S
(10-33)
~
"( In the above O' is the drained modulus matrix.
~
'--'
The procedures for incorporating this in the finite
,\
'-'
element program for an effective stress analysis are the same
\ as discussed in chapter 3, section 3.2.2.
(
(: 10.7 References
~
-,~
1
,{
t
'{
c'
'( 193
V
ç~
ç CHAPTER 11
(I'
(
SOME CASE HISTORIES
(J
,(
i'" ~ by ROY TABB
l
L 11.1 Introduction
fi
«
\
("., 195
194
Q'j changes from ice to water and water to ice then the prediction
will be made more accurate.
(,
Once the temperature gradient change has been predicteà
~ then finite element techniques can be employed. For this stage
we can use a conventional linear eIes t c program set up to
í
t /'
The axi-symmetric mesh was set up as indicated in figure
/ (11-2).
I
(( /- l After alcpse of time
/
<l 0(.
\
\..;1
l-A! end of freezing
(7' -120
q'
.:\ I .
Figure 11-1 Temperatureiprofile behind tunnel lining
I
.", i
In many Gountries, incl~ding the W.K., there are few
.
·1
commercial laqoratories suitcilblyequipped to undertake low
temperature testing of soilslor rocKs. The difficulties are
:-\ immense and for most problems the expense would be prohibitive.
~.,:
Knowledge of the physics of 1rozen materiaIs .can assist in the
selection of material properties and for further details of
the principIes involved rS~d4rs are referred to Jumkis (1966).
\ \
Cencrete Sei!
This text also presents typical material properties which, with
~ care, could be used for pr,e11minary design purposes. Upper
and lower bound values can tMen determine the sensitivity of
the analysis to the selectedlvalues.
Figure 11-2 Mesh for ground freezing problem
I
~ The major parameters. reguired are:-
The temperature c~anges from the assumed initial profile
~ (a) coefficient of linear expansion,
f.\ of figure (11-1) are specified at the element corners. These
v are used to calculate the change in stresses using the elastic
0:.- (b) conductivity, constants and the coefficient of linear expansion. These stres!
changes must then be added to the insitu stresses to obtain the.
li. (c) the elastic constants. final state of stress at the tunnel wall. The analysis will
also predict the change in hoop stress within the concrete
«~- A prelimipary analysis is required to predict temperature
gradient changes with time. This is achieved using a computar
tunnel lining.
(f program based 6n the diffusion equation in a similar way to The analysis which formed part of the real case history on
Q~ the analysis of t rens.ísrrt se,~page problems (i)• If ~he program which this illustration was based predict~d that no·sp~cial
\
has the ability to handle the latent heat aspect of the phase precautions needed to be taken since the chahge in stre·ss tva~ ,.
« (i)
i small.
f
f The theory for this is cuts de the scope of this bonk ,
í
l:'D 197
( 196
(
~d J
next se.ction in which ·case i histories are discussed in mor.e Preliminary design calculations had been completed and
detall using . both linear e.iastic and non-linear methods. the necessary wall penetration estimated using fixed earth
\j) ,
theory. The rather empirical nature of these methods for a
( cantilever wall with a large free height prómpted the more
11.3 Cantilev8r retainingiwall analysis
\:J}
detailed study.
(
,-:;) The problem, examined in 1976, involved a deep cantilever
( , wall which was to retain a cutting for the new M25 motorway. The techniques employed initially followed conventional
\...J' The motorway was cut into ~he Gault Clay hillside and the wall soil mechanics theory. The "at restO earth pressuredistributior
was to be formed using eit~er contiguous bored piling or by were calculated for the full height Df the wall.
diaphragm walling using slwrry trench methods. The cross section
() after tionstruction is show? in figure (11-3J. The techniques Prior to 8xcavation both sides Df the wall will be at a
used for this analysis are'now somewhat dated but serve to state with Ko Z Kp and the only way stability can be maintained
( ~)
illustrate how relatively unsophisticated.computer programs can during excavation is for the forces on the back of the wall to
beused to good effect. reduce since further passive resistance cannot be mobilised.
(
í1) The factor of safety against the wall overturning must be
( calcwlated in terms of the active pressures rather than by the
(1~ more conventional method in which working is in terms of the
passive pressures. This is necessary becaus~ ths pressures
on the front of the wall are already at the full passive limit
fl) and thus using these pressures will always result in rotational
t factors of safety close to unity. Any margin of safety for
such a system therefore results from an ability to reduce
r
(
\";
"
6m (approxJ
forces on the back of the wall down to the active limit.
H m thick retaining \!Iall It is quite apparent that any analysis of this type of
~ structure must involve some form of soil yielding in order to
I
\ achieve a solution and a true linear elastic approach without
C V plastic yield would not be appropriate since no further passive
( \
pressures can be mobilised.
\J)
lu The method employed was essentially that Df strategy 1
described in section 3.5. The wall was'progressively unloaded
lJ by simulating the removal nf passive support, A,t each stage
( plastic yield was determined and the elements in question had
'-.. , Figure 11-3 Cantilever retaining wall
their ability to carry further stress changes suppressed (by
G The Gault Clay, being heavily over.consolidated, is, in
using a suitably low shear modulus value whilst ~eeping the
bulk modulus unchangedJ.
L its natural stete, close to pessive yield neer the ground ,:.
surface due to the high ho~izontal principal stresses. The A more detailed description of the techniques employed
r'\1 ibility for.a cantilever w~ll. and shear on both faces as indieated in figure ({1-4),
\(1)
\
j 98 199
These would provide valuable information on displacements and
bending moments in the short term and monitor changes with time.
~
r- It is likely that wall cohssion will significantly reduce
{ wall moments in the short termo There is however the wider
,
v; ,
6m
~'<, Shear slress frem question of the effect of time on retaining structures con-
structed in overconsolidated deposits. After the initial re-
i ~
sai I 10 •••011
''-.;.i
L
( 1 duction does the horizontal pressure return with time due to
drainage ar because of creep? The drainage case can be
studied by using drained elastic constants, ie E' and V' (E'
( is often assumed to be 0.6 of the undrained value and v' is
(
taken as being close to O). The wall shear stresses reduce
\..... and displacements increase causing the bending moments to
Figure 11-4 Computed shear loading on wall increase to values which are more in keeping with conventional
!'---"
calculations.
c It is instructive to assess the finite element results
assuming shearing on the back of the wall 'as indicated in
The creep case cano in principIe. be studied using models
such as the visco-plastic model described in chapter 5. but the
figure (11-4) with the shear stress assumed ~o equal 0.3 Cu writer is not aware of its application to this type of problem.
where Cu is the undrained cohesion of the Gault Clay.
There is good reason to suppose that some of the insitu
Taking the wall thickness to be 1.2m and Cu equal to stress will return with time. With a free cantilever wall
100 KN/m2 then the moment at dredge leveI due to shear stress the problem reduces to answering the question: "t)pw large a
on the back of the wall is 1/2 x 1.2 x 0.3 x 100 x 6 for the movement can be tolerated". For propped walls then the build
6m high wall. This amounts to 108 KNm/m run of wall. To put up of stress with time must be a major consideration in the
this into perspective compare this with the minimum soil designo On selected sections of the cantilever wall for this
pressure requirements of the code for a hydrostatic force using case history measurements of insitu earth pressure coefficients
a density of 5 KN/m3 then the moment at dredge leveI would be (Ko) were made prior to construction using the Camkometer and
the results are awaited with interest. It is intended that
6 x 5 x 6/2 x 6/3 = 180 KNm/m these readings be continued for some years after construction.
Therefore the net moment making allowance for wall shear 11.4 Oil tank on normally cansolidated deposits
{' 19
\, - As part of the expansion plans at Grangemouth Oil Refinery
( 180 - 108 = 72 KNm/m a series of oil tanks were to be founced on reclaimed land near
C the banks of the Firth of Forth. Below 2-3m of fill lay an
\
\...
- only 40% of that predicted using conventional theory. extensive thickness of soft normally consolidated,very silty
( laminated CLAY. (Some reporters described the material as
\..... This conclusion from a simple analysis reflects exactly SILT). A typical borehole log together with a plot of index
what the finite element method predicted. The simple analysis
l also agreed with the f.e. method by predicting that over the
upper portions of the wall the moments were such that bending
properties is shown in figure (11-5).
( 200 201
'(' The most promising techniqu8 for the analysis which todk
into account dissipation betwe8n stages was a finit8 element
(, mo istur e co n t e o t
(
study using the critical state modelo This would not only
offer prediction of pore pressures but indicate zones of yield.
G 10 20 ,o "O 50 60 70
,(
'-' A coupled solution whereby consolidation could be calcul-
ri ated as each loading stage was applied would have been ideal,
"- however at that time no techniques were available using a
coupled critical state modelo An acceptable alternative was
<:»
to manually abstract the excess pore pressures from the crit-
t o I 1\ \ ical state f.e. analysis and use these as input to a three
I{
{
> • I
I
1-
J pias t i c
\ liquid
limit
dimensional finite difference consolidation analysis. After
dissipation was allowed over the period of one loading stage
the final pressures were fed back into the critical state
analysis for the next stage of loading. Although this "un-
,"'
--!....\.
i(
<.( -5
I
I \ coupling" of the problem violated certain theoretical require-
ments it modelled the real situation reasonably well.
(
I
I
I
\ \ The site investigation for the project was conventional
in almost alI respects with no special measurements made to
\
'( "oo ,
I
I. , \,. 1
provide parameters for the critical state analysis. These
parameters have been described in section 7.8 where the point
is made that they can be obtained from conventional tests.
( .•••
-I
,
I \ \
The derivation of suitable values is given below.
J
,~
(
'-- '
-20
I
I
I I I The analysis here is axi-symmetrical, consequently the
plane strain formulation given in chapter 7 is not appropriate.
The formulation used here, however, is essentiaIIy the same as
1
( in chapter 7 but with 0s,Od replaced with 0m and Oq respectively.
'-/ Also Ev replaces Es (they are now no longer equal as in the
It 50ft
and
lominat.,d
s i l ty clays
o r q c ruc clayey ~ plane strain case) and Eq repIaces Ed' As the clay i5
assumed to be normelly consolideted only sub-criticel yield1ng
1 is involved.
1 Figure 11-5 Typical borehole log for Grangemouth site Conversion from the plane strain form requires the con-
stant 5 = 2sin* to be replaced by "M" which in general varies
1 Prior to proceeding w1th construction it.was decided to with the third stress invariant 6(i). For the triaxial test
~ verify the factors of safety by means of staged water tests.
The preliminary analysis indicated that if loading was instan-
~ taneous then failure would occur but if excess pore pressures (ilDividing equation (5-14) by the coefficient of Oq
gives M as the c08ffici8nt of 0m' i.8
were allowed to dissipate'between stage loading a stable con-
-\ dition could be maintained. Provided stability could be M = 35in*
ensured then settlement may become the overriding factor
'\ I3Cos9 - 5in6 5in*
affecting the viability of the scheme.
'\
(
y
Y' 202
e = 300
203
r stat8 0
1
> O2 = 03 and or À = 0.182. K was assumed equal to À/4 or 0.0456.
~:' F.rom the site Lnvas t í.get on the average cjJ'was 26.25.
í
6sin4>'
M
3-sinq,'
(11-1 ) Substituting in equation (11-1) gives M = 1.03. M was estim-
ated at 1.07 by an alternative method. and the average of 1.05
~~ The form of the model used here assumes M to be the con- was used.
( stant defined by eqUatio~ (11-1). This implies the yielding.
o as in the Von Mises and Drucker Prager models described in The shear modulus G was determined from initial values of
l'0 s8ction 5.3. is ~ot indep~ndent of O2, the bulk modulus K' (which depends upon the mean effactive
stress) and an assumed Poisson's ratio v'.
A further differ~nce, is that instead of using the plastic
G
compressibility maasure i = (À-KI/(1+eol ita components (ie Wroth
(19751 has proposed a tentative relationship between
À. K and eol are sp~cified. A variable elastic bulk modulus. v' and the plasticity index (PII. Using a measured PI of
{
K'. is implied by the ass~mption of a constant K. The shear 18% this gave v' = 0.28.
'--- modulus. G. however. ia taken as constant.
To obtain G for a given effective stress o~. we first
~ 11.4.1 Material property specification compute
« The perameters required to define the model are:- K' =
(1 +el a (7-17 bisl
« À.K: the slopes of the yirgin consolidation and swelling
--K- m
((' to Ses' However. as is noted in that chapter it is unit weight. Yb' to be 10 KN/m3 gives,01 = 50 KN/m2• Using
'I
\-\ eonvenient to set S c Ses' i.e. cjJ= q,es. for normally
consolidated soil. No ambiguity therefore arisesl
the empirical relationship
Ko = 1-sincjJ
~I G the elastic shear modulus. This was assumed to vary
(J with depth. but no~ with stress. for the eosfficient of earth pressure
Ko = 0.558. This allows the initial
at re~t w~ obtain
value of O~ e~rresponding
(:J the pore fluid equivalent bulk modulus (see 3.2.11 to 01 = 50 to be ealeulated as
Kf :
A
l;iJ
a the measure of initial elastie wall size. ie half a'
m
= ~3 (1+2K I a'
o 1
= 35.3 'KN/m
eo
G the mean effective strass at the end of isotropic
eonsolidation (note that as = 0m for isotropie stress With a measured initial voids ra~io e • 1.3 and K as
~ state. eonsequently 0e is the same as in chapter 7) ealeulated then equation (7-171 gives'K' = 1780 KN/m2• Sub-
stituting this with v' = 0.28 into (11-3~ wa finally obtain
~ a ' a ' a ' T • U : the initial effeetive stress eomponents G = 917 KN/m2•
xo yo zo o o
and1pore pressure.
~.' The pore fluid equivalent bulk modulus ~as taken arb1t-
~ Ce(=2.303À) was assessed both directly from consolidption rarily as 100 MN/m2 for the undrained ene lyaí,s. AI though some-
tests and from tha empirical relationship (aquation 6.38 in
(L Sehofield and Wroth. 19681:
w~at less than tha aetual modulus of water (i200 MN/m2) t~e
results will not be affeeted 50 long as it is large eompared
Ci. with the soil skeleton stiffness.
cc = 0.83(LL-0.09) (11-2)
Q The stress oco is obtained from the condition that sinee
in whieh LL is the liquid :11mit. Both methods gave Ce = 0.42 tha soil is normally eonsolidated the initial stress point
(C
G
'-:i
/\...,.)
204 205
I,-:» 11.4.2 Results
1 I
f a~o °mo°';'
77 KN 1m2
q-
I{
~'
1ies on the yield surface. The initial stresses are fully
r" defined by Oyó = O~ (determined from the weight of overburden
X less the pore pressureJ and 0x~ = 0z~ = o~ = Ko O~.
fÇ!l Ko also has been defined (=1~sin.'J. 0m' is then ob-
-40m
-L
tained as above. With 02 • 3
0 then 0q simpli~ies to the
~rl ~~~~ator stress °1-03' Using Ko· O~/O~ it is readily shown
(;, ~ Figure 11-7 Excess, pore pressures aft,er ini tial
( °qo no 0m~ (11-4 J undreined loeding KN/mz,
() where
,( 3(1-KOJ
t:J The tank stability was caleulated using eonventional slip
{ n o = 1 +2K cirele methods with a shape eorreetion being applied. (A more
G o refined technique could h~ve expressed the factor of safety
tS With amo and 0go fixed it is an exerci se in the geometry
in cri tical state terms based on the par-cant ege yield around
a failure are. The centre and radius used for this analysis
of the ellipse to f1nd oco' The reader can verify that
L can initially be predicted using the sl í.p circle me thcd l ,
This gave a factor of safety greater ,than 2' which was cons-
a (oz )/(Mzo
L 20
co mo
+
qo mo
J (11-5 J idered adequa te for the proposed staged loading.
~,
Taking Ko = 0.558 as before, we obtain no • 0.627. Using the predicted pore watet pressures as the starting
~ lntroducing amo = 35.3 KN/mz we obtain 0qo = 22.1 KN/m2 whence point for a cdissipation analysis the Tesulting pore pressures
by (11-5J with M,c 1.05 oco = 24.3 KN/m2• lf variation in after the load was maintained for a period of one month were
({ the term (1+e) i6 neilected o 9 will vary linearly with the calculated.
vertical effective pressure w~1ch in this case means it varies
(( linearly with depth. The full water test pressure of 122 KN/mz was then sim-
ulated with the pore pressures at the end of th€ preceding
(( dissipation stage be1ng used as datum for th1s load1ng. The
This completes the derivation of the parameters needed
pore pressure contour plot is shown 1n figure (11-8J which also
f for this case study. The reader should note'that the pr9-
cedures are the same in essentials as used for example 1 in shows a well defined zone at full pIastic Y1eId.
('" section 7.9.
'\ The factor of safety under these condit10ns was assessed
f at approximately 1.75 but of greater s1gnificance was a com-
puted settIement of nearly 500mm near the centre line of
f
~\
\...."1
f'J 207
t
V
206
,( Horizontal displacements mm
\,../
50 100
-(
\:../
..0;:0"
~
1
"-
"- ,,
\
CI ,
\
(( 10 I
I
I
I
r- I
\( / Measured
Q~ E
:Ea. /
I
/
I
-40m
<t'
U> I
----L- o
f;7ff4W,?I7$A')/J,W4W»W/////h'/A'/&ffH1 ,- "
({' 20 I
I
"i:'
I i
~ Figure 11-8 Exces~ pore pressures at full water test
I
I
t
(:::) lo~dirig I i
\ I
I 1
!
Ç\ the tank. Re-analysing the problem for drained conditions by
setting the bulk modulu~ for water to zero the final settle-
I
i
t2,J
ment of approximately 1.5m and a differential of nearly 0.75m
A were calculated. Whilst these values are large they did not 30
\V require any special design changes to the proposed structure.
;~
(
,_o,
208 209
(
'-' With insufficient information non-linear methods may not
,(
l offer significant advantages over linear elastic solutions
whilst the cost of an'alysis may be 5 to 10 times greater and
'-, require computer programs not readily available to most
practising engineers. In any case it is desirable to carry
C out a linear elastic analysis first to obtain a feel for the
problemo
L x -r + -I- -!- y..
«r:
x x As an illustration of the use of conventionallinear
elastic methods for cavern analysis, a finite element analysis
X-+-+ +-t- y..
~ x for the design of the tailrace tunnel will be consideredo
X A'.-r-+- -+--+- ~ The tunnel profile is illustrated in figure (11-11).
~
1(' )<
X
----------
/~~~"
/'
....--1-.............
...•.•
. X
~
r:'( f- ;t I I,IA.~ ~~~ '" '>( ~
\-) I I 1111/ ~\ \ \
\:, I I I 11,11 \\ \ \ ..•. ..•.
.' I 11\ ,\ \
f \
, - I I I I 1I 11 I\ \ \
'( 111I I11 I
(
( t I I 1111 /11 I t
i , \ I I 11 11 I I
t I E
~ \ \ 1111 + '"
~'
.( *\ \ 1\111 ",111 I
G
~ \ \ I I 1\ ~7/,;I I
;/1/ I
f
f- f
G ~
.••.
•••••••• \ 1i
I. " " ,; I / / / I I-
f-
((
'..J
\ '\ "o .: ,,' // I I
'" \ \ \'" /,// / I I I I
')ç
~ \ " \: o ' 7"/ /
X 'l\ ", //
'f.
X X
~ 'X. 'X. ~-t ,A'/' X ~
'X. ')t... -I-. r-
( )(..
'- 'f... -I- + + + -+-+ -+--1- X x
(l )(..
'f..
"'f.. -I- + + -+- ++ + -\- X
X
Figure 11-11 Tailrace tunnel, part of f.e. mesh
rJ 'f.. x
« 'f..
+ + +++++-\- ..l(
As a first step it was necessary to obtain the magnitude
of tensile stresses due to the formation of the opening under
various values of Ko' Since at the early stages of the pro-
Ú KEY
ject no insitu measurements of Ko were availabl~; it was
Tenslle zone
l(' Compression zene
--#-
-+- important to test the sensitiv1ty of the analysis to the var-
iations in this parameter.
c~
- A second parameter study exemined the variations of crown
\.t stresses with rock bolt support pressure.
r,
,{ Figure 11-10 Principal stress vector plot for no- It was first determined that Ko = 0.5 resulted in the
\! tensiorl analysis with Ko = 1 highest value of tensile stress. A series of f.e. analyses
r,
\
210 211
( were therefore undertaken with Ko = 0.5 and various support 11.6 Acknowledgements
pressures. This set of analyses showed that with no support
~ pressure a tensile principal stress of 50 KN/m2 occurred at The analyses discussed in this chapter were alI under-
the tunnel crown. This decreased linearly with increasing
'-( support becoming zero when support pressure was approximately
taken by Geocomp U.K. Ltd., the computing subsidiary of Soil
Mechanics Ltd., and thanks must go to the membars of staff who
\...:,.
51 KN/m2•
assisted in alI aspects of the studies.
!
'-' The "support pre99ure" cauged a pre-stressing of the rock
11.7 References
/<:» above the arch. A uniform set of forces was applied to the
side of the tunnal balancad by an equal and opposite force CARSLAW,H.S. and JAEGER,J.C. (19591, "Condu.cüon 06 Hea..t -<.n
( positioned at the anchorage points within the rock mass as
illustrated in figure [11-12J.
Solidl,", 2nd ed., Clarendon Press, Oxford.
l HINTON,E. and OWEN,O.R.J. [1979], "An In;tJtodu.cüon to F-<'nUe
EtemeYLt Compu..tCtÜoYLó", Pineridge Press, Swansea, U. K.
'-
<t JUMKIS,A.R. [1966J, TheNna1. SoU Mec.ha.n-tCÁ", Rugars Univarsity
.( Fo~ce opplied to \ \ I Press .
011 onchor \
~ SCHOFIELD,A.N. and WROTH,C.P. [1968J, "CJU;üc.a1. S:ta.:teSoU
Mec.ha.n-tC6", McGraw-Hill.
{
11
-,
zente of infl~2\.~\ \ \ WROTH,C.P. [1975J, "Ins1tu measurement of initial strasse5 and
o force ~ \ \
~ deformation characteristics", State of the art report on
i
'Í '-- -,
'-..'--7 <, -,
~
\
\
\ Measurement of Soil Properties, North Carolina State
University.
1~~~
1
)
-~-
L,
(
-----...,[>
-- ----
l
'- '1
-.i
-Á
l
213
1
~i,)
'-
I
'- ,)
CHAPTER 12
L-
I
":
< FINITE ELEMENTS IN DESIGN
1\'-./ with particular reference to deep basements in London Clay
by
"
"-'
!\
o
Brian Simpson
(
"--' 12.1 Introduction
~,
The application of finite element techniques to
..{ practical problems requires a knowledge of the fundamentaIs
of the method and an understanding of the properties of the
~ materiaIs involved. In addition. experience in creating an
adequate finite element model and interpreting the output
,
I'
from the computationsis needed.
\ IndividuaIs or small groups of engineers have tended to
develop expertise for a particular class of problems. for
'I
example tunnel construction. embankments on 90ft clay.or
( offshore structures. Finite elements have also been used
\
extensively in the study Df deep excavations in the London
, Clay. and in this chapter the progress of this work is
\
described and some of the lessons that have been learnt are
discussed. Reference to Burland (1978) is also recommended.
y:J One of the first uses of the finite element method in
lu
the study of London Clay was to facilitate a back-analysis
I
of a deep basement which had been monitored during construction,
\.- -"
This provided parameters that were used in analyses of.other
,(
excavations as an aid to designo In this latter role. the
v method has proved useful in the quantitative estimation of
v deformations. and has also been very helpful to the
)., qualitative understanding of material behaviour. modes of
\.J overall deformation and the influence of construction detal1s.
~ The development that has taken place has required studies
{ of methods of modelling the geometry of excavations and of the
<:»
,( material propert1es of the clay. In parallel. with this. the
\:... types of finite element in use have improved.and the ab11ity
to man1pulate the computed results 50 that they can·readily ..
~ be used in des1gn has increased. Thus t hís t op c rs'a" gQQd
í
~
1:,
Possibly the first'major use Df finite elements in the
study of London Clay was,publ1shed by Cole and Bur Iend (1972).
'i They used the method to derive an estimate of the undrained
0:,)
I
stiffness of the clay from the results of monitoring the ~Fjll
"-'
I
IV
'(
'-
t J
I
I Di.phragrn
I W.II
~ Figure 12-2 Britannic House: section A-A (Fig. 12-1)
AJ after bulk excavation and construct1on
f of raft.
Towet
J t35storiesl
12.2.1 Computational Technigues
(j
In order to carry out this ;tudy several questions
(O' had to be answered.
(f" !
'\ o
o
a) How should a rectangular or L-shaped excavation
:l
((":, õtuee be modelled? Two-dimensional plane strain and
, Block axi-symmetric elements were available, and it was
fI e,torl ••
decided to use both of these and compare the
results with measurements along section A-A
)
(Figure 12-1) for which the most complete set of
... ) measurements was available .
Figure 12-1 Plan ,of excavation for Britannic House were placed. In view of the short time scales
it was considered that the London Clay wou1d
~ bBheVB 1n en undreinBd mennBr. Isotropic lineer
The main features ,of the scheme and the ground
~ conditions are shown in,Figures (12-1) and (12-2). An elastic behaviour was assumed, with a Poisson's
excavation roughly 110m,by 60m was required, extending to a ratio of 0.49. It was considered that even if
L depth of about 17m. A diaphragm wall was constructed around this did not give a very accurate representation
of the true behaviour it would form an adequate
<L the perimeter and excavation was then carried out very rapidly
in the centre of the site, leaving berms to support the . basis for extrapolation to other designs.
(( diaphragm wall, as showm in Figure (12-2).The computations
'- were carried out for th~s rapid excavation. Subsequently a c) What were the initial stresses in the ground
CL raft was cast over parts of the central excavation and the before excavation? It was assumed that the
r- berms were progressively replaced by struts. effective vertical stress could be obtained by
~ subtracting the measured water pressure from the
total overburden pressure at any depth. The initial
~ horizontal effective stress was obtained by
r;-
~
H
y
fl
~il
216
multip1ying theivertica1 effective stress by K o ,
217
,
the coefficient;of earth pressure at resto The 12.2.2 Results
0:1 variation of Kolwith depth was estimated from
earlier measurements for sites in which London Clay The purpose of the study was t~ abtain ~alues for the
~) apparent undrained Youngs modulus of the cl1ay,,EU' , whí, ch was
I extended to thejground surface. More recent work
by Burland, Sim~son and St. John (1979) has expected to increase with depth. Anal'yseslwere carried aut
~ both'for the end of excavation and the en~ of~construction
1:v suggested that the use of K can be misleading
because it is g~eatly influ2nced by drainage af the raft, before the'first struts,were.~laced. By ' a
, 1" ", ' ,
/:'1 and reloading of the clay. It might have been process af tr-í.e
l and errar, the pr:o<'fiil,e!'l
Çl:f' Eu wet;'ea~j usted
W better for thislsite, which was subject to until the computed displacements of ths.diFiphx;agmwall
d underdrainage, to extrapolate the horizontal matched the field measurements. The results af one such
~ effective stress directly from other si'tes,ratlier computation are shown on Figure (12-4).
( than
, assuming Ka to be similar.
I
1
'<..Y dl How could the remove l of the excavated -sodI be 15
modelled? Theionly technique available was fuMN/m'
l ~o caicúlate th~ initial stresses on 'the inter-
e:
10
5
200 4po eoo
LONDON CLAY
[ face be~ween th~ material to be excavated and ;;
f o
~hat remainir:lg,!andthen,apply these ta~ trc!!ctions
( to the finite e!ement mesh, as shown in Figure, L j.
-10
(12-3). It was!necessary to model only chengas ~
f' af stress,'and therefore' gravi ty loads ',werenot' -15
--"- ..•----- ---..
".m
IwóOLWlcR &
AEAOING BEOS
I
i
(:(7 app1ied to the ~lements of the mesh. The procedure
is that of strategy 1 described in Section 3.5.
The only finite1elements availab1e were three-node
f' constant strain'triangles.
Figure 12-4 Britannic Hause: Computed and observed
IÍF' displacementand corresponding variation
~(
G
,{
218 219
12.3 Underground car park at the Palace of Westminster.
v 12.3.1 Oetails of the project
l One of the major and,most publicised practical
applications of the finite'element method in London Clay ~as The ground conditions consisted of 10m of alluvium
,1 related to the design of t~e underground car park at the and graveI overlying London Clay to ,a'depth of 45m. The pore
"--' water'pressures were found to be in hydrostatic equilibrium
House of Commons in London. This project, illustrated in
([ Figure (12-5) required an excavation up to 18.5m deep with with the water table in the alluvium. A number of fine sand
diaphragm walls 16m from tbe foundations of the Big Ben clock and sllt partings were found in thé clay unper~ying the
({ excavation leveI. The diaphragm retaining,walls were taken
tower and only 3m irom Westminster Hall.
rr down through 'these partings to a depth of 30m to cut off
Q horizontal seepage and eliminate the possibility of hydraulic
1';-
uplift.
"(i'
The car park was constructed first by installing the
diaphragm walls and foundation piles (Figure 12-6). Steel
r'
'Í'
f A,ver
T/llm'$
t columns for supporting the floors were lowered into the
lined 'boreholes above each pile and grouted into position.
The ground ftoor slab was then cast on the groJnd surface.
~'\
I{
ç'
\ 81gB,<,\
tloClclOW,t
Excavation took place beneath the ground floor slab and the
next floor was then casto The process was repeated until the
lowest slab had been casto
q\ KEY PLAN
50 m s London
Readlng 8.a. Vl 90 Clay
U E
(;
-.
..J
Weslmlna,.,Hall W
::::'==~-:'=:=:'-:"':';:====:=':.':===
Gj
..J
80
C
W
cL U
;;,
fil 70
~v
~ '" '''',
(i Figure 12-5 Underground car park at the Palace 60
Woolwlch and
ofiWestminster.
(L !
Readlng Beda
"\
'(
220 221
Woolwich and Reading Beds (see Figure 12-4). The upper occurred at most of the monitored sections.
~ limit on E was adopted in the belief that the predicted
'~
\
~) behaviour ~ould be conservative.
J J\
\.
I /original Mo~ementtowards excavation -
The distribution Df initial horizontal effective stress ~I.evel O 4 8 12 16 20 24 28mm
r=- i ' , , , , , ,
T
Ir 104
"--' wee oelculeted teking into eccount the fact that the surface
of the cley had been reloaded by deposi tion of 10m cf sand
l I I
C
!,
and graveI; the results of the calculation are shown in
Figure (12-6). Details Df this are discussed by Burland
L • 100
~18mL
c:i 96
l
\J et aZ (1979) and Simpson et aZ (1981), and may be summarised
L as follows. L "l
-1 cri
E 92
,{ a) The horizontal stresses that ·would exist at the ãi
"-' end of erosion of the clay are first calculated.
~ 88
...J
{
\.,..- In the case of the London Clay these are governed 84
by the coefficient of passive earth pressu~e,
t K , for most of the depth Df the clay relevant ~ 80
\f tB the computation.
I 76
;\..._J
in stages, inserting springs to represent the floor s:abs
as the excavation proceeded. Using this step-by-step approach
the finite element model can represent the sequence of
the ~oolwich and Reading Beds to a value below that inferred
from the observations at Britannic House, This decision
was thought to be conservative, but Bur1and and Hancock
( excavation and construction followed in practice, and this (1977) considered that it had led to an underestimate of the
....... )
has a profound influence on the final movements and stresses. bend1ng in the wall. This type of consideration requires
i\'0! (An alternative approach to design of diaphragm walls has careful attention when parametric studies are carried out
sometimes been followed in which the wall is represented using finite elements. In a subsequent back-analysis, St.
\ as a multiply propped beam subjected instantaneously to John (1975) found that an increase in the stiffness assumed
~umed final earth pressures. Because this fails to take for the deep layers led to a significant overall reduction
. ,,--./ account of the construction sequence, the calculated pattern in computed displacements .
\ of displacemen~s is very different from that found in practice.)
x.>
"Computed" and "predicted" movements at the ground
J~ 12.3.3 Computation, predictions and measurements. surface published by Ward and Burland (1973) are compared
with the measured values (Burland and Hancock, 1977) in
~ Measurements of ground movements around the excavation Figure (12-8). The distinction between the results of a
showed that Ward and Burland's computations had led to good finite element computation and the engineer's prediction is
i predictions of ground movements. The final deformation Df an important one which is too often forgotten. The finite
« the diaphragm wall, computed using plane strain elements,
is compared with measurements for the midpoints Df the south
element method tncorpor-et ee a highly idea1ized mathematical
model of a structure. Whilst this wi11 in some respects
:J. and east walls in Figure (12-7). It can be seen that the add to the engineer's insight into how the structure will
overall magnitude of horizontal displacement was well predicted, behave, there will often be some aspects of the computation
~( but the computation suggested larger movements at depth than which do not match reality very closely. It is essential
:.;
Y'
~\
222 • 223
( \ linear elastic model with stiffnesses derived from previous
'r '.
(.,
Ward &
Burland (1973)
back-ana1ysis.
\ '
/J
'( _ '\.._ __
Distance Irom wall(m)
30 40 50 60 70
12.4 The work of St. John.
v
"
/
L:
- °1 ~
E
..§
, A·· '.' · __
... '
A....
-<:P;edicted
...~>
St. John (1975) reported the resu1ts of a research
study of both measured and theoretica1 behaviour of ground
around deep excavations 1n London C1ay. He reviewed the
,{
\:'J
~E 10
• : ~-------~--~ predictions that had been made us~ng finite element techniques
and considered both the material properties of the c1ay and
AA Computed
L ~11>
VI
A •
the techniques available for mode:ling three-dimensiona1
excavations.
,(
~. 20 •• LEGEND
12.4.1 Stress-strain behaviour.
{ • South,wall } measured
~
A East wall Eu(MPal
~ o 50 100 150
E O k r I i I
~-
E
-20
f
V
~'
L
n;
•• ••
A
• -;;;
e,
:.
!'(' ~
Õ ~ 0.2
:x: 10 20 30 40 50 60 70 UJ
a:
~ f-
~lr""""\"
(/)
Dlstance lrom wall (m)
'(
Figure 12-8 Computed. predicted and measured :rl 0.3
x movements ai ground leveI for the
u,
u,
z
<t
G that engineers review the results of computations very
UJ
:::;; 0.4
(( carefully. and they should be prepared to trust their own
\_-)
judgement in preference to the computer printout when LEGEND
<-
\---./ necessary. Too often. intelligent and experienced engineers -- Vertical field
measurements
( abandon their previous wisdom when faced with computed results osj- -- HOrizontal field
LI and accept "predictions" which are fairly obviously irrelevant. measurements
((
\J
In this case study. the computed resu1ts were based on
;'\
'-.J the unreal assumption of plane strain. It was judged that Figure 12-9 Undrained Young's modu1us for
(
this had led to unreasonable results at distances greater London C1ay'deducéd from laboratory
,"'-" than 30m from the wall. and the predicted settlements were and field dâta.
!1 therefore modified accordingly. and thereby carnecloser to
the subsequent measurements. ,It can be seen that the overall Figure (12-9) based on St. John'swor.k.,shQws c leer Iy
CL magnitude of movement was predicted reasonably well. but the the large difference between stiffnesses measured in labora-
measured movements had a rather different distribution. tory tests and those backfigured from field measurements.
'J, In more recent work. Creed. Simons and Si1ls (1980) have
In summary. this work has advanced the methods of obtained similar results from a back-ana1ysis using an
\l. ca1cu1ating in situ stressesl it has demonstrated the power anisotropic modelo The reason for this disparity has been
({ of step-by-step model1ingl and it has shown that movements of hotly debated. and many workers have concluQed that. because
the right magnitude could be predicted using an isotropic
r:
'--l
fi
'--)
j 224 225
~J of sampling disturbance, laboratory stiffness tests can give St. John carried out computations assuming constant
,1 Young's modulus with depth and remote boundaries at about
no useful information about field behaviour, at least in the
V undrained phase. Parry (1980), however, considers that by ,six times the sxceve t í.cn depth from the excavation itself.
.{
cycling the stresses on the specimen a good indication of The value of the,ratio of total horizontal stress to total
ci field behaviour can be derived from laboratory tests. Costa- vertical stress (K T) was taken to be O, 1 or 3.
1 o
v Filho (1980) has produced data which suggest that the magni- -0.1
J
,< tude of the stress qr strain increment is of great importance, /
'-l...- end Simpson et aI (1979) a~9umed this in deve10ping the / 0.1
numerical model presented later in this chapter.
~ /
«
--1"'L~~...
O~"-- .• /
'<- In the computations discussed 50 faro linear elasticity ..•.......
-'0 /
~{ was assumed, making no allowance for the limited shear S1r1ln <. /
~ strength of real soi1. St.' John suggestéd that this " P1lne 0.2
Str • .,
r simplification
of computations.
was having a significant effect on the results
He considered that a more realistic
~. distribution of movement might be obtained if passive yield
0.3
and clay softening in front of a wall were correctly modelled.
-< This would ocdur progressively at different levels as the
r: excavation proceeded.
"
\(
12.4.2 Modelling rectangtilar excavations
(a) Vertical movements (b) Horizontal movements
~'
1n order to atudy the differences between 2-dimensional (settlement p09itive)~
and 3-dimensional models of excavations, St. John stud1ed
an unsupported square excavation of depth equal to one third Figure 12-11 Comparison of computed surface movements
':( of the length of a side. He modelled this 1n three dimensions
, '1 for excavations as shown in Fig. 12-10
using linear elast1c elemen~s and compared the results with
1I
',j plane strain and axi-symmet~ic approximations. The equivalent
Figure (12-11) shows a comparison between computed
~1
,)
square, circular and infinitely long excavations are shown in
Figure (12-10). horizontal and vertical movements at the ground surface for
the plane strain, axisymmetr1c and 3-D cases, alI for
,\ KoT = 1. Good agreement was found between ~he movements
v ( computed in the axisymmetric analysis and in the 3-D analysis
"v on a plane normal to the side of the excavation and passing
( through the midside. The plane strain analysis gave fairly
<..J similar vertical movements, but the main horizontal movements
.( Axisymmetric solution were up to 100% too large;- Even though'the 'excavation' had no
\.../'
support, the 3-0 analysis gave much reduced horizontal
~ movements towards the corners of the excavation.
, (
-.../
Three dimensional computations are likely to be
unreasonably expansiva for practical designs. It therefore
l seems sensible to use axisymmetric.analyses to compute the
'(
~ behaviour normal to the wall at its midpoint. St. John
suggested an alternative procedure using both plane strain
1 and axisymmetric analyses, but this 15 not considered·
c necessary by the writer.'
\.\.
)l et graeter dietenoe. Th1e lBrgely eccounts for high stiffnass as tha next 5mall increment Df strain was
the settlement trrough apparent in the observations applied. Lewin reported a similar high stiffnass whan the
(' ) shown in Fig. (l2~8). direction Df straining wes changad.
,(
,-i
b) Tha elastic moduli used in tha computations wara Reviewing these and other data. Atkinson (1973)
sv' much largar than ithose consistently measured in concluded that for a small stress incrament following a delay
.\ high quality laboratory tests on specimens cut at constant stress tha stiffnass is much graater than would
'-./ be the case for continuous straining. Similarly. high
( ;~~;l~~~~kh::m~!~~~te~I:t~~~~~::~~ ~:::~~:~ in stiffness is found when an abrupt changa in the direction
~, large diameter (a65mm) plata baaring tests fairly of tha stress path occurs. This implies that for small
-( close to the back-figured values). stress increments in the fiald, whera tha thrashold valua
'-..Y
is not excaaded, laboratory measurad parameters will over-
-1 Three further problam~, not apparent in the analysis estimate the actua1 strain. By comparing laboratory and
v'
(( for the Underground Car Park, are considared significant. plate bearing test rasults, howevar, Simpson et at (1979)
v
I
suggested that for 1arge stress increments the stiffnesses
( c) Since tha .tiffn~ssas usad in tha computations measured in the 1aboratory might be r~levant.
G were derivad from study of other excavations of
~L similar aLze , it lis not cartain what her- r:
computations usí.rrgtha sarnalinaar proparties
The considerations outlined above led to'th~ development
Df a stress-strain model for haavily ovarconsolidated clay,
1 are appropriate t'oaxcavations of significantly
diffarant siza. A similar doubt may axist if
Model LC, with tha following main featuras:
~
TYPICAL STRAINS FOR:
Deep exavatlons
t
! ap,proximately line~r behaviour 'norma11y measured 500 Plale bearong leais
& in the·laboratory. T~e parameters of anisotropic
b,( elastiçity., as measurrd in lc:boratory tests, are
used. but the,behaviour i! not entirely reversible.
"I"
wu
~
Z
200
)x Laboralory tests
«
o
"-"
-( (d l The 'elastic' range app1ies to straíns up to 0.02% ~ 100
":1" 200
Figure 12-p Model LC: Variation of secant
Young's modulus with sheàr 'strain.
-(
--.-
(j~-crhl A typical atreaa-atrain curve der1ved from the model
-Çi"
.'- is shown in Figure (12-12). The inset to the figure
(kPa)
~ shows the higher stiffness assumed at smaI1 strains and
applied again when the stress path is changed. The concept
cri
'i 100 h of a 'kinematic yield surface' was used to achieve this
effect. If required. secant Young's moduli could be derived
1.,) Inltially rI'" D 200 kPa
CTh'_ 400 kPa Ko·2
from the model as a function Df strain. Values derived from
Figure (12-12) are shown in Figure (12-13)and are found to
) compare wel1. in the appropriate ranges of atrain, with
(.)
(
.::, ...)
O" I
4
I
6
I
a
I
10 12
I
both field and 1aboratory measurements.
C!.- -100
•...•..•.
_t .•.
I SMAll STRAIN To check the model, computations were carried out for
comparison with several sets of field measurements~ including
the Underground Car Park at the Palace of Westminster. For
(L- this. an axisymmetric idealization was used~ as suggested by
St. John's work. The main results are shown in Figure .(12-14)
ei. Figure 12-12 Model LC: Uridrained plane strain behaviour. (movements at ground leveI) and Figure (12~15)(horizontal wall
movements) . It can be seen that the shap.e bf the .settIement
1 distribution outside the site is differeRt from that comoutad
using linear elastic assumptions: the settlement 'trough
1
I.
(
y
~.-
(
230
dimension is 1arge. lhe pressuremeter ~as therefore
231
j
Ccmpulod by ,
( j Ward and Burland(197J)
v., '
\ Distance from wsll!m)
í \w w ~ ~ ~ ~ 70
C o'
(
\._,
Ê
f .\ \
/{
\
~ 10
/ Constant
(
'-' j 20
~""5("-
\. ./
South wall durlng excaY8tion!measured)
e.s! wall totllljmaa5ured)
\_ ....•.·•..••..••••
South w.1I total (measu(8d)
pressure
{ (a)
(
~
~ Southwan
South Wl!llIt0;d81 during constructlon(m •• surad)
(meBaurad) --o{
f,{ _. ,-,_.
_.... Jw Comput,dbV
ard.nd8urland1lg'31
I( ,-;:.,... "-'----
..-'-"-ç_
'/;;-~;;...._ .:::::.~._ Ea'twalltotar
Imeasured)
\J
1~ 20 J~ 40 :~---- ajo /Õ - Figure 12-16 Finite e1ement mesh used to mode1
Oistance from wall(ml
~~#~oJ;.'-::''' A~~rt~~~er~~
(~-
100 / ~ _====-lF~
1.
( 96
,IJf;/"~/~---- .,'. 17
'-' o 92
10
\ CI)
E 88
\,.
~
.!l 84
t~,
( 80
In ~r.S9uremeter
\.:/ p. preSlIl1'B
."---( 76
72
cu·undnlined sh.ar strength
trom1C2mm UU Irla.lal
tests
~
\...:.,
4 6 6 7 8 9 10
CHANGE QF DIAMETB1 (%)
<l Figure 12-15 Palace Df Westminster. Comparison Df
measured and Model LC Computed horizontal
({ wall movements. Figure 12-17 Comparison Df measurements from a
sslf-boring pressuremeter with Model
li As a further check. computations have been carried out LC results.
to model tests using a self-boring pressuremeter. lhe model
~ is designed for 2-dimensional situations and does not deal
(i adequately with situations in which the strain in the third
((
\
\J
,r
233
v 232
(
-.....;J Some of the results are shown in Figure (12-17) 12.6.1 Type of analysis
I( and it is clear that ths model is able to prediet the shape
'0 of ths complets load-displacement curve. Since both stiffness The work discu~sed in section 12.2 to 12.4 showed that
( and strsngth are assumed to bs related to mean normal strsss, an isotropic linear elastic model, with Young's modulus
the value assumed for K affscts the results, as can be seen. varying with depth, can gensrally be used to obtain a
.( It might bs tempting toOuse this as a means of dsriving K reasonable estimate of the scale of movements around a deep
'- from the pressuremeter test, and for thsss data this proc~ss excavation. Values for Young's modulus can be estimatsd
.~( from Figure (12-9).
would give a rsasonabls resulto However, such an approach
"--
would requirs unreasonable reliance on the detaile~ accuracy
? of the modelo Typical computed and measured unload/reload Results from non-linsar computations should be compared
'-'
a. cycles are shown in Figure (12-18). Agreement is again if the design is critical or if it differs significantly
satisfactory. from the case histories upon which Figure (12-9) is bassd.
J It must be rsalised, howsver, that non-linsar computations
will be more costly and will make greater demands on the
J Depth time and skill of the enginssrs.
~
oltut
,.
Iml
The techniques described in section 3.2 for modelling
effective stress behaviour have become more widely available
,-( in the last few years. 80th undrained and fully drained
behaviour can be modelled and the consolidation phass can be
'( modelled approximately by specifying the pore pressures in
the data as the computation proceeds. The effective stress
'\ technique is generally to be preferred to total stress
~i analysis, especially for non-11near work 1n which dilation,
, pore pressure and shear strength are alI related. Programs
'~ are also available for computing behaviour throughout the
consolidation phase directly, but 1t will be rare that the
'( 00 , 3 • permeabi11ty of the ground and the timing of construction
CHANGE OF OIAMETER 1:1;)
\:.......:i will be known to 'sufficient accuracy to merit such
( sophistication and expense.
~~I
Figure 12-18 Computed and measured unload/reload
( cycles for the pressuremeter. It must be emphasized that the ca11brations which led
-..:...'
to Figure (12-9) were based on total stress, isotropic
é\
\...-J 12.6 Use of the finite element method in design of deep linear elastic models using constant strain triangular
( basements. e1ements. Some of the known inaccurac1es in thsse assump-
'-../ tions may hav8,cancelled each other, and whensvsr one assump-
,-1
\
An eng near dssigning a desp basement might decide to
í tion is changsd the calibration against monitored
\../
embark on finite ~lsmsnt computations for a number of excavations must be re-checksd.
1 different reasons. Probably the most common reason is the
nead to predict ground movements around the excavation, but In ear1y computations the process of excavation was
J estimates Df prop forces, bending moments and other quantities modelled by applying tractions to elements representing the
might also be sought. In addition, an indication of the 6c11 that rsmainsd. Ncwadays althcuih th1a methcd 1a atill
3 effecte of elternetive construction sequences, rate of perfectly satisfactory for single stage excavations with
excavation and other details might be needed.
:1 simple boundaries the alternative stratsgy 2 described in
Section 3.5, whersby the unbalanced stresses on the sxeavated
The purposs of each eomputation must be clear before boundary are relaxed in a less cumbersome manner, may bs
~
•r
it is begun, since this will influence the design of the mesh prefsrred .
J, and the choiee of suitable stress-strain properties and
r: e1ement types. These in turn will determine in what respects Whilst mueh of the ear1y work on exeavationsin London
~ the computation can be expsctsd to be realistie and which used plane strain analysis, 1t was arguad 1n sect íon 12.4
.~
features of the output might tend to be misleading. that axisymmetrie eomp0tations are often more rea1istie. The
'r-Ó.
-< alternative of using full thrse-dimensional computations will
,,(
,,--- \
q
\.
'r
r
'( )
234 .
usuaIIy be too expensive, mainIy because anaIysis of
235
1~,.....
V
,(
radius for the excavation face. A plane stress analysis in 25
plan may also be used to give an indication of the
"""
,(
~
distribution of movements. In the example shown in Figure
(12-19) the properties of the elements inside the excavation
~Jhtllr--r-
were changed in three stages in an attempt to represent the
1 removal of a layer of soil and subsequent insertion of the Thanet sands
I concrete slab. lhe th1ckness of the diaphragm wall was
varied as shown 50 as to givs it high compressive stiffness
145m ·1 Chalk
i but low bending stiffness in plano
Figure 12-20 Mesh used in analyses of
~ excavation for British Library,
-{ Euston.
..( Eight node isoparametric elsments were used throughout.
Fixed It was proposed to model the diaphragm wall using only one
,( bouodaries elemsnt across its width, and a check was made on the validity
aI 250m
Diaphragm wall o f this. ' A single column of elements was subjected to
'(
(cne element thick I concentrated and distributed loads similar to the expected
strut forces and earth pressures, and the overall
1 deflections and bending moments compare well with those of a
'\ more complex mesh subjected to the same loading. A similar
conclusion was reached in Chapter 3 where the wall shown in
'1 Figure (3-10) was idealized by a single co1umn of 8 noded
elements. A separate study using a finer mesh aga1n proved the
'( accuracy of the bending moments. In both cases "reduced
~) integration" (~ee e.g. Zienkiewicz, 1977) was used. It
-} cannot be 8xpsctsd, however, that the single element width
J w1ll give any useful infonnation about the detailed stress
( distribution close to a strut .
...y'
:(
..J CORNER DETAIL Although Model LC was used to include non-11near effects
in the analysis, no special allowance was made for slip
L Figure 12-19 Mesh used in plan to study between the wall and the soil. lhis.decision was subsequently
( relative movements of mid-sides shown to be justified since the computed shear stress on the
and corners of square excavation. interface was too small to cause slipping in practice.
}
12.6.2 Oesign of the mesh As excavation proceeds there is always a stress concent-
} ration in the passive wedge in front of the wallat
Some of the main points to be considered in des1gn of excavat10n leveI. Accurate model11ng would require a fine
1 a finite element mesh will be illustrated in this section mesh at this point and for non-linear laws good 'convergence of
by means of an example. Figure (12-20) shows a mesh used the iterative processes. Unfortunately neither of these was
1 for some of the computations in the design of the poss1ble. Because the excavation leveI keeps changing, a
-l proposed British Library, Euston. lhe mesh was designsd fine mesh would be needed at alI levels and this would become
1nitially for the study of ground movements, but use was unreasonably expensive. It was found that, using Model LC,
J. also made of other features of the output. convergence in the passive wedgs is rather slow. lhe
change in the overall pattsrn of behaviour is not, however,
-\
-r-,
usually significant after the first four or five iterations.
~~
'y !
I I
236 237
X 'wall by a factor of 3. The mesh used was shown in Figure
() It can be inferred that the linear elastic model, since it (12-20). This comparison demonstrates the important point
does not provide the iterative'corrections, gives that the stiffness of ths wall has very little sffect on the
~J
( particularly doubtful results in the passive wsdge. deformations, which are controlled by the behaviour of the
soi1 and the struts. However, the stiffness assumed for the
( Many of the computations show the diaphragm wa11 wall does change the computed bending moments considerably.
c moving down, mainly due to to drag from ths soil behind it. A non-linear model of the behaviour of the reinforced
( In this process the soil around ths toe suffers a bearing concrete wall could be used to advantage to gain greater
<:»
capacity fai1ure. The mesh employed around the toe of the insight into this problem.
( wall may well be too coarse to model this propsrly.
,./
(
hc,vltlCln
J of 3 change.
The resu1ts Df finite e1sment computations usua11y
j consist of both printed and graphica1 output. The need for
Figure (12-21), p Lot t ad by the computer, shows the
graphica1 repressntations at alI stages including pre'paration
effects of changing the assumed stiffnsss of a diaphragm
~( Df the data, appreciat10n of the patterns of behaviour and
.(
(
\
'(
( 235
239
Ol5P '(
( -S!
_F!!_ -9l
-"
.r
'- 10 ...•. ,
~
10
lond""
·'ov
." f- ,
-36
."
] ",~.~.- !20 'ou"h ·lJ
.1" "\\
7m
boi,,,,
\15m
""''''''''''11 -ao
-1S ."
1;'\- ao !M,m
-10
20 / ! -H -li o"~
1/ I
.{ h/// -5~ _"
2. tV-t -.,
'd ~
."~\-u -" -li -,H ~ 11.
~p;."// ."
J 30' /
~. of reac:tloo "
01 ground ancho"
-1l
-(
a)
b)
c)
Excavation to 15m,struts inserted during excavation.
Excavation to 24.4m 1eaving 15m berm.
As f'or,(b ) 1eaving 7m be rm ,
36 , H ·4 I 1 • 4 -, - .
\' .,
-{ d) No beriÍl. .) .) .)
'( \ ·1 ·1
Dlaphragm wall
'( Figure 12-24. Use of OILPO output showing vertical
I
displacements around an anchored
'( excavation. Contours in mm hand drawn:
Floorslab
',( settlement negative.
1(
'( "" ••••• ><, tbroughout this chapter,andan example of another style,
"' ..... "- ..•" General levei 'diagrammatic line-printer output' (OILPO), is shown in
'(/ ~'" 'K" 'K" lt
Figure (12-24). This provides a rapid and inexpensive
:x,"K" ••••.• •••••• of excavation
~
! :x, ~ ~ ,
""-
~
picture on the line printer which can usually be made
available well ahead of plotted output. Any scalar quantity
'-.../ ':::I-..'/-..'/.... 'f.... -I- (inc1uding componenffiof vectors) can be represented and
( contours are easi1y drawn by hand.
~ 'I...~ ~ '+-- -+- -+-
( 12.7 Conc1usions
'-' '{.... -J- -+- -~
J 1. Avai1able finite e1ement programs and the
"'"'i-.. -I- -+- -I- understanding of modes of behaviour.of deep
J excavations in London C1ay have both deve10ped
J '""'i-... "f..... "'"'i-.. ....,.. significant1y in the 1ast ten years. Each has
contributed to the other.
J -+-100kN/m2
2. Knowledge of the stress-strain pro~erties of
J London Clay has also developed. 8ack-analyses
Figure 12-23. Computed principal stresses for case (c) have made it clear thatits undrained stiffness
1 in Figure 12-22. is much greater than is ncrrne Ll y measured tn the
..\ laboratory .
communication of design information cannot be toa strongly
~ ·emphasized. 3. This work has shown that it is essentia1 to
calibrate proposed models of stress-strain
-( Examples of various styles of output have been used behaviour, however simp1e or complicated, against
'-\
j'
(.J
240 241
I"-J fie1d measurements.
12.9 References
(/ 4. An assessment of overall magnitudes of disp1acement
.'(
around stab1e excavations ean often be made ATKINSON, J.H. (1973), "The deformation of undistupbed London
\..J
using an isotropic linear e1astie mode1. Cl.au", Ph. O. Thesis, University of London.
~
5. A non-1inear mode1, Mode1 LC, has been proposed BURLANO, J.B. (1978), "App1ications of the finite e1ement
,(
and shows promise of good correspondence not method to prediction of ground movements", Chap. 3 in
r( on1y with fie1d measurements but a1so with in situ "Devel.opmentie in SoU Nechanice'', Ed. C.R. Scott,
.~ Applied Seienee Pub1ishers, Eng1and •
and 1aboratory test results.
J 6. When mode11ing deep excavations, assumptions BURLANo, J.B. and HANCOCK, R.J.R. (1977), "Underground car
({ ofaxia1 symmetry are often more appropriate park at the House of Commons: Geotechnica1 aspects",
than plane strain. The StruaturaZ Engineer, (55), 87-100.
j
7. The purpose of each computation must be c1ear1y BURLANO, J.B., SIMPSON, B. and St. JOHN, H.O. (1979),
,( "Movements around excavations in London C1ay", Proa.
understood, and great care is needed if other
--.{ features are 1ater extracted from the output. 7th Euro. Conf. SMFE, Brighton, 1, 13-30.
) 9. Oesigners may need to check conditions other COSTA-FILHO, L.M. (1980), Discussion, Proa, 7th ECSMFE,
than those considered to be most 1ike1y to occur. Brighton (1979), 4, 124-128.
::< The finite e1ement method can assist here,
particu1ar1y when used in parametric studies. CREED. M.J., SIMONS, N.E. and SILLS, G.C. (1980), "Back
) It shou1d be remembered that assumptions which analysis of the behaviour of a diaphragm wa11 supported
are known to be pessimistic in some respects may excavations in London C1ay". Proa. ênâ Conf', Ground
'"(
be optimistic from other points of view. Movements and Struatures, Session VIII, Cardiff.
'-..>"
-(
~) 10. The use of grephical output is vital at all LEWIN. P. I. (1970), "Stmeee deformation oharaatier-iet-ice of
--í stages of finite e1ement work. eaturaiied eoct", M.Sc. Thesis, Univers1ty of London.
o
{ 12.8 Acknowledgements MARSLANo, A. (1971a), "Large in situ tests to measure the
'-.J properties of stiff fissured c1ays", Lei: Austr-N. Z.
( The work in connection w1th the new British Library Conf. Geomechawice, Melbourne, (1), 180-189.
\..../
-( has been carried out by Ove Arup and Partners for the-
~ Property Services Agency for the Department of Education and MARSLANO. A. (1971b) , "Clays subjected to in situ plate tests',
Science/British Library. Ground EngineeringJ (5), 24-31.
(
~
PARRY, R.H.G. (19800, Discussion, Proa. 7th ECSMFE, Brighton
J The Author wishes to acknowledge the help of his
colleagues at Ove Arup and Partners in writing this paper (1979), 4. k29-131.
:]
...•.. SIMPSON, B, CALABRESI, G., SDMMER. H. and WALLAYS, M. (1981),
J "The measurement, selection and use of design parameters
for stiff c leys ", General report for the 7th Euro , Conf.
~ SMFE, Brighton, Vol. 5.
0.
1.;.,.. ••
(
-;» 243
( 242
~) SOM, N.N. (19f;>8),"The effeat of stress paib on the deformation
( and aonsolidation of London Clau", Ph.D. thesis,
<:»
University of London.
(
St. JoHN, H.o. (1975), "Field and tiheoret-ical: studies of the
,(
behaviour of ground around deep exaavations in London
SUBJECT INDEX
'-
ClaY"J Ph.D. Thesis, University of Cembr-í dg s ,
I
] WARO, W.H. and 8URLANo, J .8. (1973), "The use of ground strain
msasursments in civil engineering", phil Trans Royal Anisotropy, 176-178
I Soa., London, A274, 421-428. Associative plasticit'y, ses "Norma1ity"
8i-1~near mode1, 51-62, 65-66
J ZIENKIEWICZ, O. C. (1977), "The Finite El.ement: tâet had", McGraw Body force, 11, 25, 29, 35, 40-41, 42, 44
Hill, 3rd Ed. Case studies, 49-50, 51-57, 193-240
~ Constitutive 1aw, see "Variab1e 81astic laws", "Yie1d criteria"
{ Creep, 155, 199
Critica1 state mode1, 83, 113-148, 201-204
-( Critica1 state, 113, 145. C.s. line, ll7
Oams:
'( f.e. analysis Df, 38-40, 64-65 (See also "Fills")
flow through, 171
1 oiaphram wa11, 49, 196, 214-222, 225, 226, 234-239
oilatancy, 60, 78-81, 124, 183-184
~\
Oisp1acements in fi11s, 56-57
~' Orucksr-Preger, sse "Yiald critaria"
Effective stress:
~/ PrincipIe of, 29-30
) Methods, 29-40, 189-190
1,
,,/
E1astic moduli:
Compliance matrix, (chap.7)
( Modu1us matrix, 31
,
Relations between, 32
( Values Df, 50, 55, see a1so "Material constants"
\.....-
Elasto-plastic laws, 66-69, 84-86, S88 also ~Yi81d criteria"
,( Elasto-viscop1asticity, 86-91, 181
'--"
Elastic wall: (definition of), 115
(
'--7 Embank.ments, see "Fills"
':( Excavations, 29, 47-51, 196-199, 213-242
~ Extended Von Mises, see "Yie1d criteria"
J Fi11s, 29, 51-57
Finite e1ement(s):
J Assembly process, 23-25
Basic theory, 11-27
~ Coordinate transformations, 15-18
Meshes, 13-14, 50, 56, 164, 216, 231, 234-235
1 Stiffness matrix, 20-25, 165
r=:
Types of, 12-14
J Flowru1e, 73-75, 123-125, 183-184, see also "Normality"
~,
l
l
':(
')
245
1 244
Von Mises, see "Yie1d cri teria"
'/.J Initia1 stresses, 2, 22, 29, 42-46, 47, 133, 195-197, 203, Yie1d criteria:
for Critica 1 state mode1, 121-123
204, 215-216, 220, 222
,.( Orucker-Prager~ 62, 78
Interpclation functicn, aBe "Shape function"
._../ Extended Von Mises, 62, 202
Invariants (definitions of), 4-8
~( Hvors1ev, 113
V
Jacobian matrix, 18
K-G mode1, 4, 62-64, 65-66 Mohr-Cou10mb, 3, 52, 60, 62-63, 66, 76, 146
~( Kinematic yield surfaces, 147, 229 for Multi-1aminate mode1, 182-183
.Y
-( LC model, 227-229, 235 Tresca, 4, 55, 66-67, 76, 82
~, Lode angle. 6-8, 62 Von Mises, 4, 78, 202
London clay, 49-50, 193-242 Yield function (definition of), 71, 75
J Lower Bound Theorem: Yie1d surface (definition of), 71-72, 118
} statement of, 101
application, 102-104, 109
:1 Material constants (determination of):
for Critical state mode1, 132-136, 202-204
'] for K-G mOdel, 63-64
,.( for London clay, 217, 219, 223, 236
Mohr-Coulomb, SBe "Yie1d criteria"
~" Non-linear techniques:
-,
-.{
General, 149-159
Initial strain method, 154
Initia1 stress method, 152-153
--( Secant stiffnes§ method, 158
""~'\ Tangentia1 stiffness method, 156-158
:i Visco-plastic method, 154-155
Normality, 95-99, see a lso "Flowru1e"
No tension material, 178-179, 207-208
~ Octahedra1 stresses, 7
j Over-stress ratio, 52, 55, 66, 176
:r) P1astic potentia1, 75, 124, 125
Pore fluid equivalent bu1k modu1us, 30, 34, 189
ic..: Pore pressure parameters, 33-34
Rendu1ic surf~ce (definition of), 117
~( Retaining well, 49-50, 196-209, gee also "Oiaphram wal1".
JI Rigid-plastic model, 93-95
-:-< Rock bo1ts, 209 ·210
J Seepage ana1ysis. 11, 161-172
-( Shspe functions, lI. 15-16
J
-( State boundary surface (definition of), 115
J Sub-critical (definition of), 117
~( Super-critica1 (~efinition of), 117
'-" Surface tractions, 11, 25-27, 29, 40-41, 46, 101
r(
.J
Tresca, see "Yie1d criteria"
Tunnels, 193-196, see a1so "Underground openings"
J Underground openings, 207-210, see a1so "Tunne1s"
Upper Bound theorem:
•...
~ statement of, 105
app1ications of, 106-108, 110-112
J Variab1e e1astic 1aws. 59-68, 118, see a1so "Si-linear mode1",
"K-G mode1", "hyperbolic mode1", "LC Mode1".
J Virtual work, 11, 21, 25, 26, 36
J. .
Visco-p1asticity, 86-90, see a1so u Non-1inear techniques "
J .~ j
':"d
(