You are on page 1of 170



. of the
Tl1eory of. Plasticity


Soil Mecha11ics

1. Salcl1yon

Ecole Polytcchlliquc
Ecole Natiollolc des POllts el Cllm(s.w]es. Paris

Trallslat iOIl hy

R. W. Lewis and H. Yirlogcux


hy Ille (/1111101'

;1 Wilcy-llllcrscicllce Pllblimlioll

. I



New York




First published uncler J. Salcnc,:on T1/(!ori(~ (I<~ {a p{o.llicit<; JJOU/'

(i to 1I11:c:fllliqIlC des sols by Editions

Eyrol ks-Pa ris.

Ic~s (ljJpliCOliolls




translation Copyright

1977, by John Wiley

& Sons, Ltd .

All rights reserved.

No part or this book Illay be reprodlleed by any Illeans, nor
transllIitted, nor translateo into a machine langllage without
the wrillen permission or the puhlisher.
I.illrar), of C()lIgn~s.'i Cat;lIogillg ill Pllhlicatioll Data:
Salcnc,:on, Jean.
Applications of the thcory of plasticity in soil mcchanics.
'A Wilcy-Interscicncc publication.'
Translation of Thcorie de la plasticitc pour Ic.<; applications
;\ la Illl:caniqlle des sols.
Include; bibliographics ano index.
I. Soil mcchanics. 2. Plasticity. I. Titlc.
'/"A710 .S241)
ISBN 0471 749R42
Printcd in (ircat Britain by Pagc Tlros (Norwich) Ltd, Norwich.


~-- --

Pr(~/(!ss()r .! ('(Ill 1I1017del




Chapter I General cOllsideratioll<; Oil the plastic behaviour of materials

I Yield criterion.
2 Form or the yield criteri;1
3 Plastic derormation
4 The origin of the plastic deformation
5 Flow rule .
6 Principle or maximum plastic work.
7 Consequences or the constitutive Jaw
8 Validity of the principle 0(' maximum pl;lslic work for the case or
9 Final rem;lrb .
10 Application of the pl:lsticity modclto soil behaviour

I (i

Chapter I I A shorl SIln'CY pf Iile prohlcllls of c1aslo-plastidly

I Introduction
2 Statement or the problem. Method or solution
3 Ikhaviollr or a systcm subjcct to a 10;lding proccss .
4 Importance of the geometry changes
5 Examples


Chapter I I I The prohlems of IJlICOlllaiJlcd piaslic flow and an inl'cstigalion

or the 'rigid-pla.'itic IllOdcl'
I ))cI'inition or the rigid-plastic materi;" .
2 Statemenl of the problem ror a rigid-plastic system
3 The rigid-plastic pattern and the determinalion or limit loading.
4 Govcrning equations.
5 Boundary conditions.
6 Theorem of uniqueness of the stress-lield
7 Remarks





Chapter III Appendix

General definition of the loading paral11eLer~ for a ~yslem
I Possible loadings .
2 Loading process depending on a /inile number o/" parameters
3 Case of a ~ystelll wilh friction conditions
4 Example

ChapLer I V Problcms of uncontaincd plastic flow in plane straill

I General
2 Expression of the yield criterion.
:1 A remark on the case of a non-standard material
4 Equations for the stresses
A The slress problem .
5 Transformation of the equations
() Characteristic lines
7 Relations along the characteristic lines.
X Computation of the solution.
<) Transformation o/" equations (24. 25)
10 Geometry or the characteristic network
II Matching of solutions
II Tile velocity probkll1
12 1:Iow rule .
J 3 Char;lcteristics-Relations for the velocities along the clwrac
14 Exall1pks .
I) Discontinuity of the velocity.
I() Discontilluity of Lhe stress-Jiclcl
C Study o/" an example.
17 Thc probkll1
I X Conslruction of the solution.
I() Calculation of the velocities
20 A particular case .
21 H.cmarks on thc solution.
22 The case of a Coulomb material.
D Particular uses in the study o/" a Coulomb material .
23 Method of superposition-Theorem of the corresponding states
24 An example study of a cohesionless soil with selr-weighl
Rererences .




() I






Chaplcr IV Appcndixl's


A Problems of un con lain cd plaslic now in plane strain for isotropic

non-homogcne()us material .


I Gcneral

2 The probkm ror 1he sl resses-Gcncral casc


3 The case or a Tresca material

4 The case of a Coulomb material.

5 The velocity problem.

6 The case or some non-standarclmaterials
7 Discontinuity of the stress-[Icld

B Problems of uncontained plastic now with axial symmetry for

materials wilh a yield criterion of the 'intrinsic curve' type
1 General
2 The stress problem
3 The velocity probkm .
4 Weak solutions

Chaptcr V The thcory or limit analysis (For applicatiolls to soil mcchanics)

I Presentation
2 Adlllissihk ridds . Dissipalion
3 Static approach
4 K i nema t ic approach .
5 Rem,lrks on the results of the theory 0[' limit analysis
6 M inimlll11 principles .
7 Limil analysis ill the sludy 01' plane strain pmbkms orUnCOlllain 'c d

plaslic flow
8 Interpretalion of tile solutions for the case ofa Coulomb malerial
9 Other applications of limit analysis in soil methanic.,> .
RcI'ercnces .










I I2

I 13






Chapter V Appendixes
A An. alternalive presenlation ol'the thcory orlimil analysis. L~xlensi()11

to some non-standard materials


I Introduction


2 The case of standard material


3 Case of non-standard systems

4 Friction conditions at the interface."

Referencc.') .



.. .


_ . _ _ _ 4 _ _ _ _ _ _ _

. . . . _ _ _ _ _ _ . . _ _ _ _ _ _ _ _ _ _ .. _ _ _

.. .


B Bonneau's lheorem .
I Thc problem .
2 Origin of thc prob1cm---The case or incomplete solutions
3 Lemma
4 Statemcnt of thc diffcrcnlial rclations ncces:.;arily veri Iicd a long (G)
5 Remark:.;
Refercnccs .





The purpose of this work is not the training or !ipecialists in Plasticity; il

aim!i at helping the reader 10 a beller under!ilanciing 01" Soil Mechanics !iO far
as il appeals to the thcolY of plaslicily, so as lo be able, if necessary, lo appreci
ale the meaning and validilY of available compulalional methods.
We h ave been lrying to ga ther from the lheory of Plasl icily whalever can prove
useful in Soil Mechanics, in the presenl slale of the arl; such matlers have been
exposed wilh some generality, whilst others, such as the phenomenon or plaslic
adaptalion and the dynamical problems, have not been evoked.
There are five main headings.
Thc firsl chapler is devoled lo the presentation of the cla!i!iieal model of
plaslic behaviour, and 10 Ihe possibililies ofthc applicalion ofthi!i modcl in Ihe
case orsoil!i; in parlicular,lhe prohlem 01" Ihe Ilow rille is !illidied in del ail, in Ihe
light ofrecenl research .
In the second chapler, Ihe problems of"clasto-plasliciIY arc dealt wilh bric!1y.
These are ralher difficull, and, wilh the exceplion of certain cases in which
simplificalions :Ippeared, Ihey have received sealll appliealion 10 Ihe maller!i
whieh inlere!it us. However, il i!i eS!ienlial 10 reali/.e Ihal Ihe elaslo-plaslie
problem is the basic problem as soon as we inlroduee plasticity (wilhoul
visco!iily) inlo Ihe malerial behaviour; and Ihal, as a conseqllenee, il is e!i!ienlial
lo see for which conqiliolls, under which hypothe!ie!i, and for which Iype!i of"
problem!i, we can rdllm 10 Ihe rigid-pla!itic scheme.
This queslion, in ils generality, con!ililllle!i Ihe object or Ihe third chapler.
The variou!i ways of defining and considering Ihe rigid-plaslic material arc
examined, and Ihe mdhod of staling limil equilibrium problem!i is sllldied.
The fourth ehapler is devoled to a delailed study ohhe theory of plane limil
equilibriulll, Ihe lISC!i of which arc vcry 11IIll1eroll!i in Iheordical Soil Mechanic!i.
Finally, the Iheory of" limil anaIY!ii!i i!i deall with in Chapler V. Fir!illy Ihe ca!ie
of a material with an as!iocialed flow rule i!i examined, and we clclermine
accurately the range of application of the re!iull!i obtained, particularly as
regard!i Ihe significance of Ihe rigid-pla!ilic scheme. ;\ clear underslanding of"
the calculation processes used for the rigid-pl(l:)tic material can be allained only
through the resulls of the theOlY of limit analysis. The case of a material with a
non-associated now rule is also considered, the deficiencies in this case being
shown, together with the positive points that can be used as foundations.
Chapters Ill, IV, and V :Ire ['ollowed by one or several appendixes in which
are given developments which would have made the text too heavy.
In each chapter and appendix numbers placed in square brackets in the text

- - ____~I'

- .. - - - - ...- - ..... -~ .. ~ --.......-..----.---..-. .. ~- ....... ,~Irr ........... t ....... r':"

I .


rekr to a list or rekrenees. Though not allempting to presenl an exhaustive

bibl iography, these lisL<; indicate interesting works the reader can rerer to.
J-'in:dly, it ShOldd be noted that, except for the examples that arc given roJ' a
didactic purpose, no solutions of more or less classical problems or Plasticity
linked to Soil Mechanics arc to be found in this work. Errectively, our purpose
is not to replace the well-known books of Sokolovski, I-lill, etc., but rather to
prepare the reader for their possible utilization.
I wanl, finally, to pay homage to Professor Jean Mandel, who was my master,
and whose ir'n print this book bears.
I would like to thank Professor Legrand who holds the Chair or Soil Mech
anics at the Ecole des PonL<; .at Chaussces, and who invited me to deliver the
course of lectures on whieh this book is based. I would also like to thank Pro
fessors Habib and Radenkovic rrom whom I have learnt a lot in both theory and
the practice of the subject in many friendly discussions.
Mrs. I-relcn!; Yirlogeux and Dr. R. W. Lewis undertook the task or translating
this book into English and I would like to thank them for their patience, as I
feci the readers will be well satisfied with the quality of their work.


NOTATIONS: following the ordinary practice in Continuum Mechanics, and

differing rrom that of Soil Mechanics, the stresses arc denoted
positively for tension.



General considerations


the jJlastic

behaviour 0.( 111.aterials


Thi5 chapter defines the plastic behaviour of materia15 and gives the cor
responding constitutive laws. The viscosity effects of the materials arc ignored
and hence the behaviour is independent of physical time. The subjeet5 dealt with
arc therefore elasticity and plasticity and not visco-elasticity or visco-plasticity.
Finally, a discussion on the application of this behaviour pattern is outlined
for t he case of soils.

1. Yield Criterion
Plasticity is characterized by the existence of a yield I}oint beyond which
permanent strains a ppcar.
Consider for example a simple tension test in which the stresses and strains
arc assumed to be homogeneous. J7igure l.Ia depicts the stress-strain rclationship.





( b)


Along the path OA the behaviour is clastic; i.e. the path is reversible. However,
once point A has been passed (i.e. a > au)' the path is no longer reversible.
If, for instance, the sample is unloaded after reaching point 1) and the stress (J
falls to zero, then the unloading path is given by Be, which is parallel lo OA
(provided the linear clastic properties of the material arc not altered by the
plastic deformation). After unloading there remains a slrain (;", repre5ented by





alld, termed the permanent strain. The stress 0'0 is defined as the original
yield point. If the test specimen is again loaded the path is reversible along ell,
becoming irreversible when a > 0'1/' In this case the stress 0'1/ is defined ;IS the
cllrrent yield point.
Figure represents the case in which 0'/1 is a fUllction of the permanent
and Ihlls ililistrates work-hard('"ill{/. In fo"igllre 1.Ih the slress 0'1/ is ;1
constant ami the material is said to be pel:(ecly p!os/.ic.
II mllsl be emphasized that the onset of pl;lslicity is not indicated by the
non-linearity of the stress--strain eurve beyond point A but by the irrelJersi/Jilily
of Ihe path beyond this point.
More generally, it has been proved that the concept of the yield point in the
lin iax ial c(lse may be rerlaced by a yield criterion for a small (macroscopic)
element of material subject to any action characterized by a tensor of applied
stresses 0". J is a scalar function of the state of stress of the material, such that
J(O") < 0 corresponds to the elastic range of the material and J(O") = 0 cor
responds to the appearance of the irreversible deformations.'
Usually, it is the equality



J(O") = 0

which is termed the yidd criteriol/. The fUllclion/is often called a yiddfilllC/iol1
and the su rface J(O") = 0 in the stress-space {O"} is the yield or loadil1Y sUllace 2 of
I he 111 a terial.
With a per/eclly plastic material, the yield function does not vary and the
yield surface is fixed, with plastic strains occurring only if 0" is on this surface
and .'itays on it.
With a work-hardening material, the yield function varies as the permanent
deformation continues and discrimination must be made between the original
and current yield surfaces. Additio~l(d plastic strains appear only if 0" is situated



Figure 1.2
~pcakillg. Ihi~ classical rcprcsclltalioll
slrc~s-Icnsor is nol corrcct (it docs not rcspcct

of Ihc yicld crilcrioll as a fUllclioll of Ihc Eulcrian

thc principlc or objcctivity). wilh Ihc cxccplioll of
thc parlicular COISC of an i~otropic Jl1i\tcrial (~ce [JJ]. p. 716). I-Iowcvcr.thi~ form will bc uscd sincc
only the case or isotropic materials will bc studied latcr.
1 Functioll
bcing physically dctermined only on symmetrical tcnsor~. sincc a is symmctrical
il = 1711 or aT = a). its mathcmatical cxprcssion. which depcnds on ninc a componcnts, is
somcwhat arbilrary. Thc mathematical expression cho~en is the onc symlllctrical with respect to
thc componcllts (TI) and (7)i; further on. il will be dcnoted by f.



on the yield surface and movc.<; outward. The yield 5urrace is now extended by
cr, a5 represented graphically in Figun; 1.2.
In order to take the work-hardening errect into account the yield criterion
will now be written as.

I(cr, E) = 0


where E stands for all the work-hardening parameters. These arc dc!ined as the
supplcmen'tary parameters which, ;lIong with the act \1al state of st ress, m;,ke it
possible to determine tile plastic 'behaviollr' of the materi;d. In n)()st instances
such parameters arc fUllctions of the stress-slrain history.
Various theories have been proposed in order to explicate a form for the
hardening parameters. Among others arc the hypothesis of a unique parameter
corresponding to the work expended in the plastic deformation (sec [21 J,
and Mandel's theory [32J).
2. Form of the Yield Criteria
2,1 Convexity
It will be seen ill Section 5 concerning the relation between slress and strain
that a hypothesis cOlllmonly accepted for metals and somc othcr materials,
the principle of maximum plastic work, implic.<; that the yield surface is convex
in the space {cr}. It follows that I can then be chosen as a convex function of cr.
This means, from a mathematical viewpoint, that if cr l and cr 2 verify
J(cr l ) ~ C)


f(cr 2 ) ~ 0

then for VA. E (0, I),

Then the yield surface is a convex surface in the stress-space {cr}. Tile allthor
bclievc.<; that the convexity of tile yield function' can be considcred as a gcncr;dly
valid feature for normal materials, including those for which the principle of
maximum rlastic work call be neither proved nor even admilled.
In fact, it is sufficient that each or the elementary ~nechanisms necessary for
plastic?,~ow in a small macroscopic clement corresponds Lo a convex condition
for cr. \(he clastic range of the clement is then determined by the interscction
or the plastic \anges or each mechanism and IS therefore a convex domain In
the space {cr}. )
2.2 Denling wilh I11Cllerial sYl11llletries
Il is possible ,to be more cxplicit concerning the form of the yield criteria for
the case of material symmetries. In fact, lhe yield criterion must deal wilh these

Or 'loading runclion.

(,h'"; , 1/( ..
~ t .',

.. \ )

/ .\.


, r ,I

.:, 11 I'



.\1 .1
' . t



Thlls, in lhe case of a maleri;" w'I\i<:h"was origin;dly isolropic,I in eqllalion (I)

depends only on the invariants of the tensor (J' [56J, or may be expressed in the
fOl"m of a symmelricaL Junction oj lhe principaL stresses 0'" 0' 2' 0'.1"
If isotrOI1Y is I1reserved throughout the work-hardening I period the function
J in equation (2) also depends only on the invariants of the tensor (J', but its
expression varies with the load. This assumrtion of isotropic work-hardening is
a theoretical conception that may be admitted provided the cldormations are
nol excessive. The Illost important yield criteria used for isotropic materials
will be stuclied later.

(6 '.6 Y;(6;6';)' {6,;-{;',y

2.3 Von Mi,c;es' criterion





' '77

,- ' '/)"


"' 1, r.r '1

~ UJ), l\\r~! 1



"'1 ./ \ vv ,


For cluctile materials experience has shown that if the tensor

to scparate its deviator s in the form



is decompose<.\

. LI

'rU I,

'~' H


~V \

(Jjj "

,\" \

(~ 1

~ ~i6 6,111" .) {' ;)
l,\-yU "
(r'ijj being compone\lts of Kronecker's unit tens(.>r, l~, / clepclld~ oilly o~~ s"p ,
The 1l1;lterial plaslic behaviour require." no moddicalloll 011 adding any IsbC'
lropic stale of slress.
For ;111 isotropic, ductile ll1aterial /.<.Iepends only on the invarianls of s:

0i : :

D1 0

"'J 7 -

\ (J '\'

" ') - )/

I }


.'I 0
= , .1 7, =



S;j -

I co co

,,/ '


' .


I .

The simplest criterion of this tYI)'e ll1ay be stated as .',

J,fl/' J), .. ( J 1\) l '!
j '7
.I 1. - k = 0 ,7 I'





IJ 1,f I

1 \I ,




V' :

where /{2 is a constant. This is known as the Von M ises criterion. It can easily
be checked that in the stress-space 0' 1'0' 2' O'J the yield surface is a cylinder
or revolution about the vector (I, I, I). The yield point in pure shear is given by k.
'fhc limit in pure tcnsion is kj'J.
~ ~ 6,'" 6) 1 -


2.4 TrescH's criterion

( .

Trcsc,,'s criterion also is valid for isotropic, ductile materials. It Il)ay

stated as a function of the principal stresses, ordered according to (J, ~ ~I

The criterion is


, 0' I -

0' I I I

where the intermediate principal stress

6wli1'f ' , ,b'Y\"~ll





plays no part.










I " ,'.



/ _

_'- I

. (5)




" ~~:

This is, in particular, lhc case ror 'Taylor's isolropic work-hardcning', which dcpcnds on one

scalar paramclcr.

Unless lhc conlrary is spccificd, lhc summalion convcnlion ror rcpcalcd subscripls is uscd.

With respect III lhe IIIHlI'dered principal slrcsses. Ii,.

assumes the symmetrical form
I(a) =


{Ii; -

lij --

Ii.~, n.I'

lhe yield rllnc!ion


; " I. 2. ;I
"'" I. 2 \



It will be veririedthilt in the space i i i ' a 2 , li J lhe yield surf:lce is;1 hex;lgonal
prism parallcl to the ilxis ( I, I, I). The yield point in pure shear is givcn by k.
The limit in pure tension i.-; 2/';.
According to the results of Section 2.2, the Trescil yield-funclion (equillioll ()))
can also he expressed hy means of lhe invariants of a (and, in this C:ISC, the
invi\l~iants ofs). In the presenl case this raises some problems since a closed I"orm
in J~ .I, eCJlliv:delll 10 (<1) does not exisl,l and presenls no pr:lctic:d inlerest.

~~ft~ Jr fA
u).\- 2.5 Intrinsic Clln'C (Mohr, Ca<Jllot)
Mohr. followed by Clquot. proposed
'\.."\'.;\)~I~\~ r
for Cln isOll'(lpic material:

II ",


generalii'.iltion of Trcsca's criterion



the principal strcs.-;es Ii I' Ii l' Ii.! being ordered according to Ii, ~ Ii J. ?:: Ii.\" (AS no
ambiguity can be expec!ed the distinclion bctwccli Ar;d,ic alld l{onl;11l indices 1
is no longer maintained.) Thc func!ion {j must be determincd eXI)eriment:dly
and is found 10 hilve the following properties:
{j ..... if (Ii I -I- Ii))': it bccome.-; zero ror :1 positivc vCllue of (Ii, + Ii .,) whcn
(a I + a.l)/2 is lhe isolropie tensile strength: and when (a, -I- a.l) '" -. rfJ, fJ tends
10 a limit so that Tresca"s criterion is round asymplotie:"ly (sec [:U:n.
The characteristic or this type of yield critcrion is th;lt thc intermediate
principal stress has no illOlicnce.
Equation (7) gives a relationship between the radills (Ii, - 1i.l)/2 ;Ind the
abscissa (iii -I- a;l)/2 of the cellire of a Mohr's circle ill accordance with the
6' /6" ?-i 0,;~ ' 6 , state of stressc.<; at lhe point under consideralion. This verifies lh:ll Mohr's
circles represenling atl the limit states have :In envclope, termed lhe intrinsic
() '.
eu rve .

Adopting Hitl's notation gives the expression

- {J

= (Ii I -I- Ii J )/2

For the stress representation given by Mohr. Figure 1.3 shows Mohr's
circles corresponding to various stales of Ii mil equilibrium and the illtrinsic
curve eIlveloping these circles.

Ir I =



is the equation of this curve, where Ii and rare thc normal anc! tangential
components of the stress 011 a plane.

..,. .


Thc c10scu polynomial cxprcssion in .1 2 ' J 3 givcn in [44] is not conccl.

Thc cxprcs$ion for I(a) as a funclion of Ihc unordcrcd principal s(rcssc.~ is analognus




I,'igllrc 1.:1

F"or a material having a yield criterion given by equation (7) the following
condition applies.
In order that the stress slate at a poinl M docs not violate the yield criterion,
i.e. it s;\tisfies I(a) ~ 0, it is necessary and sufficient that Mohr's circle does not
intersect the intrinsic curve: thus,

is valid for all the planes through point M.
If, and only if, the yield criterion is reached, there exisltwo planes on whieh
=--: 11((1').1 These planes are symmetrical to each other with respect to the
planes of the major and minor principal stresses.


2.6 Coulomb's criterion

Coulomb's yield criterion is very often used for soils and is of the 'intrinsic
curve' lype in which the intrinsic curve is constituted by two symmetrical
slraight lines inclined at an angle (/) towards the axis of normal slresses,
Wilh the usual notation, !-I = C cot </) and equalion (7) becomes

, .-'.
and (9) becomes

(I 1)

In the stress space (Ci p Ci 2 , Ci:\) the yield surface is a hexagonal pyramid which
admits the axis (I, I, I) as a ternary axis of symmetry and the bisector planes as
planes of symmetry (a consequence ofisolropy). The eross section has t he form of
;\ hexagon. Complementary dala in this respect will be found in classical works
(for example [JO]).
II will bc a~sllmcd Ihallbc rclalionlrl = /rIfT) i~ real; i.c.lhc family ofeirdcs has a rcal cnvclopc,
I;.qllation (X) reprcscnls two :Ires symmelrieal ahout OfT. The problem or the shapc of Ihe inlrinsic
cllrve h:lS bcen discussed (,flal'or 'angular' sunl111il); aec(lrding 10 Ihc aulhor'.~ opinion,there is nil
real difliculty. Onc mllst not forgel, moreovcr, that most oftcn, the physic.! phenomenon cor
respontling 10 thc vieinilY of thc slim mit or thc inlrinsic Cllrvc h;ls a diffcrcntnaturc: brillic fracture
occurs wil honl plastic yielding.
1 Thc somcwhat lin usual minlls sign is duc to the adoptcd sil~n cOllvcnlion. 1-1 ill's convention and
othcr common notal ions arc choscn 10 make further reatling casier.


2.7. Drllcl{cr--Pragcr criterion

Drucker ;In(\ Prager [IIJ proposed a criterion for soils that is related to
Mises criterioll just :IS Cou!omb"s criterioll is re!:lted tn that ofTresca. The yield
stll-face is ;1 rig'ht cir~ular cones with axis (I, 1, I). It is rcgular, a fact which
C;1I1 prove acivant;lgeous for some calculations.
The loading function for this criterioll is written as a fUllctioll of t he inv;lriants
of the strcss tensor ;\I1d of the deviator. If

it ;ISSUl11eS t he for m

= (/1 1

+ .I ;'/2


in which (/ and k. arc positive constants.

Other ;Iulhors also proposed a yield crilerion for which the yield surface
forms a regul;lr hexagonal pyramid with an axis (1, I, I).
2.8 Plasticity and fracture

As observed in Seetion 2.5 il is useful in this sludy of yield criteria lo lake

fracture illto cOllsideratioll_ It is Ilot illlellded 10 de:d :It great lenglll with Ihis
subject, which is now in full development.
Usually a fracture---i.e. lhe separation of a solid into sevCl"al pieces--can
appear in eilher of two ways.
(I). Without previous plastic deformation it is termed hriU/e,li'ac/llrc', which
occurs by fracture aloJlg a surf:lce normal to the direction of the greatesl
tensile stress.
(2). Afler previous plastic deformation it is termed dllcli/e I"/I[I/Ilre.
Considering the phenomenon in a simplified manner I.he el:lstic range of the
material can be conceived as the inleraction of two ranges which are probably
convex. aile of these corresponds 10 Ihe limil with regards to brittle fracture,
and the olher to the appearance of plaslic deformations.
The yield surface is the boulldary of Ihis range all.t! it appears to cOllsist ofa
summit zone (if it exists) corresponding to Ihe brittle fracture, a tr;lI1sition zone
in which bot h phcnomell;l C:111 be mixed, and I he pl:lstic yield boulldary.
Experience indicales Ihat when the spherical part of Ihe slress-tensor is
relativcly large the matcrial is no longer in the brittle-fracturc range, and ductile
rupture occurs. Moreover. experiments show that for m;tteri:lls composed of a
single ph:tse (this excludes porous materials--see [~9, 40J), when the stress slate
is isolropic, however high the pressure, t here is no plast ic yielding. The loading
surface is thus open (with the usuall1leanillg oftllC word) ill the directioll of the
isolropic pressures. 1

For porolls materials. the clastic range is .:Iosed in all directions. The IItili/.alion rIll' soils, which
arc granular matcrials. or Coulomb's yield critaion or or olher criteria or the 'open' intrinsic
curve type corresponds to a <Illite jll~tiriGd approximation ror the range or actual stresses.



. - - . - _ . _ _ 0 0 . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ . _

.... _

........ . . .

For ductile materials, there is no brittle fracture, no sUll1mit zone, and the
loading surface is open towards the two directions corresponding to the iso
tropic prcssurcs and tcnsions.
2.9 Anisotropic results
The previous paragraphs have presented the main yield criteria used for
isotropic materials. Real materials arc often anisotropic (metals, soils or rocks-
sec, for example, [IJ and [53J fort)Qth laller types of materials).
Vilrious authors have dealt with the problem of anisotropic plasticity in
cluding Hill [19-21 who obtained interesting results for a metallurgy problem,
Caquot and Kerisel [5J, who introduced the notion of a tensor or anisotropy
and the authors of [2-4, 22J, etc"


3. Plastic Deformation
As. stilled in Section I, pilssing thc yield point or crossing the yield surface
corresponds to the appearance of permanent deformations. Plastic deformation
is dcfined as follows. In the normal case, where the elasticity of the material
COllcel'lled is lincar and not modified by the plastic yielding, the plastic deforma
tion is the deformation after complete unloading of the material,
assuming the unloading to be clastic. After a given degree of work-hardening it
may hilppen that the real total unloading cannot be completely clastic, so that
lhe zero loadi;lg no longer belongs to the clastic range of the clement. In this
case, I he pl;lstic deformation is the permanent deformation obtained after a
virtual elltirely clastic total unloading. (Sec Figure 1.4 for the e<lse of a uniaxial
loading process.)

P ',


Figure 1.4

The exact definition of the plastic deformation, valid in all cases, refers to
strain-rates. If I) is defined as the strain-rate tensor then, at each moment of the
loading palh, the tOlal strain rate /lij (the result of the test) and the clastic strain

The loading functiolls proposed hy these authors arc exprcssed \lsillg Eulerian slrcsslcllsor;
as already stated, this is approximate.

rate ,,~'J. ;II"C known (;111 infinitesimal unloading is suITicicnl). Tile pl;Istie slrain
rale is Ihe difference between these two: i.e.



The plastic clcformalion is the integral of

-I- ,,!'.

( 12)

along the loading path. '

4. The Origin of the Plastic Deformation

The plastic deformation of solids is explained by Mandel [33J as follows .

In the solid, which is an assembly of crystalline grains, plastic deformation
can /eslIlt from either of Iwo phenomena.
(I). The relative mOlion or the gr;lins themselves. In Ihe case of soils. tile
irreversibility of the deform,\tion is expl<lined by the friction between the grains.
(2). The permanent dcformalions of the grains. This applies to metals. The
deformations arise by sliding in the crystal lattice of the grain. along the atomic
planes. The phcnomenon is explained by the theory of dislocations (sec for
eX;lI11plc I: 12]).

5. Flow Huk

The previous definitions makc it possible to determine thc onsct of the plastic
deformation, and its nature. In order to have a complete knowledge of the
constitutive law of plasticity, the questions of mechanism and magnitude must
also be considered. The answer constitutes the now rule.
Figures 1.1 and 1.4 clearly show thai the constitulive law canl10l lead (as is the
case in elasticity) to a unique relation between the actual stale or stresses and
the actual strain. (Thus, Ihe points 0 and C correspond to the same loading
state but 10 different deformations). The actual deformation depends on the
10;Iding path followed before reaching Ihe ;Ictllalloading st;lte. 2
I f the actual state of st ress and t he work-hardening state, i.e. the ;lctual values
of the work-hardening parameters representing the loading history, arc known.
then the increment of deformation may be determined from thc incrcment of
I Rigorously spcaking. (irccn's strain-tcnsor OUl!ht to havc bccn uscd: IllorCOv\:r, thc notion of
untoadcd statc Ill' neutral state is fundalllental hcrc: thc I:hllil:c of this rckrcncc conliguration,
invcstigatcd in (2\), 4J. 52J for the isotropil: I:asc, W;IS I:ardully looked at by M;lIldcl [J6, J7] in
morc gencral cascs.

lit doc.~ not dcpcnd on timc sincc viscosity phcnomcna havc bccn excludcd . Thc limc lakcn into
account in statil: or quasistatic plasticity is thcrcforc no morc than a kincmatil: paramctcr. the
sl:ale of whidl I:an hc modified as desired. Thcrcfore timc can play no role whatever in A,J.',A in
equalions (14) and (15). In :Ictual fact, a matcrial's irrcvcrsiblc deformations :Ilways contain 01
viscous parI, ;lIld physical lime musl be prcscnt in thc now rulc, ;111(1 cven in thc very nolion or
yicld point, which dcpcnds on thc loading spccd. (This is indicatcd in particular by thc theory of
dislol:ations for mctals [511].) Thc study of visco-plasticity, after a purely phenomenological OIp
proach that Icdto limitcd rcsults, now sccms 10 bc leading to a morc rund;lIllcntalthcory. (er. [15J
whcrc thcrc arc also refcrcnccs to formcr studic.~ of thc subjcl:t).
Obviously cnough, thc utilil'~ltion of thc modcl or plastic behaviour without visl:osity will he
justilicd for phcnomena in whidl visl:osity docs not prepondcr;ltc.


slres.<;. (I n the uniaxial case of Figures l.J a and lA, /: plays the role of the work
ll:lrclening parameter!). Such a stress-strain relation is termed an incremental
constitutive law.
III the general case of muItiaxialloading the tensorj<~[ formula is given by
(II: = fM(G, E, dG)

which is independent or
dirrerential form

a because or lhe absence of viscosily. It assumes the

df: =

(cr, E) dcr


d /:ij = fJd jj,lIk(G, I!.) dcr ille

Jnterms of the strain and stress rates, the relation becomes

v = .16'(0", E, 0)

'-A{t~ 'i k


r /\~:(

which Illllst be homogeneolls with respect to (kinematic) time, so that


= !!4 (0', E)a

In the decomposition of Section 3, (d{;C)jj follows, by delinition, the clastic

constitutive law
I r A is defined by



(d/;Jl)jj =

9. E)e

( 14)

terms or stram'- and stress-rates,

( 15)
Because orthe material behaviour being different during 'loading' and 'unloading'
characteristic of plasticity), Ajj.llil will assume two different
These expressions may be specified as rollows. According to the result of the
tests indicated in Section I, the plastic deformation ofa work-hardening material
for which the yield criterion is satisfied occurs only if loading continues (Figure
1.5).3 If 'the differential of J at constant work-hardening' is denoted by



d .{ = --'- da




, 1111


For a perrectly plastic materi:1I (Figure 1.1 b), when fT = fT II and dfT = 0, Ur. is undetermined . We I
can say that any material exhibits some work-hardening, lillie though it be, that makes this result I
physically insignificant. This work-hardening is, in some cases, small enough to be neglected in ,
problems likely to be simplifieu by :;0 doing.
1 Strictly spea king, objective definitions (derived from Truesdell, for instance sce [33]), must be ,
takcn ror UIT and a in formulae (14), (15).

It is known that in this easc the yield surraee, according to the very uclinition of work-hardening.

is extendcd with the load point; whcnce/(IT + da, E + dIn = O.

this dcrormatioll condition, simple in the case of loading with only one para
meter, may be expressed by d,/ > O. Then
(dr.") I..)

1= 0

is given by equation (14) where A 1= O.

If the load point remains on the current loading surface, i.e. if

d ,/ = 0

then the phlstic deformation is zero:

(dl: I ') I..) = 0

((0; [) =0



Figure 1.5

Finally, during IIII/oar/ill{/. i.e. if

d ,/ < 0
there is no plastic dcformiltion
(dl:I') I..) = O. I
Moreover, the passage is continuolls from one case to the other: i.e.



(dl: I ') I..)


Therefore, each linear form (c1r.I')ij in dCT llk or equation'( 14) is zero wh~1 the linear
form d 1/ in dfT j . kofequa tion (I G) is zero. These forms are therefore dependen t and


~J;Y;;,-E}t 1/


Thus, the increment\~of l(,lls()/~l<; present in dr. 11 only via the scalar ~/.
Because of this, the principal directions of the plastic deformation increment
as well as the ratios of its principal values do not depend on the incrcments of
strcsses. This fact was verified by Morrisson and Shepherd's experiments (1950)
on the tension and torsion of metal wires.
In the case of an isotropic material subject to isotropic work-hardenin@s an


A11'0, there i!i no modification at the yield !iurraee.

/If/11c /f~r;;(p/f1/ .'



- - - - - - - - - - - - - - - = - -- - - - -- - -- -

i.;otropic tcnsorial function of cr and has therefore the same princip:d directiolls
as cr. It follows that dt:'" and cr have identical principal directions. 1
To summarizc, for a work-hardening material, the general argument makes
it possible to srecify the stress-strain relation for the plastic deformation ill the
J(O'/ol<' E) = 0


(dr....).. = H ..(cr, E) -:"I.)




<I 0'1Ie

oj .


{ -.-- dO';;' 0



(11;") .)
. . = h in the other cases

No more can besaid about the flow rule without any comrlcl11cntary hyrothcsis.
For perfectly rlastic millerials, the stress-strain relation is simply



-.-:- dO' . . =

nO'. .


(d/;");} = h in the other cases

Moreover, if the material is isotropic, the tensors

prillcipal directions.


and cr have identical

6. Principle of Maximum Plastic Work

It is often assumed thaI malerials obey Hill's principle of maximum plastic

work (1950), which can be staled as follows.
II cr is (l plost.ically ac/missihlc s(rr~ss tCl1sor, i.e. slIch lhal J(cr) < 0, vI' (l slroill
r((lc tensor c{)rrcsjJ()/uiil1{J to this strcss-stal.c (lcc()rdill{J lo the plastic cOllstitutiIJC
law, and cr* is another plaslically (uilllissi/J/e stress tellsor (J(cr*) ~ 0), thclI

( 18)

which maya/so he written

(cr - cr*)v'"

Gcometrically, lhe inequality (I S) shows lhat if tcnsors v" and cr (both sym
mctlriCal), (llrCttW;~l.te~ bO~fve~torstol fa Si X-<li I1l(cn)siOllal sdP~fcc.(~CC [32])t' ~hle
sca ar pro( lIC fff1JI-.'{jIl IS:? I cr IS 011 le sur race J cr = oan I cr IS no 1 ou Sl( e
il. This implics thc convexity of the loading surfacc and v'" is an outward normal
10 lhis surface at point cr. z
Mllthematically, taking into account the symmetry of lensors vI' and cr, and
the fact thaL J is symmetrical in O'ij and (Jji' the following argument applies.
If the surface is rcgular at point cr, then





More pre\:is\:ly, II and cr, (II:" amI cr, Vi' and cr have at least one sy:;tclII or principal dircctions in
" Figure 1.(, wa.~ drawn for thc case in which thcre is isotropy, aiming al greater simplicity.


r: i g \II" C Ui


or, in order to include the case or" conical point,

A ~ 0. 1

"I' E }J)I(a),


is thercrore the plastic IJ(}tC'l/tial.

With a work-lwrdening material, the principle or maximum plastic work may
be assumed al each slage or Ihe work-hardening.
[n 1.l1e c;lse or melals, I he sliding along alomic planes is governed hy Schmid's
law: Ihe sliding occurs ir Ihe shear slress on Ihe plane in Ihe direclion or Ihe
sliding ;1l1'lins a critiea[ vallie. The principle or maximum work c<ln be proved as
a consequence or Schmid's law (sec [33J, p. 720).
It must be nOled thaI some aUlhors (e.g. [27]) prerer to :lccepl Drucker's
'q uasi-I hermodyna m ie' pos lula Ie [X] as a basis. A.s I he ma Ihema Iica I eq II iva lcnee
or Drucker's and Hill's rorillulalion can be proved, Ihe 'lhove resulls concerning
the yield criterion and the now rule still hold.
Yet olher authors ;lssume (/ priori both Ihe properties or COII/I('xily
or the criterion and or lIor/l/ality or Ihe now rule, ralher Ihan to usc HiJ/'s
principle or Drucker's posllllale, whose physical roulldalions are rar rrolll


of thc COlIsliluti"c Law

Assuming the validity or the principle or Ill<lXilllUIll plastic work (i.e. the
convex loading runclion to be also Ihe plastic pOlenlial) leads, in the case or
work-hardening, to
v!'. =

-~ _(?l~ (f}.~~ n-)





(frT M,


= 0 in the other cases

.f =


cJI .

-;._.- rT/Ok ~ ()
(JrT/O k

is lhc slIhdiffcrclllial offlillClioll / al poilll er. ,WI(er),..t ~ 0, is Ihc

/ al I his poin\.

I IJ/(er)

(21 )





. ..i' :' .,

where M is the work-hardening modulus which depends (/ priori on the loading

history. In the case of a non-hardening material


= A-oa j




---0= 0



..t ~ 0, arbitrary positive factor

v}j = 0 in the other cases

For such material, the now rule is said to be associaled, and the material to
be slal/dard (Radenkovic [45, 46J).
As :U1 example, for a Yon Mises standard material without work-hardening,
the flow rule is

= Aro, A ~ 0 if J 2 = k 2
v" = 0 in the other cascs.


.J 2


For a Tresca material, the now rule, referred to the principal axes, is

VI'I --



= 0' :V"I =

-..t' A ~



= A+

II, V~



= - A, A ~

O,.u ~ 0





0" J -






= 21c
= aI


= 21<


8. Validity of the Principle of Maximum Plastic Work for the Case of Soils
From experimental results, the principle of maximum plastic work appears
to be valid for materials whose yield criterion is independent of the average
pressure (- aj3). This is the case for ductile metals, and also for clays with
4) = 0 (undrained soil). For thcse plastic deformation occurs without volume
Nevertheless, investigations on such materials have been carried out using ;
Trcsca's criterion and the Ilow rule associateu with a Mises' potential (e.g. [2IJ).
For a soil following Coulomb's criterion (10), the principle of maximum
plastic work would lead to the following now rule: .
V~' =


I)~ =


+ sin


= -..t(l - 'sin



A change of volume would result from the plastic deformation, equal to

O~ vr + v~ ~ 2A sin 4> ~ [1 -

tan'(i-i]v\' ~ sin4>(vr - v~)


Such a significant volume change is not in agreement with experiments. Thus,

the principle of maximum plastic work i~ Ilol admissible for :-ioils obeying
Coulomb's criterion (I'i-' 0).
Some authors (see, ror eXlIlI1ple, [17J) have adopted for soils the hypothesis
tilat the plastic derormation occurs withoul volume change; i.e. v); = O. As the
material is isolropic, tensors \,1' and (J must have idenlical principal directions.
The now rule or Tresca's standard material (i.e. obeying the principle or maxi
mllm pl:lstic work) may be relevant in this case, i.e.


vi = 0


= -A.


According to [2.1]: this hypothesis is admis:-iible ror Coulomb-type soils irthcy

are at the critical void ratio ror the established plastic deformalion [24].

vr =

A.( I -I- sin 1')



= - A.( I

- sin v)


in which the angle )' is the so-called angle or dilation or the elemen!. (I' is positive
if there is dilation.)
In fact, angle II must vary as the plastic deformation or the clement proceeds,
and therefore it would be dirfieult to use rule (25). However, aecording to [23J.
satisractory results are obtained by considering an idealization of the malcrial
with v constant, 0 ~ )' ~ r/I, as proposed in [I X], [45J, and used, for example,
in [50].
Thus, it may be considered that soils have a Coulomb yield function (angle (/I)
- and a Coulomb plaslic polcntial (allgle I' ,I (/I) dirrcrcnt rroll1 the yield fUl1clion.

9. Final Remarks
It mllst be emphasiz.ed thaI the definition of plastic behaviour ror a material
consists of two paris:
(I). The yield criterion.
(2). The flow ru Ie.
The yield crilerion (or, more accurately, the yield function) intervenes in the
flow rule only for a work-hardening material. In this case it is used to 'measure'
the intensity or load supported by the clement of material. Only when the
hypothesis of the principle of maximum work is effected does the yield criterion
also define the now rule.


I '

These aspects have Ilot it/ways beell well distinguished, as indicated by Roscoe
[49], for 'M ohr-Coulomb's rupture criterion'. This point will be dealL wiLh when
sludying plane plastic Jlow.
10. Application of the Plasticity Model to Soil Behaviour
The prediction of soil behaviour via plasticity models does raise numerous
problems. However, the resllits obtained by this means have proved of interest
for many engineering problems. This point will not be discussed here, but
reference will be made to relevant literaLure.
As stated in Section 9, Lhe eondiLions for applieation of the yield-point and
yield-criterion concepts, concerning their significance for soils on the one hand
and the question of the now rule on the othe, must be examined. For this kind
of analysis reference will be made, for examfJlc, to [47] and [51].
With respect Lo the yield criterion the experiments performed with true
triaxial apparatus [14,28] indicate that the Coulomb criterion in the form of an
intrinsie curve models true behaviour fairly closely.
Regarding the now rule, there has been much discussion in the case of soils.
Various authors [7,13,25,26,31,34,38, 54J have attempted to determine a now
rule by consideration of the kinematics of the granular medium, the basic con
cept being that the plastic deformation takes place by sliding. Othcrs [15, 17,
18, 4lJ prcferred the approach based on the notion of a plastic potential different
from the yield criLerion, and have proposed now rules such as those indicated in
Section 8.
Finally, we must note that Mandel's works [36, 37J, which introduced a
director lrihedron for each clement of the plastic medium, might be useful for
granular media. In particular, light eould be shed on the problems connected
with the nOIl-coaxialily of tensors vI' and (J for isoLropic materials.


[I] J. Bi"rcz (1962) Contribution ;" I\~tudc c1c)i proprietc)i mecaniqucs dcs sols ct dcs
materiaux pulverulcnts, Thcsis Dr. Sc., Grcnoblc (1962).
[2] .I. P. Boehlcr and A. Sawczuk (1970) E'lllilihre limitc dC)i sols allisotropcs, J. M('c.,


[J:1 J.




R. Bookcr and E. H. Davis, (1972) A gencral trcalmcnt or plastic anisotropy ; .

undcr conditions or planc strain, J. Meell. l'hys. Solids, 20, n" 4, pp. 2:\9-250.
[4] J. Boschat and D. Radcnkovie, (1962) UIlC generalisation dc la loi limitc dc Trc~ca
aux rnaleriaux anisotropc)i, Z. AIl~/. M allr. MC'ch., 42, pp. 90-9\.
[51 A. Cl(lliot and J. Kcriscl (1956) 'l1'(lil(: tic M (:C(ll1i'l"e des so/.~, 3r<l. cd., GauthierVillars, Paris.
[6] A. Drcschcr (1972) Some rcmarks on planc now or granular mcdia, Arclrives (J
M cellall;c.';, 24, 5-6, pp. 837-848.
[7] A. Drcschcr and G. dc Jossclin dc Jong (1972) Photocla)itic vcrification or a mcch
anical model ror thc Ilow or a granular matcri.d, J. M eeh. l'hys. So/ids, 20, pp.

[R]-D:-e:-Dnlckcr(l95l)AT110rc rundamcntal approaclilo-plaslicstrcs -slrain rclations

1',. c._lsLU .S,-N aLCungr.. Appl.. ..M ec.!!:.d?:..i~7.


I()I [).

C. Drllcker (I())(l) On uniqlleness in Ihe theory or plasticity, (Jllliri. All/d. M(llil.,

I. pp. .15 "2.
[ 10] D. C. Druckcr (1961) On st rcss-sl ra in rcla Iions ror soi Is a nd load ca rryi ng COl paei ty,
I'roc. 1st 111(. Cm!( (}II (ill' M('c/I(//Iics I!f Soil Veilic:/( Sysl('IIIS, Torino, It;dy, 12 -- 1()
Junc 1961.
[II] D. C. Druckcr and W. Pragcr (1952) Soil Mcchanics and Plastic Analysis or
or Li III il Dcsign. QI/arl. A 11111. M (I( II., In, pp. 157 -165.
[12] J. Friedel (I <)()(,) /)isillca(;mls, Pergamon Press, Oxrord.
[13] G. J\. Gcnicv (1958) I'rohi(,lIIs <!f DYIIOI/I;CS (?f Grollllior M ctlio, (in R lIssian),
Gostckhizd, Moscow.
I: 14J M. (joldseheider and G . (iudclllls (1InJ) Some sectinll;dly proporliol1al rl:dilillear
cxlcnsion tesls on dry sand, I'rllt:. SYIIIJI. 1111 lire H(}le I!f' l'l(lslicily ill .\'lIil M cr:illlll;CS,
CIIllhridgc (G.Il.). lJ--15 Scpt. II)7J, PI'. 5() -- ()0.
[ I 5] G. G udch us (197 J) Elas tOI'l asl ischc Slorrglcich \lngcn nir t rockenen Sa nd, 11I{J(,lJiellr
Archill, 42, pp. 151-169.
[16] P. Habib (1953) Eludc dc I'oricnlalion dll plan dc rupture ct dc I'angic dc frolle
mcnl intcrnc dc ccrtaincs argiles, I'roc. 3rtllll/. CII/!f Soil Mecl,., Ziiricll, I, pp. 2~-31.
[17] J. I3rinch Hanscn (195J) I~arlll I'rc's.~lIr(' Ca/c:"lali(}/I, D;lnish Technical Press,
[I~] Bcnl Hanscn (1958) Linc ruplurcs regarded as n.u'I'ow rupture zones, I'I"IIC. COIlf,
Sll,. I('.~ 1'l'fIhihllcS d(' 1IIIIISSC:<'s tics lerr<,s, Bnlxellcs, I. PI'. 39--49.
[19] R. Hill (194X) A Ihcory or Ihe yielding and plaslie now of anisolropic l11elals.
1'1'11('. HIlY. Soc. A. 19.:1. pp. 2X 1--2XI).
f{iCi] JC Ilill .II ?~'2fTIlCi hel) r of )1:1 Ilc';I:lst ic st r.,in for a 1l1sol r(~pll; nl~(;I;r1S:-'-'I'Ii~
~:- __ r~:(Jc..A,_.J 9H, n1).::;<t_2X-1J 7
[21] R. Hill (1').')0) nl(' Malircl//(//ical -n'CIII',1' I!/ I'laslicily, Clarendon Press, Oxford .
[22] I-I. Igaki, M . Sugimoto and K. Saito (1970) Anisotropic yicld crilcrion IIndcr Ihe
maximum shcar slrcss thcory. J.S./I'I.I:.., D, n" 61, pro R2S-~36.
[23] R. G. James and P. L. Bransby (1971) A velocity field for some p:lssi'le carlh
prcssurc problems. (ic:llicc/illie/llt', 21, n" I, pp . 61 .. :0.
[24] A. W. Janikc and It T. Shicld (195<)) On Ihe pl;,slic now of COlllomb solids bcyond
original failurc. J. Apl'l. Mec/r., TrOllS. ASME, 26, n" 4, Pl'. 599-602.
[25] G. dc Jossclin dc Jong (1964) Lower bOI."HI collapsc thcorcm and lack of norma lily
of slrain-ratc to yickl ' surfacc for soils, J'roc. I.U:I :A.M . 'iY"lp. Oil IU,colou)' l~ Silil
Mec/lallics, Grcnoble, pp. 69--75.
[26J G. de Jossclin dc Jong (1971) The dOllhle sliding, frec-rotating model for granll!;.r
asscmblies, Chlleel/llie/IIC', 21, nn 2, pp. 15S-ICI.l
[27] W. T. K oilcr (I 'JoO) Gcncral t hcorcms for clast ic- plast ic so lids. I'rllyrcss ill Solitl
Mec//(//Iic.s, Ed. Sncddon and Hill, Vol. I, pp. 165-221., Norlh-Holland Pub. Co.,
Amslcrdam .

[28J P. V. Lade (197:1) Discussion, scssion II. 1'l'IIe. .\'.\;;111'. 111/ lile role 4 1'/lIslic:ily ill
Silil M Cdll/llics, Ca m bridgc ((j. B.), I :1-15 Sept. 197:1, pp. 12')--I)S.
[29] E. 1-1. Lcc (1969) Elastic-plaslic dcformalion allinitc strains,.I. Apl'l. M cd,., S('ric I~,
36, n" 1. Pr>o 1--0.
[30] J. Legr:II111 ( IW)')) C (lins dt' M c:collicille dcs Sills, E. N. P. C, Pa ris.
[JI] J. M:llldcl (1947) Sur Ic.-: ligncs dc glisscmcnl ct Ie calcul des dc.':placements dans la
dc.':forlllatioll pl:l~tiqlle. C./LAc:.,'ic:., I'I/ris, 1.22.'), pp. 1272--- 127:1.
[n] J . Mandel (1964) Contribulion thc.':orique .\ l'c.':tude de l'c.':cnlllissage cl des lois de
\'ccoulcmcnl plasliquc, Proc. 11(// 11Il. CCJII{/I'. Apl'l. Mec:lr., Miinieh, pp. S02-S09.
[D] J. Mandcl (1966:1) /\Ilc:c(/lIiqll<' des Milh'lIx ClI/llillllS, Vol. II, G:llilhicr-Villars, P:\Iis.
[J4J J. Mandel (1966b) Sur Ics c.':qualions d'ccoukmcnl dc.'i sols idc.':aux en ddorlllation
planc ct Ie conccpl du doublc glisscmcnl,.I. M('c:I,. Phys. Solids, 14, pp. :103--30X.
[35] J. M:lI1dcl (1969) COllI'S de Sciellc(' ties Mol.(:,.imlx, Ecole N:llionalc Supc.':ricul'l; dcs
Mincs dc Paris.
2(1, n"



[:lo] J.

. "-..",


M andcl (1971 a) Sur 101 decomposition d'une transformation clastoplastique,

C.lCAc:.Sc., P(/ris, 272,1\, pp. 270 -279.
[37J J. Mandel (197Ib) Plaslicilc! CI(lssiquc el ViscoplasLicilc!. C.I.S.M., Udine (Italy),
Springer- Verlag.
[.11Q G. Mandl and R, Fernandez-Luque (1970) Fully developed plastic shear now of
g.ranular materials. Geol('clllliqllc. 20, n" 3, pp. 277 307.
[:W] P. Morlier (1970a) Plastieitl: et ecrouissage d'lIn metal rrittc. M (:/11, Sc. H('". de
M (:lal/llroie, 67, No.6, pp, 401-412.
[tlDJ P. Morlier (197011) Comportelllent des roehc.o; sous eontrainte en ronetion de leur
(clleur en cau, IAI lIouil/e IJllIIlCh(~, S, pp. 471 -475.
[41] Z. Mroz (1963) Non associated Ilow-Iaws in Plasticity. J. Mec(/I/iqlle, 2, No. I,
pp. 21-42.
[42J V. N. Nikolaevskii (1971) Governing eq uations or plastic deformation of a granular
medium, J. Appl. Malh . &. Meclr. (trans. P.M . M .), 35, No.6, pp. 1017-1029.
[tI:lJ M. Piau (1970) Description du comportement d'lIne c1asse de miliellx continus . .. ,
J. Mcc., 9, No.3, pp. 375-401.
[44J W . Prager (1951) Theory of PerfccLly Plastic Solid.~, John Wiley, New York .
[45J D . Radenkovic (1961) Thcorcmes limitc.<; 'pour un materiau de Coulomb ;\ dilata
tion non standardisce, C.R.Ac.Sc., Paris, 252, pp. 4103-4104.
[46J D. Radenkovic (1962) Theorie dc.<; chargc.<; Iimites, S(llIIilluirc! de Plasticil<I, Ed.
J. Mandel, pp. 129-142.
[tl7] D. I{adenkovic (1972) Equilibre limite des milieux granulaires. Modeles de COIll
portement rigide-plastiqlle, Plaslicile cl Viscoplaslicile, Eel. D . Radenkovic and
J. Salenc;:on, Ediscienee, Paris,l97t1, pp. 379--394.
[4HJ K. I;r. Roseoe e al. (1967) Principal axc.o; observed during simple shear of a sand,
I'roc. (ie'o. Con!, Oslo, pp. 231 -23X.
[49J K. H . Roseoe (1970) 10th Rankine Lecture: the innuenceorslrains in soilmcchanks,
Geolechniquc, 20, No.2, pp. 129-170.
[50J J. Salenc;:on (1966) Expansion d'une cavitc dans un milieu clastoplastique, A 1111.
Pl.~. Ch., 1966,3, pp. 175-187.
[.~ 1:1 .I . Salcn<,;on (1974) Plaslicilc pour 101 Mec:lnique des Sols. C.I.S.M., Rankine Session,
.luly 1974, Udine (Haly).
[52] P. SidorolT (1970) Quelquc.<; rcnexions sur Ie principe d'indilTcrence matcriel1c,
pour un milieu ayant un ctat rel;'tchc, C.RAc.Sc., p(/ris, 271,1\, pp. 1026-1029.
[53J P. M. Sirieys (1966) Contribution I'etude dc.<; lois de comporlement des slruclures
rm;heuses. Thesis Dr. Sc., Grenoble.
['54 '1 A. J. M. Spencer (1904) 1\ theory or the kincmatics or idcal soils under plane strain
conditions, J. Meclr. Phys. Solids, 12, No.5, pp. 337-351.
[55] P. Stutz (1972) Comportcmenl e1astoplasliquc dc.o; milieux pulverulents, PL(/sticil(:
el Viscop/crslicile, Eel. D. Raden kovic and J. Sa lenc;:on, Ed iscience, Paris, 1974,
pp. 395-420.
[50] A. S. Wineman and 1\. C. Pipkin (I %4) Material symmctry restrictions on consti
lulive equalions, Arch. Ral. M"c/r. AII(l/., 16. pp. IR4-214 .
[57J It N. Yllun~ and E. McKycs (1971) Yield and failure ofa clay under lriaxial stressc.o;,
1. Soil Mcch. & Found. Div., Proc. ASCE, 97, SMl, pp. 159-176.
I.~xl .I. Zarka (1970) Sur 101 viscoplasticitc dc.o; Il1ctaux, M t;lf. Arl. Fr. (Sc. Tech. Arlll.),
fasc. 2, 1970, pp. 223-292.
[59J H. Zielger (1969) Zum plastischell potential in der ilodclllllcchanik, Z.A.M .1'.,
20. pp.059-675.



A short survey q( the jJroblel11s o.{


1. Introduction

The so-called problems of el:!"sto-plaslieily arc problems in which the con

slilulive equalion adopted for malerial behaviour is of the lype indicated in
Chapler I. The deformalion consisls of an elaslic and a plaslie parl. [0'01' example,
in the case of all isolropic linear elaslic material obeying M ises' yield criterion
with wo'rk-hardellillg ;1l1d the prillciple of maximum plaslie work, the eOIl
slilulive equation is:

I) .


E 1 (5 I)..

I/.IS ..

if.i 2 ~ 0






II =

11 .

= -E a..

II' =



[J(.l1,.I."/~).I2 ~ 0

I + II
= - -E0 - IJ.. - -eLI,

k 2 ()

(1 )

in the olher cases

which is lermed the Pralldt.I-Reliss law.

Problems oflhis nalure arise as soon as plasticity (wilhoul viseosily) becomes
. signilicanl in the deformalion of malerials. As will be ::ieen, lhey arc difliculL lo
solve, and few can be solved expliciLly. lL is IlOt inlended to examine them in
delail, but only to invesligate the form in which lhey appear and to make some
general observalions. (Reference may be made lo classical tcxt books, parlicu
1<lrly [6J, for a more delailed presenlation: theorems of uniqueness of the solu
tion, minimum principles, etc.)
2. Statement of the Problem. Method of Solution

The objective is lo delermine the actual state of stress and strain of an elaslo
plaslic syslem. The problem is to be solved incrementally following the loading
path. This differential approach is dictaled by the incremenlal form of lhe


cOllstilutivc cquatioll. In gCllcral,the slate ofstrcss alld slraill withill the syslem
dcpcllds on thc /o(lI/iIlY /m", followed to attain thc actual loading.
Thc incremental method of solution is as follows. I\t any time I the stress and
displ;lccmcnt fields arc known, and by using the equilibrium equations, the
constitutivc equation. ((I) for instancc), and thc boundary conditions imposcd
On the velocities and on the rates of the imposed forces, it is possible to deter
mine the stress-rate field O'jj <lndthe velocity field u j throughout the whole solid.
It is then possiblc to determine thc stress and dis(Jlacemcnt fields at time (L + dt).
It must be emphasized that for each time increment, the boundarics between
thc clastic and plastic zoncs vary and must be redetermincd. Point!'; situated
within the clastic zones (I(rr) < 0) during the prcvious time step now lic within
plastic zoncs, and (Joints situated within the (Jlastic zoncs during the previous
step (/(rr) = 0) now lie within clastic zones (an occurrence known as 'local
unloading'). This movement of state boundaries is the main source of praetical
Only a few problems, in which im(Jortant sim(Jlificatiolls arise due to geo
metrical symmetry, can be solved cX(Jlicitly by analytical means (see [12] and
[13J). Usually, numerical methods (based on minimum principles) are used.
The:\c in vol ve d iscretizat iOIl in space (for example, by means or the finite clement
lllclhod [{J, [15], [17J etc.), as wcll mi inlilllc. The approximalion in time may
Ie:ld 10 problems of convergence of the Ilumeric;tl schcme, tiS indicated in [9J.
Numerical mcthods havc been used (assuming slllall deformations) to solve
problems concerning Tresca or Mises materials with positive work-hardening
(or evcn negative), or Coulomb material using either the associated now rule,
or olle or Ihe rules indicated in Chapter I.

3. ncllllviour of a System Subject 10 a Loading Process

3.1. Loading depcnding on one paramcter
Let liS consider an elasto-plastic system, constituted by a solid or a set of
solids. This system is subjected to loads proportional to a parameter Q and
. undcrgocs the following loading process: starting from the natural state, with
zero stresses for Q = 0, Q is made to increase.
During the solution of this eiasto-plastic problem, assuming small deforma
I iOlls (i.e. CiJClII[jC'S oj yC'oJllelry being ncglected), the following stages arc met.
Firstly, as long as Q is Jess than a value Qo' there arc no plastic zones and we
11:1 vc I(rr) < 0 everywhcre.
For Q = Q() plastic zones appear and the yield stale is reached at a point (or
several points simultaneously). Qo is termed the initial elastic limit of the system. I'
I\s Q increases beyond Qo' the plastic zones spread. Plastic deformations are
possible within these zones, but they are contained by the deformations of the
clastic zones. This means that, at any point within the plastic zones, the (Jlasticj
dcformation is limited for each value of Q, whether the material is work-harden


- -I

ing or perfeelly plaslie. owing 10 Ihe eonlinllily o/" Ihe malerial Ihrollgh Ihe
boundary bel ween the el:lstie and plastic zones.
Finally, whcn Q rcachcs thc valuc QI' the plastic zoncs havc dcvelopcd
surlicicntly to makc Ill/col/tail/ed plC/stic .flo IV possible, i.e. plastic ddormatiolls
arc no longcr containcd by clastic zoncs.
From this point on, the bchaviour of thc systcm dcpends IIpon whether the
matcrial is work-hardcning or pcrfectly plastic. In thc latter casc, with geomctry
changcs bcing neglcctcd, thc deformation in the plastic zones bccomcs unlimited
for Q = QI' This unlimited deformation of a part of thc system, undcr constant
loading. is known as collapsc. Thc value Q. is tcrmcd the lill/it lood of t he system . .
In the case of a work-hardening matcrial the deforlll;ltion in the plastic zones is
stirl dctcrminatc. With gcomctry changes bcing ncglcctcd, thc loading must
incrcasc bcforc dcformation can continuc.
As an cxamplc, thc planc problem of a uniform prcssurc with variablc in
tcnsily Q, applied to the surface of a half-planc consisting of a pcrfcctly plastic
Tresca matcrial, I may be considcrcd. Qo is found to bc equal 10 n/.;., and the
plastic zone thcn consists of lhc scmi-circle with diamctcr A'A (Figurc ILia):
for lhc limit load Q. found 10 bc cqual 10 (n -I- 2)1.:, lhc plaslic zone may be
rcprescntcd schematically as in Figure 11.1" and thc malcrial can flow towards
lhe sllrface in bOlh direclions.



"'- .......

(a )

, ,





Figure 11.1

3.2 Loading depending IIpon se\'eral paramctcrs

For lhe case of a load which is depcndent on a finitc number of paramclers

Q} thc conccpls of Seclion 3.1 arc ;'11plicd ;as follows.

LeI liS dcnole by Q lhe poinl with coordillales Q .. Q2"" Q" in a II-dimensional
space (as no confusion may arisc Q will also denole the veclor with lhe samc


eoordinalcs in a !;-dimcnsional linear space).

The illilial clastic dO/llaill of the systcm is lhcn defincd as the domain for lhc
loading-poinl Q, in which lhe system remains Cl;lslie along ally loading-palh
starting from thc natural state wilh zcro-strcsscs for Q = O. This domain is
limited by the init.ial C!lasl.ic boundary. As a consequencc ofthc rcvcrsibility off.he

This malerial is homogeneous, wilh linear isolropie e1aslicilY.

For example subjeel 10 /I forces or "clive pressures lhe respeclive inlcnsilies of which arc
Q/(i = I .. . , II). Tile nol ion of load ing para melcrs will he e)( pia i ned ill del a i I in ella pIer" I. so
as 10 include more difficlIll e;\sc.." of boundary eondilions.


l~~______ _

.. --"',

___ _

clastic deformations, that boundary is proved to be the locus of the points Q()
defined on each of the above-men tioned loading-paths by the first appearance
of plasticity in the system. It is a convex surface for the hypothesis of a convex
yield function and of linear elasticity.
If the load is increased beyond this boundary, the clastic domain of the system
changes and an aClual elastic boundary of the system, limiting the actual clastic
domain is defined. The domain is, so to speak, hauled by the loading point Q.
This is the 'work-hardening of the system' due to the work-hardening of the
plast ic clements of the system and the incompaLi bility of the plastic deformations
(see [5J). Obviously the actual clastic boundary depends on the loading path
The load being increased along a given loading path a point Q 1 is reached
for which uneontained plastic now becomes possible. The locus of points QJ
corresponding to all the loading paths is termed the yield boundary 0/ the syslem.
/\s an example, P'igure II. 217 taken from reference [5J, shows the different
boundaries for the structure represented in Figure II.2a.





Figure 11.2

4. Importance of the Geometry Changes

The description of the behaviour of a system subject to a loading process,

given in Section 3.1, is realistic only if the geometry changes are actually negli
4.1 GeoJl1etry change..; during IIncontained plastic flow

r"irslly, it is obvious that if the stage of uncontained plastic flow has been
reached then t he deformations are no longer Ii mited to the order of c1astie
derorJmllions and it may now bc necessary to take geometry changes into
account. This has the following eonsequenees: for It perfectly plastic material,
either the load necessary to continue the deformation increases endlessly, or it
increases and reaches a maximum, thus giving rise to instability and collapse.

However, in both cases Q. docs have a practical meaning as it corrcsj)oncls

approximately to the appearance of inadmissible dcrormatiolls in lite structure.'
For a work-hardening materi~d, the increase of load required for continuing
derormal ion is regulated both by geometry changes and by the work-h:II'dening
of the elements. The eventual collapse oceurs through the phenomenon of
instability which must be investigated in each case. The value of Q 1 has no
applicatio.n in these systems.


4.2 Geometry changes before uncont:lineo plastic now

In actual fact, the problem of geometry changes arises cven before the stage
of unconlained plastic 110w. For many problems, if the elastopbstic solulion is
carried out neglecting the geometry changes as indicated in Section 3.1, un
contained now occurs only ror very large or even inrinite derormations. Hence,
the changes or geometry leading up to the stage ofulH.:ontained now at a loading
value of QJ cannot be neglected.
However, from the practical point of view this load is 01" interest ror those
problems where Q. is 'attained quickly' that means that the value of Q is already
very close to Q. (for example 0,<)5 QI) when the deformations arc still two to
rive times as large as those obtained at the initial clastic limit of the system and
therefore still negligible. 2 It seems likely that if the accurate solution were
carried out, i.e. the geometry changes were takell into account. the results would
become noticeably different from those for the case of fixcd geometry only at
values of Q close to Q I; and consequently, the loael for which important deforma
tion occurs would be elose to that given by QI' The problem of elasto-plastic
Ilexure of a beam with rectangular cross section subject to uniform bending
moment (sec [2J anel [6J) gives a simple example of this case.
It is not always true that Q. is attained quickly, as in the case or a spherical
envelope consisting of a perfectly plastic Tresea material, subject to an internal
pressure. Here it has becn found, by means oran clasto-plastic analysis neglecting
geometry changes, that the deformation corresponding to a fixed percentage of
Q" referred to that obtained at the initial clastic limit, increases considerably
as the envelope is made thicker. Thus, it is not possible to give Q I a signiricant
meaning with regards to a thick envelope.
The accurate elasto-plastic solution of lhis problem [6J gives Qo as the initial
clastic limit of the system. Then, in the elaslo-plastic phase, il is found lhat the
load reaches a maximum for Q equal to Qc called the critical load in [IOJ and
[14]. If the initial thickness of the sphere under investigation is decreascd 10
arrive (It I.he prolJlelll (?f the thill splwric(// ('II 11c/0pc, the geo met ry cha nges beeo me
negligibfe, and Qc tends to Q 1 as delined above.
Some problclll1i e;o;ist in which illlporlanl gC(lmclry cll:lnl~es can he admillc(l: I hcy prnducc :l
considcrablc 'gcomclrical work-hardcning' ;1I1d lhe collapsc load is vcry differenl from lhal givcn
byQ .
l On .hc assumption Ihal deformations ohlaincd at thc limil of Ihc slr\l(;lurc arc ncgligiblc
IhCI11.~cl vc.~.


- - ----l

Gellerally :>pc,lking, the aceur:lte c1aslo-plastic solution of each problem is

required (i.e. the geometry changes need to be taken into account) to decide
whether the limit load Q 1 as defined in Section 3.1 has a practical significance
for a perfectly plastic-ela:>tic :>ystem. In actual fact, this complex study could be
carried out only in exceptional circumstances. It can be seen, by analogy with
the problem of the spherical envelope under increasing internal or external
pressure, that either one or other of the following cases arc allained:
(I). The load increases and reaches a maximulll as the pl:lstie dcrormatioll

(2). The load increases continuoLlsly as the plastic deformation develops.
In the first case the collapse or the structure is defined either by the critical load
parameter Qc' which co~responds to plastic instability, or by a criterion of
maximum admissible plastic deformation. In the second case, a criterion of
maximum deformation is used. When the geometry changes are negligible, all
these criteria will lead to the value QI' Thus, the problem lies in estimating a
priori whether it is possible to forgo the accurate elasto-plastic investigation and
to deal with the collapse of the structure using the limit load QI'
The efficiency of the methods available for the determination of Qp dealt
with in Charter V, compared to the difficulties of the accurate or even simplified
elasto-plastic solution, is the main reason for the imporlance of the theory or
limit loads and or the usc or the rigid plastic scheme associated with it. It is
essential, however, to be aware of the limits of the application of this theory.

S. Examplcs
5.1 A simple structure
To illustrate the concepts or clastic limit and yield limit, an example is taken
rrom [5] and [7J, as represented in fo'igllre Jr.).
)9Ijr('( [v.~

~1'\ U)l

(8- T) b,..

7 ()" .




AC : :- 1,


(~.- (~o

: L. \J -L

(fl ({ . /lC
C~ -=~Q,




. yD
( nr \

cl=--L 'f.
:z L

~ ~D %"1


Figure J 1.3

. '

The system under investigation is a square frame with apices A, B, C, D as (

hinges. The bars arc clastic perrectly plastic and all have the samc cross section!
and properties, the plastic limit having the magnitude L ill both tension alld l

f - - - . : . . . . - - - - - - -



cOll1pression. The structure is subject to the action or two rorces -I- Q :lllti - Q ilt
A and C. Such loadillg is dependent upon only olte paralllc!cr. The s),stelll Itas
one degree of redundancy and is assumed to be initial1y unstressed.
Ir T denotes the tension in bar AC, we can obtain through statics the rorces in
all the bars:
AC: T, CD and CIJ: (Q - T)J2/2, nD: -(Q - T).

For elast"ic behaviour

T= QJi
Thus, the greatesl rorce is exerted in AC, and the clastic limit or the struclure
is:attained ',::hell this force ~lai,.1s the plastic limit ora bar; hence,

V\,'t5(j ~ t/Jy,Q



If Q exceeds Qo' the rorce in AC remains egual to L, i.e. T = L. The forces in the
ot he'I:t),li:s: irc- tl"1C"rCTorcknown :
CD and Cll: (Q - L) ;;. IJD: -(Q - L):


/~~F L

the derormations in lhese bars arc purely elastic ;Inc! delerminate.

In AC, the derormalion is the sum 01" the ehlsti~'lll<.L!)lasJi~ex..l~nsion. It IS
the arbitrary value or the plastic part which allows the compalibility or ddol'lna
lion. The plastic deformations are contained since the derormation or the whole
structure is;...QU1posed or the five bars 1..lL..-'JC, CD,
DA and n~~ble_cl_Lo forces (Q - L) (ind - (~L). This structure has a degree
'orrcctundancy one lower than the initial structure: it is isostatic.
The preceding calculations are valid only until the plaslie limit is rcached in
:Inother bar. @ is the.JleALbar to b~come_plaslic. when Q = 2L. Now there is
uncontained now: the lengthening or AC and the shortening or IJJ) are not
limited by the derormations or the remaining structll re as it has now become a
mechanism. If the changcs or geometry (variaLion-o angles) are still neglected,
the now will conlinue under a constantloatlQ, = 2L which isthe lill/it load.
It is important to note that:
(I). The absence or work-hardening is an essential condition: the rorce in a
bar in limit equilibrium rell1;tins const;lIll in spite or ;Iny incremcnt of ddorma
(2). The geometry changes are neglected up to the appcarance or the uncon
tlined plastic now. This is possible as the elastic deformalions or A J~ CD and
nD remain small ir the Young's modulus of the malcrial is sufficiently large
(with respect to L).
5.2 Cavity in an infinitc clastic plastic mcdiulIl
Two problems or elasto-plasticity lhat arc or intcrest in soil mechanics have
becn investigated and solved analytically :IS a result or the symmetry involved.



- - - - - - - - - - - - ---J

(I). The expansion of a spherical or cylindrical cavity subject to iln internal

pressure, jll all elasto-piastic infinite medium. The case of a standard Tresca
material with or without work-hardening was dealt with in [6J; that of a n011
hardening Coulomb material was investigated in [12J for an associated or non
associated 110w rule.
(2). The contraction or a cavity under the same conditions [13]. It is essential
10 a\low ror geometry changes in these solutions (since a very thick spherical
or cylindrical envelope is dealt with).
A finite limit to the pressure is approached asymptotically for the case of
ex pansion, whereas the inner radius of the cavily and the radius of the boundary
between the plastic and clastic zones (a ring arollnd the cavity) tend to inlinity.
For the case or contraction, the complete crushing of the cavity requires a
negative pressure (tension); inlinite for the case of a Tresca material, and equal
to - fl = - C cot (/) for the case of a Coulomb material. (Sec, ror example, [3J
and [14J for further details.)
The solution of this hypothetical problem has numerous applications in
soil mechanics, e.g. the equilibrium of very deep subways and subterranean
cavities, [IJ and [16].


[ I ] F. l3aguclin cl al. (1972) Expansion of cylindrical probes in cohesive soils, JI. Soil
Mull. & Foulld. Div .. A.S.C.E. 98, No. SMI, Nov. 1972 pp. 1129-1142.
[2) J. Courbon (1%5) ComplCmellls de Rbislallce des Maleriaux. PI(Jslicile. E.N.P.C.,
[J] P. l-Iabib (1973) Precis de Geoleclll/ique. Dunod, Paris.
[4] S. K.obayashi and C. H. Lee (1970) Elastoplastic analysis of plane strain and axy

symn1etric Oat punch indentation, llll. J. Mecil. Sc., 12, pp. 349-370.

[5] .I. Mandel (\964) Contribution thcorique a I'ctude del'ccrouss:lge el des lois de

f'c!coulemelll plastique. Proc. I II" COllgr. 1111. Mech. Appl., Munich, pp. 502-509.

[G) .I . Mandel (1966) Cours de M ecallique dr.s Milieux Contil/us. Ga uthier- Villars, Paris .

[7] .I. Mandel (1969) Cours de Science des Malcriaux, Ecole Nationale Superieure des

Mines de Paris.

[XJ P. V. Marcal and 1. P. King (1967) Elastic-plastic analysis or two-dimension,,1

stress systems by the lillite elenlent method, 1111. JI. Mech. Sc.. 9, pp. 143-155.

. [9J Q. S. Nguyen and J. Zarka (1972) Quelques .methodes de Resolution Numeriqueen

Plasticite Classiqlle et Viscoplaslieite, Plnslicile el Vi.l'copl(l.l'/icil(:, Ed. D. Raden

kovic and J. Salenc;on, Ediseience, Paris, 1974, PP. 327-357.

[10] \) . Radcnkovie and J. Salenc;on (1971) Equilibr.e limitc el ruplure ell M ccanique dcs

Sol.l', I.e Comportenent dcs Sols avantla Rupture, Journces Franc;aises du C.r-.M.S.,

n" Sree. Bulletin de Liaison des Laboratoires des Ponts et Chaussees, pp. 296-302.

[ I I] .I. Salcnyon (1969) La theorie des charges limites dans la resolution des problcmcs

de plastieite en d{;rormation plale, Thesis Dr. Se., Paris.

[ 12] J. Salcm,:on (1966) Expansion d'unc eavitc . . . dans un milieu clastoplastique, Ann.

PIs. CII .. 1966,3, pp. 175-187.

[ lJ) J. Salenc;on (1969) Contraction d'une eavitc ... dans un milieu clasloplastique,

AIlIl . PIs. CiI .. 1969,4, pp. 231-236.

[14] J. Salcn<,:on (1974) Plaslicitc pour la Mceaniquc des Sols. C.l.S.M., Rankine

Session, July 1974, Ucline, Italy.



P. M. Slrellls doerrc r (1973) Les melhod es incrcm enlalcs en

Applic alion ,\ I'clllllc dcs ca viles de Slocka gc de gai'. Cll c()\lchc de scI:
U. F. M., No. 47.
pp.29- 40.
[161 A. S. Vesic (\972) Expan sion or cavilic s in an inlinilc soil
mass, JI. Soil Mcch. &
Foulld. Div., A.S.C. E., 98, No. SM3, March 1972, pp. 265-29 0.
[\71 O. C. Zicnki ewicz, C: I-Iump heson, and R. W. Lewis (1975)
Associ aled and non
associ' lled viscop laslicil y and plasticilY in soil mecha nics, G(I()/cc
/lIliqw!, 25, No.4,
pp. 671-68 9.


J" . '


The jJroblelns 0.[ uncontained plastic .flo'vv

and an investigation 0./' the 'rigid-plastic

1. Definition of a Rigid-plastic Material

A rigid-plaslic material is defined as a material in which the only deformation
occurring is plastic. The constitutive equation is, therefore,
( I)
It is plain that sllch a formulation of the behaviour may be related to real

problems dealing with c1asto-plastic materials only in the limiting case when
phenomena arc studied for which either the clastic part of the deformation is
negligible compared to the plastic part, or the clastic properties have no influence

2. Statement of the Problem for a Rigid-plastic System

' . ....

On considering a rigid-plastic material system, the behaviour of which is

followed dllring the loading process, the following remarks may be made



(I). The rigid-plastic pattern will, as a rule, lead to indeterminacies of variolls

kinds; for example, stress fields cannopbe wholly determined, in consequence of
the nOIH.1eformability of the non-plastic regions.
(2). The rigid-plastic pattern allows solution only of problems of uncontained
plaslic deformation, as non-zero contained plaslic deformalion cannol occllr
in a rigid-plastic system. This will have an important influence on the way the
problem must be set (sec Section 5). Thus, when dealing with a rigid-plastic
system it is normal to study first the problems of incipient plastic now and then,
following the changes in geometry step by step, the continuation of uncontained
plastic now.

The unknowns in the problem arc no longer the stress-rates and velocities,
as they were in Chapter II for elasto-plastic problems, but rather the stresses

:llld velocities. It is shown in [5:!that such a problem is well-ddined. The dif

ferem:e of viewpoint between the rigid-plastic and elasto-plasti(: problems may
be explained as follows. for the rigid-plastic system the step by step solution is
carried out from the onset of llncontained now. Thus, the velocities (since the
actual displacements :lI'e known to be zero) and the stresses must be determined
at this moment. The same procedure is repeated at the end of a lime-step, in
which the displacements may then be determined and arc henceforth known.
As already mentioned, the geometry changes Inust be taken into account
during the solulion, which is generally dirricult. Thence it follows that allention
is paid essentially to the problems of:

(I). Incipient uncontained now, where the geometry is the initial geometry,
(2). Permanent lIncontained now, where the geometry remains ullchanged,
i.e. a perpetually incipient 110w,
(3). Self-similar lIncontained now, where the form of the geometry remall1S
similar to the original.
In actual fact, these three types of problems arc of the same nature and cor
respond. in one way or another. to the study of incipient uncontaincd now. The
system is then said to be in a lill/it eqllilihri/lll/ state.

3. The l{igid-plastic Pattern and the Determination or Limit Loading

3.1 An example
On studying again the example of the structure dealt wilh in the previolls
charter. it may be seen that the following. data parlieipate in the determination
of the clastic limit (Qo ='LJ2):

(/). Thc initial slresscs, which in this example :Ire zero.

(2). The elastic propcrties o/I/Ie II/(/(erial (though their erreet docs not appear
explicitly in the result, as the assumpt iOIl that these propert ies arc the same for
all bars leads to simplifications).
In the determination of the limit loael of the strlleture, neither the initial
stresses nor the clastic properties appear. Effectively, for T = L in AC and
- L in ED the value Q = 21.. is obtained simply by stellics. Member CIJ has a
stress equal to LJ2/2 and is lherefore safe and the structure, which was iso
static for Q < QI' becomes a lIIec/wllislll for Q = Q\.
Thus, it might seem that, according to this example, the limit load of the
system is independent of the clastic properties, in this case the Young's modulus,
E; of the constitutive material; and that, in particular, E can be suppo!\ed
infinite; i.e. it is possible to work on the rigid-plastic system. II mllst not be
forgotten, however, thnt the study was carried out under the assumption that the
geometry changes could be negleeled.


3.2 General case; rec()urse to the rigid-plastic material for the dclerll1inatioll
of the limit loadings
/\ system (2:::1) is constituted by an clastic perfectly plastic material (M I)
of which the elastic moduli al each point M are given by

A second system (1: 2) has the same geometry but is constituted by an elastic
perfectly plastic material (M 2) of whieh the clastic moduli arc

)...2. , .(M)
' ) , ,,,



= p)...I. , .. (M)

(p> 0)

I). '"

The yield criterion and now-rule are the same for the two materials.
For the system (1: 1) a history of loading (I-I I) is given by the values of the
body forces Fl and the boundary data (T'!)I on STI and (ut) on SUi at moments
in time, t.
The loading history (I-I 2 ) is given for system (1: 2 ) at the same moments, t, by
the data

(Tf)2 = (Tf) 1


S J"j

Then, if the geometry changes may be neglected for (1: 1) following (H I)'
and for (1: 2), following (fJ 2)' it is obvious that, at each moment t, the stress fields
arc identical in (1: 1) and (L 2 )
and the velocity fields u l and u2 are related by



= ul/p

In particular, in the plas(ic zones the values of parameter ).. ~ 0, left arbitrary
in the !low rule, are)..1 and )..2:)..2 = )..I/p .
The plastic zones arc the same in (1: I) and (1: 2) at each moment. Doth systems
will, therefore, reach the uncontained flow state at the same 1110111ent, i.e. for the
same limit loading. If, in particular, p tends to infinity, it can be seen that system
(1:,,,), following history (11 fY.)' and constituted by the rigid-plastic material (M ..,J
defined by fI approaching the limit. In system (L:J the imposed vclocitic.<; arc
everywhere zero except on the point of u ncontained flow.
Thus, the following result can be stated: for the assumption of negligible
geometry changes, the limit loadings of an clastic-perfectly plastic system can be
determined by progrc..'>sing to the limit in the corresponding rigid-perfectly
plastic limit liystem.
Similarly to what was stated in Sections 3.1 and 3.2 the limit load (and limit
loadil1g in the case of several parameters) may be defined as the load correspond
ing to the appearance of uneontained plastic flow in the corresponding rigid
perfectly plastic limit system.



This rigid-perfectly plastic limit system, for which, moreover, no indctcr

. minacy appears in its resolution, and the rigid-perl'ectly plastic system (Ln)
derined by
arc not identical. However, the solution (stress rield, velocity field)
corresponding to the incipient uncontained plastic now for (L,J is also valid for
(LO)' In other words,a limit loading ror (L,) and (L",) is also a limit 10adingfor(Ln).
The thcorem of the ulliquC"IeSS (?ltl/( limit IO(l(lill[lS, later stated in Chapter y,
will show.the converse of the above property under the hypothesis of the I'ril/
cil'lc (?[ I//(ixil/IIIIII plostic work Thlls. from the viewpoint of the limit loadings,
it will be proven that it is possible [0 work directly upon the (Ln) system inde
pendently of any progression to the limit. I


4. Governing equations

As stated in Section 3, the problem of a rigid-perfectly plastic system of the

type (LO) is reduced to the study of the incipient uncontained plastic now and to
the determination of the corresponding limit loadings. With regard to the
governing equations, two types of regions may be considered:
(1). The rcyiOlls ill lilllif equilibrilllll, in which the yield criterion is satisfied at
each point;

axJ] +


(2) yield criterion

pX. = 0

(3) equilibrium equations

= Vli
v. . = ).(iJ[/iJa., A.

(4) now rule

(for instan(.;e I)
;;: 0 in the case ofa standard material).
(2). The rigid rcgioll in which there is no deformation occurring;
I~j =


[(a I..) ) :s; 0

aa.. +




pX . = 0


It can be seen that in the plastic zones exactly sl,fficient eqll<ILions to deter
mine the unknowns aij' Ll j (stresses and velocities) arc available. However,
in the rigid ;!.ones three furl her equations arc necessary to determine the stress
fieltV whilst the velocity ficJd corresponds to motion without deformation
(rigid-body motion).
The plastic zones (I) arc the regions of the system whidl C;ln be deformed and
the rigid zones (IJ) arc the llmlcformecl regions. The di~tinction bel ween these
two types of regions may not appear logical since it is possible to belong simul
taneously to both types. As a mailer or fact it come..'> from certain peclliiarilie..<;

II will be noticcd thaI in thc c)t a III pic of Section 3.1 uniqueness was implicitly taken into ilccounl.
This is not surprising since in these regions the strcsses do nol intcrvene in the constitlltive equation.



.: '


of the prohlel11~ of lIncontained now for a rigid-pla~tic matcrial, which will be

() III I i 11 cd 11 () w.
The problem of 1I ncon tained now for a rigid-plastic ma terial is a frce bounuary
problem since the boundary between the plastic zoncs (a) and the non-plastic
zones (non-(/) is not k nowll bcrorehand, as wel1 as the boundary bet ween the
deformed zones (non-h) and the lIndeformed zones (h). The particular feature of
the problcm is that there is no uniqueness of these boundaries: for the same
problem different solutions of uncontained plastic now may happen to be
evidenced where the boundaries between the plastic and non-plastic zone..<; arc
different as well as the boundaries between the deformed and uncleformed
regions . I The only result of uniqueness to be available from that point of view,
will be given in Section 6.2 and state..'> lhat once a point of the system lies in :1
(non-b) zone in a solution for the problem it will belong to (a) type zones in any
It follows that the method of solutions of problems of uncontained now for a
rigid-plastic material is rather intricate. As will be seen in detail in Chapter IV
(for the particular case of plane-strain problems) the procedure is as follows.
'Solutions' are found, after assuming a priori the deformable zone..<; of the
systcm, (a). the remainder of the system being lIndeformed, (b). Then, using
appropriate theorems (given in Chapter V), the interpretation of these 'solutions'
is sought and the actual solutions of the problem can be charactcrizcd amongst
t hCI11.

5. Boundary Conditions
5.1. Classical prcscnt:ltioll


I\s stated previously, with the rigid-plastic scheme it is possible to deal only
with uncontained now problems. The boundary conditions must take this
fact inlo account; i.e. they must be consistent with uncontaincc\ now. This leads
to the introduction of consistent boundary conditions, as introduced by
Mandel [0,7]' Howcver, as defincd in this text, the problem always reduce..<; to
onc of detcrmining thc limit load, ami may be stated as follows. ~(tI/(' :'iys(elll is
s/l/~ic'c('d (,0 (II/ (lction (?t' (l UiIJe!1l type, tllell (.lie IJa/ul: (!( (,his ac:t.ioll for which UlI
, collwilled.f7ow (okes pLaCe!

is the soLution SOIJ[Jhl.

Geller,dly speaking, the boundary conditions consist of data rclating Lo the

strcsses (surface tractions) and velocities at each point on the boundary of the
solid. Three mutually-orthogonal componcnts are prescribcd for each of two
vectors--T ror surrace traction, and 1I for velocity-as is classical.
Howcver, duc to the particular rorm of the problem, the..<;c data must rulfil
ccrtain requirements. It is clear, ror example, th~lt the three Coi11poncnts of thc
surfacc tractions could not be prescribed everywhere. The dynamic conditions

Onecantakeas an exalllple Ihe probtem of the indentation ofa rigid-pla:;ticTrc.~C:1 hatfplane by a

smooth rigid punch, with the lwo well-known solutions by Shield [10] and Ui:;hop [t]. lind the
veloeilY fields given Prandtl and Hill.



(i.e. those conccl"Iling the forces applied to the system) Illust be of a somewhat
variable character. Thlls, the bOlllldary cOllditilllls will llecess;II'ily contain
some data concerning velocity, which in t he expression for the work done by the
external forces provide the variable dynamic conditions. With the problem being
delined with bOllndary data for the velocity, the solution will necessarily be one
of uncontained plastic flow, and the limit loading will he obtained as the cor
responding value of the loading.
An cxainple is shown in Figure 111.1, involving the indentation of a rigid
perfectly plastic half-plane by a rigid pUllch moving vertically.


- - - - - - - -______~~~~~WUL-__ "------------~

Figure 111.1

The boundary conditions arc as rnllmvs: At (\),

n,. ,.. / /Y = ( )

and at t he surface (.I' = 0)

Ixl >
Ixl <









The variable dynamic conditions are, thererore, the distribution or normal

stresses along AA'.

5.2 Discllssion of the classical prewnlalion

The presentation or Ihe bOllndary conditions whereby. al each point, three
mutually orthogonal components ror the set or two vectors T and II :Ire pre
scribed is well-known in Continuum Mechanic.o; (elasticity, visco-elasticity,
elasto-plasticity, etc.) It has also been adopled by numerous authors for prob
lems concerning a rigid-plastic material. However, it is not always well suited
to the case, since in practical problems the dynamic conditions present lheir
variable character in such a way that it is hardly possible to distingllish bel ween
the strc.o;s and velocity data.
Another example of a plane problem, similar to that of Section 5.1., is the
case of a uniform pressure of variable intensity w applied to the surface of a
hal f-plc:1 ne (on the segment A'A in Figure 111.2). The limit value of (I) is required.
The condition along A'A in the direction Oy cannot be considered as stress or
velocity data since the value of OJ is not known.




\ r:

~lLL lJ-1.-JA_._;'


~ ~ ~ 11 y'V'
\~ \;
cr/H~' Iii ~.<z. Vl)/~ H010 r/f~t~~[~1.2, b a (t:r (/Vr7
Similarly, l~le problem of the indentation of a half-plane by a smooth
!) ( "V





pUlIch under the action of an axial vertic:d force F is another example (Figure 3).
The condition of A'A in the direction Oy is neither stress nor velocity data.
I t is known that the distribution of u). must be that of a rigid motion. and thc
(1 corresponds to an axial re~;u1tant force.




Figure 1/ 1.3

il f'

'"II "

It is obvious that without invoking artificial transformations, which would

:Irrect the clarity, it is not possible to insert these boundary conditions within
the sta nclard rramework of Section 5.1.
Filially, it should be noted that all the boundary conditions in these practical
problems ;lree;lsily expressed by means ofa finite number orIoading parameters. I
1':01' example, in the case of Figure 111.2, the boundary conditions depcnd only
on the parameter Q I = w; in the case of f7igllre 3, F is a loading parameter.
Given laler as an appendix is a complete presentatioll of the concept of
loading parameters or generalized forces. 2
A survey of this presentation is proposed now.
The boundary conditions arc composed of dynamic and kincmatic compo
. nents that allow the reqllircd variable character to the dynamic data.
For any statically admissible stress field, (1, satisfying the dynamic boulldary
conditions, and any kinematically admissible strain-rate field, IJ, satisfying the
kinematic boundary conditions,3 the theorem of virtual work yields the equation
Because of this Ihere is no insistence on the minimum principles for the strcssc.~ and veloeitic.~ in
which the hOllndary conditions must be written under the form of Section 5.1. Moreover, even
when this is possible these 'prilleiplc.~', in spite of their apparent generality lire of practical lise
only in the case of a !inite number of loading parameters when they arc reduced to the theorems
of Ihe theory of limit analysis, as remarked by Hill [3J.
1 Note that in [2] Ihe Q denote generalized strc.~sc.~, Le. the loading. par:lllleters of one clement of the
sy:;tem .
.1 When this is necessary, parlieularly in Chapter Y for the exposition of the theory of limit loads, a
distinclion is llladc hetween the local valuc-<i, JabcUcd {1, Y. U, and the corresponding fields labelled





9J-v"' -


() ,

Y' ~ , "
\1"/)"IN t

.,1 , ' t "1 .


\ '6, ("If' \I
,6(, (() I,

~, -O~(,

J. ,


r/o/), 1"1""11 -'f





(<T' y)d V



T u liS 1





The classic al rormu iation assum es that the reduct ion





dS -I-






cnn be errected (T;l and 11;1 denot e the given value.<; or the comp
oncnt s or the
stress or veloci ty on S), which is not alway s trlle. Thc formu lal
ion in tcrms of
, loadin g param eters aSSllllles thal

with the follow ing correspondence.<; hein(J lincn/, .

a -. Q(a)


u -. (HII) E H"
This is alway s possib le in the practi cal cases where the comp
onent s Q.(a)
of Q(a) are 10t1ding param eters of the system . The vector q(v) Illay
be termed the
strain rate of the system . 1
The proble m therer ore rcsides in determ ining the limit value.<;
or veclor Q,
togeth er with the corres pondi ng stress and veloci ty fields.

6. Thcor cm of UniCjllCIH'.<;S of thc Stress -field

A theore m of \Iniqu eness for the soluti on of the proble m of uncon

lained now
will now be given which is valid on the assum ption ofa sl.tIIlt[ nrd
mater ial and a
conve x yield functi on (i .e. obeyin g thc princi ple of lllaXim~lIn plastic
work) .
6 . 1 Funda mcnta l incqua lity

The proofs are found ed upon the inequ. llity of the princi ple
of maxim um
work (Chap ter I, Seclio n 6):

from which
(0- - 0-*)'

If the velocity rield corrc.~ponds to a rigid body Illotion . (HI') ==


This inequality imrlic,<; the convexity or the yield fUllction which i:\ also Ihe
pla.<;lic rotential.
The inequality (10) is not rigidly binding and It IS Important to investigate
under what circumstances the rollowing eqllality holds trlle.
(0' - 0'*).

= 0

( I I)

For equality, when v ;;p 0, both (J' and 0'* must belong within or on the yield
sllJ'face. Since 0' is sitllaled on the surrace in accordance with the as:\umplion
Ihal v ,/0,0'* cannol be internal :\ince the surface is convex and v is directed
along an out ward normal at point 0'. Thererore, owing to the conv(".xity, the
segment O'*cr lies wholly within or on the yield surface. On the contrary, accord
ing to equation (II), the segment belongs to a place either external or tangential
Lo the yield surface. Consequently for (I J) Lo be true for v ,p 0, Ihe segment 0'*0'
mLlst be part of the yield surface. It follows obviously that if this surface is
striclly convex, 0' = 0'*.

6.2 Theorem

or uniqueness

. Hill's classical theorem of uniqueness of the stress field is established for Ihe
assumption of boundary conditions of t he type described in Secl ion 5.1, for
which the reduclion given in equalion (8) is pos:\ible. II will be seen later that
an analogolls theorcm is valid for the case of boundary conditions expressed
in terms of loading parameters.

0.2. 1 Classical stalemenl

For consistent houl1dary condit.ions of lhe type descri/Jcd il1 Scction 5.1 lhere
exist. several solul.ioIlS, Imt lhere is uniqueness o.f thc slress field in the ref/ions
/orll1ed hy comlJinil1[j the deformed zones of these various solutions (.for ccrtain
COII1 pleme/ilary (LSSUll1plions).




I.he case/or l1-paramder loadi/,lfl

SeIJeml solutions C(ln exist for boundary conditions (ijlhe type rJiven ilL Section
5.2, cor/"(~spOl1(linrJ to /oadin{J depending Oil a finitc l1umber o.f paramelers Q;
lVi til Q heillfl tile loadin{1 /or which I1l1colilailleri .flow takes place al/(I ci tile
slroill r(lle 0/ lhe syslem at unconlained ]low. However, therc is uniquel1ess 0/ the
slressfield ill lhe reuiol1,'ilornwd hy c:omhillillU the r/c;{orl1lcd zones (ijlhesc v(/rious
soit/liol1s (for cerlain complemenlary (i.'iSUll1ptioIZS).
6.2.3 Proo.f
For the case of each of the previolls statements, Iwo possible solutions are
(ai, Vi), (a 2 , I?). Applying the principle of virtual work to fields ((JI _ (J2)
and (Vi - 1)2) yields, in the case of Section 6.2.1,



<J ')(v'

~ \.,) d I' =

t,. ('/';' -

T;')(,,: - ,,;) dS

r (T
Iv p(F -



J.. . .



F)(11 1


d JI

and in the case of Section 6.2.2,


(a l - ( 2)(yl - y2) d II

(Qjl -


Qn((ii - (i;)

since it is assumed that QI = Q2 (resp. cjl = ri 2).

The remainder of the proof is then similar in both cases.
It has been shown that
( 12)
whence, applying equation (10),
(a' - ( 2)(yl _ v 2)


( 13)

at any point of the system (JI). The left-hand side of equation (13) is zero in the
following cases :
(I). If Vi = ,,2 = 0 (i.e. in the zonc.<; which are rigid in both solutions),
(2). In the zones that are deformed in at least one solution (i.e. in the regions
formed by the combination of the deformed zonc.c; in each solution).
If only one ~ensor

is non-zero, e.g.


thell equat ion (13) becomc.<;

(a' _( 2 )y'


which is equation (II), studied in Section 6.1. If both tensors are non-7.ero,
equations (13) and (II) imply that
(a' - ( 2 )y' = 0


(a 2

a')v 2 = 0

again equation (11). From the conclusions of Section 6.1 we call deduce that
a' and (}"2 belong to the yield surface:

In addition the segment (}"'(}"2 belongs to the yield surface, from which it follows
that if this surface is regular (i.e. there is only one tangent plane at each point),
then Vi and v 2 are colinear.
Thus, if we define D as being the combination of the deformed zones cor
responding to each solution, the following conclusions can be made.



I/ .; .

, ,J " .. .


(I). In (lny ~olution, f) is in a limit equilibrium state.

(2). Irthe yield surrace is regillar,there is at each point of f) a proportion;dity
betwecn the strain rates or the different solutions (in particular, the principal
directions arc the same). In the case or isotropy, this implies the coincidence
of the principal directions of the stresses.
(3). Ir.r is strictly a convex runction or cr, the ~trc.-;s lield is determined in a
unique way at any point or D.
(4). Ir.r is strictly a convex function of the deviator s (Mises' criterion for
example), s is determined in a unique way at each point of f) (since cr 1 cr 2 is a
generatrix of the cylinder). cr is determined in a unique way at each point of
f) ir the normal stress is known in at least one point on the boundary or f).
fnti1e case of the Trc.')ca or Coulomb criterion (2) may be replaccd by the follow
ing statement.

,- ;. .

(2a) . The now regime at any point of D ;s the same in all solutions (same face
or same edge). In the case of the face -regime, there is proportionality between
the strain rates in the different solutions (in particular, they have the same
principal directions).
The statel11ents corrc.'>ponding to (3) and (4) arc more complicated (sec [5J).

7. Remarl{s




On reading the previous paragraphs, a sense of doubt may arise as to the

userulness or the rigid-plastic behaviour pattern. Indeed, the modelling of the
real elastic-plastic medium by the rigid-plastic medium can bring many com
plications, as demonstrated in Section 5. Moreover the question of the exact
meaning of the rigid-plastic system (2:) with respect to the elasto-plastic one
(LA) has not been thoroughly dealt with (Section 4). Only will it be settled in Chap
ter V when the identity or(L,) with Clny(2: will be proved under the assumption
of the principle or maximum work. It i.<; our opinion however that the rigid
plastic behaviour pattern has to he associated with the principle of maximum
work in one way or another in order to be given a significance.
In spitc or difficulties, the rigid-plastic pattern has proved to be very useful for,
as we shall see in the following chapters, it made it possible to solve a number of
problems engineers had to overcome. Also it must not be forgotten that the
adoptcd rormulation sh.ould be adapted to the problem under investigation and
to the method of computation available. It may be that at some future date
either the elasto-plastic, visco-plastic, or the pseudo-plastic material will become
the Inost userul pattern, as a consequence or the development or new cOl11puta
tat ional means.


I: I]

J. F. W. Bishop (1953) On the complete solutions of deformation of a plastic rigid

material, J. Mcc:h. Phys. Solids, 2, pp. 43-53.

[2J J. Courbon (1969)

N()(ioll.~ de PI(/sticit(',

Cours E.N.P.C, Paris.

[J] R. J lill (1951) On Ihc slale or stress in a plastic rigid hody al the yield point,

Mag., 42, PI'. 868-875.
[4J W. T. Koiter (1960) General I heorems ror clastic plastic solids, In Prowcss ill Solid
M cchallic:s, Ed. Sneddon a Illi 1-1 ill, N ort h-I-I olla nd Pub!. Co., Alllsterda m.


[5J J. Mandel (1965) Sur I'unicite dll champs des contraintcs lors de l'cquilibre limite
dans un miliell rigide-plastique, C.U.llr.Sc:., Paris, 2GI, pp. 35-37.
[6] J. Mandel (1966) !!'{(:C(/I/it/IIC rlc.~ Mi/;(,lIx COl/til/liS, Vol. II, Gauthicr- Villars, Paris.
[7J J. Mandel (1972) Unicitc et principes variationnels en viscoplasticitc, Pla.~lid(f
d Vi.~('n/'[{/s{ir:i{(:, Ells. I). lb(\cnkovic and J. Saknyoll, EdisciclH;c, Paris, 1'774,
PI'. 186-202.
[8J J. Salcn~.on (1969) La theorie des charges limitc..<; dans la resolution des problcmc..<;


dc plasticilc cn deformation planc, Thcsis Doc\, Sc., Paris.

J. Salcn~on (1974) ['Iaslit:;{(: /lour la Mh:alliCfllc d(~s Sols, C.I.S.M., Rankine Session,
Udine, Italy, 1974.
R. J. Shield (1954) PI'lstic potential theory and Prandtl bearing capacity solution,
J. IIppl. Mcch. lrall.~. II.S.M.E., 21. pp. 193-194.




._- - - -- - - - -






1. Possible Loadings
1.1 Boundary conditions

A deformable body (V) is considered with a volume V and a boundary

DV = S. The boundary data concerning both the stress vector T and the velocity
u consist of
(I). Three mutually orthogonal components for the set of these two vectors
at each point of the boundary S;
(2). The body forces throughout V.
(i = 1,2,3) at each point M of S, are the orthogonal axes corresponding
to the directions of the given components ofT and u.
Attention i~ given to the problems of equilibrium of the body (quasi-~Ialic
deformation) under all the loadings compatible with equilibrium for which the
axes M XI remain the same (Vi) as well as the nature (T or u) oflhe datum along Mx j
The~e loadings will be called possible loadings of (V).
The part of S on which the component I; is given will be called Sr,; and the
part of S on which the component II, is given will be called SII,' It is obviolls that


S./. (\ S



S'T , V Su, = S

I .

I .

1.2 Definitions
The suit. of possible dynamic dala, lei' has components?; over S"/"l (i = 1,2,3).
F (body forces) throughout V (such that equilibrium is possible).
The suit of possible kinematic data, J c' has components u, over SUI (i = 1,2,3).
Ii suit of possible data, J, is formed by the union ofa suit orpossiblc dynamic
clala and a suit of possible kinematic data:
1 = lei


It is easily seen that the sets of all suits of possible dynamic data, of all suits


of po~~ible kinematie data, and of all suits of possible data, may be given
structures of linear spaces in It The first two of those spaces, generally of infinite
dimension, will be denoted by ~ and
A st.atically (ulllli:o;siIJle strcss-jidd i~ a stress-lield a associated with :l suit of
possible dynamic data, and expressed as




1(\ E ~,

if, for this suit of data, it satisfies

(I). The equations of equilibriulll (in the weak sense),
(2). The boundary data for the stresses.

JA kinematically admissible strain-rale field is a strain-rate field v associated

with a possible suit of kinematic data expressed as
uK.A. ass . .Ie E rt::,

if its components


all .)


aX j


derive (in the weak sense) from a velocity liekl

ditions in this suit of data.


satisfying the bOllndary con

1.3 Theorem of yirtual work

It is assumed that

Va S.A. ass. 1,1'

1(\ E!2},

Vv K.I1. ass . .Ie'

.Ie E


Noting that a'l1 = T at each point of S and denoting by

from which v is derived,


the velocity lielcl

r TudS+ r (lFUdV=J, (ay)dv=.r(a,IJ).'




where .r(a, u) is bilinear form (runctional) of a (Ind v.

2. Loading Process Depending on a Finite Numuer of Parameters

2.1 Definitions

Of all the possible loadings of Section I, consideration is given only to those

that arc actually feasible within the frame of a particular loading process. The
body is then said to be SUbjected to a loading process that depends on a finite
1.1umber of parameters Qj if

The notation is (0" v)

= n ,,' lJ'r

.J '


- -J

(I). The set of all the permitted suits of dynamic dala constitule a linear sub



(2). Tile seL of ail the permilleu suils of kinemalic daLa is a linear sub-space




(3). Two linear applicalions can be determined,



Va S.A. ass. .1,1'


a -~ Q(a) = [Qj(a), ... , Q,,(a)]





. J.'''

bOlh being such thal


:Y(a, li)

Q(O"). <"I(V)



== 1

The Qj(O") values are the loading parameters of the body.

2.2 Properties
The vectors Q(a) corresponding lo all the stress fields 0" S.A. associaled wilh
all the feasible suits of dynamic daLa, constiluLe an II-dimension linear space,
{Q}. A veclor Q(a) will be called a loading ofLhe body. Likewise, the q(/}) vectors
,corresponding to all the strain-rate fields associated wiLh all the sets of kine
malic dala constitute an n-dimension linear space {it}. {Q} and {q} are dual.

2.3 nClllarJ<s

For a given .1'1 E :.0., Lhere are usually several possible sLrcss-fields, and several
Q(a) vectors corresponding to these fields.
Actually, if

a I S.A. ass. i , IE

0) I'


a 2 S.A. ass. .II, E

~ I'


a, - a 2 is S.A. ass. 0 E

~ p'

but it docs nol necessarily follow that Q(O",) = Q(a z). An analogolls property
holds for J C E r(/ 1" , U, 4(u).

A single Q vector can correspond to several S.I\. fields a but all of them arc
associated with II singk .III E !'i'll" An :lllalogous property is valid for (i, II (the
J c arc then identical apart from rigid-body motion.) A ej(v) veclor is called a
strain-rate of a body.' It has been asslIllled in the definition in Seetion 2.1. that
fIJ" and ('(1" are linear sub-spaces. This implies that, in the loading process,
the constant data must be zero. for the ease in which constant data would
exist (for example, dynamic data representing body forces or load on partof a
surface), these should be considered as variable. and only given their prescribed
vailles at the end.

J. Case of a Systelll with Friction Conditions

Statements in Sections 1 and 2 concern the case of a body with boundary
data belonging to the type indicated in Section 1.1. The applical ion to the case of
a body, or system or bodies, in which friction conditions intervene (non-smooth
contacts) requires further consideration. The loading parameters, in the case of
body with a friction contact on part of its boundary ~Ire defined for a larger
system including both the body and the friction interface. (The system's boundary
passes outside this interface.) The friction condition only appears as a now rule
within the system. The same applies for a systenl comprising several bodies
coming into contact with friction .
4. Example

Indentation by a slllooth rigid plate of a half-plane with a uniformly dis

tributed surface load (figure fll.A. J).




Figure III.A.I



for S "1 ,"I = 0;

for S "l ,II,,. = O.
On x';1' and ;1s,


.)T I ' '1'1

= 0;

for ST~' 'l~ = - 1'.


This nomenclature makes it apparent that if the velocity field

q(v) = o.

([1 is arbitrary).

is thaI of a rigid-body lIlotion,



On A'A
for S.,.., '1"1 = 0;
for Sell , u 2 = - U

+ Ux.


are arbiLrary).


(jl -



dx -

.[n u



= N,

Q3 = M.

(N is the resulLanL of the forces applied by the punch, measured positively in

the - Oy direction, M is the moment with respect to 0).



Proble111s o.f uncontained

plastic .floHJ in plane strain

1. General

This chapter deals with a class of uncontained plastic now problems the
solution of which incorporates important simplifying features. For this reason
the form of many fundamental Soil Mechanics problems has been reduced to
that of plane plastic now.
These problems will be studied very closely, as they are rather difficult, espe
cially in the logic involved which is sometimes inlricate.
The probel11s of axial symmetry can be solved similarly for certain hypotheses
(e.g. the criterion of the intrinsic curve type, Haar-Karman's hypothesis [2IJ).
These problems are dealt with in Appendix n and readers requiring more com
plete informalion are advised to read references [I, 12,21,39,54,60]' .
Only the case of isot ropic m:llerials will be deal I wil h lhough anisot ropy has
been studied in plane plastic strain [2, 7, 13,25, 26].
2. Kxprcs.o.;ioll of {he Yield Crilcriol\

The plane str;lin normal to the Oz axis is clcrinecJ by the following condit ions.
/I .~

and ")' arc independent of z;


= ()


/I.\' :

=/J:: =0


In the deformed, i.e. plastic zones, in which the factor A -=P 0, equation (2)
represents three equations which generally make il possible Lo express (1., "C.,,:,
"C .1 as funclions of (1 .~, (1)" t.~)'. The yield crilerion satisfied in thcse regions can,
therefore, be wrillen as a function of only lhree components (relating lo the plane
(x, y)):

is no indicalion of Ihe dependence of P and f on x, y if Ihe malcrial is nol homogcncous

(homogeneous only along Oz): Ihc resulls ofScclions 2, J and 4 arc valid in Ihis C;ISC.

I There


._- - - - - - -

Since oilly lhe case ofisolropic materials is deall with the yield criterion may be
expressed as a fUllction of lhe principal stresses:

Ullder plane strain conditions , Oz, being a principal direction of v, is also a prin
cipal direction for <T.l Selling rT: = 0") (with unordered principal stresses) allows
equation (2) to be written as

IJ )


= 0

Assuming the material obeys the principle of maximum plastic work it follows
from equation (5) that


-.-.. = 0



Thus, equations (4) and (G) deline the yield criterion in plane plastic strain, viz.
F(rT" (J 2> (J)) =






This criterion, expressed as

ill the plane (x, y),


function of the two principal stresses

(T I



is obviously, as a consequence of the convexity of F, the equation of the projec
tion in the (T) direction of the apparent boundary of the surface F = 0 onto the
plane of (T I and rT 2' This is a real curve symmetrical with respect to the bisector
((J. = (J2' (J.l = 0). Thus,.r is symmetrical in rT. and (J2' and may be wrillell in
the form

i.c. in thc plane (x, y) a criterion of the intrinsic eurve type exists.

With the notation of Chapter I equation (8) reads;

g( - p, R) =


where R is the radius of the Mohr circle and - p the abscissa of its centre. Later,
the intrinsic curve is assumed to be real.
The theorem derived may be summariz.ed


II is sialed in Chapler I Ihal in lhe case or isolropy v and rr necessarily have the same principal
Jire!:1 ions. A very dela iled dis!:lIssion or Ihis reslill a risinl~ rrom Ihe !:onsl illll ive la w lin king v. rr and
/)rr/D( (in which Ihe relalionship belween v and f)rr//)( is linear) will be round in [25]' (See also
[2] and PO]).


For (Ill iSOl1'Opic, ri[Jid-plas( ic lI1(/teri(/1 oIJeyill!/ the principle

(~r Ill(lXillllll/l pl(/s( ic

work, (/IIY yield crilaio/l ill pl(/lle pl(lstk sl/'{/ill /"(~dllc(!s to a criteriOIl (~rthc intrillsic
curve type ill the pla/le 0/ the straill.

In parlicular, thc critc'riawhich depend only on the deviator strcss (ductile

materials), always result in the two-dimensional Tresca criterion, i.e.

3. A Remark on the Case of a Non-standard Material

For an isotropie material, with a yield criterion of the intrinsic curve type, e.g.
a sDil obeying the Coulomb criterion, it may not be necessary to refer to the
principle of maximulll plastic work to arrive at forllluia (R) or (9): it is surficient
that (1:, a principal stress as defined previously, is the illtermediate principal
This may be a consequence of the assumption of plane plastic strain and a
particular now rule, not necessarily that of a standard material. It occurs, for
example, if a flow rule similar to equation (25) proposed in Chapter I is lIsed, or
if the constitutive law of a standard Mises material is adopted.
It is also true for an isotropic material with a yield criterion of the general form
F(al' a 2 , a:\) = 0 (F being convex and symmetrical), and plastic strain obeying
the flow rule of a standard Mises material. Again, in the plane of the strain the
criterion is represented by a real curve symmetrical with respect to 'the bisector
(a, = (12' a J = 0), i.e. a criterion of the intrinsic clI\'ve type. 1'1'0111 hereon it will
always be assumed that the envelope of the Mohr circles is real.
Under the conditions stateu the results to be developed ror the stress field arc
valid for such non-standard materials.
4. Equations for Ihe SIrcsscs

Under the specified conditions for the validity of equations (8) or (9), three
equations are available for the thrce unknown stresscs in the plastic zones, which
are independent of the velocities.

/(a.\" a,.,.,. )

aa + -'~'~
+ pX
ax oy
orx c)a,.
--' + -,- -I- f) Y
ax ely



= 0

= 0

(I I)

= 0

The main interest in plane plastic strain problems is due to this particular
circumsLancc. If the boundary conditions are appropriate, it is possible to
proceed in two stages, starting with the determin(Jtion of the stress field in the


..1 ';', ,



- ---


plastic ZOIlCS, without illvolvement of the velocitics, and only then determining
tile velocity field.
However, this procedure proves to be more complcx than indi cated here, and
it is onen necessary to make usc orthe veloeities for the determ ination ofthestress
rield, e.g. to select one out of two possible stress solutions. Nevertheless, this
possibility of dealing firstly with the stresses and then with the velocities, does
make the work considerably simpler.
It should be noted, moreover, that the system of equations (3,11, 12) and the
accompanying conclusions are normally valid in zones where rT z is the inter
mediate principal stress and a state of limit equilibrium prevails, when the mate
rial obeys a yield criterion of the intrinsic curve type. Therefore the results of the
following Section A can be applied i.n elasto-plasticity problems to zones in limit
equilibrium (J = 0) which meet these conditions.
In the rigid zones (denoted as b-type zones in Chapter III), three equations
arc needed to determine the stress field ifit is assumed that rT z is a principal stress.
One equation is lacking in order to determine the three stresses
rT ,rT ,t' ,as the
yield criterion supplies no more than an inequality.


S. Transformation of the Equ:ltions

The method adopted rollows Mandel's approach [35, 36,38]'

The principal stresses (J 1 and (J 2 are ordered in the plane O.~y following rT 1 ~ (J 2'
A new variable is introduced, the angle 0 = (Ox, (1), which is the angle between
the direction of the greatest traction and the axis Ox.
The stresses rTx' rT y' t' xy are expressed as functions of the three variables R, [1, 0,
by mea ns or Figure IV.I :
(1x - -p + R cos 20

= - fJ - R cos 20

't' ... y


( 13)

I< sin 20

- Txy

Figure IV.1

From hereon the material is considered as homogeneous, and the case of a

material which is non-homogeneous in the plane of tile strain (when J is a
runction of x and y) is studied in an appendix.

The yield criteri on (indep enden t of x and y) is satis!i ed at any

point of the
plaslic ZOlle and is solved in the form

R = R(p),


(the soluti on of equati on (9)).

Accor ding to relatio n (13).


ap + _
dR ap
- Cos 20
ax ax ell' ax

. 20-;ao
- 2R SIt1

( 15)


and likewi se for

or .. ar
ax ay






The deriva tive dRjdp is easily estima ted from Figure IV.2. Here,
denote s the
angle betwe en Ox and the tangen t lo the intrins ic curve at the
point of contac t
with the Mohr circle having a centre of abscis sa -I'



( 16)



Figure IY.2


The system of equati ons for the stresse s in the plastic zone is
thus transf ormed
by the introd llction of new variab les. Eqllat ion (3) is eql.,ivalei1
t to (13) and (14),
and equati ons (11) and (12) becom e

DfJ (1 - -a


SIl1 ./)

. 2 0 -()()

cos 2)
0 - 2 R SIl1


I -a
()I' SI11
. 2 ()
. ao

Sill ()

-I- 2R cos

. "'.
+ op
oxsln' l's1l12 0' +

0 00
2Rco s2 ' ax

- oy(l

20 -




- sl11 cos20 )

. 2 ao
+ 2 R SIl1
0 oy -+ (l Y

= 0



--- ---- - - -- - - - - -- -

This is a quasi-I inear system of two eCj\la tions of the first order for the ul1known
fUl1d ions, f1 and n, of the tw() va ria bles x a nel y. It ca n be shown t ha t if the en vclope
of the circles defined by relation (14) is real, i.e. if the intrinsic curve is real sin
qJ ~ 1), the system of equations (17, 18) is always hyperbolic. Also, (here arc
two families of real characteristic lines.


6. Characteristic Lines
Any classical melhod of delermining the characteristics shows 1 that, al any
poi 11 t of the pi astic zone, t hey arc defi ned by

~~ = tan [ 0 (~+ ~)]

Thus the characteristics make an angle + (n/4 + qJ/2) with the direction of the
greatest traction. By convention, the slip line inclined at - (n/4 -I (/J/2) to (11
is labelled 0:, and the other IJ (sec Figure IV.3).

f'igurc J V.3

At M, the Mohr circle is a tangent to the intrinsic curve, as this point is in a

sta te ofl i mi t eq uilibrium (plastic zone). Eq uation (19) shows that the characteris
lie directions at point M correspond to the points or contact of the Mohr circle
with the intrinsic curve. These arc the two surfaces on which t = 11((1) (sec
Chapter I, Section 2).


7. Relations Along the Characteristic Lines

I t is known that along each characteristic line, the solution (p, 0) of the problem
must satisry a differential relation. This relation states that Cauchy's problem
along tl1e characteristic line, with the given values of(p, n) ror the solution, is not
impossible but indeterminate.
These relations can easily be established. Assuming that axis Ox follows the
tangent to Mo:ut point M, inclined at - (n/4 -I (/J/2) to (11 (Figure IV.3),equations
(17) and (18) become

For example. wriling lhal, alonga characlerislic, Cauchy's problcm is impossihlc or indctcrminate.




- 2H sin

op .

5111 </>



4) -

2R sin

1) - (J.y

cos 2


4) -fIX


-I- I' Y = 0

(21 )




2 H cos

1) --(}y

On mulliplying (20) by cos 1) ~lIld (21) by sin 4) and adding, a relalion is oblained
in which only partial derivalives wilh respecl lo x :Ippear:

+ p X cos 4) + p Y Sill 4) = 0


- -;- cos (/) - 2H --




This is the differential relation along the characteristic line a,

- dp - ----.-. dO -I- I)(X 1 X tan 4) dS ......
cos 4)

X + Y tan


= A is the




oblique component of the body force in the a, {faxes,


up + - dO
cos 4)

= pA ds



and likewise, along the {f characlerislic line



up - -_
.-- dO
cos 4)

= plJ


= O.


If the material obeys the COlilomb criterion, then cqll:tlions (24, 25) dlle to
Mandel [35J, in the case ofa homogeneolls material and for any intrinsic curve,
reduce lo Kotter's equalions (31, 32).
In the Appendix to Chapler Y, which deals with BOilIleall':; theorem, Kotter's
equalions are oblained by considercltions thal clearly indicale lheir physical
8, Computation of the Solution
The mathemalical properlic.<; described in Sections 5, G, and 7 are of greal
practical inleresl, as lhey make it easy to compute the solution of the stress
problem in the plaslic zones. As the problem is hyperbolic, lheclassical method of
characlerislics (Massau's melhod [4IJ) can be used lo lind the solulion.
1 In

Appcndix A, dcaling willi non-holllogcncous malcrial (and cqually applying 10 honwgcncous

malcri:1I). cqualions (24) :In([ (25) arc wrillcn wilh lhc dassi(:al lCllsorial nOlalion.



The va 11Ics of fUlldi OilS jJ Clild (J ,dong ;l.lloll-c!la rac!cristie a rc A IJ ;1 rc assullled

I n he k 11 n wn. The cOllti n 1I OllS sol U tioll of equations (17, 18) is \I n iqucl y detcrmined
on each side of AB in the curvilinear triangle bounded by AB and the inner
characteristic lines issuing from A and B. Thc area is the so-c;dled domain of
. determinacy of AB. (The proof of this classical result of the theory of partial
differential equations is outlincd by the method of charactcristics).
The method of characteristics allows computation of the solution by discreti
zation. The curve AB is divided by 11 points 1,2,3 ... 11 (Figure IV.4). At each of
these poi n ts, the slopes of the cha racteristic lines C( and II are known from rcla tion
(19). The assimilation of these characteristics with their tangents at points 1,2, ...
11 (which becomes more acceptable as the division of AB gets liner) allows the
detcrmination of points (J, 2), (2,3) . .., (n - 1, 11), which approximate to the
intersections ofthccharactcristic lines C( and II issuing from points 1,2,3, ... , II
At lhese intersection points p and 0 may be calculated. At each point (k, k+ I)
the relations necessarily satisfied by the solution along the characteristics
C([k+ 1,(k,k+ l)]andf3[k,(k,k+ 1)] supply, by means offinitedifferencerela
lionships, two linear equations for the two unknowns.

I-;"igurc IV.4


The process can then be repealed by making the points (k, k + l) (amounting
to a number /I - I) play the rolc of the points k in the preceding stage. Thus, the
solution is calculated along the characteristics; whence the result already stated
concerning the domain of detcrlllinancy. Clearly, the liller the mesh, the more
accurate will be the approximation of the solution obtained.
This method can also be used to compute the solution from data of jJ and 0
on two convcrging arcs of characteristics, as well as in other cases (see, for
example, [26]')
It is not intended to become involved in the details of numerical analysis but
merely to give the principles of the method of characteristics, using the simplest
pattern of [inite differences. Obviously, more complex patterns can be used 1 both
to determine the intersections of the grid and to discretize the relations along
the characteristics. The essential point is that the solution is found by using the
characteristic lines and the relations along them.

NOle tha t both the mesh refinement and the refinement of the discretization pattern contribute 10
an improvement or Ihe :\ceuraey of a Ilumerieal solution.



The method ofchar<lcteristics is the basic tool for the solutions of pI <Inc now
problems and is usually carried out numerically. Only in exceptional cascscan
analytical solutions be found, as indicated later in a classical example. The
method has been much used by Sokolovski [55-58]'

9. TransforlU:l(ioll of I~q\lll(i()ns (24, 25)

In the c<ise of no body forces, equations (24) and (25) may be wrillen


+ - - dO
cos (P

= 0

along an

dp - - - dO
cos (/)


along a






In accordance with Mandel, L is c1e!inec\ by

= - ~

cos c!>(p) c\p



whereR = R(p)isrelation(14),and</) = 1)(p)istherelationbe(weentheabscissa

of the ccntre of a limiting Mohr circle and the angle .of its tangcnt at the poin t of
contact with the intrinsic curve. Then (24) and (25) arc integrated to give


along an



const. -


along an



L - ()

= cons!.

L -I- 0


For Tresca's criterion

L= -l'j2k
which results in Hencky's relations [24J:
I' -I- 2"0

J1 - 2"0

= canst. along an CJ. line;

= const. along a II linc.


For Coulomb's criterion,

Ir l=C-atan</1,
R = (H

p) sin (/'


11 = C cot


On the other hand,

( 16)

dp = (iHjsi n (/'
which yields


cot </' fdH

cot (/' I

= - - -- = - - og I~

cot 4)
L = - - - l o g (f/ -I- 1')



l~' - -9.2 Case of a Trc.c;ca criteri oll with cOllservntjvc hody forcc.<;
When dealin g with Tre~ca'~ criteri on, (/) = 0 and the charac
teristi c linc~
con~titlltc a net of ort!t0f /o/w[ CU/'l}('S.
If the body forces are derive d from a potent ial V, as is most
comm on, then

A = - --.



JJ =


l:,qllat ions (24) and (24) thell becom e,


P V) -I- 2k dO = 0, for an ex line


d(p + (1 V) - 2/C dO = 0, for a fJ line

from whieh are derive d cquat.ion~ (27) and (28), modif ied by
the replac ement of
p by (I' -I- (1 V)_

10. Gcom etry of the Chara cterist ic Netwo rk

The charac teristi c lines consti tute a netwo rk of two famili
es of curves that
interse ct at an angle (ex, jJ) = (n12 + (fl).
10.2 Hencl<y's theore m
For the above cases, where the relatio ns along the characteristic~
are integr ated
inlhe form of equat ion (27) (exclu ding the case ofa Coulo mb soil
with self-w eight) ,
I he inlerse ct ion or pairs or Cl and II lines (al M, N ,I',
Q in Figure I V.5) is 1l0W COIl
sidere( L

Figure IV-S
From cq\l:\lion~ (27),

L.M - OAt = LN -- ON

L r - 0/. = LQ - 0Q

LM -I- OM = L./. -I-


() I'

LN -I- ON = LQ

whenc e,


0M =


(31 )


This rclationship is known as I-Icllcky's first thcorcm. In thc casc whcrc 4) is

indcpcndcnt of fJ (for the Tresca or Coulomb criterion), the theorem leads to a
simple gcometrical propCrly: thc variation of thc tangent anglc along an CI.
charactcristic line two given f1 characteristic lines is indcpendent orthe
considcred a charactcristic linc (and similarly if the rolcs of CI. and f1 arc revcrscd).
A nctwork having this property is callcd a Hcwky lIC!

10.3 Conseqllences of I-Iellcky'.<; theorem for some simple fiel<ls

Semi-homogcl/colls field: In thc conditions for which Hcncky's thcorcm is

applicablc the strcss ficld is said to bc scmi-homogcncous in a domain if CI.
in cquations (27) is constant in that domain.
By reference to Figure IV.S again, it may be seen that () and L are constant
on each a (resp.
line in the domain; also, the CI. (resp. (i) characteristics arc rccti
linear in that domain.
Homogeneous field: In thc conditions for which I-lcncky's thcorcm is appli
cable, the stress-fielel is said to be homogeneous in a domain if both CI. and f1
arc constant. Then Land 0 are constant in the domain, and both families of
charactcristic lines are rectilinear.

lOA Some properties or semi-homogeneolls fields

It is of interest to dcal in more detail with semi-homogeneous fields for thc

Tresca or Coulomb material, as thcy appear to be very useful for obtaining
The CI. (resp. II) characterist ic lines of such a ricld arc segments of:1 family of
straight lines depcnding on one parameter. I [Reciprocally, it can be proved that
any family of straight lines depcnding on one parameter may be taken as a
family of CI. (or
lines.] These straight lines have an envelope, and it call be
shown that points where characteristics come into contacl wilh the envelope
canllotlie within the body.2
An important case is that whcre the ellvelope converges 10 a single point,
forming what is known as Prandtl's f;1I1 [47]. This point appears as a dcgener:lted
f1 (rcsp. a) characteristic. If the Coulomb criterion is used, thell Prancltl's centred


Figure I V.6

Onc may lakc, ror instancc, Ihc curvilincar abscissa along a givcll fI (rc.~p. (X) linc.

rc.~ull is rrom a particular C;ISC of BOllncau's thcorem Cl)], givcn in Appcndix Il orChaplcr V.

z This





- ---J

-- ----~

- - -----

filiI is constituted by logarithmic spirals (circles III the case of the Tresca
criterion) ;1/HI radii (see J7igure IV. 6).
In the general case for the Tresca criterion, where the charactcristic network
is orthogonal, the fJ (resp. a) characteristics are the involutes of the envelope
(E) Wigure [V. 7). The abscissa on (E) of the point of contact T of each a line
with (B) is denoted by s and the distance TM from this point of contact to the
corresponding point on a {I line by r (Figure IV.7). It is known that fI = s -I- r
is a constant along each fJ involute of (E).


Figure IV.7

Hence, allthe rectilinear segments comprised between two slip lines of the other
family have the same length.

ft. M a tehing of Solutions

As the characteristic lines arc rcal, it is possible l to match two solutions
(//011(1 such a line which correspond to the same values of p Clnd 0, satisfying (24)
and (25), but dirrering in the values of their derivatives. Thus, a continuous
sollition is obta ined, with discon ti n uous derivatives normal to the characteristic.
Ir Jlencky's theorem is applicable, it is easy to see that only a semi-homo
geneous field can be matched with a homogeneous field. Ir the matching is
carried out ,\long an a (resp. fJ) line it is thea (resp~ fJ) parameter that is constant
in (he semi-homogeneous field: for the Tresca or Coulomb criterion, it is
(herefore the rJ. lines (resp. fJ) which are straight in the semi-homogeneous field.
The results obtained so rar refer to the stress field in the plastic zones. Although
Ihe slress problem has not yet been fully explored, an investigation of theprob
lem of the velocities will now be commenced. The problem of discontinuous
slrc.'\s fields ill [hc plaslic /.olle will be dealt wilh al Ihe end or P"rl B (Section )0).
AI1 example will be given in Part C, and in Part Dsome peculiarities of problems
involving a Coulomb material will be mentioned.



"row Rule

The participation or Ihe flow rule has been dealt with in Seclions 2 and 3.
For materials with a criterion or the intrinsic curve type it was found to be

Thc indetcrminacy of thc solution of COlI/chis problem on a slip linc arc with charactcristic data
arises from (he indctcrminacy in the calculation of the normal derivativcs.


unllecessary for the now rule to be that of the standard material ill order (0 have
(he stress problem in I he plastic zones formulated in the manller indicated. Thus.
it was possible to deal with (his problem without fully specifying the now rule,
provided thal cerlain conditions wcre satisfied . On the contrary it is obvious
that the flow rule musl be thoroughly defined for the study of the velocity
As a simplification. only the case ofa standard Tresca material is to be investi
gated. It is known that the behaviour pattern in such a case is acceptable for
metal and undrained clays (4) = 0).
The case ofa standard material with any type of intrinsic curve is dcalt with
in Appendix A. Although the hypothesis of the principle of maximulll plastic
wor~ is not physically realistic in the case of cj) f: 0, it is lIsefulto deal with the
problem under this hypothesis:
(1). On the one hand, the solution for the standard malerial can supply
information with respect (0 the bchaviour of the rcal material (cr. in Chapter
V, Radenkovic's theorem, [48J);
(2). On the other h~lnd, as indicated in (he Appendix, (he study of thc velocity
problem for a. standard material with any type of intrinsic curve supplies all
the information required to dc(erminc solutions for nOll-standard matcrials
which, in plane problcms. have hoth a yield criterion alld a plastic potell(ial
of the intrinsic curve type (sec, fur eX:\Il1plc, [t.!{j).
12.2 The s (andard lila terial

In the case of a standard material. the now rule is



0,.. T.,.,.. oJ = 0 is the yield criterion;

/(0." a .. r'.lJ = 0 is the projection of this criterion onto the plane (s. y)

so that

where the function

0:(0.,. 0)"

T.".) is obt:lined by resolving






- - - - - ---


u.. = A-:'I-forA
()(J . .

O,i,j = 1,2



Thus ./: the 'criterion in the plane (x, yr, is also the two dimensional plastic
r;or lhe case where F depends only on the dcviator,
((7' . __ (7')2

...... ~ ___......... L .

.2 .2

+ ...D'

-I- .. 1-~~





= A(J.\. -




Uxy =

A((7'y -


A. xy

13. Characteristics-Relations for the Velocitics Along the Characteristics

'./ ~ '

With thc stress problem being solved as indicated in Part A, i.e. the stresses
being dctcrmincd in the plastic zones, the corresponding veloeitics arc now
The zones in which <T was not determined arc assumed a priuri to be unde
formed i'.ones, in which the yicldstate is not necessarily rcached. The motion is a
rigid-body motion, and the velocity field is determined from the boundary
conditions. Intercst will now concentrate on the velocity distribution within the
plastic ".ones.
. As only II Tresca material is being investigated, the CJ. and Ii characteristic
lincs arc orthogonal.
At a point M oftlle plastic zone, let Mx and My be axcs langcntialto Ma and
M/f(J7igure IV. 8).Then,atM,(7'."'.. = -I} = (7'.
= k.
y' .\')'


Figure IV.B

The flow rule (Jo) yields

()u)tJx) = (auJay) =



+ (au/ax)

= A~ 0


From this we derive I he rolk)\villg resull. A( (11}()i//( M (1,(, ('x(('//sio// r(/(cjidloll'i//f/

(l Mr;. will "'I II is zero {/I,d (hl' slw(If s(r(/ill rol(' ///IIS( Iw J>()si(ip(~.
With the introduction of v" and /)(1' the componellts of the velocity following
directions CJ. and fJ. eqtlation (J7) can be expressed, by equating the projections
on Ms of the velocities 'a t /1'1 alltl N (also at M ;\lld J1 on My), as

(hI' direr(iolls

dUll dV(I



dO = 0 along CJ. lines




= 0 along II lines

and the lIecessary condition (). ~ 0) as




From a mathel11<ltical standpoint, the problem of the velocities in the plastic

zones where the stress-field is known, as defined by (JG), is hyperbolic and linear.
The CJ. and II characteristics for the stresses are shown by (J8) to be the character
istics for the velocities. I These equations, referred to as Geiringer's equations
(17,18), are the relations along the characteristics. Geiringer's equations state
tha tt he deforma tion in t he plane (x. y) occurs wi t hOllt vo'llllne ch:1I1ge. a n(\ has :IS
(orthogonal) directions of zero extension the directions of the CJ. and II stress
cha racterist ics.

14. Examples
Normally the determination of the velocity field in the plastic zone is carried
OLlt using the method of characteristics. It is possible in some cases to determine
the explicit form or the soltltion. An exa Il1ple is given by the velocity fields associ
ated with the simple stress fields met in Section 10


14.1 Homogeneous stress-field

The characteristics constitute a net work of orlhogonallines tilal can be taken
as lines of carlesiilll coordinates: II" = " ", "" = 'J,. (see Figllre IV.<J). l:-:qU;ltion
(38) shows tlwt

~.'~.~ = ~~~x = 0



Hence, the form of the general solutioll for the velocity field in a homogeneous
stress field is

"x = .rev). "..' =



and expres:;ion (39) yields

/(y) -I- U'(x) ~ ()
I The

eh:lraeleri:;lie:; of :;lIch a problem of lhe fir:;1 order are lhe

(41 )
ellrvc.~ along which :\ differelllial

relalion exi:;l:; for lhe :;Ollllillll.


I,~' ~ ',


- - - - - - - - -_ .

figure IV.9

14.2 Semi- homog eneou s strc.<.;s-ficld

Thc {J charac teristi c lines are assum ed to be rectili near
segme nts whose
cnvelo pe is (13:) Wigur c lV.IO).

Figure IV.I 0

The Geirin ger equa tions becom e

dUll -


dO = 0 along a. lines

c!vll = 0 along {J lines


The a. and {J charac teristi c lines can be taken as a system of

curvil inea r ortho

gonal coord inates , the corres pondi ng param eters being 0,

consta nt 011 the {J

lines anc! variab le on thea lines, ,Inc! 1', consta nt on thea lines
and variab le on the

{I lilies (cr. Sectio n 10.4).

Equ;ll ion (42) implic s that VII is a functi on of 0 only; for instan



= f'(O)






il follow s lhat

Equat ions (43) ;IIH.I (44) give the genera l form of the soluti on
for the vc10ci tics

in a semi- homo geneo us stress lic1d.

The necess ary condi tion (39) is, theref ore, (since dS" = -clp
and ds", = r dO)

wri llcn as

- {I'(p)



- r - -I- [f(O) -I- [J(p)]

r1 ~ 0


15. Discontinuity of the Velocity

The problem as dealt with so far has concerned continuous velocity Iields
in a plastic zone for which the stress Iield is known and is continuous. Next
comes the problem where the velocity Iield is discontinuous, although the stress
lic1d is continuous (the deter'mination of the weak solutions).1
As the stress-field is known, the characteristic lines network is Iixed. This
corresponds to the fact that, with the stresses being determined, the problem for
the velocities is linear.
Il is known (sec Courant and Hilbert [IIJ II pages 4~6 and following) that
for any linear hyperbolic problem the discontinuity lines of the weak solutions
arc ne~essarily characteristic lines. A jump condition and (In equation for the
propagation of the discontinuity along the characteristic line arc obtained.
The main results may be summarizec!"as follows.
(I). All the velocity jump lines arc necessarily characteristic lines.

(2). As the velocity flux through the discontinuity line must be conserved,
because of the incompressibility of the material, the velocity jllJIIP is necessarily
tangential. This is the very jump condition foreseen above.
(3). If, for instance, the line of discontinuity of \I is an Ci. line, then in the close
neighbourhood of this line Geiringer's equation can be applied: with dv"
-up dO = 0 following the Ci. line and v,. being continuous ~Icross the ex line, the
propagation eqllation results; i.e. 2


= O.

The velocily jump is cOllslanl along the discontinuity line.

(4). Finally, the necessary condilions corresponding to J.. ~ 0 mean that,

if the Ci. (resp. Il) line is crossed in the dircction of the Il (resp. ex) line, lhe dis
conlinuity (?f Va (resp. Up) Imlsl.he [!()siliue.

Figure IV. I I

It will be noticed that the characteristic lines which arc the maximum she'lr

stress lines appear as the possible lines of sliding. From this comes the fact that
and fJ lines arc usually termed slip-lilles. This is in agreement with some experi
mental results in Soil Mechanics [2, 22J (Figure IV. I I).


The need for laking inlo accounl Ihe weak Solulions of lhe malhemalical problem appears in
praclieally all examples. This is clearly sugge:;led by cxperimenls as one orten observes lhe locali
7;llion of lhe deformalion in very lhin zones which may be considered as sliding alollg a surface.
1 Square brackels indicale lhe disconlilluily.


16. Dis con tinu ity of the Str ess -fie

sta led in Sec liol \ II, lhis sec tion
dea ls wit h the wea k scl ulio ns ror
slre sse: ;
lha l arc som etim es nec ess ary in
a par ticu lar typ e or pro ble m . (Se
c, for exa mp le.
[ ()3j .)
Th e stre ss pro ble m is qua si-l ine
ar and t(lke!) lhe ror m of a con
ser v;lt ion law .
Th e ma the ma tica l stu dy is giv en
in \" II] II pag es 488 to 490 , the
resu lts of wh ich
arc as foll ows .
(J). Til e dis col llil luit y lillcs
arc /lot ell: lrac leri slie s.
(2). We ak sol utio ns can be obt
ain ed eve n if the dat a are con tinu
ous .
(3). Jum p con diti ons arc obt ain
ed, wh ich cor res pon d to the
tinu ity of
the stre ss vec tor app lied to the
dis con tinu ity sur fac e. The se arc
not sur fici ent
to det erm ine a uni quc dis~on
tinuous sol utio n, and in pra ctic c it
is nec ess ary
to hav e som e ind ica tion of the
sha pe or the dis con tinu ity line
Wi th reg ard to the vel oei lies ,
a vel oci ty dis con tinu ity can noL
exi st alo ng a
stre ss dis con tinu ity line . Thi s
line mu st be con sid ere d as the
bou nda ry of an
infi nite ly Lhin rigi d zon e. (Cr.
Ref ere nce [26J and App end ix
Sec tion 7.2, of
the pre sen t cha pte r l .)


17. Th e Pro ble m

/\n exa mp le is noW giv en to dem
ons trat e the con cep ts intr odu ced
int he pre
ced ing par agr aph s. Th e pro ble
m con sid ere d, as in (38), is the pas
ssu re ofa
wei ght less pla stic soi l on a sm
oot h rigi d pla te. Th e soi l is ass
um ed to obe y the
Tre sca crit erio n.
The bou nd. ,,y con diti ons arc
as foll ows (Fi gur e IV.12). At
inli nity , the
vel oci lies .1 rc zer o .




) s

On OX ,

= - q,


= 0

> 0: l' = 0,

s < 0:

t =



= n(a -I- s)
= 0

'/M == a+ s

Thi~ is wha l


Fig ui'c IV. 12

mak es lhc wea k solu lioll s ror slrc

sse~ sigpific:l1I1 wilh resp eci


Ihe phy sica l prob lem .

18. Construction

or the Solution

It i$ intcnded to construct a ~olution corresponding to an upward now or soil,

i.c.the way Ihc plaslic zonc bctwccn AD and 01' would move.
IJy slarling along ())' :lnd applying Ihe method or Section ~ it i~ seen that in
thc angle (OC, 0 Y) = n/4 lhere cxists a homogcncolls ficld, as t =' 0 and (T
= constant on () )'. Two sllch cqllilibriulll situations an; possihle, bec:llIse OC
can eithcr bean a or {J characteristic line (two possible Mohr circles). Thercfore
I hc velocitics I11l1sl i ntcrvcllc (a nel, thcrerore. t he now rule): :IS the motion of the
matcrial must proceed from 011 towards the surface () )', the principal stress If
must be the grcatc~t traction and thercforc DC is a If linc. I
Only a scmi-homogcneolls field can bc connected to a homogencolls ficld,
whcrc'thc (J charactcristics will bc straight lines. Any IJ linc of this scmi-homo
gcncous field that cuts 110 must make an anglc of n/4 with 110(T = 0 on 110).
Thcsc lincs will, thcrcforc, bcparallel straight lincs and wc klvc undcr AD
anot her homogcncous field, 0 A l3.
Thcrc remains thc angle BOC, which connccts both homogcncolls ficlds.
If this is assumcd to bc continuou~ (in ordcr to h'avc a continuolls solution),
thcn the arca may bc sub-scctioncd u::;ing a Prantill's fan ccntrcd 2 in 0 with
an apcrtllrc (/).
Thcstress on 011 is normal and uniform and is simple to calculate. Its vallie
is dctermincd lIsing the rclation along ex characteristic lincs from o}r to OX.
On O)~

/' = -

If I

-I- k


whcncc, on 0 A

= (q +



andthc pressurc on 011 is givcn by


If 2

P -I- k -. (I -I- 2k( I -I- w)


19. Calculation of the Velocities

Undcr AIJCD,the wcdge considered as rigid lllust be immohik ill accordance

with thc conditions at infinity. Gciringcr's c<]uations (38) show that ufl is COil
stant along cach IJ line. Also vfl is continuou:; whcn crossingalincs, in particular
thc A nCD linc. As vfl = 0 undcr A IJC n'''fl = 0 on A IJC 0 a I~o, in thc pl:l~t ic zonc.
Thereforc, vII = 0 in thc wholc of thc plastic zone.
l ,! 011.
Thc calclllalion of IIa COnHl1CIlCc." :t1ont
DAn II :::-.: constant
This reasoning is nol pcrreclly rigorolls. ;lIld IIses S(Hlle inl lIilinn: hilI il is nol indispcnsahk. The
problem of defining Ihe CI a nd {I cha radcrisl ic lines ca n be lefl aside and Ihe I w(, possi hie l~qll iii hri II III
siluali(1ns Ihal supply Iwo dislincl v;III1c.<; for Ihe pressure on OA e;ln be sludied . The queslion is
selllcd when Ihe velocilY problelll is considered. Eqllali(1n (J9) Illllsi be salisficu al ;1I1Y poinl. and
Ihis nppears 10 be possible for only one Sclecli(1n (1r Cl and {I (al Ihe lll(1sl, since illllay harren Ihal
neilher seleeli(1n makes il p(1ssible 1(1 salisry (:19) everywhere).
lOis 11(11 wilhin Ihe plaslie range, bUI is on Ihe bOllndary. ill at.;reelllenl wilh Seelioll 10.




~ ------




each a line (rrom equati ons (]8)) and 011 OA. Il" = n(lI -I- s).
Since "II = 0, there
is soil sliding under nealh OA (which is admis sible as 0 A
is smool h) rrom A
towar ds 0, and

V/1 = n(a



With the notati on of Sectio n 14.1 (Figur e IV.l3) ,


D.(aJi: +

2y) = J(y)

from (40) and it is verifie d that I'(y) ~ 0,

rn one, v,. = consta nt along each (J. line accord ing to equat ion
(38), and the
contin uity of v/1' when crossi ng OB, determ ines its value.

AI.-L- ---'---L -...1-

Figure IV.13

Hence v/1 = n(a-f i + 2(R - r)) = g(p) = g(r) with the notati
ons of Sectio n
14.2. A check is made tha t g(p) - rg'(p)' ~ O.
In OeD, the field is identi cal to that in OAB.
Finall y, there is a discon tinuity of the veloci ty along ABeD .
[oJ = D.a-fi is
positiv e when A /JeD is crosse d in the direct ion of the {J lines.

20. A Partic ular Case

In Figure IV.14, the partic ular case is shown where (J) = n/2,
(I = 0 and I is at
infinit y. i.e. the plate is horizo ntal and is given a vertic al transl
ation. The plastic
flow call occur all the right of A as well as on the Icft of A' since
I is at inlillil y.

Figure IV.14

The pressu re is given by


(J 2

= (n



This soluLion is valid regardless or the rricLion condiLions beneath the plaLe.
The symmetry of the problem requires tilat the equilihriulll under I.he plate is
homogeneo-us.(lf Lhere is fricLion, the velocity distribuLion is symmeLric, but if
there is no friction, the velociLy disLribuLion depends on an arbiLrary non
decreasing funcLion or y, namely the sliding velocity of the soil underneath the
21. Rcmarks


thc Sollltioll

For the general case of Section 19, the solution is valid only if q is not too
great. Intuitively, it is seen that if q becomes too great the material tends to
collapse on the left-hand side of A towards AX.
The above solution is not complete as stresses in the rigid zone under AJJCJ)
have not been considered. To fully solve the problem, it is neccssary to deter
mine the stress-field that satisfies the equilibrium equations, and the boundary
conditions ror the stresses, and does not violate the yield criterion. I t is sufficient
to prove that it is possible to find sllch a rield (which, furLher on, will be called an
admissible stress field). I
The theory or limit analysis, which will be presented in the rollowing chapter,
will make it possible to give a clear significance to the results obtained.
22. The Case of a Coulomb Material

The same problem is now studied ror the case of a weightless Coulomb
material. In accordance with Sections 2, and 3, only the stress-field is considered,
as this does not require a fully specified now rule. (Appendix A contains all Lhe
information necessary for studying the stresses and the velocities. It should
be noted that in the dealing with a Coulomb material (eg. [55J
and [58J), the velocity field is usually nottakell into account.
The reasoning required ror the construction or the solution will not be re
peated in detail. The drawing for the problem is modified as indicated in Figure
I V.15. The characteristics are not orthogonal ill this case.


Figure lV.IS

II ha:; been proved Ihat Ihi:; can <lclllally be done ir q i:; not


great [51]'


., .



- - - - - - - - - - - - -- -- - - - - - - - - - -- -- ---..1

As ror:l Tresea material. the stress applied on 0;1 is a principal stress (as on
It is Ilecessary to know whether this stress is major or minor (i .e. <11 or (1 2 ).
In Section 17, this was made known by the participation or the velocity rield.
In this problem the movement of the plate is into tbe soil, and hence the stress
on OA is necessarilY<1z
Mandel [J8] proposed that in the case ora Coulomb material when the now
rille is lert unspecified, the selection of the solution sbould be based on the
solution for the same problcm in the case of a Tresca material with associated
now rille. In classical Soil Mechanics lextbooks eilher tile question is never
po.';eci or it is solved by 'physical' considerations via the concepts of 'ma~iml1lll'
or 'lI1inimum equilibriulll' of active and passive pressure. As exrlained later in
Chapter V the choice can be made ollly by using the theorems of limil allalysis
for sl.(lndard (lnd lIoll-slone/(lrr! mOlerials. The real significance of the two above
interpretations is that the Coulomb material is assumcd to be a standard one


In tbe present case, we derive that in OeD the field is homogeneous and
rr = - fJ = <1 I' whence (OC, 0 Y) = n/4 - #2. In OAI3 Ihe field is homogeneous
and Ihe pressure II under the plate is the greatest pressure, i.e. /I = - <12 alld abo

(OA, 013) = n/4 -I- (/J/2

This zOlle is in a state of [Ulssive limil eqllilihrium.
The tield in ol3e is semi -holllogeneow; (logarithmic spirals). The slress on
0;1 is Ilormal and uniform. It is calculated using the relations along the charac
t\.t .S on 0 Y, - p = rr I - R. I-lcnce P = fJ -I-- (f-J -I- p) sin </) and, frolll equation
cot 1)
cot </)
H + q
2 . :..,. - - .. -Io!:, (11 -I- p) =7 .- ----- Io!:'--..--- - -
2 ' I -- sin t/)
At P on OA -I' =


-I-- R. I-lcnecp =

II -

(H -I- p) sin rP and

cot (/)
[-I -I-- /I
L = - .- -- . log - -- -
I -I- sin (/)

Applying equation (27) (2:: -


(n+ I-I) = (q

constant) along the



I --




lines, gives

1: 2 "'IAI\0/'


For the c'lse where (J) = n/2 (the rroblelll of the bearing capacity of a footing on
a weightless soil) the solution is given by


= -

[-[ -I- (q

I + si n (/)
[-I) -__.__




I;"IAn ,"

which contains the terms for surrace load and cohesion.





The previous arguments ill Section 20 are still valid herc.

In order to have a complete solution for the strcsses, <In admissible stress-field
under ABeD is required.
The velocity field is not cOllstructed, and as will be l'hown in Chapter V it is
impossible to interpret the results ohtained through this type of solution. As
indicated previollsly, it is possible to give <III interpretation by taking as a basis
the solution of a similar problem for a Tresc:1 material with associated /low
rule for which 'the velocity field is known. In actual factti1is procedure is equiva
lent to assuming the Coulomb material under consideration to be standard.
The theory of limit loads for non-standard material [tlRJ will he cxpl:lillcd in
Appendix A of Chapter V and will then make it possiblc to draw some conclu
sions for this type of matcrial.


23. Method of Superposition-Theorem of the Correspondil1g States
23.1 Method of superposition



The solution process developed for the gcneral case (Sections 7 ami R)
makes it possible to study directly by a single calculation the problem of a
cohesive Coulomb material with self-wcight :Ind with or without surface load.
Usually, the calculation of bearing capacities or coeflicients of active and
passive pressure is executed in parts, applying the so-called ll1el.hod of super
position. For instance, if the bearing capacity is required for a footing on a
cohesionless soil (c = 0) having 1, f.:. 0 with self-weight and wilh a surface load,
then these loads arc studied separately.
(:I) The prohlC'1I/ (~r the /Jcoril1(! copocily of (lll/olainlwil h se(r-II'(~i{/hl (//ullVilhollt
a sUI/ace load.
A stress-field () I is found which satisfies the equations of eqlJilibrium with the
body forces and the boundary conditions, with imposed stresl'CS being zero at
the surface. Also,
f(cr l ) = 0 in the plastic zone
/(cr I) ~ 0

outside the plastic zone. 1

(b) The prohlem clthe hcarill(j copocity o/a IVci(jlztlcss /lUlI.erial IVir.h a sUI/ace
A stress-field cr 2 is fOllnd which salisfies the equations of equilibrium withoul
body forces and the boundary conditions, with stress on the surface being equal
to the surface load. Also
/(cr 2 ) = 0
/(cr 2 ) ~ 0

in the plastic zone (which can be different from Lhat of cr I);

outside the plastic zone. t

A.s nlready slaled. lhe slresses arc nol sludied oUlside lhe rlal;lil: zones. The heurislie hypolhesis
is made Ihal lhe slress-field can be eXlended in each rigid while rc.~pecling lhe equilibriulll
equalions and Ihe boundary condilions. wilhoul violaling lhe yield erilerion. This omission is, nnd will he sludied Ialer in Chapler V.


Then, the stress lield cr = cr l + cr 2 salislics the equilibrium equations with the
body rorces, the bOlll1dary conditions with the surface 10(ld, and, as a consequence
of the form of f,


f(cr ' -I- cr 2 )




(The clastic range of the material, f(cr) < 0 is a convex cone with a summit 0).
At a given point, f(cr ' -I- cr 2 ) = 0 only if f(cr ' ) = f(cr ' ) = 0: and, ifboth tensors
cr l and cr Z have the same ordered principal directions and therefore the same
marginal plane facets, /"/ = h(cr). Here cr is known as a field in 'safe' or limit
eC] II j Ii bri lim.

Thus, lhe hcaril1{f capacily oiJtaillcd by superposition is an approximation to

lhe real iJearin{J capacity and borders on the conscrualiue side.
The method of superposition is also valid for a Coulomb material with
co hesion. fn this case the criterion /" acting ror cr l , is the criterion of the corres
ponding cohesionlcss material, and for cr 2 f is the criterion of the material with
cohesion. Then,
/,(cr l )



f(cr 2 ).:::;; 0


f(cr i

cr 2 )

= f(cr)

.:::;; 0

the equality f(cr) = 0 being obtained only ir f(cr ' ) = /(cr'J.) = 0, cr l and cr 2
having the same ordered principal directions.

23.2 Theorem of corresponding states

If th\e material has cohesion, the theorem of corresponding states is applied,
making it possible
to reduce the problem to that ofa cohesionlcss material with a
confining pressure I-I at the boundary.
Introducing a tensor cr' derived from cr according to
,r.- . .

cr '

= cr

- HI

reduces the problem to one in which the same equilibrium equations, with the
body forces and the boundary conditions modilied by adding a normal pressure
equal to 1I, must be satislied. The Coulomb yield criterion applies without
cohesion .
For the calculation of bearing capacities, this amounts to the application of a
fictitious surface load which is taken into account just as any surcharge. The
confining pressure H is then subtracted rrom the calculated pressure.
The utilization of the theorem of corresponding states is evidently not essen
tied. It is possible, as demonstrated in Section 22, to carry out calculations with
both surface load and cohesion for a weightless material. Theconnection between
the two terms is self-evident in the formulae. Nevertheless, this theorem is very
useful to perform reductions in the formulae, a priori, before making use of
dimensional analysis considerations, for example.


23.3 Consequences
The application of the method of superposition leads, in the case or calcula
tions of bearing capacity, to the classica I relationship due Lo Terzaghi:

where N ,N ,and N are, in the case ofa homogeneous material, functions of </)
only. With cllarts ofcN y ,N " and N as functions of </) availa ble, it is possible to
calculate Pull for any values of q, C, y and U.
from the practical viewpoint, this relationship has advantages. j n fact, in the
global calculation, the different parameters arc mixed, giving through conside
ratioll~ of dimensional analysis, a relation for Pull of the form,


fyl3, qfyl3, Cfyl3, (/J) = 0

By applying the theorem of corresponding states

Pull=q+(C+qtan(/ . N 1 ( C



/.</) )

tan ()

(5 j)

or, equivalently,
/'''11 = - Ccot</) -I- (I -I- C~ cot (/J).N

- .------ ---- . </) )
(I + C cot </)

For I.he application of these formulae it would be necessary to have at one's

disposal the curves representing N I or N 2 as functions of their arguments, for
each value of (p. However. computations by the global method carried out under
the direction of the author [16, 50. 5J] have shown that, through a convenient
reduction of the coordinates. it is possible to usc equation (51) by referring to
a single curve for 4" ~ (/) ~ 40", apart from the charts giving Ny and N". In fact,
for 4 ~ </) ~ 40 the curves representing

+ C coJ </)

N,,(I+ Ccot</)

as functions of


N" . 2(q


C cot </J)

constituted (I very narrow band. Thus, the utilization of the glolxll calculation
ror the b~lring capacity of a footing no longer appears impossible from the
pr:lctical viewpoint.
23.4 Rcm:uks

The proor or the theorem of corresponding slales, in order to be rigorous,

OUGht ;llso to consider the velocity fields _For this purpose it would be necessary


-- - -- - --

- - - - - - - - - - --

--- --J

for the constitutive law for a Coulomb soil to be speeilied, which is not normal
practice in Soil Mechanics. Likewise, attention mllst also be given to the friction
conditions (rollghness) bet ween the soil and, for example, a footing.
Generally, the stated results are true if the flow rules of the materials, with or
without cohesion, are identical for corresponding states. (This is actually the
case for all the flow rules currently proposed for soils.)
On the other hand, as regards the proof of the superposition method, the latter
is supportecl by the concept that, since the 0- 1 -I- (f2 field is 'safe', the corres
ponding bearing force (for the example of a footing problem) is necessarily an
linder-estimation (in the sense of safety) of the real bearing capacity. It will be
secn in Chapter V I hat this apparently intuitive idea is but the static theorem of
the theory oflimit analysis, the validity of which has been proved only in the case
ofa standard material. Stated otherwise, the proof of the method ofsllperposition
implicitly assumes that the'considered Coulomb material is standard .
.' r

24. All Exalllpic Study of a Cohcsiolllcss Soil with Sclf-wcight

I n order to demonstrate some peculiarities of the problem of a cohesionless

Coulomb m;ltcrial with self-weight, the example chosen is the passive pressure
caused by a smooth pia te acting 011 a wedge (Figure I V.IG). The ground surface is
stress free and inclined at an angle IJ to the hori/.onlal. The wall has a batter
cqual to A..
This problem must be associated with that of Section 22 in the application of
thc method of superposition . 1


( .......

Figure IV.16

The boundary conditions are that

on 0 Y,


on OX,

= 0, a




Strictly speaking, conditions on the velocities cannot be usefully imposed as a

Coulomb material is assumed. However, as the equilibrium is that of a passive

In the case of Section 22, the inclination of the stress-rree surrace ({J # 0) would involve no compli
cation wilh respect to the proposed sollilion.



the motion of the smooth wall OA i~ assumed to be towards the right

hand side. A solulion analogous 10 Iha I of Sect ions 18 and 22 is sought for, sta rting
from 0 Y.
A peculiarity appears at once: (J = r = 0, whence fJ = 0, and, therefore, the
Mohr circle oflil11it eqllilib~ium h<ls a zero radius R along 0 Y. Thus, on 0 Yonly
the value of [J is known, as 0 is not given. In facI, it has been proved [15J that the
value of 0 is fixed.
The stress-free surface is a line of singular points at which the determinant
involved in the solution ofCauchy's problem is zero. The problem has;1 solution
only if a condition which determines the value of () is fulfilled. This condition can
be explained as follows.
As~uming that 0 is known on 0 Y, and two point~ P <lnd P' infinitely close to
one <inother are considered, an attempt is made to apply this method of charac
teristics (Figure I V.16). Equations (24) and (25) are c1iscretii'.ed, taking the value
of Rat the surface to be 7.ero. and this leads to two linear equations for only one
unknown, the value of I' at M. Such equations mllst be compatible.
M ore rigorous Iy. a long 0 }' dp = 0, or

tJp .

- --- sin



tJ" cos II

= 0

-I- -_.-


From (17) and (18),


ax =


+ sill c/J cos 20

sill c/J sin 20

- 0)1 = }' -cos 20

cos (IJ



sill /1

(20 - /1) = -;-



The possible values of 0 are either

(Ox.(J,) = 0 = n+


. sin







o = -n2 +

/1 I
. sin II

-- - - Arc Sill - 2

si n elJ


The choice between these two possible values corresponds to that which has
already been operative in the cases of a Tresca material and a weightless
Coulomb inaterial. Thus, the indeterminacy is removed as illdicated in Section
22. The rule. is tha t for a passive pressure the solu tion is tha t correspond ing to the
greatest force on the wall, i.e. expression (52), in which (J, is closer to the normal
on the stress-free surface than in rcla t ion (53). The la tter ex pression corresponds,
therefore. to the case of active pressure.
The field is now known in the domain of determinacy of 0 Y. There appears to
be an indeterminacy as the value of the normal derivative of 0 along 0 Y can be

I ;.

I:' "

'- ----

..---- , '---_.-.'- . ' . ~~ . ' , ,~,-,

arbitrarily chosen, since in expressions (17) and (18)

ao/ax and ao/ay have zero

In fact this valuc is fixed, and is nccessarily I'.cro. To prove this it is convenient
to write (17) and (18) using axes 0 X and 0 Y. Then oO/oy, op/ay and the derivat ivcs
of higher order with respect to Yare zero on 0 Y. Equations (17) and (18) must
besatisfied within the whole zone in limit equilibrium. Thercfore, from the equa
tions dcrived from (17) and (18) by difrerentiation,



'--' = 0


In this field, 0 is constant, the characteristics are straight lines and the stresses
depend only on X: the zone within the angle (0 Y, OC) is in Rankillc cquilibrium.
The solution for the angle (OC, 011) must now be determined. This can be
done using the known values of p and 0 on the OC characteristic and a known
reilitionship ror 0 on the non-characteristic line OA :



A. (for a smooth plate)

It should be observed that 0 is necessarily a point of discontinuity or the

solution, as 0 has two distinct values at that point, given by expression (52) and
by 0 = 1t + A.. For Tresca and Coulomb weightless materials this singularity
of the boundary conditions has been satisfied by the introduction of Prandll's
ran, centred at a point 0 in the solution. Here a property of invariance by
homothctia with respect to point 0 will bc used. A solution in the form,

p = yrS(w), 0 = O(OJ)
where rand



denote polar coordinates with respect to Ox, will be shown to be

Sli i la b Ie.

For the cohesion less Coulomb material in the plastic zonc,

R = p sin


whenee, by means of ( I J),


- p(1 - sin



p( I


4) cos 2 (0

- (J)))

4) cos 2 (0

- (J~))


p sin (/) sin 2 (0 - (J))

I\s fJ :111<..1 () have the rorm (55), the boundary conditions on OC and 011 are
sa I isficd, being homol hetic with respect to o. Here the angle () is independent of r,
and the stresses 011 OC are proportional to r (as shown by integration of(24) on
OC). The equilibrium equations in polar coordinates yield






-' + _ -...!:!!! + -'---'" + ')'




= 0

-. ~-

.......--------.---.--, ..._ -

--- . ~

..... - ~


. .. ..... --,, ~-


" ...4- .",,"' . ~._ ...... _,.

)', -

I' cos

1',., = - ')' sIn



It can, therefore, be seen that"the form of solution (55) is admissible. The terms in

r vanish, and (57) is reuucecito the system (58) of two differential equations of the

first order for S(O) and 0(0)


S sin 2(0 - (I)) - sin (20 - w)


- cos w - sin 1) cos (20 - (0) + S COS 2 ,(/)

----------------------------~---2S sin (/) [cos 2(0 - (I)) -I- sin (/)J



cos 2(0 - (I))


sin (/)


Equation (58) is to be ~olved with the data of S(w) anel O(w) on OC, and of

O(w) on 0/1, which gives, apparently, one condition too many. Actually, as OC

is a characteristic and as the data of p and 0 on OC are characteristic, equations

(58) are indeterminate on OC as a starting line with these data, and the data on

OA must be taken into account so as to determine the solution (by thc so-called

trial and error method). Thus, S((I)) and hence [1, and O(w), are obtained.

The solution derived is the only continuous solution of the problem. It is

homothetic with respect to 0 and the distribution of stresses on the wall is

triangular. The characteristic network presents the appearance shown in Figure

IV.17. It is seen that in this case, unl ike tha t of Section 22, there is 110 fa n in 0, only

one {J characteristic in 0 (i.e. OC), and 0 is also a point of discontinuity (49).

From the Soil Mechanics viewpoint, the solution of this problem gives the

value of the weight factor in the problem of the passive pressure.


Figure I V.17

For the calclllat ion of bearing capacity by application of the method orsliper
position, the Nc and N q coefficients or(50) are given by (49). For Ny the solution
must be determined for a footing all a soil with self-weight, having 4) '1= 0 and
without cohesion. To do this, the a bove solutio 11 is used for the case of the aperture
equal to n. The corresponding characteristics network consists, on the right
hand side, of the network of J7igure IV. 17 (ror the aperture n), bounded by both

.J . ... .



/1 charactt:ristics which are biseCted by the axis of the footillg; and the ldt

11;111(1 side of I he net work is formed by sYlllllletry aboul Ihal axis (:\4). lJnlikt: lilt:
value of Nand
N q that of N y depends on the friction condition under the fooling.

The global method calculations of bearing capacity carried out by tundgren

and Mortensen for rp = 30 [34J showed that the solution for a material with
weight, without cohesion or surface load, and in which a point of discontinuity
occurs at the corner of the footing, appeared as the limit of the global solution
when the cohesion and the surface load tended to zero". This was confirmed in
(10) and (50) for other values of 4). Such mallers will not be developed further as
many text books deal with them in detail.
New problems are continuously being published as the appearance of efficient
computer methods opens up new horizons in this field.


[IJ I3. G. 13erezancew (1952) Prob/ernc dc l'I1qllilibrc Limitc d'ulI Milieu PullJh'lIlclII ell
Symetrie Axialc. Moscow.
[2J J. P. f30ehler and A. Sawczuk (1970) Equilibre limite des sals anisotropes, 1. d~
Mec(/l/iqIJc, 9, No. I, pp. 5-33.
[3J M. I30nneau (1947) Equilibre limite et rupture des milieux continus, AIIII. Pts et ClI.,
Sept.-Oct. 1947, pp. 609-653, Nov-Dec. 1947, pp. 669-801.
[4] J. R. Booker (1970) A property or limiting lines ror a perrectly plastic material,
Vniv. Sydl/ey, Civ. Eng. Lab., Rcs. Rcpl., No. R ) 34, March) 970.
[5J J. R. Booker (1970) An improved method ror thc numerical integration or hyper
bolic equations, Vlliv. Sydlley CilJ. Ellg. Lab., Rcs. Rcpl., No. 154, Nov. 1970.
[6] J. R. Booker (1972) A method of integration for t he equations of plastiei ty of a
weighlless cohesive frictional material, Quar. 11. M(!ch. Appl. Math., 25, No. I,
pp. ()3-H2.
[7J J. R. Booker and E. H. Da vis (1972) A general trcatment of plastic anisotropy under
conditions of pJanc strain, 11. MCc!I . Phys. Solidr, 20, No.4, pp. 239-250.
[8J A Caquot and J. Kerisel (1953) Sur Ic terme de surface dans lc cakul des fondations
en milieu pulverulent, Proc. 3rd. Int. Con! Sci. MCc!,., Zurich, Vol. I, pp. 336-337.
[9] A. C"q lIot and J. Keriscl (196(,) Twile de M(:caniqllc des ,<"01.\', 4th cd., Ga u thier
Villars, Paris.
[I OJ 1. F. Collins (1970) A slip line field analysis of the deformation at the conOuence of
two glacier streams, 11. of Glaciology, 9, No. 56, pp. 169-193.
[11] R. Courant and D. Hilbert (1962) Mcthods of Mathcmatical Physics, Vol. II,
Interscience, New York.
[12J A. D. Cox, G. Eason and H. G. Hopkins (1961) Axially symmetric plastic derorma
tions in soils, Phil. Trans. Roy. Soc. London. A, 1036, 254, pp. 1-54.
[13] F . H. Davis and J. T. Christian (1971) Dearing capacity of ani'sotropic cohesive soil,
11. Soil Mech. & Found. Div., ASCE, 97, No. SME, pp.753-769.
[14] Y. d'Escatha and J. Mandel (1971) Prorondeur critique d'cboulement d'un soutcr
rain, C.R . Ac. Sc., Paris, A, 273, pp. 470--473.
[15J J. Ferrandon (1951) Techniques dc I'!ngcnieur, Conslructioll I, Mccanique des Sols,
[16J P. Florentin and Y. Gabriel (1974) Force partante d'unc fondation sut sol verticale
ment non homogcne, Trav.fln d'eludes E.N.P.C., Laboratoire de Mceaniquc, Ecole
Poly technique, Paris, Junc 1974.
[17] H. Geiringer (1930) 13cit zllm Vol1st~indigen eben en PlastizitiiL<; Problem, Proc. 3rd
1111. COllgr. Appl. MCc!I .. 2, pp. 185-190.


r - - - - - - - - - - - - - - -- - - -

[IRJ H. Gciringer (19)7) Fondclllcnls malhcmaliqlles dc la lill:oric de." corps plaslique.o;

isolropes, MOil. Sr. M(/th., lUi. (jallthier :Villars, Paris.
[19J J. P. Giroud, Tran vo Nhiem, and J. P. Obin (1974) 'rab/cs 1'0111" Ic C(/Iml dcs
FOlldatiollS, Vol. 3, Dunod, Paris.
[20] A. P. Grcen (1962) Two-dimensional problems in plasticity, !landbook 0/ Ellfl.
Mcc"., Ed. W. Fliigge, McGraw-Hili Book, Co., New York.
[21] A. Haar and Th. Karman (1909) Zur Thcorie der SpannungsZ\lstiinde in plasliehen
und sandarligen Medien, Nnchr. Gc.~. Wiss. Giillill!J(,II, Moth. !,hys. Kl., pp. 204-218.
[22] P. Habib (1953) Elude de J'orientalion du pJan de ruplure et de I'angle de froltell1ent
interne de certaines argiles, Proc. 3rd. [lit. COllf Soil Mcc"., Zurich, I, pp. 28-31.
[23] Bent Hansen A Iheory of plasticity for ideal frictionless material, Thesis .
[24] H. Hencky (1923) Ober einige stalisch bestimmle F;\lle des Glcichsewiehts in
plastischen Korpern, Z. AI/gcll'. /I1Mh. Mcc"., 3, PI'. 241.-251.
[25] R. Hill (1949) The theory of plane plastic strain for anisotropic melals, l'I"oc. Roy.
S~c., A, 198, pp.428-437.
[26] R. Hill (1950) Thc Mathc/llatical Tllcory 0/ Plasticity, Clarendon Press, Oxrord.
[27] R. Hill (1967) On the vectorial superposition of I-Ienehy-Prandl. nets, Jl. Mcch.
P"ys. Solids, 15, No.4, pp. 255-267..
[28] Hon Yill1ko and L. W. Davidson (1973) Bearing capacilY of foolings ill plane strain,
JI. Soil Mcch. alld FOlllld. Diu ., ASCE. 99, No. SMI.
[29] A. 13. Jenike and R. T. Shield (1959) On the plastic flow of Coulomb solids beyond
original failure . JI. AJIJlI. Mcell ., TrailS. A.C.M . J~., 26, No. II, pp. 599-602.
[30] G. de Josselin de Jong (1959) Statics (//1(1 KillC/I/(/tics il/ th('.r(/il(/ble ZOl/e 0./(/ Gr(/II/I
lor Matcrial, Uitgeverij-Delft.
[31J F. Kt)tler (1903) Die 13estimlllung des Druehes an gebriimmctcn IJlalllbellen, cine
Aufgabe aus der Lehre vom Erddruck, Bcrl. Akad. lJc:ricltt, p. 229.
[32J F. Kouer (1909) Ober den Druck von Sand, lJcr!. Akad. lJericllt.
[33] J. Legrand (1970) COllfS dc M(~cal/iqllc des Sols. E.N.P.C., Paris .
[34] J-I. Lundgren and K. M ortell1sen ( 1953) Dclerll1ina t ion by (he I heory of plasl icily of
Ihe bearing capacity of continuous footings on sand. Proc. 3rd Jill.. Co,!!. Soil M cclt.,
Ziiricl" I, pp.t109-412.
[35] J. Mandel (1942) Equilibres par (ranches planes des solilles;\ limite d'ecoulelllenl,
Thesis (Louis Jean, Gap.), V.z. Travaux, Juin'::'Juillel-Decembre 1943.
[36] J. Mandel (1961) Problclllcs de derormalioll plane (el de eontrai plane) pour les
corps parfailement plasliques, SCII/. Plasticite, Polyleeh. Paris, PST No.1 I G, pp.
i37J J. Mandel (1966) Sur les equations d'ecolilemelll des sols idcaux ell dUorlllalion
plane ellc cOllcept de double glisselllellt,.I. Mcc/1. Pltys. Sol., 14,6, PI'. 303-308.
[38] J. Malldel (1969) COllI'.\' dc Scicl/cC dcs Mat(:rifllIX, Ecole Nalional Superieure des
Mines de Paris.
[39] J. Mandel and F. Parsy (I96J) Quelques problc/Ilcs Iridill1ensionncl de la thcorie
du corps parfaitelllent plastiqlle, SCII/. Pla.rt. Ec. PolytCc/llliqllc, PST No. 116, PI'.
[40] J. Mandel and J. Salcnyon (1972) Force porlanle d'ulle fondation sol line assise
rigide (etude thcorique), Gcotccltl/iquc, 22, No. I, pp. 79-93.
[41] J. Massau (1899) Mcmoire sur I'inlegralion graphique des equations aux derives
partielles. Chap. IV. Equilibre limite des lerres cohesion. repro as Editiol/ dll CCII
lenairc, Comite National de Mece Bruxelles. 1952, from AI/I/. As.\'. [I/rl. I:'c. Gal/d.
[42] G. Meyerhof (1950)-1951) The IIl1imate bearing capacity of foundalion, Gcolcclt
nique, 2, No.4, pp. 301-332.
[43] A. Nadai (1950-1963) Theory 0/ FlolY al/d Fracturc of Solids, Vol. II, McGraw Hill
Book Co., New York.
[44J R. Negre (1968) Conlribulion;\ ,'elucle de l'equilibre limile des sols et des materiaux
pulverulents el coherenls, Thesis Doel. Se. Grenoble.


145) Nguyen Chanh (I96H) Etude expcrimentale de la pOllssce et de la butce dcs terres.
AIIII. des 1'0111.1' el ClulII.r.\h~s. No.3, 1968, pp. 225-236.
[4()] J. P. Obill (1972) Force portante en deformation plane d'un sol verlicalc!l1(:nt non
homogcne, Thesis, Univ. Grenoble.
[47] L. Prandtl (1933) Anwendungsbeispiclc ZlI einem Henehysehen Satz liber das
plastische Gleichgewieht, Z. Angew. Malh. Mech., 3, p. 401.
[48] D. Radcnkovic (1961) Theorcmcs limitcs pour un materiau de Coulomb ;', dilatation
non stancJardisce, C.R. Ac. Sc., Paris, 252, pp. 4103-4104.
[49] H. RcisSller (1925) Proc. lsI Inll Can/. Appl. Mec":, Delrt, p. 29
[50J C. Roche (1973) Force portan te d'une Condation su perliciclle. TrailS. fill d'cllldes
E.N.P.C., Lab. Mccanique des Solides, Ee. Poly tech. Paris.
[51 J J. Salenyon (1972) l3utee d'une paroi lisse sur un massif plastique solutions statiques,
.I. Mer.alliqllc, II, No. I, pp. 135-146.
[52j J. Salcnyon (1972) Prolongement des champs de Prandtl dans Ie eas cJu materiau de
Coulomb, A rchiues of Mec"anics, 25, No.4, pp. 643-648.
[53) J. SalelH,:on (1974) Plaslicilc pour /{/ M cca/lique des Sols, C.I.S. M., Rankine Session,
July 1974, Udine, Italy.
[54] R. T. Shield (1955) On the plastic flow of metals under conditions axial symmetry,
Prac. Roy. Soc., 233, A, nr. 1193, pp. 267-287.
[55] V. V. Sokolovski (1960) Slalics of Soil Media, l3ullerworths, London.
[56] V. V. Sokolovski (1962) Complete plane problems of plastic flow, 11. Mecll. Phys.
Solids. 10, pp. 353-364.
l57J V. V. Sokolovski (1963) Limit equilibrium or granular medium with variable
wcight,11. Meclt. Pllys. Solids, JJ, No.6, pp. 395-410.
[5H] V. V. Sokolovski (1965) Slalics of flranular media, Pergamon Press, Oxford.
[59] A. J. M. Spencer (1964) A theory of the kinematies of ideal soils in plane strain
conditions,1. Meclt. Phys. Sol., 12, 5, pp. 437-351.
[60] W. Szc7..cpinski (1967) Wstep do analisy procesow obrobki, Insl. Podsl. Prahl. Polish Ac. Sc.
[(il] .I. P. Tournier (1972) Compartelllent d'une cOllche compressible limit par un
substratlllll rigide cl soulllise ;', unc change vertieale applique par line semclle
filante, Thesis Ph. D., Sherbrookc Univ., Canada.
[62] G. P. Tschebotariorfand J. R. IJayliss (1948) Determination of the shcaring strength
or varved clays ilnd their sensibility to revolding, Proc. 2nd IIIL. COil! Soil M echo
Rollerdal71, 1, pp. 203-207.
[oj] A. Winzer ami G. F. Carrier (1949) Discontinuities of stress in plane plastic now,
.fl. Apl'l. Medt., TrailS. A.S.M.E., 16, pp. 346-348.



App end ixes



1. General

The proble m orunc on tained now in plane strain is consid ered ror an
isotro pic,
non-h omog eneou s, rigid-pJas lic ma leria I, assum ing tha I there is
homog eneity
norma l to the plane or dcrorm alion.
As in the case or homog eneity , it is shown that ir the mater ial is
'stand ard'
(i.e. it behaves accord ing to the princi ple or maxim um plastic work)
, any yield
criteri on may be reduced, ror plane strain proble ms, to an intrins
ic curve
criteri on in the plane or the strain .
Alternatively, if the criteri on is or the intrins ic curve type, it surficc
s that the
plane of stra ini ng conta ins the princi pa I stresses so tha t the preced ing
conclu sion
still holds. This condit ion can result from relatively broad assum
ptions for the
now rule.
Furthe r, the proble ms or plane strain , for the case or mater ials
having an
intrins ic curve criteri on in the plane or the strain , i.e. a criteri on whose
varies with the locatio n, will be studiedo
2. The Proble m of the Stress es-Ge neral Case
2.1 Stress charac teristi cs

For the stresses in the plastic zone the system of equati ons is

va + _~
ax ay +


flX = 0


at: ,0,' ao.

ox vy







ill which the criterion

explicitly depends on point (x, y) . The existence of
cOlltinllOlls rirst derivatives ()f.r with respect to x and y is assumed.
On inlroducing the parameters R, p, 0 and ordering the principal stresses in
the (x, y) plane according to 11. ~ 11 2 , 11.IC' 111 , L.lCY arc given by
11 x

= -

P -I- R cos 20



P - R cos 20

1: .tCl'

I< sin 20

The yield criterion, which is an intrinsic curve in the (x, y) plane, is solved as



x, y)

a"nd the angle </), defined in Figure IV.A.i, for the intrinsic curve at point (x, y)
makes it possible to evaluate the derivative:



::;- = S111 (/)


Equation (1) is equivalent to (4) and (5); and (2) and (3) arc transformed into

o[J (I _

''/(P cos 20)


. ) -ao -I- sin

. (/)
- 2R s1l12(




tJ [J.
. 20 +
+ __
sin (/) Sin

cos 20

2 I<' cos 2 aT
00 --;- -0p (1




oR .




21< cos 2 0 00


+ pX = 0

. (jJ'/ cos 20) +

+ S111

DR .







2 0 -DO '

-I- -s11120 - -cos20 -I- pY = 0


1<, </1, tJH/Dx, tJR/<'Jy arc known functions of p, x and y. p and 0 arc unknown func

lions of both variables x and y.

Figure IV.A.l

(.'1:. y) is a syslem of orthogonal coordinates. X and Y ;Ire the components of the body foree F

along the axes.


The system (7, S) is of the same type, and has identical properties lo lhal ob
tained in the homogeneou:i case.
The slopes of the characteristic lines are once again


e1y/clx = tan[o



(lhe equation of the intrinsic curve).

The relations along the characterislics are obtained by means of a standard
melhod. For example, by siting the axes x and y along Ma and its normal and
combining (7) and (8) the equation for line a becomes



ax +


P (X cos

(P +

Y SIl1


= 0


in which only thc derivatives with respect to x of thc unknown functions fJ and ()
playa role (oR/oy being known).
2.2 Coordinate system associated with the characteristic lines

xa. and xP form a new system of coordinates, the coordinate lines of which are
thea and {J lines (Fiugre IV.A.2), and Ea.. Ep at each point M are the local vectors
linked to this system:

A vector at M is written as elM = dX a Ea.


X = const.

along each

c( l i l')c


X = cons!.
along each
f3 line


and these ,directions, which correspond to the points of contact of the Mohr
circle with the intrinsic curve al point (x, y) are denoted by (J (resp. a) . These are
the surfaces on which
11'1 = "(a,x,y)

- ax cos 1) -

Figure IV.A.2


.: ';',J

Thcn, the


unit basis is given by



e = E E
= E E
a a'"
-'" "

and dM = dx"'ea . (The xtZ arc Iloll-holollomial coordinates.) ell, efl rorm the duid
basis (associated by the Euclidian inner product) and
dM = clxa .ea 1
Thc rollowing notation

i~ u~cd .

rJ_ =
- aXa

2.3 Rcbltions along the characteristics

f .J

Equation (I I) may now be written as

rJ p -I- - tZ
cos cf)

[p Fa -I-

aa 0 -

() R] =

( 12)



+ --dO

( pP + -a
aR) dx

is the relation along the



' .. __ .1 .

= 0.

( 13)

characteristic. Similarly, along a (I characteristic,

. [ pF" -I- -aR] dx"

dp - --dO cos cf)

,r - ,)


cos (/)


= 0

( 14)

. These cquations are, ror a non-homogeneous material with any intrinsic curve,
thc homologucs or Kotter's equations ror a homogelleous Coulomb material.
It can be seen that these equaLions arc deduced from those obtained in the
homogcneous case, by replacing R(p) and cf)(p) by R(p, x, y) and (/J(p, x, y) and by
adding fictitious body forcc~.J
I The


COlli ra va ria III coordi lIa Ics dx 4 , d .... p a rc I hc obi iq IIC com POIlCIl Is or <1M 011 thc lin it basis,tangclI
the lillc:, a, {I; dx., dX II are the orthogonal projcction:. or Ihis very vector on Ihi:. basi:..


oblique componcnts along M. M p or a vcctor wilh X, Y coordinatc.~ rollowillg ;a ny orthogonal

axcs arc

2 The



--~.-- [X sill {o
cos '/' _

= _ I [_ X sin



whence thc rormulae givillg 1J/{/clX.


+t)} _. y
{a - (~ + !)} +



= ir /{




Y cos

tJH/()x p = all f{

(n.:4 -I- .i2




{a - (~4 + !)}]
as runctions or au/ax and oH/tJ.\'.

AlIclltiolllllllst hc paid 10 thc inversion orslIgscripts : the fictilious body rorcc.~ arc not derivcd lIsing
I{ as a potential.


3. The Case of a Tresea Malcrial

In the case ofa Tresca material with" variable she~lr ~trength k(x, y) (~ce "Iso

[ 18]),

Da R

a.(/ R



= -ak =. -ok cos ( 0 - -TC)



= -ok


-I- ok
- si n


ak . ( 0 - -TC)
= - -Sin


( 0 - TC)
-I- -cos

= dXa' /-,"

lines arc OrlhogonC\l. and cis"

( 0 - TC)

= Fa' ctc. Equations (13) and

(14) arc wrillcll


2k(x, y) dO - ( -~

dp - 2k(x,

y) dO - ( ox:


for an


cix = 0


pF(/ dx(/ = 0

As~umingthcbbclyforccs"rcclcrivcd froma potential JI,P








= - aJl/ox




P V)

P JI) - 2/\ dO - -

2k dO - -



ds = 0,

for an

( IS)
for a

dx(/ = O.



4. The Case

or a Coulomb Material

For a non-homogcllcoU:i Coulomb matcrial.

R(p.x.y) = psin


-I- C(x.),) cos



In particular, if the matcrial has a constant intcrnal (Ingle or friction and a

variable cohc~ion, thc nctwork of charactcri~tics is isogonal and rclation~ (13)
and (\4) arc writlcn

2R dO
+ --I
cos ()

- ( pF




4) -a


dx = 0,

for an





2R dO - ( n p(/ -I- cos (/) .~- d -,;.(/

d [J - ~
cos ( p "
(IX J . '

for a




5. The Velocity Problem

The stress problem can be solved as indicated without having the now rule
rully defined, provided that it satisfies certain conditions. In order to solve the
velocity problem it is necessary to have the flow rule completely defined. The
case of a standard material will be considered, and then it will be seen how to
deal with the case of some non-standard materials.
Asslllning the stress problem is solved wilhin the general framework of Section
2, the velocity problem may be formulated as follows. The two-dimensional yield
critcrion in the plane of the deformation also appears as the two-dimcnsional
plastic potential. Therefore,


= A

V12 =V 21 =O


The expression of J as a function of the principal stresses is easily obtained: in

fact, = R - R(p) with the definitions of Section 2.1., i.e.,



whencc, according to (6),

( 18)
IJ 2 2

= - A( I - sin



Thc value of (/) is known at each point (x, y) from the previous stress solution and
the velocity problem is seen to be linear.

5.2 Velocity characteristics

Again the coordinates x,
of tensor v'J
. . arc



defined in Section 2.2, arc used . The components

(J 9)


or, alternatively,

Da IJ..



} (D",IJ fI


+ Dflo)~O

where the D's denote the covariant derivatives, the expressions of which arc as

.. a

DfI /I

= () U


'" a . .



DUfl = ()

Dflufl =









cos 1) fI


_1- afl

'" eos

- ' -</)) 2



1.) (0 + 1.) +

4)a (0 -

I , ()

u ,

,/I rJ ( 0



rJ 4> =



() - </))

_ 4))

4) a{I (0





tan ,/JrJ

(21 )

(0 -\- '2I

(0 + 4))2 -\- ofltan (0 -\- ,I))2

</)a fl

(J",O and(JflO are known (rrom.the stress solution), as are

decomposed into two terms:

</) a


fI cos

a.. '/J

and afl'l) which me

a.. p' acl)
- + ap ax"



(the term or lhc inlr'insic curve plus the term of non-homogeneity).

are fixed orlhogon<ll coordin<ltc.~ in which Ihe non-holllogeneily i$ dcrined . The exl's denote the
eosines of the <lxes x' with respeet to the t'lllgent$ to the C( and {1lines at the eonside~'cd point :

I x'


, - eos (7t
0 - - - .-.


n; = cos (0 ~ + ~)

(/ 2
= sin

nIl =




--' -7t -

+ 4'7t + '2

5.} Interpretation of (he results

/)aY,. =


0 implic~ that the extension rate along the a. line i ~ I.ero. Therefore,
line is a characteristic for the velocities and
/) IJ
" "

= U IJ

a a

v tan


4)ua ( 0 -


- "(I-- U 0 - cos 4) a

= 0


is the relation along this characteristic. The same applies for the (J line, with




co~ 4) <I{I (0 -I- ~) -I-

v{ltan 4)U{I (0

(~) =



OaV{I '- (Va tan


1) +

_I_ -Icos 4>

(0 _4))2
tan 4)) a (0 + 4)) ~ 0

V _I- ) a
p cos 1)


i~ the condition of positivity (of A).

In the ca~e of a Tre~ca criterion, the Geiringer relations are recognized in

equal ions (23, 24), al~o in (25), the conditiol) of positivity for I he homogeneous
case (the non-homogeneity having no inOucnce on the form of the relations in
this t;asc).
I\s a rule, the determination of the velocity field in the plastic I.'.ones is effected
using the method of characteristics (relations (23) and (24), the condition (25)
lhen being verified at each point).
. 5.4 Hodograph
I t is convenien t to express these results by introducing the velocity plane (a
lllelllOd developed by Green [13] for the case ofa homogeneous Tresca malerial).
I~clations (23. 24) ~how that at one point. an ex (re~p. II) line and il~ image (/
(re~p. h) on the hodograph (the a, h network being the image of the ex, [I network,
in the plane u.<' II) are orthogonal. Therefore, the situation is represented by one
or the four configurations of Figure I V.A .3.
If M -1 - dM is a point of the first ~ector (IX, [I), i.e.


dx{l > 0

the condition of positivity (25) then gives

(D a."// -I- /) (II)a) c.\x a dx{l ~ 0

Also, DalJ{I dx a = Dvp, (24) being taken into account and D{lva dx P
being taken into account. Therefore, (26) may be written, '

DVa dx a -I- DIJ/, dx{l


Du (dM). elM ~ 0


DVtJ' (23)

r---------------------------------------------------- --


"P '

... . _


~_ ........ ~,..:. . . . . . . .:..~_

. . - . . . .;.


_ ._














b........ -




""\ ~











Figure lV.A.3

and the rcciprocal can easily be seen, i,e. ir (27) is sati:died VdM in the sector
(0:, fJ) then (25) is verified .
Obviously, ir configuration (1) is relevant (27) is certainly verified, but for (c),
(27) is not verified. There is somc doubt concerning conIigurations (b) and (d).
The construction of the hodograph orten provides a convenient means to
check if the condition of posilivily is fulfilled (a rcsull indicated by Ewing and
Hill [7J in the case of a homogeneous Tresca material).

5.5 Discontinuity of thc Yclocity

As the velocity problem is linear hyperbolic, the discontinu ity lines o[ the
velocity are the characteristics of thc problem, i.e. thc (J. :lnc! [J lines.
Wriling relations (23-25) in terms or discontinuilies gives for (23), whcn cross
ing a {J line (i.e. [ollowing an CI. line),

[vJ = 0
and the discontinuity concerns


only: and [or (24), when crossing an




= 0

and the disconlinuity concerns va only. These are the discontinuity conditions.
The propagation equations or the discontinuity are obtaincd by applying (23)
or (24) on both sides o[ the corresponding discontinuity line. Thus, rollowing an
0: line,


aJvJ -




(/)) =

0 -"2


and rollowing a



equations which are easily integrated if cp = constanl.

Finally, rrol11 (25), come the conditions of positivity. On crossing a II line in
the direction of an a line, [ufJJ ~ 0 and likewise, on crossing an a line in the
direction ofa /J line, [uJ ~ O.
crossing ora line:

if 4>


= 0


~ 0,


= constant [vJ = [u..]o exp [0 - Do) tan 4)J

crossing of {J line: '

if4) = Constant


[ufJJ ~ 0, Dp[ufJJ = 0
[ufJJ = [ufJJo exp [-(0 - 00)



On the hodograph, the vectors representing the discontinuity are necessarily

oriented according to M" or M rrom the (0) configuration. [t is to be emphasized
" material is the velocity discontinuity a tangent
thaI only in the case ofa Tre:lsca
to the discontinuity lines.

5.6 A comment


the interpretation of expressions (23), (24), (25)

In order to prepare ror the study of a non-standard material, a comment is

made on the interpretation or (23), (24), (25). The equations (23) and (24) imply
that the strain rate in the plane allows zero extension along the a and II lines
which form the angle (rc/2 -I- 1)) and are bisected by the principal directions of<T.
Thererorc, the strain rate y has the same principal directions as <T, and satisfies
the now rule
U I = A( I -I- sin (M

= -

l(1 - sill 1))

and (24) indicates that A is positive.

6. The Case of Some Non-standard Materials

A non-homogeneous, non-standard material is now considered, having a
yield criterion which assumes the intrinsic curve form,


R(p, x, y)




= sin </) (p,

x, y)


The now rule is such that v has the same principal directions as
/J I



= A( I -I- sin 1'(/', x, y))

v2 = 0

..1.(1 - sin v(p, x, y))


A~ 0
v(p, x, y) ~ c/J(p, x, y)
(a I




a J)

Thererore, these are non-standard materials with a yield criterion or the intrinsic
curve form, ,:nd also a plastic potential or the inlrinsic curve lype, both functions
dcpcnding on x and y due to lack or homogeneity. This is agcncralization or a
Coulomb material wilh a Coulomb or Tresca plastic potenlial, as studied by
numerous authors (e.g. [2,4-6, 14, 16, 17,29, 31, 32J).
According to the flow rule the deformation plane contains the extreme
principal stresses in the plastic zoncs. Thus, the problem for the stresses is set
and solved as in Section 2.
As the stress field is determined in the plastic zones the velocity problem can
also be solved in the same regions.
According to Section S.O the strain rale \' al each poinl mllst admil as direc
tions of zero extension the directions rand (5, which form with a I the known




v(J}, x, y)


and thc factor A. must also be positive (J7igure IV.A. 4). Expressed differenlly,
the velocity problem , being linear and hyperbolie, admils as its characteristic
lines the (y, (5) lines, with slopes given by
dy = tan [0





The relations along those characterislics arc

J) yVy = 0 along a r line

(31 )

= 0 along a (5 line



Figure IV.A.4




,J '; ' ., ,

The l:ondition of positivity of A i~

f) yV"


~ 0


Eqllalions 01, 32. 33) are ohlained from (23-25) by replacing (IX, II, (/J) by
(y, f~, v). I
EXAMPLE: Taken as an example is the problem of the indentation of a half
plane or a homogeneous. weightless Coulomb material, with an interna I angle
of friction </' and zero angle of dilation v (deformation without volume change).

Figure IV.A.S

Figure IV.I\.5 represents the networks or the characleri."tic~ (<'l.. fJ) and (y, /j).
The (a:, II) characteri~tic net work is classical. consisting or homogeneolls fields
with rectilinear G1l<lracteri~tics, and Prandtl's fan (vector radii and logarithmic
spirals). The (y, c) characteristic network consists of orlhogonalline~: straight
line~ in APEF'A', AGI1. A'G'H'. and logarithmic spirals in AFG and A'F'G'.
The pressure on A'A i~ given hy the u~ual reslllt


= - f-I +

I + ~in (/)
(q + /-1) .------.-- e" Ian
I - ~in (/)

The velocity field is obtained by integrating the Geiringer eqllations numeri

cally (or c()l\~trllt:ling the hodograph) in A' I~FGll A. It is ~yml1letrical.
The same problem has heen dealt with in [2J for a Coulomb inaterial, t:lking
weight and surface load into :Iccollnt, with various assumptions on the value of
t hc cOllstant angle v. The constructioll of the characteristic networks (a, II) and
(y, r' is more difficult than before, as the determination of the fOllr families of
lines mllst be carried Ollt jointly. The construction firstly the charac
teristics (IX. If) but underneath the footing it nXluires the participatioll of the
(y, /i) characteristics. According to the numerical results obtained in (2) the value
of I' seems to exert liltle innuellce 011 the bearing capacity (for the case of a
weightlcss matcrial, it has no innuence. since the stress field is determined
independently of v).
The checking of the condition of positively of factor A. (or relation (2R)). i~
carried out numerically. In [2], Davis and Booker, having carried out calcula
tions for various v:dues of 4' and Y, ~tate that the dissipation i~ always po~itive .

For the Illitization or the hodograph (Seclion 5.5). ((1./1.


</I. n. h) must be replaced by (1'5, \', c'. d).

f .

Howevcr. Drcschcr [4.5]' on dcaling with thc cxamplc or Figurc I V.A.5. which
is a particular ca~c or thc problem studicd in (2), shows that thc dissipatioll is
ncgativc in somc !.OIlCS whcn \' '1. 1>.1
7. Discontinuity of the Stress-field

Thc solutions considcrcd in Scct ion 2 ror t hc strcss prohlem arc 0111 y a pplica blc
whcrc t hc strcss-ficld is con tinuous. In somc problcm~, wca k soil! tions ror strcsscs
must bc con~idcrcd. i.c. solutions admitting lincs or discontinuity or thc strcss
Thcsc solutions will bc cxamincd rirstly rrom thc vicwpoint or stresses. :111<.1
thcn or the strain rates. in ordcr to dctcrminc thc ncccssary conditions ror thc
vclocity rield ;liong a linc or discontilluity or strcsscs.
7.1 Conditions for stresses

Thc stre~s problcm as dcfincd hy (7. R) is hypcrbolic and ql1(1si-lincar. His

known that
(I). Thc discontinuity lincs or thc wcak solutions ror such a problem arc not
thecharactcristic lincs:
(2). Wcak solutions can bc obtaincd cvcn rrom continuous data;
(3). Discontinuity conditions cxist. but arc not surticicnt ror the detcrmilwtion
or a wcak solution. Rccoursc must bc madc to :1 supplcmcntary condition in
order to dctcrminc thc di~continuity.
Discon( il1ui( .I' rmuli( iOl1s:

The discontinuity conditions ror thc ~trcss-ricld arc obtaincd by applying thc
continuity or strcs~ to thc surracc or discontinuity (Figurcs IV.A. 0(/ and /I).
With thc axi~ Ox placed. ror convcnicncc, along thc normal (0 thc sllrrace or
discontinui(yat M.

R(/,\ .Y.y)cos20\ = -1'2

1/'2.\'.\')cos'20 2

R(/,\.x. .I')sin'20\ = R{['2'x.y)sin 2 2




Figure IV.A.6


For v = O. there is equivalence between the condilion J. ~ 0 (generally more restrictive) and the
thermodynamic condition of nOli-negativity of the dissipalion.


In the particular case of a Coulomb material ((p independent of p), equation
(J4) leads to
sin (Ot - 02) [sin 1) cos (OJ - 02) -: cos (01 + 02)] = 0
whence, if the solution corresponding to the continuous field is ignored, the
relation between 01 and 02 in the case of the discontinuity is
sin 4> cos (01 -

cos (01

()2) -

and therefore,
tan0 2 =cot()1

(}2) ::.~


I - sin 1)
. 1>
I + SIn

Figure IV.A. 6h clearly shows that in the case of any intrinsic curve (convex
and 'opened' in thedirection a < 0) the values of I and 02 corresponding to the
discontinuity are situated, with respect to the values


= 1>(pl'x,y)




= (P(P2'X,y)

as indicated in the following table (3R).




() I




- - - - - - ---- ---------- -------_ .- _._-----._. ._-_._---,

rk~ > () > __ + 1~1.
>0> - - _.
- .-n +--.
> 0 > n





-- -I- -- .


'! _ 1'2 < () ..- ~~ 1:;-J-'~;-:1)~-< () <

. n


._--------------- - - - - - - - - - - -

7.2 Interpretation of the line of discontinuity for the stresses

The line of discontinuity of the stress field may be considered as the limit of an
infinitely thin transition zone which serves to connect two continuous limit
equilibrium stress fields-producing opposite stresses on the faces of the zone
Wigure IV.A.7a). In this zone the stress-lield cannot be in limit equilibrium
everywhere: if this were the case, a family of stress characteristics would neces
sarily have an envelope in this zone as in Figure IV.A.7a, and according to
Bonneau's theorem, such a field cannot exist (see Appendix 13 of Chapter V.)
The transition zone, therefore, includes a part in which the criterion of limit
equilibrium is not satisfied. This is (according to Hill [15]) a centra) core in
which most of the transition occurs. The stresses -r Xl' and a vary little when
crossing the transition zone and the Mohr circle representative of the strcss
state always passes closely to the point (a, -r) corresponding to the stress applied
to thc faces of the 7.One in the states (I) and (2) (Figure IV .A.7b).1

Moreover,the discontinuity lines or the stress field do appear thus in the solutions or some elasto
plastic prohlems ror a standard Tresca material. e.g. indentation oran acute angle wedge [20)'


(a )




Figure IV.A.7


.J~ .... .


- - - -- -- - - - --

7.3 Conditions for velocities

7.3.1. OiscollLinllily of velucily a/oilY {/ linc (e) ill



I\.s already state (Section 5), once the stress problem has been solved the
velocity problem is linear and hyperholic.
I\. classical theorem implies lhat, in each domain where the stress-rield is
continuous, the discontinuity lines for the velocities are the characteristics for
the velocity problem.
[f(C) is a line of discontinuity in the stress-field then, according to the previolls
statement, a discontinuity of the velocity along (C) in tile domain (I) (resp . (2))
can cxist only if(C) is a velocity charactcristic in the domain.
Thercrore, the following conclusions may be stated:
for (J stamford mdtcrial,it is known that the velocity characteristics are
identical to the stress characteristics (Section 5.4.) and it follows from Section 7.1 .
that (C) is not a velocity characteristic. There can be no discontinuity of velocity
along (C) either in domains (1) and (2).
(2). For (j non-standard mo[aial, a now rule of type (28) is adopted. I\.s the
material has any intrinsic curve (convex, open in the direction a > 0) it is
assumed that, beside the condition v(p, x,y) ~ (M", x, y), II is a non-increasing
function of " . In these conditions, ror



v(P I ,




v 2 = v(p 2 ' x, y)

it is possible to designate the values of (] I and

illdicatcd in the following table (40).


with respect to (/) I' (/)2'


and v 2


- ---- - - - ----


> 01 > - -

_. __ . _ __

> () > .. - - - .

--- - -


- > 0 > - -.- -1-



--]----__ _ .



. _ _____ __ . _

_. 1

. ___ _

> () > _._n



- : -(-

_ .. .. _..._ _ ...___. ___.. _

+ .I'.1.

- - ---- -- -------
- - + -- - > n > - .


+ :! > () > _

- : -(-


j ~ -;



According to Sect ion 6. the velocity characteristics are no longer identical

to the stress char:lcteristics. The line (l). without bein~ a stress characteristic.
can be a velocity characteristic in either domain (I) or (2). This occurs in domain
( I ) if
(41 )

and in dom<iin (2) if



() 2 = -4


I' 2


Table (40) shows that (41) and (42) cannot both be trlle simultaneously. Thus.
in the case of a non-standard material of the type indicated. there can exist a
discontinuity of the velocity along (C) in the domain (I) (resp. (2)). if condition
(41) (resp. (42)) is satisfied along (C). Then (C) is a velocity characteristic in the
domain (I) (resp. (2)) and the velocity discontinuity is governed by the equations
of Section 5.6.



Discol/lil/llily of Ihe pc/ocily frOIll ol/e dOIll(/;n 10 (lIIolhel'. aossillfl


By interpreting the discontinuity line for the stresses. as in Section 7.2. it

can be seen whether this zone also makes a transition possible for the velocitics.
corresponding to a discontinuity from domain (I) to domain (2).
This transition can occur only in thc plastic region of the zonc. which accord
ing to thc conclusions of Section 7.2mcans that therc is a discontinuity of the
vclocity along (C) in domain (I) or (2). as in the previolls problcm. Thcreforc,
there is no discontinuity of thc velocity from onc domain to anothcr. crossing a
d iscon t inu ity Ii nc of the st rcss-field .
7.3 .3.

II/(~.Y/(,I/sihili/.l' of Ihe slress discolllinllily IiI/C'.

Since [IIJ = 0 and the velocity is continuous. thc derivatives tangcntial to thc
stress discontinuity line are also continllolls (Had'llnard's relations) or,


== 0

[11.... ,.1

= 0

Hence. for the strain-rate tcnsor


= 0


A now rule of type (2R) is adopted. and on both sides of the discontinuity line

("1) sin 0

(1l 2 )

cos 2 0 = -t[sin V({,. .Y. y) - cm; :>'{JJ}

-t ~ 0



r--- - - - -- - - - --

and (43) bccom cs

A [ si:1 v(p),.'(,:':) - co<;2( ) = A [sin v(P2'x ,y) 2
Al ~ 0,
A2 ~ 0



Accor ding to the values or PI and Pz (whence 01 and (}2 by (34)), it mayo
r may
not be possib le to determ ine values or A > 0 such that (45) is satisfie
d. Jr this is
not possible, the only solutio n will be AI = A. = 0 and hence,

The strain -rate tensor s arc zero on both sides or the stress discon
tinuity line,
so that the line is inextensiblc.
Firstly , the case or a standa rd materi al with any (convex) intrins
ic curve is
exami ned. For each value or p.
v(p, x, y)

q>(p, x, y)

It is possib le to determ ine values or A, and A2 > 0 ir the soluti ons

0, and
equat ion (34) corres pondi ng to the discon tinuity arc such that
(1( 0



COS 20, - sin 4>,

= -------- ---" ~ 0 2 - - Sl\1
. {)2
I >0




I t is immed iately obviou s rrom table (38) that

p{O I' ()z) ::; 0

The Oldy solutio n of (4() is, thcrdo rc,

and hence

= (V)2'

Thus, ror a stal1dard materi al with any (convex) intrinsic curve, the
discon tinuity
lillc or stresses is inexte nsiblc .'
_The case or a non-st andard materi al or the type indica ted in Sectio
n 7.3.1. is
now consid ered. The sign or the ratio,
P(OI,0 2'V\,V 2)

cos 20, - sin VI

- S 1\1 V z

= cos '0
~ z


is seen, by rderen ce to table (40), to be not always negative. Therd

ore, in the
case or a non-st andard materi al, condit ion (45) is equiva lent to
the incxtensi
bility of the discon tinuity line only for some values or 0 and 0z
among the
solutio ns or (34).

This result gencr"li7.c-~ Geiring er's [12].


() 1

For \I Coulomh 11l\llcrial (1) and \' indcpcl\dl:nl of Il)

;lI1d 02 of cqllation (J(i) verify

)'1' A,2 /.;

0 if I Ill: solutions


< - - () I
4 "In

- -I- -


[I J D. Llerthel. J. C. Hayot, ;111<1 J. Salenc;on (1972) Poin\:onnement d'un miliclI scmi
infini ell materiau plastique de Tresca non-homogcne, Archi[)es of Mechanics, 24,
No. I. pp. 127-138.
[2] E. H. Davis and J. R. Booker (1971) The bearing capacity or strip rool rrom the
standpoint or plasticity theory, Univ. Sydney, Civ. Eng. Lab. Research rept. No. R
[3J E. H. Davis and J. R. Booker (1973) The elrect or incrca~ing strength with depth on
the bearing capacity or clays. C(:ot('('hniqllc. 203, No.4. pp. 551-563.
[4] I\. Drescher (1971) A note on plane now or granular media. Proh/C'I1IC de' 1(/ Rh<;o
10Uic. SYl1lfl. Fmn('(}-Polol/(/is. Warsaw, 1971, pp. 135-144.
[5J I\. Drescher (1972) Somc remarks on plane now or granular media, Ar,hives I~r
Mechanics, 24, No. 5-6, pro S37-S4H.
[6] A. Drescher, .K. Kwaszc7.ynska, and Z. Mroz (,1967) Statics and kinemalics or the
granular medium in the case or wedge indentation, Archil}e.\ (!t MCc//(//lics, 19,
[7] D. J. f. Ewing and R. Hill (1967) The rlastic constraint or V. notched tension bars.
J. Meel!. Php. Solids. [5, No.2, pp. 115-124 .
[8] G. Favretli (1965) Impront:l di lin punzonc rigo SIl un materiale non omogeneo,
Il/yc{llIiaria McccaniC(l, 14, No.9, pro 37-50.
[9] G. Favretti (1965) Dipendenza rra durc7.7..1 e prorondila di ci Illentazionc--Appli
cazione della t co ria d ella ria st ici t;i dello s t lid io tco rico del pro hi ema, II/(l CO II icra
Mecctlnica. IS, No.6.
[10) G. Favretli (1966) Indentation ora rigid punch on a plastically non-llOll1ogeneow;
material, Mccctll/iw. I, No. J/4, pp. 83 ..94.
[I J J P. Florentin and Y. Ga briel (1974) Force porlante d'lIne rondation S\II' sol vert iealc
ment non-homogcne, Tr(/l}.jill a'etudes E.N.p.e.. Lab: Mccaniqlle des Solides, Ec.
Poly tech. Paris,june 1974.
[12] H. Geiringer (1953) Some recent results in the theory or an ideal plastic body,
Adpallces ill Applied Mechallics. Academic Press, New York.
[J 3] A. P. Green (1954) 011 the use or hodographs in problems or plane pla~tie strain,
J. M(',h. Ph.l's. Solids. 2 No.2. pp. 73-80.
[14] Dent 1lansen. 1\ theory of Plasticity for ideal frictionless material. Thcsis.
[15J R. Hill (1950) The Mathematicalthenry of Nasticity. Clarendon Press, Oxrord.
[16} R. G. James and P . L. Bransby (1971) A velocity field for some passive earth pres
sure problems. Gcotecllllique. 21, No. I, pp. 61-83.
[17] A. W. Jenike and R. T. Shield (1969) On the plastic now or Coulomb solids beyond
original failure, JI. Appl. Mech .. TraIlS. A .S.M.E.. 27, pp. 599-602.
[18] A. 1. Kuznetzov (1958) The problem of torsion and plane strain or non-homogeneous
plastic bodies. Arch. Mech. Stos., 4, pp. 447-462.



..I ~ "

': \.

11')j E. II. Lee (1950) On stress discontinuities in plane plastic now. Proc . .l,.d S)'l1Ip.
Appl. Mallt .. McGraw Hill cd., pp. 2IJ-22R.
[20) J. Najar, J. Rychlew~' ri and G. S. Shapiro (1966) On the prohlellls of the clastic
rlastic state of an infinite wedge. Bull. Ac. Pol. Sc. 14, No.9, pp. 515-522.
[21) J. P . Obin (1972) Force rortante en dcrormation plane d'un sol verticalcmellt non
homogcne. Thesis Univ. Grenohlc.
[22) W. Olszak and J. Rychlewski (1962) Geometrical properties of stress fields in
rlastically non-homogeneous bodies under conditions of rlane strain. Pr9c. Int .
Syll/p. 2"d ord. In Elaslicily. Piaslicily and Fluid Dynamics. Haifa.
[23) W. 0Is7'<lk, J. Rychlewski and W. Urbanowski (1962) Plasticity under non homo
geneous conditions. Advances ill Applied Mechanics. Academic Press, New York,
pp. 132-214.
[24J J . Ostrowska (196R) Initial rlastie now of semi-space with a strong layer non
homogeneity. Arch. Mech. Slo.\' .. 20, No.6, rr. 651-668.
[251 W. Prager (1955) The sign of rlastic rower in the grapilicaltreatlllent of rrohlems
of r1ane plastic-Ilow. Quart. A1'1'1. Malh .. 13, No.3, pp. 333335.
[26] J. Rych1ewski (1966) prane plastic 110w for jump non-homogeneity. In/. 11. Nun
Linear Mech . 1, pp. 57-78.
[27J J. Salenc;on, M . Barbier and M . Beaubat (1973) Force portante d'une fondation sur
sol non-holl1ogcne. Proc. 81h 1111. COil! S"iIMecli . & Foulld. Ellq .. Moscow. 1.3,
rr. 219-224.
[28J J. Salenc;on (1974) Bearing capacity ofa footing on a (/) = 0 soil with linearly varying
shear strength. Geolec!lIliqlle, 24, No.3, pp. 433-446.
[29J I\. 1\. Serrano (1972) EI metodo de los campos associados. f'roc. 51h Elir. COil!. Soil
Malt .. pp. 77 81\ .
[30J Ie T. Shield (1953) Mixed boundary value prohlems in soil mcchanics. Quarl. Appl.
Malh . ll,pp.61-75.
[31) W. Szczepinski (1971) Some slip-line solutions for earthmoving processes. Archives
of Mechanics. 23, No.6, rp. R85-896.
[321 W. S7.czepinski and H. Winek (1971) On some problems of large flow of soils.
Symr . Franco-Polonais, Problcll1cs de 101 Rhl:ologie, Warsaw, 1971, pro 353365.
[33J 1\. and G. P. Carrier (1948) The interaction of discontinuities surfaccs in
plastic fields of stress. .11. Appl. Mecl!. Trans. ASME. 15, pp. 261-264.
[34) 1\ . Winzer and G. F. Carrier (1949) Discontinuities of stress in plane plastic now.
11. ApI'/. Mcch .. Trans. ASME, 16, pp. 346-348.

,' ~ ,


1. General
This appcndix indicatcs thc typc of plasticityproblcms occurring with axial
symmetry for materials obeying a yield critcrion of thc 'intrinsic curvc' typc,
under thc hypothesis of Haar-Karman currently adopted in soil mechanics.
The theory already explained in Chapter IV and in the preceding appendix
will not be rcpeated but some problems regarding discontinuity lincs, that have
becn ignored so far, will be studied.



r-------------------------------------------------- - --
2. The Strcss Problcm
2.1 Statement of the Problem, Haar-Karman hypothesis

With r,O}. z as the cylindrical coordimltes. consideration will be given to

problems of uncontained plastic now ror which the stress distribution has the
following properties.
(1). There is axial symmetry around 0;;:
( 2). (J ru)

= (J


= 0: (J

The stress-field
lar require that



is independent of 0).


is a principal stress.

must satisfy the conditions of equilibrium, which in particu

(1). There is no tangential component or the body force: F", = 0;

(2). The boundary conditions are compatible with (J r t l=
) : (J
W = 0 (Tm = 0 on

the surfaces perpendicular to O. and to 0);

(3). The body forces and the boundary conditions are axially symmetric
about 0:: F, and F: are independent of (I) , and data '/~ and 7~ on a plane per
pendicular to 0:. are independent of (I).
The stress-field must therefore satisfy both equations or equilibrium, i.e,




__ . -I- .... ..







_!.-. . . !"

-I- I)F = 0




pF: = 0


The isotropic material is assumed to have a yield criterion of the intrinsic

curve type. It is not necessarily homogeneous, but it is necessary, so that (I) is
truc, that the yield criterion is independent of (I).
With the usual notation


{) = -





the criterion aSSUllles the form


R = R(p. r. ;;)

According to condition (I), (J,,, is ,i principal stress. The Ibar-Karman hypo

thesis (6). which states tha t the floll' rcqimc ill (l plastic WIlC OCCI/rs alo/l{/ all cd{/c
(or corner), (J", /)Cill{/ e(ll/ol to OIlC (?f the pri/lcipal stresses ill tI,C II1eriliio/l plallc,
= (J

(eJ 2

= (J.


CJ 2



This hypothesis is discussed further in [3. 8, 12].




ForO = (Or,a l ),
a r = - p -I- R cos 20

ar :

R cos 20

- p -


R sin 20

a '0' = - J1 - d~


r. = - 1

corres ponds to

a = a l'


a = aJ

= -I- I


(Sce [2J, which has differe nt sign conve ntions .)

In the p\as tic zones


R = R(p, r, ;;)



nl</np = sin 1)(p, r, z)

lienee , ror the stress fiekl in the plastic, the system of
partia l differe ntial
equati ons ohtain ed hy applyi ng (6, R, 10) to (2) is

.___ ._(I - sin


4) cos 20)


-- 21< sin 2(} ---. -f- sin


1) sin 20 -- -IDz


2R cos 2(} -.ti:


cos 20 + I:
+ ___ cos 2() +-sin 20 + I< - - - - -- + pF


__ sin


1) sin 2(} +


DO ' Dp .
ar () z (1 + Sin 1) cos 20) -I- 2R sin 2(} --

2R cos 20 -- - ()I<

-I- __- . sin


= 0

2() ..- .. -. cos 2(} .\. R sin
---'--'- ' -I- /,F.




( 12)

. The system (II, 12) is identic al to that obtain ed in the Appen

dix ror the plane
proble ms, thoug h two supple menta ry terms have appea red,
which arc due to
axial sYlllm etry and are taken illlo accou nt as body rorces .
The systcm (II, 12) is hyperb olic and quasi- linear. At each point
of the meri
dian plane, there arc two charac teristi c direct ions:
dz/dr = tab [0 =+= (n/4

(<X, (J lines)


The relatio ns along the charac teristi cs arc

Uf' + '--d O - { pP


+ --iJx(I

r cos

[. , (




( 13)

dp -


cos </J

dO -



~~~ + -~[Sill (0 + (~ + 1!.))

r cos (j>


3. The Velocity Problem

The compo nents of the velocity are denote d by ""
tionsh ips of radial symm etry are




1 all
all - .-~
= _ ___r +. --.!!!




r OW


= _f~

II"" Uz '

The norma l rela



= -


( 16)

The stress proble m having heen solved

lem will be solved in the same re!,!ion .


the plastic ZOIlCS. the velocity prob

3.1 Case of a standard material

The materi al is assum ed to obey the princi ple of maxim \lm plastic
work. and
v. which has the same princi pal directi ons as 0'. has. as princi pal values

( 17)


O. 11

O. r.

( 18)

3.1.1. Ca1clllat iOIl or

With v bein t'"o a princi pal value.


tJ rm =V Zfl)


( \9)

Assum ing that the velocity field is also axially symm etric (on condit
ion that this
assum ption be compa tible with the bound ary condit ions), (16) and
(19) yield




== 0



- --- J
(21 )

whence, II ,wilh the form u = exr, representing a rigid-body rotation ahout Oz.



3. 1.2. III th" f,I(/IIC (r, z)

For the unit basis tangent at each point to the lines ex and IJ, equation (17)

resli Its


= r. -2 cos 1)


/I fIf1

= 1:"2J1 cos 2 (/)





(A + ~) cos



from (17) and the expression for v"HI' from (16) into account gives the


= -

(I +


2~ (1



sin (p)


r. SIn 1)

which show that the ex und IJ lines ure velocity eharucteristies. Equation (26) is a
differential relation valid along the ex (resp. 27, fJ) lines. These relations can also
be written in terms of ur and uz from the classical formulae expressing v and
IJJl/1 as functions of vrr ' vr :' l'~z and from (16). Equations (26) and (27) arc the
relations along the velocity characteristics.

J I.J. C(}nditions 4 I}(}si{ i1Jily

As illdicated ill (I H), A alld It must be non-negative. Therefore, equatioll (25)
implies that,

~ 0,

bill this cOlldilion alone is not sufficient. r;rom eqllations (17) ancl (22) is ob



A~ 0



(A+"2)1) cos


~ i cos 2 1)

or, from (25),


2 cos 1>

V II ~ -a

Lind also

2vall ~ - 1:_'(1

+ I:sin eM


Equations (28) and (29) are tile condit ions of positivity.

3.2 The case of non-standard materials

In the case of non-standard materials those considered here have a plastic

potential g of the intrinsic curve type (as in the Appendix on plane problems),
which is assumed to be independent of (I) . The angle of dilatancy v is a function
of the stress state and of the point


"(p, r, z) ~ t/>(p, r, =)

Then the strain-rate tensor has the same principal directions as cr and its
principal values are

/J I

= ( A. + -I+r.)
2- /1 (I + SIl1

= - (A. -I-

/J ( l ) tI)

= -

2 /1) (I


si n \')


r./1(1 - r. sin \')

A. ~ 0, /1 ~ 0, r. = :I: I

I-Icnce, following the hypothesis that the velocity field is axially symmetric


Ct. ,

The problem in the meridian planc is linear hyperbolic and

characteristic directions are r and (~ :
(0" Y)}
(Or' (5) = () +

(IT"4 -{- :2")


c;lch poillt Ihe


The relations along the velocity characteristics are given by (26, 27) in which
(Ct., p, 4 are replaced by (y, D, 1'). The same argument applies for thc conditions
of positivity (28, 29).

- - - - -- -- - - - - - - - - - - -
4. Weak Solutions

4.1 Weak solutions for stresses

The properties of the weak solutions for stresses are analogous to those for
plane problems. The discontinuity lines I are not the characteristics. The dis
continuity condition expresses the continuity of the stress. (7COM is discontinuous.
4.2 Weak solutions for velocities

The lines of discontinuity of the velocity are the velocity characteristics, iX, {l
in the case ofa standard material, and y, 0 in the case of the non-standard material
of Section 3.2. Expressed by means of equations (26, 27, 28, 29), this statement
implies that
(1). For crossing of an '(l (resp. p. or y resp. b) line vp is continuous (resp. (l or
o resp. y) and the discontinuity concerns only vlJ (26, 27);
(2). Crossing an (l (resp. {l or y resp. (~) line, [vlJ] ~ 0 (resp . {l or y resp. 0) (29),
and [va] must be such that (28) is verified on both sides;
(3) . The equation of propagation is also slightly modified with respect to the
plane case, and becomes


= -

-~2r [v] cos [0


- (~+
1:.)J (1 +


sill (/J)


[uJ = [uJ cos [ 0

(~ +


with an analogous equation along a {lline (resp. }' and (~) .

It follows (see [7,12]) that a finite velocity discontinuity on the axis Oz is not
II] J. F. Adie and J. M . Alexander (I %7) A graphical method of obtaining hodographs
for uppcr bound solutions to axisymmetric problems, Inl. J/. Mcch. Sc., 9, No . 6, p .
l2] B. G. Bcrczancew (1952) Prob/cl1Ie de /'Equilibre Limile d'lIl1 Milieu Pull}crulellt CII
SYl1lclrie Axiale, Moscow .
(3] A . D . Cox, G . Eason, and H. G . Hopkins (1961) AxiaJly symmetric plastic deforma
tions in soils, Phil. TrailS. Roy. Soc. Londoll, A. 1036,254, pp. 1-45.
14] M. Croc, G . Michel, and t\. Peeker (1972) Quelques problcmelide non-homogcncitc
en symctrie axialc, '['ruu. fin d'C/lIdc.\ E.N.P.C., Lab. Mccanique des Solides, Ec.
Polytechn. Paris.
[5] G . Eason and R. T. Shield (1960) The plastic indentation of a semi-infinite solid by
a perfectly rough circular punch. Z.A .M.P., 11, pp. 33-43 .

These arc the Iraces in Ihe meridian plane of the surfaces of discontinuity of the axially symmetric
stress-field .


[6] A. Haar and Th. Karman (1909) Zur theorie der Spannungszustiinde in plastischen .
und sandartigen Medien, Nac/" . Ge.f. Wiss. G(juingell. Malh. Plrys. KI.. pp. 204-218.
[7] D. D. Ivlev and R . I. Nepershin (1973) Impression of smooth indenter into a rigid
plastic half-plane, /zv. AN. SSSR, Mrkhallika Tverdogo Tela, 8, No.4, pp. 159-163,
Ellgi. Trallsl. M ccizanics of Solids. pp. 144-149.
[8] J. Mandel and F. Parsy (1961) Quelques problcmes tridimensionnels de la thcorie
du corps parfaitement plastique, Scm. PlaslicilC. Ec. Poly technique, P.S.T., No . 116,
[9] Z. Mroz (1967) Graphical solution of axially symmetric problems of plastic now,
Z.A.M.P .. 18, pp. 219-236.
[10] R. Negre (1968) Contribution a I'ctudc de I'cquilibre limite des sols el des materiaux
pulverulcnts et cohcrents, Thesis Dr. Sc., Grenoble.
[II] J. Salenc:;on, M. Croc, G . Michel and A. Pecker (1973) Force portante d'une fonda
tion de revolution sur un bicouche, C.R. Ac. Sc., Paris, serie A, 276, pp. 1569-1572.
[12] R. T. Shield (1955) On une plastic now of metals under conditions of axial sym
metry, Proc. Roy. Soc., 233, A, 1183, pp. 267-287.
[13] R. T. Shield (1955) Plastic now in a converging conical channel, JI. Mech. Pllys.
Solids. 3, pp. 246-258.
[14] R. Sibille (1944) Ca1cul approche des solutions de Prandtl dans les cas de revolution,
C.R. Ac. Sc. Paris, t. 258, gr. 2, pp. 2017-2019.
[15] A. J. M . Spencer (1964) The approximate solution of certain problems of axially
symmetric plastic now, JI. Mech . Phy.\". Solids. 12, No.4, pp. 231-243.
[16] W. Szczepinski (1967) Wstep do analisy procesow obrohki, /II'SI. Podsl. Prob/.
Techn., Pol. Ac. Sc.
[17) W. Sz.c7.epinski, L. Dietrich , E. Drescher, and J. Miatowski (1966) Plastic now of
axially symmetric notched bars pulled in tension, JII/. .It. Solids (l1Ii1 Structures
2. pp. 543-554.



I. '; .

- --

- - -

- -J


The Theory 0.[ Limit Analysis (For

Applications to Soil Mechanics)

I. Presentation
This chapter deals with the theory of limit analysis and its applications to
Soil Mechanics . No mention will he made (except for didactic purposes) of the
prohlems related to the application of this theory to the design of strllctures,
i.e. the so-called limit design of structures. For this purpose reference may be
made, for instance, to [7, 8, 40].
The chosen exposition is classical and simplified (similar to that of [37J).
A slightly more axiomatic and detailed presentation will be found in the
Appendix , with the possible extensions of the theory to the case of non-standard
materials [44, 45J.
The concept of limit loading of a system was introduced in Chapters I I and
II I, an d here only the essen tials of the previolls argumen t will he repeated .
1.1 Definition of the limit loadings
13y considering a system made of an clastic-perfectly plastic material, sub
ject to a loading process depending on n parameters Qi (Q = loading vector),
the so-called il1itial elastic limit load QO was defincd , corresponding to the
appcarance of plastic deformations, for any loading path starting from the
,neulral state of stress. Theset of all these loadings Q O is the initial elastic houl/dary
n(rlw system . For the loading path going beyond QO, the limit loading is defined

boundary of the system

initial elastic boundary of the system

Figure V. I


as the loading which generates uncontained plastic now for the (/ssumptio/l of
negligible gcometry c!IlI/lgcs. The set of limit loadings has been callcd the yield
boundary of the system (Figure V.I) for reasons to be explained in this chapter.
It was shown in Chapter III that, since the geometry changes are assumed
negligible up to the appearance of the llncontained plastic now,the limit loading
on a given loading path can be determined by considering the system as con
stituted by a rigid-plastic material (defined by passing to the limit) provided that
this system follows the same loading path as the elasto-plastic system considered
initially. Hence, the possibility arises of defilling the limit loadings on the rigid
plastic system (obtained in each case by passing to the limit) as the loadings
required for the appearance of uncontained plastic now in the elasto-plastic
Finally, introduction of the rigid-plastic system obtained from the initial
system by eliminating the elastic deformation of the material (without passing
it to the limit) results in the limit loadings defined above being loadings for
which there is a non-zero deformation of the associated rigid-plastic system.
In this chapter a study will be made first of the loadillgs for which there is a
non-zero deformation of the associated rigid-plastic system. For convenience,
these will be called limit loadings without specifying 'of the associated rigid
plastic system.' This omission will appear justified ([ posteriori in the case of
systems satisfying the principle or maximum plastic work :1t any point.
1.2. Determination of the limit loadings. Variational approach

The determination of the limit load (in the case of a loading with one para
meter) or of the yield boundary of the system (in the case of a loading with
several parameters) requires the construction of complete solutions of the
problem of uncontained plastic now for the rigid-perfectly plastic system. This
is often difficult, and hence a variational approach is used. Two theorems will
be demonstrated, which make it possible to obtain under- and over
2. Admissible Fields. Dissipation

The system is assumed to undergo a loading depending on /I parameters

Qj' the boundary conditions being both dynamic (body forces and stresses)
and kinematic (relating to the velocities) (see chapter III, Section 5.2).1
2.1 Plastically admissible strcss tClisor

At a point M of the system where the loading function is f, a stress tensor eJ is

said to be plastically admissible (P.A.) if
f(eJ) ~ 0


In this formulation, there arc no constant data except those equal to zero. If some data arc non
zero constant. they will be deall with as though they were variable, and will he given their pre
scribed values atlhe end of the solution.


- - - - - - -1
2.2 Plastically admissible strain rate tensor
At a poin t M of the system where the loading function is [, and where the
plastic flow rule assumes the form (t)-which is not necessarily that of the
standard material-a strain rate tensor v is said to be plastically admissible
(P.A.) if the flow rule (L) may be solved for this tensor; i.e. if

3(1, f(1) = 0



(1 - - -


This inversion is not always possible, and when possible, is not

always unique. The example of a standard Mises material is a good illustration
of this result. For the inversion to be possible, the given tensor v must satisfy
tr(v) = 0, as the plastic deformation occurs without change in volume. If
tr (v) = 0 the inversion is indeterminate and only the deviators of (1 is determ ined.
I n the case of the Tresca criterion, the degree of indeterminacy can be yet greater.

2.3 Dissipation
A stress tensor (1 is associated with a plastically admissible strain rate tensor v
by relation (2) in the expression

Under the hypothesis of the principle of maximum plastic work (equivalent to

being convex and the material standard), the expression (3) depends only on
v and is independent of the tensor (J associated with v. Therefore,


= n(v)


which is the dissipation for this strain rate tensor.


Let (11 ard (12 be two stress tensors associated with v by (2), in which
convex and (L) is the flow rule for the standard material.







:f(12) = 0

v E A(Jf(1I)






I t follows that

In the case ofa Mise!> material,

n(v) =





In the
0", 0",


of a Tresca material, which has the now rule

0" J

= 2k




0"1 =(12)(1J


v, = l


v2 = 0

= -),


= 2k

v1-- l


v2 = Ii

Ji ~ 0



Side face regime

Edge (or corner) regime

n( v) = k{ / VI/

+ /v 2/ + /vJ /}


2.4 Admissible strcs,,<;-field

A stress-field is said to be admissible if it is sta tically and plastically admissible

S.A.: the stress-field must satisfy the equations of equilibrium and thc bound
ary conditions for the stresses (dynamic conditions or the problem).
P.A.: a tan y pain t oft he system, the stress tcnsor must be plast ica1Jy ad missible.
2.5 Admissible strain rate field

A strain rate field is said to be admissible if it is kinematically and plastically

admissible (K.P.A.).
K.A.: the considered strain-rate field
such that


311, vij =~(Uij

must derive from a velocity field u I


li j )

and this velocity field must satisfy the kinematic boundary conditions of the
P.A.: at each point of the system, the strain-rate tensor must be plastically
2.6 Dissipation in an admissible strain-rate field

Under the hypothesis of the principle of maximum plastic work, the dissi
pation in an admissible strain rate field is defined by the integral


n(v) d V

which is an univocal function of the field v.

I From

hereon, the notation used is thaI a, y denote tensors, whilst


v denolc fields of these tcnsors.


~ ', .

- - - - - - - -- - 3. Static Appro ach

3.1 Admis sible loadin gs

A loadin g Q, which can be equ ilibrat ed by at least one admis sible
stressfield. is called an admissible loading.
The set
(I I)
{Q(IT)\IT admissihle}

will be dellot ed by K . K c W.
It should he noted, in partic ular, that any limit loadin g is admis
sible. (More
over. for the loadin g paths of Section 1.2, of an e1asto-plastie struct
ure for the
assum ption of small deform ations , all the loadings, from 0 to Q
I inclusively,
are admis sible loadin gs.)
3.2 COllvexity of K

TJlEolu;.M: If I is convex, then K is convex.

The proof is trivial. Let Q I' Q be two loadin gs of K, and a I' IT 2 two
admis sible
fields associa ted with them. Let Q be given hy Q = ..1.Q I + (I ~ ..1.)Q2'..1
. E [0, I J.
From the definition of loadin g param eters (Chap ter I II, Section 5.2),
the stressfield
is S.A. associ ated with Q.

Also, from the convexity of J, the stress at any poinl is limited by

J((f) = J(..1.(f, + (I - ..1.)(f2) ~ 0




J.3 Points of K at infinity

TIII'.()IU~M: For materials h:lving, a convex loading. surface 'open' only in the
direct ion of isotro pic pressures (see Chapt er I, Section 2.7), and with
zero hody
forces applie d to the system. K has a point at infinity in only one
direct ion, at

From the definition of the loadin g param eters (Chap ter III, Sectio
n 5.2)an d
the fact that IT = 0 is plastically admis sihle- the materi al being non-h
arden ing
the load Q = 0 E K.
It is assum ed that K does have points at infinity, and (I..) denote s the
direct ion
of one of these points . Since 0 E K and K is convex, the whole semi-a
xis OL E K.


- --- -- - - - - - - -- - --.--1
Lei Q* be prescribed E Ot.
To say thai OJ. E K up 10 inrinilY implies thai, oul of all I he admissihlc slrcss
fields associaled wit h Q*, thcre is (al least) one which remains admissible if it is
multiplicd by }.(positive),howcver great it may bc. Thus, this lield must consist
of an isotropic pressure at any point of the solid.
If there are no body forces this pressure is constant, and Q* is a loading which
corresponds to a uniform isotropic pressurc throughout thc solid. If it exists
then it is a loading proportional to one p;lrameter, and proves the foreseen
result. 2
In the case of ductile materials, for which the loading surface is open in the
directions of thc isotropic pressures and tensions. K call have a point at infinily
in two opposile dircctions.
3.4 Limit loadings

By definition, a limil loading Q is a loading for which a solution exists to the

problem of uncontained plastic now of thc associated rigid-pbst ic system. TIlliS,
there exists an admissible stress-lield a associated with Q, and also an admissible
strain rate field v to which the strain rate vcctor of the system (j corresponds.
(Also, as v #- 0, q f.:O.) The flclds (1 and I) are associ<lted by the now rule.
3.5 Theorem of maximum work

It is assumed that the mat.erial constituting the system obeys the principle of
maximum plastic work .
TIIEOltEM: Let Q be a limit loading and it the corrcsponding strain rate of the
system: Let Q* bc an admissible loading; we hav~:
(Q - Q*).


( 13)

As a and a* are the admissible fields associated with Q and Q* and v denotes
an admissihle field associated with q. by definition.
(Q - Q*)



(0" - 0"*)


which is positive. from the ass\lmption concerning the principle of maximum

plastic work.

II may be that the loading proeess under study admits 'no unirorm pressure field in the solid as
being statically admissihle.

The result is true in the case or non-7.ero constant hody rorces. For body rorces variable as runct ions
of III parameters, the directions or Ihe points at infinity. ir they el\ist, are inside the pyramid built
upon the corresponding posit ivc semi-al\es.



3.6 COllsequences

UnJer the hypothesis of maximum

belongs to the boundary or K.


wo rk, any limit loaJing

Let Q be a limit loading point and q the corresponding velocity vector. It is
known that Q E K.
I r it is assumed tbat Q is inside K, then 3" > 0, so that 13(Q, '1) c K. [B(Q, tJ)
= sphere with a centre Q and a radius ".J There can be found in 13, and hence
in K, a point Q* such that (Q - Q*) <i < 0, which is obviously abslircl . There
fore, Q belongs to the bou1l(/ary 4 K.
Also, the theorem of maximum work indicates that at the point of limit
loading Q, ci is orient(lled (I/ollg (III outward norl/lo/to the yield bound(lry. which,
aceortling to Section 3.2, is a convex surface.
3.7 COJlverse of the preceding theorem

Under the hypothesis of maximum plastic work, let Q be a point on the

boundary of K (a convex surface). I\s shown in Section 3.3,0 E K. For a pro
portionalloading process (i.e. a radius issuing from 0) defined by the direction
OQ, there exists a limit loading point 1 situated on the boundary, according to
the theorem of Section 3.(,. This point must be Q. as a consequence of Ihe
convexity. Thus, any loading belonging to the boundary of K is a limit loaJing.
3.8 Conclusions. Gvozdcv's theorem


. .

Under the hypothesis of the principle of maximum plastic work, the boundary
of the convex K is the yicltl boundary of the associaled rir/ill-p/aslic syst.cIII and
it also is a conlJ('x sUI/ace.
This result is I be so-called 'theorem of uniqueness of the limit loads' and
implies, amongst other lhings, that a loading situated on the bOllntlary of K is
the limil loading of any loading path that arrives there. 2 Concerning the elasto
plastic and the rigid-plastic systems (the Ialter being defined by passing to the
limit in each case), the preceding theorem gives the following results.

Here an :\ssumptioll is made which can be l:onsidered as a fact or experience. For the eX[1erimenl
or proporl iOlla I loading or an clasill-plastic syslem (under the assllmpl ion or small ddorllla t ions).
accounl being laken Ih;\t the limit 10adillg.~ of this system arc limit loadings ror the associated
rigid-plastic syslem, it is evident lhal either:

(I). It is possible 10 go on loadi;\g endlessly, which implies lhat lhe set K has;\ point at inrinity
in the loading direelion, or
(2). I t is not possible to go on loading endlessly. and it must be admilled that this always h;\ppens
as a consequence of the appearance or uncontained pl astic now; i.e. there exists a limit loading.

When dynamic data arc fixecl at their prescribed values. a yield houndary is obt;\ined for the other
loading parameters which is the intersection of the boundary of K with a (n-p) hyperplane of R".
where p is the number of fixed 10ilding pilrameters. Hence the properties or this yield boundary
arc known .

1J 0

. " ... . .- - . /..........'-,-.

. .......


.,..... ~




... -- ".,... ......""""'- ......


..- . . . . ....-,....... ... ,


(I). From the viewpoint of limit loadings, there is no longer any reason for
distinguishing between tile rigid-plastie systems, as they are all equivalent to
the associated rigid-plastic system.
(2). The limit loadings are, as a consequence, independent of the clastic
properties of the in itial rna terial.
(3). The limit loadings being independent of thcloading paths, it is, therefore,
unnecessary to specify the initial stress state at the beginning of a loading
The following theorem forms the basis of the under-approximation of the
limit loading.
GVOZDEY'S THEOREM: A loadillg which call bc cquilihratcd by all allowahle

stress-jield is bcyolld or

thc yield bOllndary.1

In the case of a proportional loading, A = 0, being allowable, the statement

'Any value of A, such that an allowable stress field can be associated with it,
is smaller than or equal to the limit loading'.

Al is thc great cst vaillc of Athat an allolVahle strcss-jield call bc associated with.
3.9 Static method
Taking Gvozdev's theorem as a basis. together with the convexity of K, gives

a static method for the determination of the limit loads and the yield boundary.
The construction of allowable stress-fields permits the observation of allow
able loadings; and the largest convex polygon with those loadings as summits
is an inner approximation of thc yield boulldary (Figure V.2).

- - unknown boundary of

OI .

approllimalion from inside

(a )
known admissible


( b)

Figure V.2
I This

theorem is more general than that given by Gv07.tlev [24J for the OInalysisofstructures, taking
as a basis the model due to Kazinczy [33J and Kist [34] (see [56]). Thc present statement, linked
to the principle of maximum plastic work, seems to be due to Hill [26], in the form of a maximum




- - - - - --

-. -~

4. Kincmatic Approach
L I Th.:orcm

Under the hypothesis of maximum plastic work, for a given permissihle

strain rate field u (ha ving a corresponding strain rate of the system (j), and a
permissible loading of the system Q*, the power of Q* ill q cannot exceed the

dissipation within the system oj the st,rain-ra/ejicld u.




be an allowable stress field associated with Q~. Then, by definition,

Q*q =

cr*v d V.

( 14)

From the principle of maximum work, at each point of the system

,,.... ..

<T*V ~ <TV


= n(v)

where cr is a stress tensor resulting from the inversion of the now rule for
Hence, the result,


n(v) d V


( J 0)

4.2 Dual dcfinition of K

For all the allowable strain rate fields, let K. denote a convex surface of {Q}
space, the intersection of the half-spaces defined by


n(v) d V

(J 7)

From Section 4.1 K c K. , ConsiJcr Q* K and let Q bc the point of K's

boulldary (convex) situatcd on OQ* . Q is a limit loading, according to Section
3.7, to which a permissible strain rate field u and a permissible strain rate of the
'system q(u) =1= 0 correspond. According to the convexity of K
Q*q(u) > Qq(u) =
so that Q* rt K .
It is possible to derive lhat K '"'" K
cllvelope of the planes given by
Qq(v) =


n(v) d V

and the yield boundary is the over;!ll

n(v) d V


corresponding to all the allowable velocity fields .


_ _ _ _ _ _ _ _ _ _ _ . _~_ _ _ _ _ O_

_ _

0_ ,

_ -_ _ _ _ _ ___

If WI (/l/o\\'(/hlC' stl"(/ill r(ltC'/ielclfor the syst('1/1 is

Q(ltv) =



O __ _ ~ _ _ O _ _ _ _ _ _ _
-_ _ _ _ _ _ _

th('1/ the ('(II/alit .I'

( 18)

n(v) d V

determilles (/ plalle which is elltirely ('xterior, or tallyelltialto the yield hOlllul(/ry

oj the system.
For a plane which is a tangent to the boundary surface, the point of contact
is a lilliitloading. corresponding to the strain rate (i which arrears in (IR).
In the case of a proportional loading, the vallie of -) obtained by balancing
the work done by the exterior forces and the dissipation in any admissible
velocity field is greater than or equal to the limit load:

L, n(,,) d JI

J. = '-- - -Of----.U


Here, )'1 is the slIlallest vall/e oJ). sl/ch that the work dOlle by the exteriorJorces
is eql/alto the dissipatioll ill at least aile allo\\'ahlC' velocityfie'ld.
4.3 Kinematic approach
As in the static approach, the preceding results are used for the killematic
method of determining the limit loadings and the yield bOIlJldary. !\llow<lble
velocity fields are constructed and the convex polygon limitillg the intersection
of the half-spaces defined by (17) is an exterior approximatioll of the yield
hOUlul(/ry (Figure Y.3).
unknown yield boundary


approximation from



o lIm


Figure V . .l

5. Ih'marks


Ihe neslllis

or (hc Thcor)' or Limi(



From the logical viewpoint. the results of Section 3.6 prove that the yield
boundary is either the whole or part of the K boundary, which is fundamental



0 __

to the static method. Similarly, the results of Section 4.1 contain the basis of
the kinematic method. However, it seems desirable to invoke the existence
hypothesis in order lO arrive at simpler and more 'natural' statements.

The results of the theory of limit loads, also termed limit analysis, can be
expressed in the following form.

0). If a loading is such that it is possible to determine an allowable stress

field which produces equilibrium, then the structure will resist this load.
(2) . If an allowable deformation mechanism and a loading sufficient to cause
collapse are known, then the structure will not resist this loading.
These results may appear intuitive, but as seen previously their proof requires
tha t the principle of maximum plastic work be verified. I It should be noted that
counterexamples of the theorems of Sections 3.8 and 4.2 have been given In

In the case where the principle of maximum plastic work is not verified,
whether because of the constitutive materials of the structure or because of the
interface conditions between the different constitutive solids (e.g. Coulomb
friction), the yield boundary can no longer be proved to be of (11 - I) dimen
sions. Stated otherwise, the 'theorem of uniqueness of the limit loads' is no
longer applicable.
The independence, with respect to the initial stresses, the loading path and
the elastic properties, of the limit loadings of the elasto-plastic system is no
longer ensured. Also, it can no longer he stated that all the definitions of the
rigid-plastic material are equivalent with respect to the limit loadings.
Non-standard systems have been studied extensively and Appendix 1\ gives
the results [5, 15.31,32,42,44---46,50,52,55]'

6. Minimum Principles
Using the principle of maximum plastic work, the classical minimum
principles are given. They are valid for the case where the boundary conditions
may be expressed in the classical form:
T given = Td 011 S .. ;
F given
u given = ud on SII

The principle must be verified also at the interfaces of the solids that constitute the system.

_ 114

_ _ _


6.1 Minimum principle for stresses (Hill)

Let a be an allowable stress-field. to be used in the functional
.Jf'(a) = -

T(q) . u'dS.


Then, i/o soll/tioll oJrlle IlI/collt(/illCd plastic:.flolV prohlel/l eXists/or the give/l dottl.
the accompanying strcss-fleld rcnders the fl/Ilcriol/al .)'(' a mil/imum amol/{j all the
allowable srrcssJields.
6.2 Minimum principle for the strain rates (Marko\').

With v being an allowable strain rate field. derived rrom the velocity field
fUllclional is defined by
.'M(u) =


[n:(\') - IIF '

IIJ d V .-


r' 1\ dS r




Then, ({ a sollltion oj 'he /llIcolltained .flo\\' prohlelll exists Jor these dala, rite
aCCOIIIJl(/II),in{j strain ratc field rellders tlte/illlct iOIl(/llJ a minimum (lIllOI/{j (lllllte
ollmv(/!Jle srraill r(/re (iC'!rls.
The proof of these two principles is simple as they are supported by the
theorem of virtual work and the principle of maximulll plastic work. I t can also
be proved that for a solution of the uncontained now problem
!Ja(ll) = - .1('(a)

(21 )

7. Limit Analysis in the Study of Plane Strain Prohlems or Ullcontaillcd

Plastic now

In the study of plane strain problems of uncontained plastic now the exact
significance of the solutions ohtained was left aside. In particular, the fact that
attenlion Was paid only exceptionally to the undeformed zones is important.
It is now possible to deal with these points, assuming the principle of maxi
mum plastic work to be valid. by means of the theory of limit analysis. Also, the
uniqueness of the stress-fields proved in Chapter TTl (Section 6) makes it
possible to derive further conclusions.
7.1 An example study

The example considered is the passive pressure of a wall on a wedge of

Tresca material, as studied in Chapter IV. Two loading parameters are the
components given by reducing the external forces acting 011 the rigid wall so as
to act at a'fixed point (e.g. the middle of the wall).

(normal force-the tangen tial force is zero, as the wall is smooth)


- - -- - - - - - --.J.j


1\ third pammeter, <!1' which is maintained constant. is

by the vallie or
slir/';Ice load OJl ())' (I;igure VA).
This problem l1;tS hcen partially solved to give allowable strain rate and
stress-fields . The strain rate is known ror the whole solid: "hove AfJCf) Ii f n,
and under ABCD IJ = 0, with a velocity discontinuity along ABCD. A stress
field (J (in limit equilibrium) is known above ABCD. At each point or the
dcrormed zone <T and yare associated by the now rule.
I he



Figure V.4

The rollowing comments may be made about such a solution.

(I). It is not complete as the determinatioll or the slresses under AllCD has
not been dealt with .
(2). It is possible to use the kinematic method, as an allowable strain rate
ficld has been provided for the system. The calculation of the dissipation
integral is simpliried by the knowledge or the allowable field (J associated with
p in the deformed zone.


The work done by <T in v is equal to the dissipation n(v).

l3y applying the theorem or virtual work I it is shown tllat the integral or the
dissipation (including that along the velocity discontinuity line) is equal to the
work done by the pressure Q 3 and the stresses under the rigid wall corres
ponding to (J .
. By applying (18), the equation or the corresponding plane which is tangential
or exterior to the yield boundary is obtained. It appears as an upper bound ror
the moment, with respect to I, or the rorces applied to the plate.
(3). IL is possible to show [4H] tha t ir the value or the surrace load Q.I is not
too great, then an allowable extension of the stress field into the zone situated
beneath ABCD may be found .
Under these conditions, the static method can be applied, as an allowable
stress-field (J is known . The corresponding loading is also allowable, and its
representative point is not exterior to the yield boundary.

Thi~ application is po~~ible ror the whole ~tructure. although the ~tre~~field in lhe undcformed

zone remains unknown, a:; the equilibrium or the boundary rorces is salisfied here.




lly combining thcse rcsults. it may bc secn that

(I). An upper bound has been obtained ror the moment. with rcspcct to J, or

the rorces applied to 011:

AI, = M -I- N((/

h):::; 2h[QJ -I- 2"(1 -I- (1))]((/ -I- II)


(2). frQJ is not too great a limit loading has been ohl:lined corresponding
the rolation of Ihe plale around the point I:

Al = 0


N = 211[QJ


+ Ctl)]


Forthecasewherew = n/2andQJ = O,lheproblem becomes Iheindentation

of a half-plane without surrace load. Thrce dirrerent cxtensions or the stress
field arc known in this case [3, 57. 5RJ. The limit force or indentation bccomcs


211 x (n -I- 2)"


In the case wherc QJ is not too great, and where it is possible to cxtend thc
stress-field, a solution or the problem or uncontained plastic now is obtained.
Then thc thcorem of uniquencss ror the strcss-field, as given in Chapter III
(Section 6). can be applied.
The stress-rield in the dcrormec\ zone is shown to be uniqlle ror the givcn value
or Q. i.c. (23), or the given value or cj. i.e.

eil = n

ei 2 = n((/ + h)

L'(ei ' = 2h (i 2 = 2hn((I

-1- h)] .

Obviously, the results have more practical interest in the latter form, which
implics that the stress distribution in 011 /JCD, and in particular undcr 011. is
the exact distribution. Thus. the exact distribution or stresses bencath 0;1 is ,I
unirorm pressure cqualto [QJ + 2k(f + (I))j.
7.2 General case

For plane strain problems, assuming the principle or maximum plastic work.
the rollowing general statements apply,
(f). If the proposed solution provides only a stress-rield in limit equilibrium,
in only a par! or the .~ystem, then no conclusion can be rcached concerning the
result obtaincd.
(2). Ir thc proposed solution provides a strcss-field in limit equilibrium in a
part D or the system. such that the static equilbrium or the boundary data for
the remainder or the system is ensured, and :til allowable velocity field is
associated with the stress-field in D,leaving underorlllcd the exterior of D, then
the solution is said to bc ;1l(:OII/I'/c'{(' [3]. This corresponds to the application or
the kinematic method, as demonstrated in the example.

The result is an upper bound of 1.he limit load for the case where there is only






one loading parameter, or more generally, a hyperplane exterior to the yield

(3). If the solution provides a stress-field in limit equilibrium in a plastic zone,
and an allowable extension of this stress-field outside this zone, then it is a Sl(/lic
solution, making possible the application of the static method.
The result obtained is a lower bound of the limit load or, more generally, an
allowable loading represented by a point internal to the yield houndary.

(4). If the proposed solution corresponds to both (2) and (3), the solution is
colllpletc, and provides the limit load, or more generally, the limit loading, and

an associa ted deforma t ion mecha nism.


In fact, only complete solutions warrant the name solution. However, usc or
the above terminology, due to Bishop, has the advantage of being very explicit.
For complete solutions, the theorem of uniqueness of the stress-field is
applicable. and demonstrates that the stress distribution in the deformed zones
is the exact. distribution for the given problem. This docs not entail the unique
ness of the complete solutions. e.g. for the same incomplete solution several
allowable extensions of the stress-field may be found to complete it.
The construction of the allowa ble ex tension of t he stress-field of a n incomplete
solution is of no interest in itself. What is important is its feasibility. It is sufficient
to have a theorem which proves that the extension is possible. At first, it might
seem tha t such a theorem would be simple, since the stress-fields of the ex tension
have only to verify the equilibrium equations and fit the stress boundary
conditions, whilst not viol;lting the yield criterion. However, a theorem docs not
exist, and it is usually necessary to actually construct an allowable extension.
In fact, theorems arc available which sometimes show that the extension is
certainly not possible, e.g. Hill's theorem [28J, and Bonneau's theorem (see
Aprendix B).
Following the work of Dishop [3J and Shield [58J, research interest has in
latter years been devoted to the static approach. A number of non-trivia I static
solutions and complete solutions have been demonstrated, though no general
method for their construction exists [1,9-11,18,25,47,48,53.54,59]'
. Finally, although being obviolls, it must he stressed that there is no unique
ness of incomplete solutions. These can give quite different results; and (2)
indicates that the one that gives the best upper bound of the limit load (i.e. the
lowest one) must be taken into consideration, but should not be thought to
give the actuill limit load. Also, it docs not necessarily follow that the solution
can be completed .
Incomplete solutions constitute most of the 'solutions' proposed for plane
now problems with a Tresca material.

7.3 Conclusions
Limit analysis acts as a guide in the research for 'solutions' of problems of


. "' -- .... .-... .- - - ... ,_..0...-_ . .--__

Ie .. '

... . ~_ . _ _ .... .... .... ___

' . _' 0 _

. . .. . .

uncontained (and particularly plane) now for rigid-plastic materials. It is also

the only method of choosing between several incomplete solutions'.
The mode of construction of the 'solutions', which may at first ha ve appeared
intuitive, and the distiilction operated in Chapter III (Section 4) between a-type
and b-type zones can now be better understood. Usually, the construction of an
incomplete solution is planned by firstly forming an appropriate deformable
(a-type) zone, and then ensuring the overall equilibrium of each /Hype zone.
Such a 'solution' is interpreted completely by the theory of limit analysis, as
already stated. One may try then to complete the solution by an allowable
extension of the stress-field into the h-type zones.
It should be noted that problems do exist where purely static or kinell/otic
'solutions' are developed . An example of those will be giveti in Section 8. This
last type of solution, which has not been presented ill Section 7.2, sometimes
utilizes the theory of plane limit equilibrium and provides only an allowable
mechanism of deformation, which may be interpreted by means of the theorems
of Section 4.2.

8. Interpretation of the Solutions for thc Case of a CouIomh Matcrial

Soils obeying tile Coulomb criterion do not satisfy the principle of maximum
plastic work . The theory of limit analysis in its classical form is, therefore, not
The uniqueness theorem of the limit loadings is no longer applicable, as the
set of limit loadings of the associated rigid-plastic system is no longer identical
to the boundary of the set of allowable loadings.
Thus it may be possible for the same loading to be a limit loading for the
elasto-plastic system along one loading path, but not along another. Similarly,
it is possihle that the same loading along a single loading path, mayor m:ly not
be a limit loading for different elastic properties of the material. Finally, there
is obviously no longer an equivalence between all the definitions of a rigid
plastic material.
With regard to the different types of solutions to the problems of plane
deformation, and their significance, the analysis of. Section 7 is no longer
applicable. In most cases, the proposed solutions are analogous to those of
Section 22 in Chapter IV, giving a stress-field in limit equilibrium in a part of
the structure under investigation. They belong to type (1) of Section 7.2 and no
conclusions can be made aboul the results obtained.
A position thn t may be adopted [38J for the interpretation of solutions in
the case of a non-standard Coulomb material is to refer to the corresponding
solutions to the same problems in the case of a standard Tresca material. For
thde, the interpretation can be supported by the theory of limit analysis. It is
assumed that the interpretation for a Coulomb material is similar, without
specifying the flow fIIle, as has been suggested previously in the construction of


- - - --

The results obtained should he considered as approximate but. as often

shown by experience. can he l/.';cful sources of information.
8.1 A Standard Coulomb material

The study of the case ofa non-standard material with a convex yicld criterion.
presented in Appcndix A of Chapter I V (also [55J). suggcsts ideas of great
Ir/is thc COIlVCX yield crilt:rion or the material (not necessarily homogeneolls).
thcn the allowahle loadings arc defined as in Section 3.1. and constitute a
convex scl K I'" For any system wit h a criterionI, reg:\Hlless or its clastic proper
ties or now rule. it is known that the loads which the system can hear and the
limit loadings for all the loading paths necessarily lie within Kr Only in the
case of a standard system is it possible to state that the system can bear all the
loads within K/. and Ihal the boundary of Kr is the loclls of the limit loadings.
For the determination of K I' and its boundary. which reprcsent 'the maximum
the system can do'. all the available methods will be used: the static method
founded on thc convexity, and the kinematic method. assuming the material to
be standard. which premise has no other significance than that of being a means
or calculation. Following this reasoning. the solutions of plane problems will
be interpreted assuming a standard material.
If a static extension of the stress-field is known, a static solution for the
standard Coulomb material is obtained, and hence. an inner approximation to
the boundary of Kp. If a velocity field can be associated with the stress field, an
incomplete solution is found for the standard Coulomb material. and hence. an
outcr approximation to the houndary of K,... Finally, a complete solution for
the standard Coulomb material provides a point on the boundary of K I' and
an associatcd tangent plane.
8.2 Gcncral remarks

Although not proved. it is thought that the method of interpretation pro

posed ill Section R. I by reference to a standard Trcsc" material is in fact e(jlliva
1~1l1 In that indicated previollsly. i.e. assllming that a COlilomb material is
As indicated in Chapter IV. Section 23.4, the proof of the method of super
position considers static solutions ror a Coulomb material. and is supported
by the 'static theorem' and thc static method. Thcrcf()re, it implicitly assllmcs
Ih;It a COlllomb material is stand;Irtl.
The allowable extensions or the stress fields uscd in the example have been
givell in [5]. 5X] for the case ofa weightless material with cohesion and surface
load. and in [10] for the c;Ise of a 'material with weight and without cohesion or
surface load'.
It is possible 10 refcr to a kinematic method for a non-standard material [55J.
since the construction of kinematic solutions for a slandard material provides


(In outside approximation to all the [Jossib\c limit loadings for a non-standard
material. Ilowever. it would not he reasonable t{) speak of a static method for
a non-standard material. especially relating to st,ltic solutions constructed ror
a standard Coulomb material with a criterionf
8.3 nadenkovic's theorem

A theorem origina ted by Radenkovic [44, 45J. and :l1so proposed under
va rious forms ill [5. J I. J2, 42. 46. 52. 55J III a kcs j t possible t () refer to sta t ic
methods in the case of nOll-standard materials whose now rilles have p:lrlicul:tr
properties [55]. This leads to all inner approximation of the of possible
lim it; load ings.
The detailed proof of this theorem is given in Appendix A of the present
chapter and in reference [55]. It supposes that the now rule of the non-stClndClrd
material derives from a plastic potcntial {I which is COllvex, alld h:ls ccrtain
properties with respect to the criterion! It is proved that the limilloadings ror
the non-standard material cannot lie within the set K(; of allowable loadings
for the standard system constituted by a (fictitious) material with:1 criterion fl.
Thus, the boundary or K(; provides an internal boundary for the zone of
possible limit loadings for the Ilon-standard material. and hence, by means of
the st:ltic method for a system made of the standard m:lterial with criterion {I,
it gives an internal approximation to all possible limit loadings for the non
standard material.
This theorem is applicable in partiCUlar to a Coulomb material which has a
flow rule defined by an angle of dilation 1', where 0 :oS; I' :oS; 4) (sec Chapter 1).
The potential g is then a Coulomb potential, with an angle 1', which must be
internal to the criterion/in the region where it is used.
Unfortunately. the restllts obtained by means or this theorem are two con
servative to be of practical usc. particularly in the casc of I' = O.
8.4 The concept of a possible solution

When considering a Coulomb material with an angle of internal friction 1>

and a dilation angle 1'. it is sometimes possible to construct a static solution
with which a velocity field is associated via the now rule of a non-standard
material. A complete solution for the non-standard malerial is thell available.
It is said in this Case tilat a /llJssih/e soll/tio/l for the non-standard material is
obtained, with the corresponding loading heing a possihle limit loading.
The example de;t1t with in Appendix A of Chapter IV (Section (1) is a COIll
plete solution for a standard Coulomb material/[U. 14.53, 5XJ: whence the
interpretation given in Section R.2. However, if v :f-. 4), there does not exist any
velocity field associated with the stress-field. 1 (The condition A ~ 0 is not
satisfied I J. 14].) An example ofa possible solution for a non-standard material,
for at least some values of the parameters. is to be found ill [9]'

This docs nol proye Ihal Ihe corresponding loading is nol a J1(lssihlc limil loading

fllr l'

of </"


... ...

-. - ......

- .:..-~

...... ._....


9. Other Applications of Limit Analysis in Soil Mechanics

The use of the theory of limit analysis in Soil Mechanics is not exclusively
linked to problems of plane limit equilibrium such as those dealt with in
Chapter IV. The theorems of both static and kinematic methods are variously
applied in engineering practice.
9. J Example
An an example, the determination of the maximum height of a vertical bank,
consisting of a Tresca material with self-weight, is considered.
(a) U t ilizal ion of lhe .'Italic method
The static method is used as follows .
Any height h such that' an allowable stress-field in the whole medium can be
fOllnd is less than or equal to the limil height h l' The allowable stress-field must
satisfy the equilibrium equations,
" '"

aO" + -.---.:Y.
a. + V =
oy .


~~ + oO"y



the 'plastically admissible' condition,

IfTl - O"zl ~ 2k
and the stress boundary conditions,


=0y >

= h y < 0:

0 : fT .<
0" X




O<x<lty=O : O" Y



Conditions at infinity apply only to the velocities, which must be equal to zero.
The discontinuolls stress-field proposed in [17J is represented in Figure V.S.



----l'x i-..- --

t .(Lh)

Figure V.S


.. ..: .:::1 ....

It is defined by
y ~ 0 0 ~ x ~ h: a l = 0,
y~0h~ x
: a~. = - y(x - h).
Y~ h ~ x
: a). = - r(.x - h),

t x )'

= 0, a.

t .\")'

= 0'


= -yx
0, ax = -yx

a .\"


This field satisfies the equilibrium equations, the boundary conditions, and
the continuity of the stress applied along y = 0, which is the line of discontinuity
of the field. It is, therefore. statically admissible. It is also plastically admissible if





(h) Ulilizalion o/thc kinclIlatic method

The kinematic mel hod is applied in the following way.

A height for which an allowable mode of deformation exists, where the power
of the exterior forces (gravity) balances the dissipation, is greater than or eCJual
to the limit height of the bank .
An allowable velocity rield must be sllch tilat
(I). The deformation occurs without volume change,
(2) . The boundary condilions for the velocity arc satisfied, i.e. Ii.~


= 0 at

For the mechanism shown in Figure V.G, a rigid body slides along an isolated,
(necessarily) circular line with a centre 0 and a radius h.

Figure V.6

This constitutes an allowable mode of deformation, and dissipation of the shear

strength k occurs along the circle (rriction), being equal to

k x D.II x Te/'2h
The work done by external (i .e. gravity) forces on the rotating block is equal to

1"'2 du

f: yn,.

(cos u)' d,.

yn II J /3


/.-~ ' ,

-- ~


Iknce . on balanc ing the two expres sions,


In k


= -' 2

giving an upper bound for the maxim um height .

Anoth er type of mecha nism is repres ented in Figure Y.7, where
sliding with
a veloci ty U occurs along an isolate d straigh t line passin
g throug h the foot of
the bank.

Figure Y.7

The dissip ation is equal to k x IJ/eos CJ. x U and the work done by the
extern al
forces is


ta n CJ. x }' x U cos CJ.

lIence , on balanc ing,


= ----



sin 2CJ.

giving an upper hound to the maxim um height .

This mecha nism can be optim ized; i.e. it is possih le to determ
ine a value of
CJ. which provid es the lowest upper hound . I n fact, CJ. = n/4 and It
= 41<fy gives
a beller upper bound for the limit height than the previo us
Gener ally speaki ng, by search ing for the best mecha nism
with ,In isolate d
line of sliding (circle ), the upper bound is reduce d to JXJ k/y.
Jossel in de longl
and I kyma n PSJ, by constr ucting an aJlowa ble stress- field,
have prove d that



?- 2 J 2 ,

A Is(), 1'" 1mcr2 has proved tha t

I knee


- '- ~"

~ J.HJ -

which appea rs to be the best bound s obtain ed by analyt ical

means for the

prohle m.

1%5, \lnpubli shed .

Private cOllllllunication .


The exact vallie of h I for a bank made of a Tresca material is :;till unknown.
For u bank made ofa Tre:;ca malerial with ten:;ion clIl-ofT(i.c. no len:;ile sIres:;
allowed) the critical height is hi = 2k/y (see [17J).
9.2 Coulomb's wedge, Fellenius' method

The kinematic method with an isolated rectilinear line or sliding, as used

above in the case of a Tresc\! material, has an equivalent in Soil Mechanics for
the case of a Coulomb material.
The so-called 'Coulomb wedge' method is in fact a kinematic method,
despite its static appearance. It lIses an isolated, rectilinear line or velocity
discontinuity for a standard Coulomb material [6]' I t should be noted that
along this line the velocity discontinuity is constant but not tangential (forming
an angle ), in accordance with Appendix A of Chapter IV (and also [17J).
Thus, Coulomb's wedge method, supplying an exterior approximation, will
lead to an under-estimation of the active pressure and an over-estimation of
the passive pressure in the case or a standard material, and a!orliori for a non
standard material.
Similarly, using the kinematic method with slip circles for a Coulomb
material, gives solutions involving all isolated line of velocity discontinuity
assuming the form or a logarithmic spiral. This is inclined at (n/2 + ) to the
vector radius coming from the rotational centre of the rigid block. The velocity
discontinuity i~ not tangential to the line of discontinuity (which must be
considered as a transition line in which dilatancy occur:;).
The slip circle methods (originated by Fellenius) arc orten used in Soil Mech
anics for the problems or slope stability. These methods cannot, ror rea:;ons
given above, be interpreted as kinematic methods for a standard Coulomb
material (the discontinuity velocity line not being a spiral), which explains
their shortcomings in practical applications.
9.3 Finite clement method

The theory of limit analysis makes it possible to lise the Cinite clement method,
either in the static method ror the construction of allowable stress rields [35, 36J,
or in the kinematic method for the construction of allowable velocity rields
[20,21 J. The author believes that the kinematic approach is to be preferred to
thestatic approach in problems involving large dimensions, e.g. stability oran
embankment. It orrers the po:;sibility or limiting. (/ priori, the di:;cretil.ation to
the regions known intuitively to be under greatest strain, the remainder of the
system having a rigid-body motion determined by the boundary conditions.
The reader should refer to [20J for an example of the thrust of a smooth wall
on a plastic infinite wedge ofTresca material. It is shown how the combination
of static and kinematic methods and of the various processes for the construc
tion of allowable fields, including the finite clement method, canlcad to remark
able results for the determination of the yield boundary of ~ system.

For the problem of the critical height of a bank I made of a Tresca material,
lIsing the finite clement method in the static approach yields 31 k/y as a lower
bound for hi L43]' Various attempts made with finite demcnts through the
kinematic method did not succeed in yielding a better upper bound for h I than
383 kfy.
Concerning three-dimensional continuous media, with the exccption of
axially symmetrical problems very few works are available dealing with the
determination of limit loadings [18]' The finite method element could be very
useful in this case.

[I] J. M. Alexander (1961) On complete solutions for frictionless extrusion in plane

strain, Quart. Appl. Math., 19, No. I, pp. 31-37.


[2] E. Anderheggen and H . Kn6pfel (1972) Finite element limit analysis using linear

programming, 11I1. J . Solids & Structures, 8, 12, pp. 1413-1431.

[3] J . F. W. Dishop (1953) On the complete solution to problems of dcrormations of a

plastic rigid material, J. M echo Sol., 2, No. I, pp. 43-53.

[4] S. J. Button (1953) The bearing capacity of footings on a two layer cohesive sub

soil, Proc. 3rd Inl. ConI Soil Mech. & Found. EIIOg., Zurich, I, pp. 332-335.

Collins (1969) The upper bound theorem for rigid-pl as tic solids, J. Mech.

l'hys. Sol., 17, No.5, pp. 323 - 338 .

[6] l. F. Collins (1973) A note on the interpretation of Coulomb's analysis of the thrust

on a rough retaining wall, Geotechnitlue, 23, No.3, pp. 442-447.

[7] J. Courbon (1971) Plasticite Appliquee em Calcul des Structures, E.N.P.C., Paris.
[8] M . Croc, G. Michel and J . Salen<;on (1971) Application de la programmation

mathcm a tique au calcul Ia rupture des structures, Jill. J . Sol. Strucl., 7, No. 10,


[9] E. H. Davis and J. R. Booker (1971) The bearing capacity of strip footings from the

stand-point of pla$ticity theory. Ulliv. Syciney, Civ. Ellg. Lah., Res. Rep. R 170.

[10] E. H. Davis and J . R. Dooker (1972) A note on a plasticity solution to the stability
of slopes in inhomogeneous claY$, Ge:otec/lIJiqul'. 22, No.3, pro 509-513.
LII] E. II. Davis and J. R. Booker (1973) The effect of increasing $hear strellgth on the
bearing capacity of clays. Geotechnique, 23, No . 4, pp. 551-563.
[12] L. Dietri c h and W. Szczepinski (1969) A note on complete soilltions for the plastic
bending of notched bars, J. Mech . Phys. Solids,!7, No.3, pp. 171-176.
[ 13] A. Drescher (1971) A note on rlane now of granular media. I'roh/e}II1(!s cle III Hheo
louie', SYlllp. Frwlco-l'%llais, V(lrsovic 1971, pp. 1)5-144.
[14] A. Drescher (1972) Some remarks on plane now of granular media, Archives (?f

Mecltallics, 24, No. 5....:6. pp. 837-848.

[15] D. C. Drucker (1954) Coulomb friction. plasticity and limit loads, J. Appl. M ech.,
21, rr. 71-74.
[16] D. C. Drucker (1956) On uniqueness in the theory of plasticity, Quart. Appl. M alit.
26, No. I, pp. 35-42.
[17] D . C. Drucker and W. Prager (1952) Soil mechanics and plastic ana lysis or limit
design, Quart. Appl. Math., 10, pp. 157-165.
[18] D : C. Drucker and R. T. Shield (1953) The application of limit analysis to punch

indentation problems, JI. IIppl. MecJ,., Trans. ASM E, 20, pp. 453-460.

[19] Y. d'Escatha and J. Mandel (1971) Profondeur critique d'cboulement d'un sOllter

rain, c.R. IIc . Sc., Paris, II, 273, pp. 470-473 . .

[5] I. F.

11 is to be remembered Ihat Ihe problem is considered as a plane strain problem .


----------------------------------------------------------------------------- - ..

.. ~ ... -, .

-.. -

[20J M. Fremond. A. Peeker Clnd J. Salenc;on (1974) Methode variationnelle pour Ie

matcriau rigidc-plastique, S)'II/fl. Fr(//I(:o-J'oirJllilis de J~/I(I()lo(/;(' ('/ M c1c(/II;I/U(' til'S
Sols, Niee (Fr.), July 1974.
[21] M. Fremond and J. Salenc;on (1973) Limit analysis by finite clement methods,
Symp. 011 the Role 0/ Plasticity ill Suil Mech., CO/llI,,;c/gc (G.B.), Sept. 1973, pp. 297
[22J 1. P. Giroud and Tran Vo Nhiem (1971) Force portante d'une fondation sur une
pente, Ann. I. T.E. T.P., No . 283-284, pp. 130-180.
[23J J. P. Giroud, Tran Vo Nhiem and J. P. Obin (1974) Tables pour Ie Ca/cul des FOII
datioTL~, t.3 ., Dunod. Paris.
[24J A. A. Gvozdev (1936) La determination de la charge de ruine pour Ics systcmcs
hyperstatiques s'ubissant des deformations plastiques, Proc. COllf 011 Plasc;c
De/ormatiolls, Ac. Sc. USSR, p. 19, v.z. transl. by R. M. Haythornthwaite in lilt. J.
J. M ech oSc., 1, (1960), p. 322.
[25J i Heyman (1973) The stability ofa vertical cut, lilt. J. Mech. Sc., 15, pp. 845-854.
. [26J R . Hill (1948) A varia tional principle of maximum plastic work in classical plasti
city, Quare. J. Mech. Appl. Math., 1, No. 18.
[27J R. Hill (195 \) On the state of stress in a plastic-rigid body at the yield point. Phil.
A.J 1.J{j., 42, p.. 868-875.
[28J R. Hill (1954) On the limit set by plastic yielding, J. Mech. Sol., 2, No.4, pp. 278-285.
[29] P . G. Hodge (1970) Limit analysis with multiple load parameters, lilt. J. Solids &
Structures, 6, pp. 661-675.
[30] D. D. I vlcv and R. 1. Nepershin (1973) I mpression of a smoot h spherica linden tel'
into a rigid-plastic half-plane, Jzv. AN. SSSR. Mekllallika Tverdo{/() Tela, 8. No . 4,
pp. 159-166, Engl. Tra nsi. in M echall;cs 0/ Solirls, 8, No.4, PI'. 144-149.
[31J G. de Josselin de JO!1g (1964) Lower bound collapse theorem and lack of normality
of strain-rate to yield surface for soils, Proc. I UTAM Symp. UII Rheolouy & Soil
Mechanics, Grenoble (Fr.), pp. 69-75.
de losselin de Jong (1973) A limit theorem for material with illternal friction,
Proc. S ymp. UII the Role 0/ Plost icit y ill Soil M cc:hall;C:s, Ca mbridge (G .13.), pp. 12-21.
[33] G. de Kazinczy (1914) EXfJeimeces sur les POlltres Ellcastrecs, l3etonszemle, Vol. 2,
p. 68.
[34] N. C. Kist (1917) Dissertation inaugurale. Ec.. Poly!. Delft.
[35] J. Lysmer (1970) Limit analysis of plane problems in soil mechanics. J. Soil M ecll &
Foulld. Dill. ASCE, 96, No . SM4, Prot. Paper 7416, July 1970. pp. 1311-1334.
[36] G. Maier, A. Zavelani-Rossi and D. Benedetti (1972) A /inite element approach to
optimal design of plastic structures in plane stress, III/. J. NUII/. Meth. Engll(J., 4.
pp. 455-473.
[37J J. M andcl (1966) !vi ecallic/uc des !vi i1ieux Cont inus, Tome 2. An n. 20, Gau thier
Villars, Paris.
[38J J. Mandcl (1969) Cours de Scicllce des Ma{(lri(Illx, Ecole Nationale Superieurc des
Mines de Paris.
[39J J. Mandel and 1. Salen~on (1972) Force portante d'un sol sur une assise rigide
(etude thcorique). Geotechnique, 22, No. I, pp. 79-93.
[40J Ch. M assonet and M. Save (1970) Calcul Plast;quc ties COllstructiOlls, Vol. I et 2,
2eme cd ., Ed. CBLIA, Druxelles.
[41] J. P. Obin (1972) Force portante en deformation plane c'un sol verticalemcllt nOll
homogene, Thcse Univ. Grenoble (Fr.).
[42] A. C. Palmer (1966) A limit theorem for materials with Ilon associated now laws,
J. M ecanique, 5, No.2, pp. 217-222.
[43] J. Pastor (1976) Application de I'analyse limite I'ctude de 101 stabilite des pentes et
des talus, These Univ. Grenoble (France).
[44] D . R<idenkovic (1961) Thcorcmes limites pour un materiau de Coulomb, C.R. Ac.
Sc. Paris, 252, pp. 4103-4 I04.



~r '

'. .

- - - -- - - - - -


[45J D. Radenkovie (1962) Theorie des Charges Limites, in Senrinaire de Plasticilc:,

Ed. J. Mandel, P.S.T., Min. Air N o . '1o, pr. 129-142.
[46] G. Sacchi and M. Save (l96S) A note on limit loads of non-standard materials,
Meccanica. 3, No . I, pp. 43-45.
[47] J. Salen<;on (1969) La thcorie des charges limites d:1I1S 101 resolution des problclllcs
de plasticite en deformation plane, These Dr. Sc. Paris.
[4H] J. Sa lenc;on (1972) Bu tee d 'une paroi Jisse sur un massif plastiq uc: so lutions sta t iq ucs,
JI. Mecanique, II, No. I, pp. /35-146.
[49] J. Salen<;on (1972) Un exemple de non validite de la theorie classique des charges
limites pour un systeme non -standard, Proc. Int. Symp. on Foundations or Pbsti
city, Warsaw 1972, North Holland Publishing Co., Amsterdam.
[50J J. Salen<;on (1972) Theorie des charges limites, Seminaire Plaslicile el Viscoplaslicile,
Eds. D . Radenkovie and J. Salen<;on, Ediscience. Paris, 1974, pp. 207-229.
[51] J. Salenyon (1972) Charge limite d'un systeme non-standard. Un contre-exempJe
pour la theorie dassique, Seminairc PIa.~licile cl Viscoplaslicile, Eds. D . Raden
kovic and J. Salen<;ori, Ediscience, Paris, 1974, pp. 427- 430.
[52] J. Salen<;on (1972) Ecoulement plastique libre et analyse limite pour les materiaux
standards et non-standards, A nnalcs I. T.B . T.P., No. 295-296, Ju Iy-August 1972,
[53] J. Salellyon (1973) Prolongement des champs de Prandtl dans Ie cas du materiau
de Coulomb, Archives of MeclJanics, 25, No.4, pp. 643-648.
[54] J. Salen<;on (1974) Quelques resultats theoriques concernant la butee d'une paroi
sur un coin plastique, Annales I.T.B.T.P., No.3 I3, January 1974, pp. 185-194.
[55] 1. Salenc;on (1974) Plasticitc pour la Mccanique des Sols, C.I.S .M. Rankine Session,
July 1974, Udine (It .).
[56J M. Save and Ch. Massonnet (1971) L'inOuence dc la plasticite et de la viscosite sur
la resistance etla deformation des constructions, Rapp. Inl. Congr. A .I.P.C .
[57] M. Sayir and M. Ziegler (1968) Zum Prandtlschen Stampelproblem, I ngC!l1icur
Archiu., 36, No.5, pp. 294-302.
[5H] R. T. Shield (1954) Plastic potential theory and Prandtl bearing capacity soilltion,
JI. Appl. Mcch., transl. ASME, 21, pp. 193-194.
[59] W. Szczepinski (1966) Indentation or a plastic block by two opposite narrow
punches, Bull. Ac. Pol. Sc., seric Sc. Tech., 14, No. 11-12, pp. 671-676.







1. Introduction
is not in
t of the
't analysis, but attempts,
adopting a more mathematical mode of
more clearly the
. will make it easier to
on to non-standard materials, and
has already been adopted m [12, 1

2. The

of a Standard Material

2.1 Theorem 1
f be 3. convex
which are all
of the convexity off,
directions of a convex cone.

{vlv E ;"of(cr), A

0, f(cr)

assumed to be
as a consequence


0, io-I < oo}


is a convex cone with a summit 0 of R6. This cone is the complementary of the
J(cr) = is
cone of directions in

Mechanical aspect:

a standard material with a loading function f, G' is

convex cone of
the plastically adllllssible strain rate tensors. Usually, if
'open' in
one direction of
isotropic pressures,
is a half-space,
is 'open' in the directions of isotropIc pressures
is a plane.
tensions, it is necessarily

2.2 Theorem 2. Dissipation

In all that follows, J is a convex loa~~ function, and it is .assumed that

J(O) :::; 0.
The bilinear form is considered:

where v is a strain-rate tensor, and () is a s~s tensor. The convex set.

G = {()IJ(cr) ~ O}


is used to look for


.. ...

;. :



Sup{F((), v) (J" E3 G}


(1). If v E G' then F(cr, v) has a finite maxim.urn, reached for

0': J(())

= 0, V E J.~n()), }. >



The value of this maximum will be denoted. by

7[(v) = Sup {F ((j, Y) lcr E G}

Since cr = 0 E G,
(2). Ifv G'
Sup F(cr, v) =

-r- OJ

In this case,

rr(v) = SUP{F(()l Y)I(T~G} =


Mechanical aspect:
(1). For a material with a loading function}: G is the convex domain of plasti

cally admissible stress tensors.

(2). For a standard material with a convex. loading functionJ, 7[(,,), in the case
where v E G', is the dissipation for the strain-rate tensor v. Equation (5) corres
ponds to the inversion of the flow rule for y.
2.3 Theorem 3. Convexity of


v 1,

G' are gt\'en.



}.V 1

(1 -

A) v 2 ,

). E



According to Theorem 1, v E G 1
Let () be a stress tensor associated with v by (5), so that
n(v) = cr '--v = Acr' v 1

(1 - J.) cr . v 2

_.. :.


. _'..


C"___ -/ . . ~, '.-:'__ .~ : " ' ,'

. ::. .'

the definitions of rc(v 1 )

n(v 2 )

() . v 1

() V 2

n( VI)

rc IS,

According to


rc(v) =

rc is convex on the





of all

positive homogeneous function.

2.4 Definitions
ote the
The symbols (J and v
stresste fields () and v
tensors, i.e. values of
a particular pomt)
It is not intended, in the following, to describe from a
in tegrals have their
. te functional spaces in which the
defini .
used is


10- E G; 31d E 92 p such that

: H forms a convex
is assumed to be a linear
G a convex
It is i


(1 -

(see Appendix
verified that, if (J' E

),)(J2 E

that since 1(0)

d defini tion is



0, H


(J- E



K.A. ass.



is a convex cone wi a summit o.

I t suffices. to state that
is a linear
and G' is a convex cone with a
summit O.
lv! echanical
H is the convex set

the allowable

It is understood that the rna terial constituting the

each point, suitable G and G' spaces will be

is Dot


homogeneous. At



,.: I.

admissible) for the system under study, subject to a loading process with n
H' is the cone of the allowable velocity fields (kinematically and plastically
admissible) for the system under study in the same conditions.

2.5 Functional I
A functional J is defined by


In(v)~v- Q(a)q(v) = 1 [n(v) -

. v



for any statically admissible stress field a and any kinematically admissible
strain rate field v.
This functional is non-negative on H x H', i.e.:


VVEH ' ,

we have

J(a, v)


This result follows from (7) and the definition of n(v).

2.6 Theorem
THEOREM: A solution to the problem of uncontained plastic flow for a system
of a standard material renders J(G, v) a minimum on H x H'.

Such a solution consists of:
(1). An allowable stress-field, a E H, and
(2). An allowable strain rate field, v E H'.
Both fields are associated by the constitutive law for a standard material; i.e.
if v #- 0, then f(cr) ' ~ 0 and v,= ),af(cr), ), > O. It follows that for both fields

I(G, v) =

[,,(v) -


v] d V = 0


Therefore, J is a minimum according to (8).

2.7 Converse
THEOREM: Any solution (a, v)to the problem
Min J(a, v)

(0", v)



such that v =I=- 0, is a solution of the problem of uncontained flow for the system
of a standard material.

1 ._

It i~ known that

Min l(a, v) = 0,

(a, v)



as the field v = 0 belongs to J-l'. Therefore, any other solution of the
tion proble
lea, v) = 0

of 71',
A t any poin t where v/;O it follows, accord ing to the definition



,{ > O.
that J(q) = 0 and v E J..8J(cr),
of a
A soluti on to the proble m of uncon tained plastic now for the system
standa rd material is, therefore, obtain
2.8 Other definit ions

Let the set K be defined by:

When a describes H, Q(a) describes K c W
It is immediately verified that K is a convex set.
Let the K' be defined by:
when v describes H', q(u) describes K' c W
It is immediately verified that K' is a convex cone with a summ it O.

M ec/zaJlical aspecc:
is the convex
K is the convex se( of (he allowable loadings of the system. K'
cone of the allowable strain rates of the system.
2.9 Coroll ary 1: theore m of maxim um work

standa rd
Let (a, v) be a ~olution to the problem of ullcontaincd now for a
Let a' E /-I, so that
l(a', v)

I(a, v)



n(y) d V - Q(a') q(v)


71'(v) d V - Q(a)q(v)

[Q(a) - Q(a') ]' <iCl]) ~ 0

of K and
As a consequence, the Q(a) limit loading belongs to the bound ary
q(v) is an outwa rd norma l to this bound

M echallic(I/ aspect:
The ineq\lality (II) exrres ses the lhcore m of maxim um work (Sectio
n V.3.5)
from which the static theore m and the static metho d may be derive
o tSectio ns
V.3.8 and 3.9.).1

2.10 Coroll ary 2: convex K I' concave L 1

2.10.1 Convex K,
The convex set K, in space {Q} = R" is deline d by

n. (Q Ir n(v) dV Jv


Q. q(v)

~ 0)


(t follows from Section 2.2 that the same set K I is obtain ed by consid
ering in (12)
all the K.A. strain- rate field~, wheth er or not they are P .A.
Consi dering now any Q' E K and a corres pondi ng allowa ble stress
(J' E H, it follows from (8) that:

Vv E H'

J((J', ti)

n(v) d V Q'q(ll)


There fore


Conse quentl y, given v E 11', the half-space

n(v) d V -- Qil(ll)

contai ns the set K


~~ 0

and, thcrefore, also the set K.

M cclw/1ica/ aspect:

The result obtain ed is the basis of the kinem atic theore m and the
kinem atic

metho d (Section V.4.3).'

2.10.2. CO/lC{lI JC L I
1..1 is conca ve since its compl ement ary set in space {Q} is eviden
tly convex.
Marco ver

CK , C L, c CK ,

K I and L, have the same bound ary. As n is a convex functio

n of v E Rtl, it
follows that Jy n(v) d V is a convex function or v K.A. This entails

n(v) d V Iv K.A., 4(v) =



Withou t making use or the existenc e hypothe sis as indicate d in

Section V.S.1.


is a convex function of cj E R". This function is the supporting function of K \'

meaning that any plane tangcntial to the K I (and L I ) houndary assumcs the

q E R".
2.1 t Properties of the limit loadings

If Q1 is a limit loading, then

(1). 1<.. is not open at point QI'

(2). Q 1 ELI' which is not opcn at this point,

(3). K and K I arc tangcntial in Q I

In fact Q I E K and, according to Section 2.9., Q 1 is on the boundary of K.

On the other hand, irv is a strain-rate field corresponding to this limit loading,
it follows (Section 2.6) that

n(v) dV - QI(j(t:) = 0

and, therefore,



It results rrom Section 2.10.2 that at this point L, cannot bc open and Q I
belongs to the LI (and K I) boundary.
Furthermore, the plane

n(v) d V - Qq(v)


passes through Ql' and is a tangent to K (since (Hv) is an outward normal to K).
I' by definition; therefore. it is a tangcnt to K I also.

It cannot intersect K

2.12 Existence hypothesis (of the solution to the problcm of unconlnincd

plastic flow)

The hypothesis assumes that

== K I (as a conscqucnce, K is c1oscd),

(2). Ll = CK 1 (stated otherwise, Ll is closed).

(1). K

Mechanical aspecl:
This hypothesis is equivalent to the following proposition, which has been
presented as the existence hypothesis in Section V.3.7.
For a proportional loading process given by

= ).Q*,


- .



either the loading call increase endlessly, i.e. ).Q* is allowable, V)' > 0, or there
exists a limit loading Q = ).Q* corresponding to the solution of a rrohlern of
uncontained plastic now.
Effectivc/y, either A.Q* is the direction of a point of K at infinity, or there
exists in this direction a point Q l on the common boundary of the closed sets
K, K I' and L , . In this case,

Q 1 E K => 3<1 1 E Ii:




Q 1 = Q(O:I)


71 (V , )dV

3v 1 EH',

- QJI(V I )


1= 0:

(giving properties of the planes which are tangential to the boundary of the
closed set L I)' or
1(<T1' VI) =0
so that (<T" vJ is a solution oftbc problem of uncontained plastic now. The proof
of the reciprocal property is derived from Section 2.6.
It follows from this result that any point at a finite distance, of the common
boundary of K, K I and Ll' is a limit loading.
The necessity of the existence hypothesis may appear merely mathematical
since, when stated in a mechanical form it seems physically self-evident. But
it is to be remembered that the concept of a rigid-plastic material is purely
mathemaLical by itself; an example has been recently given [17J of a problem
where a limit loading docs exist without uncontaincd plastic flow being possible.

Let Q* be a loading such that

).Q* E K,



Let K.A. ass. with J (' E C I' be such that

Q*(j(v) > 0

( 14)

Theil, eilher V rf: /1', or v E H', and the plastically deformed zone
has an infinite extent.'


this field


According to Section 2.10, K c K I' so that (J 2) is verified by ).Q*, and as v

is kinematically admissible, allowable or not, it follows that

Iv n(v) d

V -



VA. > 0


In practical problems, this circumstance will often be excluded by the very form of the boundary
conditions: generally, u = 0 at 00; while T = 0 al 00 is much rarer.






It(Y) d V



which proves the predicted result.

2.14 Consequences of the existence hypothesis; practical methods j theorems

2.14.1 Slalic melhod

The yield boundary is orten determined with the help of the ~latic method by
consid~ring a loading process where the vailles of Q j v~ry proportionally.
LeI Q" be a given loading. The uncontainecl plaslic now solutions corre~poncl
ing to limit loadings assuming the form Q = AQ" (possibly with the restriction
X> 0) are now sought. This requires solving the problem
Min {l(a, v) Iv E H'; a E 1-1, Q(a)

= AQ"}

(I G)

I<1 El-I. Q(a) ~ .l(a)Q'J Iu E I-I'}


and keeping only the solutions in which v =f- O.

Using the notation

= Sill (Q"(j(o))

expression (16) can be wri tlen as

M in {Iv n(v) d V - ,,,q(u)Q" Max {r.".l( a)
Solving the problem
Sup {1:"A(a)}.
separalely for the ca~es COl = ' + I, and COl = -I, gives two values A~ and A~
(resp. > 0, < 0).
The range for a in (17) is a convex set c 11, on which are found two values
A~I' and A~ (resp. >0, <0) corresponding to SliP {c){a)}according to the field v.
The loadings A~ Q d and A~ Q,/ are 011 K's boundary at the intersection with the
stn~ightline Q = AQ" (Figure V.A.I).

Figure V.A.!


As a conse quenc e of the exisle nce hypot hesis the loadin gs dealt
with are limit
loadin gs (if Lhey are at a linite distan ce). This means that they
corres pond to
fields u~ (resp. u~J of H which realize the Max term in expres
sion (17) (Min in
ain(1 6)).
The existe nce ofa soluti on v'+ f:: 0 of the proble m



(j(V)Q'A~ Iv E U', r., = + I}

n(v) d V -


is ascert ained. The soluti on of (18) will be useles s if, as is the

genera l case, only
the limit loadin gs are requir ed.
The practi cal static metho d involv es determ ining the limit lootiin
y hy solvillg

the problem

Max {).(o-) u E N, Q(<T)

).(u)Q", ).(u) >



Thus, an attem pt is made to invest igate the conve x set or( 19),
so as to maxi
mize). . K is explor ed, follow ing vector radii, and by dealin g with
a sub-se t of the
con vex set of (19), an approximation of the valid boundaryfro 111
inside is obtain ed.
2. 14.2 K il1ematic method
The kinem atic determ inatio n of the yield bound ary is effecte
d by setting
4(v) in (12).


Figure V.A.2

Let ('1" E K', q" ~ O. The soluti ons of the uncon t a ined now proble
m are now
sough t, such that the corres pondi ng strain rate of the system
(Figur e V.A.2)
be given by

Thus, the soluti on of the proble m

.M in {I(<T, v) u E H;


H', q(v) =



is a solution to the uncontained now problem if Min is zero. Using the notation


n(v) dV

equation (20) may be writt'en as

Min {P(v)lvr.H', (Hp) = it"} - Max {Q(a)i!"laEH}.


Here, v is obtained independently of a by solving the Min problem in (21) and

a independently of v by solving the Max problem.
The existence of Sup {Q(a)q" Ia E H} is ensured and the corresponding loading
Q I is the point of K's boundary where (i" is an outward normal. According to the
existence hypothesis this is a limit loading. Therefore, it corresponds to a stress
field a IE H which realizes the Max in (21); and there exists VI T- 0 which realizes
the Min in (21), I such that
Stated otherwise, the plane
passes by Q I and is a tangent to the yield boundary. The classical kinematic
methods is Sli b~equen tly cieri vetl. The Iw if-splice
Min {p(v)lv

Jr, q(v)

q"} - Qq" ? 0


contains K and is a (angelltlo the yield bOllndary.

The convex set of (23) is explored ~o as to minimize P(v). Thus, L I is explored

by means of half-spaces of fixed direction. In practice, a sub-set of the
convex set of(23) is used to obtain an approximaUolI oflhe yield boulltlaryfrolll

Thus, a static dual approach, also, is available.

The loading obtained by so/villg the prohlem

Max {Q(a)(/ Ia Eli},

is a limit loading for which





Finally, the condition (J" E K' stated at the outset prevents 'the convex ~ct of
(23) from being void, but it is sometimes difficull to know, a priori, whether
it" E K'. In fact, in (20) and also in the statements of lhe classic;). I kinematic
method and its dual approach, any iI" whatever in R n may be considered, both
latter problems being sta ted

Inf{P(v)lv K.A., q(v) = ('I'/} - Qi(' ? 0

Sup {Q(a)qdl a


H}, for any qd whatever, and the results will be unchanged as,

(1). If

q'd E K' ,

P(v) ==

+ 00

Effectively. cjl is an outward normal at Q I to the boundary or L .. which is closed.

, 139



and, thererore,
Inf {P(u)} = -I- co,=> Sup {Q(oYJ"} =

+ co

by application of the exi:>tcnce hypothesis, and there is no limit loading.

(2). Jr




Inf {P(v)} <

+ <Xl

(\ Min is reached for v E H' (as v rf- H' => P(v) = -I- <Xl) and again the initia I for
mulation results.
La:>tly, the absence of a solution can occur, even if <ii' E K, Inf{P(v)} = + 00
can be rcalized in this case, when all the field:> f) of Ir, such that <HI)) = qJ, have
an unlimited deformed zone (this agreeing with the results of Section 2.13).


2.14.3 Theorem oj association

When using the static and kinematic method, it is a:>sumed that firstly a
slress-field a* E I-I is found and Q* is made equal to Q(a*), and secondly a strain
rate field v E H' with q* equal to q(v*), such that

P(v*) - Q* 'q* = 0
The following can then be stated.
(I). Q* is a limit loading.

(2). <i* is a strain rate of the system corresponding to this loading.

(3). The fields a* and v* form a limit equilibrium solution corresponding to

Q*, <'\*.

The proof of this proposition is immediate, since I(a*, v*) = O.

The uniqueness theorem of the stress field (for certain assumptions, in the
\ll1ion of the deformed zones) presented in Chapter HI, develops as a conse
quence of the theorem of association (Figure V.A.3).

2.14.4 'f71C case oj a /oadiny process with one parameler

I n the frequent case of a loading process depending on one parameter Q (> 0),

Figure V.A.3


the statements corresponding to Scctions 2.14. 1 and 2.14.2 assumc simplc

classical rorms.
Static method:
Kinematic mcthod:

QI~m = Min {P(v)


H', q(v)

= I}

and its dual rormulation,


= Max {Q(a)



which is idcllticalto that orthc usual static method .

J. C:lse of NOll-standard Sys(clll.<;

3.1 Definition
The classical theory of limit analysis prcsentcd ill thc prcceding paragraph
explicitly uscs both properties of convexity and normality, cCJuivalent to thc
principle of maximum plastic work.
As already stated in Chapter I, the convexity of the yield criterion is a properly
with almost general validity. On the contrnry, the normality of the flow rule is
often non-existcnt. In particular, this is the case for soils obeying the Coulomb
criterion, and also, as will be seen in Section 4, in many cases for the condition
of contact at the interfaces between the solids which constitute a system.
Thus, the case dealt with will be that of systems with a convex yield criterion, 1
but with a flow rule which does not satisfy the condition or normality at any
point. These are 1/0/l-s[(/lId(/rd systems.
This type of system was first investigated by Drucker [3J with regard to the
conditions at the intcrfaces; then rollowed a general survey by Radenkovic
[7,8J and studies by Josselin de .long [4,5], Palmer [0], Collins [1, 2J and
Salen<;:on [II, J 2, I


3.2 A kinematic method

Thesymbol M denotes the constitutive material 2 of the non-standard material
(M). The yield criterion of M (assumed to he convex) is represcnted by.f. The
definition ofa plastically admissible stress tensor for M is obviously unchanged
with respect to the case ofa standard material:


and hence the set


{erlf(er) ~ O}

I The

expression 'yield criterion' is taken in it:; widest sense: it includes not only the yield criterion
of the constitutive materials, but also the friction condition at the interfaces between the solids
that constitute a system.

M includes the constitutive material itsclf and the con tact interface between solids or the system.


Similarly defined is lhe sel of allowable slress fields (or S.P.A.) for (M),

and the set of allowable loadings for (M),

K M = {Q(<l)


I-I M}

As a consequence of the convexity off, the sets G M' H M' Kh{ are convex, following
the same arguments as in Section 2. It results from the definition of KM that all
the loadings that can be borne by (M) are necessarily included, and, in particular,
the limit loadings for (M).
If F denotes a standard material with a yield criterion f, and (F) a system
geometrically similar to (M) and constitu ted by F, the sets GM' I-I M' K Mare
identical to spaces G, [-I, K, as defined for (F) in Section 2, and will be denoted by
GI" Up KJ" In particular,

KM = KJo'
Whereas for (F) the boundary of KJo' is the yield boundary, the same property
cannot be proved for (M), for the normality rule states that the standard system
(F) 'proceeds to the limit of its possibilities'.
The boundary of KM appears for (M) as the boundary oftolcrable loads, and,
in particular, as an outer boundary of limit loadings. Its determination is
interesting for this reason. As a consequence of the identity KM = Kp this
determination is carried out through the intermediary of the standard system (F),
(IS (ill auxiliary for lhe calculalio/1 by either
(J). The static method for (F), or
(2). The kinematic method for (F).

This supplies an approximation of the boundary of KM from the outside,

which. in itself is an outer boundary of the limit loadings for (M), and justifies
the tiLle of kinematic melhodfoJ' (M) (Figure V.AA).

~'"----- kinemolic method

for lFl

Figure V.AA

3.3 Radenkovic theorcm-a static mcthod

3.3.1 Type ojllol1-slCllu/ard systellls stlldied
The non-standard sys,tem (M) considered has a now rule with the following
properties at each point.
There exists a function g, convex on R Ci , such that
g(O) ~ 0
Vcr verifying j(cr)

3cr'(cr) verifying g(cr')



and .'

v /'( cr)

E ). iJg (cr'),

A~ 0


3.3.2 Remarks
(1). Standard systems are obviously a particular case of the non-standard
systems presented above, and for these f = fJ.
(2). A Coulomb material, with a flow rule defined by a constant dilatancy
angle v, satisfies the conditions of Section 3.3. 1. The value of g is given by the
Coulomb loading function with an angle v and a cohesive pressure equal to
that of f: cg/tan v = c/tan c/>.
(3). For a non-standard material satisfying the conditions of Section 3.:1.1,
there is no uniqueness of function g or of the correspondence



Thus, if g and cr -+ cr' satisfy the conditions, then {J", defined by {J",(CJ.) = {J(IIICJ.)Vex,
with'm > 1 and cr -+ cr" = cr'/III, also satisfy Lhem.
Similarly, some translations can be effected on cr' and on g, etc.
(4). It results from (25) and the convexity of f and {J that if G c is defined by
G c = {crl{J(cr) ~ O}


, GG


(5). In Chapter I the concept of a plastic potential that was differen t from the
yield criterion was evoked, VI' appearing as a sub-gradient of a function of cr.
This function is not usually the function g that appears in Section 3.3.1.
In the case ofa non-standard Coulomb material, with a flow rul~ defined by a
constan t dilatancy angle v (0 ~ v ~ ) or with a standard von Mises flow rule,
the function g and the plastic potential can become the same. It was from this


[7J that the concept ofa plastic potential different from the yield criterion
first gcncr;t1ized [H], The important concefJl in the general case, for whal
follows, is tilat of the function g.

3.3.3 Syslem (G)-Rarlenkouic theorem

For a material M of the type indicated in Section 3.3. I, and for each function [j,
an associated material G is defined. This is the standard material with a criterion
rl. The system (G) is that constituted by C, which is geometrically similar to (M) .
A solution of the problem of uncontained plastic now for (M), constituted by
the fields a and I), gIves

(J E 11 F




K ,: = K M

(Section 3.2)

and as a consequence of (24), v E J-I~, the set of the all allowable velocity fields
ror (G).I-Icnce, <HI)) E K~, the set of all allowable strain rates of (G).
For the field a' defined at any point of the zone formed by (24), it follows
from (25) tha t

Iv (0' ,-




or, also,

The first term of (27)
and thus



v d V Q(a) q(u)



the dissipation P G(v) of field v for the system (G),


which proves, by application of the results of Section 2, 10.2, that Q(a) is

illlanallo KG' lhc scl 0/ a/lowalJlc loadings/or (G).l


3.J.4 COlJscquenccs
The boundary of K(i' therefore, appears as an inner boundary for the limit
loadings of the system (M). Its determination will be carried out by either the
kinematic or the static method of the classical theory, as (C) is standard. The
static method leads to an approximation from tbe inside of the boundary of KG'
justifying the name of sUllie mcl/lOci/or (M) (see Figure V.A .S).
Thus, by grouping the results of Sections 3.2 and 3.3, for the conditions or
validity of the Radenkovic theorem, it is proved that the limit loadings arc to
be founu in the 'ring' comprised between KF and K(i, the boundaries being

The importance or this result should be emphasiz,ed. In a solution Ca. u) or Ihe problem or uncon
tained now ror (M). the yield criterion f = 0 is necessarily reached al some points. Thus, a is on
the boundary of H r Now G(i c GI' and therefore H G C U y and KG c K f .; but this is obviously
lIot sufficientlo prove thal Q(IT) is not inside KG'



Figure V.I\..5

3.3.5 Remarks
As stated in Section 3.3.2, there is no uniqueness of the function 0 specified
in the definition of the now rule for M. A.s the theorem is valid for each function 0,
it follows that the limit loadings are not inside


KJ' -


This amounts to defining the 'best' funclion 0 by a property of the envelope

as indicated by Palmer [6].
3.4 Variants

The Radenkovie theorem retains all the generality of the classical limit
theorems with respect to the nature of the system and the loading proeess.
Starting from a property of the now rule for M, the theorem is proved inde
pendently of any particular condition .
This generality leads to the possibility that the inner boundary may be trivial
or present lillie interest. This is the case for a Coulomb material with a now
rule corresponding to v = 0, and for which Gil = 0; i.e. the G material is a
If the proof of the theorem is again considered, it is seen to require only that
(26) be true globally for a limit equilibrium solution and that a field a' verifies
(24). This observation is not of lise as stated, sillce the solution (a, v) is assullled
to be known. However, if it were possible to estimate, (J priori, at each point of
the system and for a given loading process, a likely range for the stress <T, then
(24) and (25) would only have to ~)e imposed for <T varying in this range. This would
make il possible to use 'beller' 0 functions and would improve the results of lhe
lheorem. Such procedures appear to be the aim of the works of Josselin de long


~ ~~~~

4. (,'riction Conditions nt the Interfaces

In t:1C Cl:;C o r a :; ~':; ~::: !i! involving sevenl bodies, the rrict!Cn conditions at the
different interfaces are represented by particular forms ofyicld criteria and now
rules. They must, therefore, satisfy the principle of maximum plastic work so
tha t the results of Section 2 are applicable. If the necessary conditions are
satisficd. the theorcms in Scctions 3.2 and 3.3 may bc used.
4.1 Variotl..<; types of friction conditions
For an interface separating two materials M 1 and M 2' II and 12 are their
loading functions ncar thc intcrfaccs and rJ I' rJ2 their plastic potentials, in the
sense indicated in Section 3.3.
The friction conditions currently known will be examincd.


~ )

4.1.1 Smooth interlace

A smooth contact at the interface corresponds to the plastically admissible

range for the stresses,



= 0

. and to the flow rule,

< 0,


a = 0,


[IIJn =0,

any [IIJt

[lIJ 'n

~ 0,

any [lIJ t

figure V.A.6

17igll1'c V.I\.6 rcpresents thc 'load ing surfacc' of this 'con tact' and thc corrcspond
ing strain ratcs, in this case vclocity discontinuities, [uJ = 1I 2 - u'. The prin
ciple of maximum plastic work is vcrified, and with the notations of Section 3.1 I
is convex and I == g, for the contact


1 - - - - - - - - - - - - -- - -- - - - - -

4.1.2 [III ('I:facc w; I II cI ry CO lli(Jlllh

Ir; Cl ;(III

The plastically adm issi ble ran ge

for the stresses is given by


(lnd the now rule by

0< 0,


0= 0,

\t\ =

< -a tan 4)0: [1I]

= O. t = 0

: [1I] . 11



any lI' t

He nce the form of Fig ure V.A.7.

Th is con diti on is not sta nda rd: J
of g (IS J cor res pon ds to the cur ve
of fig ure
V.A .7, wh ere as 9 cor res pon ds to
tha t of Fig ure V.A.4.
J(<T) = Sup {a, t 2 a 2 tan 2 <Ilo
}' {j(<T') = SliP {a', t,2}

Fig ure V.A .?

Th e stre ss sta te and the velocity

field at the inte rfac e are gov ern ed
not onl y
by the friction con diti ons at the
inte rfac e itself, but als o by the pro
ties of
M I and M 2. Th e pla stic ally adm
issi ble ran ge is the refo re the inte
rse ctio n of
(32). of JI ~ 0 and of J2 ~ 0; and
the velocity discontii1Uity is gov ern
ed by (33).



Fig ure V.A.S



... ~~

by rJ I or by r/, according to the stress state. 1 Figure V.A.8 represents the case
in which M I is rigid and M 2 is a standard Tresca material.
It is assumed for the example of Figure V.A.8 that tan (p ~ 1. The condition
(32) then limits the P.A. range of Figure V.A.8 to small values ofl al. As a limiting
case the scheme of Figure V.A.9 represents the so-called 'perfectly rough'
interrace for M I rigid and M 2 a standard Tresca material. A similar procedure
applies for any other M 1 and M 2 materials.

Figure V.A.9

4.1.3 IlItcl/ace without relative motion

Some theorems [3J assume an ideal model in which the interface itself dot:s
not allow any relative motion. The P.A. range and the rules for the velocity
c1iscon tin uities then depend only on /1, /2, g 1, g2.

4.2 Application of the theorems of Section 3


4.2. 1 Standard and non-standard intci/aces

The study in Section 4.1 shows that only a smooth interface is standard. The
'perfectly rough' interface is not standard, regardless of the type of M I and M 2'
(The case is similar for an interface with any dry friction condition.) The type of
interface with no relative motion depends only on M 1 and M 2' and if M I and
M 2 are standard the interface is certainly standard also.
For the 'perfectly rough' interface, or the interface with dry friction, g,
regardless of M I and M 2' is the function corresponding to the smooth inter
face (see Figure V.A.G).

more precisely, I. and 12 being the yield functions of M and M 2' both regarding the stress
Icnsor <1, induce restrictions concerning the (J and t components of the stress vector acting on the
interface. Let 4i. ~ 0 and 4'2 ~ 0 denote these rc.~trictions, the plastically admissible range for
(IT, t) is therefore the intersection of (32), (P I ~ 0 and cfJ 1 ~ O.

I Slated


4.2.2 System (G)

As the convex plastic potential for the interfaces is that ofa smooth interface,
the application of the Radenkovic theorem is immediate. A system of standard
material (G) with smooth interfaces is considered. The limit load for this system
is smaller than or equal to that of the real system. 1 In particular, if the system is
constituted by a standard Tresca material, then Drucker's theorem 13 [3J is

4.2.3 System (F)

The theorem of Section 3.2 leads to a maximization by cOIl~iderillg the system
made:ofa standard material (F) and admitting as the !low rules at the interfaces
those deduced from the normality of [IIJ at the boundary of the plastically
admissible range for each global interface. The limit load of this system is
greater than or equal to those of the actual system.
This is the best upper bOlli1d that can be supplied by the theorem. As an
alternative, as this interface flow rule is not very prepossessing. it may be
preferable to consider a system composed of the same material but with per
fectly rough standard interfaces (i.e. with [IIJ normal at the boundary of the
P.A. range). This method is represented in Figure V.A.! 0 for the case of two
materials, M l ' rigid, and M 2' a standard Tresca material (as in Figure V.A.lO).


Figure V.A.IO

An upper bound of the limit loads of the real system is then obtained which
cannot be better than the previous value, as the convex surface of allowable
loadings for this second standard system contains that of the first standard
This theorem is equivalen t to theorem A' stated in [11 J. The Drucker theorem
A [3J uses the condition of an interface with no relative motion, and supplies
an upper bound which cannot be beller than the previous one.

For simplicilY lhc terms of the proportional loading case are used.


~_.~~'~'~H~OMe.~ .~~~






LIJ l.

F. Collin;.; (I%X) An OP!iJllUnl loading c rilc:-io;l [or rig id pl;lsli c ::l;;l c rials, J/.
M echo Phys. Solids, 16, No.2, pp. 73-80.
[2] I. F. Collins (1969) The upper bound lheorcm for rigid plaslic solids generalized
10 include Coulomb friclion, JI. Mech. Pllys. So/ids, 17, No.5, pp. 323-338.
[3J D. C. Drucker (1954) Coulomb friction plasticity and limil loads, J. App/. M ecll.
TrOllS. A.S.ME, 21, No. I, pp. 71--74.

[4] G. de Josselin de Jong (1964) Lower bound collapse theorem and lack of normality
of strain rate to yield surface for soils, Proc. I UTA M SymfJ. ()/I Rheology & Soil
M echo/Jics, Grenoble (Fr.), pp. 69-75.
[5] G. de Josselin de long (1973) A limit theorem for material with internal friction,
I'roc. Symp. OIl the Ro/e of Plasticity ;11 Soil M ccll., Cambridge (G.n.), pp. 12-21.
[6] A. C. Palmer (1966) A limit theorem for materials with non-associated now laws,
.I. M ciCllllique, 5, No.2, pp. 217-222.
. P'I I). I{adenkovie (1961) Thcorcll1es limites pour un material! de Coulomb ,\ dilala
tion non standardisce, C.R. Ac. Sc., Paris, 252, pp. 4103-4104.
r!f] I). Raden kovie (1962) Theorie des charges limiles, Scm;lIo;rc rle I'l0slicitc, Ed .
J. Mandel, P.S.1'. No.1 16, pp. 129-142.
[9] D. Radenkovic and Nguyen Quoe Son (1972) La dualitc dc.<; Ihcorcmes limites pour
unc structurc en matcriau rigide plastique standard, ArchilJes of Mechanics, 24,
No. 5-6, pp. 991-998.
[10'1 G . Saeehi and M. Save (1968) A note on the limit loads of nOIHitandard materials,
M(~ cc(/lI;ca, 3, No. I, pp. 1\3- 45.
Salenc;on (1969) La lheorie des charges limiles clans la resolution des problclIles
[I I
de plaslicile en deformalion plane, Thesis Doc!. Sc., Paris.
[12] J. Salenyon (1972) Ecoulell1enl plastique libre et analyse Ii mile pour les malcriaux
slandards et non standards, Anl1. 1.T.B.T.P., No. 295-296,pp. 90-100.
[13J J. Salen<;on (1972) Ull exemplc de non validitc de la theorie dassique dc.<; eharges
limites pour un systcme non standard, Proc.llll. Symp. on Foundations of Plasticity,
Warsaw 1972, North Holland Pub . Co., Amslerdam .
[Ill] J. Salcnc,:on (! 972) Thcorie des l:hargcs lilllites. Sem. PI(lsticitc c:t viscop/oslicilc:,
Ed. D . Radenkovic and J. Salell<;oll, Ediscience, Paris, 1974, pp. 207-229.
[15] J. Salcnyon (1972) Charge limite d'un systcllle non-standard. Un con Ire exemplc
pOllr la lheorie classique. Scm. l'i(/sticite c:t Viscopl(/st;citc, Ed. D. Radenkovie and
J . Salcnc,:on, Ediscienee, Paris, 1974, pp. 427-430.
[ 16-1 .I . Salenc;on (1974a) Plasticitc pour la MccClnique des Sols. C.J.S.M., R;lIlkine
session, .Iuly 1974, Ulline (It.).
L17:1 J . Salcnyon (I n4b) Bearing eapacily of a i'ooting on a 41 = 0 soil wilh linearly
varying shear slrength . Gcotcchllique, 24, No. 3, pp. 443-446.
[IX] M. Save (1961) 011 yield condilions in gelleralized stresses. Quart. Appl. Math.,
19, No.3, Pl'. 259-267.


The general proof of Bonneau's theorem [1] will be given for the case of a
non-homogeneous material with any intrinsic curve. The existence ofcontinllous
first derivatives of the loading function with respect to x and y is assumed.

----------------~- -



1. The Prohlcm


A plane problem is considered.

A curve (C) drawn within a solid separates the regions (I) and (11). The normal
and tangential components or the slress on (C) are given by a and L. These
functions (J and 't along (C) (Ire assumed, for reasons explained later, to satisfy
the yield condition (see Chapter I, formula 8),

IT I =

lI(a, x, y)


at any point of (C), with a constant sign (i.e. 't = II or 't = -II). Thus (C) is an
envelope of'marginal' surfaces of the same type.
It is intended to show that for an allowable stress-field to exist in the solid
(i.e. for stress-fields lo be in equilibrium in (I) and (I I) while satisfying the con
tinuity or the stress components (J and 't on (C), without violating the yield
criterion at any poinl), it is necessary that the stresses along (C) veriry the dif
ferential relation for slresses obtained in the Appendix of Chapter IV, along the
characteristics corresponding to the considered lypc or surface. (The relations
are denoted by Ea and Ell') Thus (C) must be characlerislic of the stresses.

2. Origin of the Problem-Thc Case of Incomplete Solutions



The problem has several origins. The firsl is the question of the characlcristic
envelopes [2J, whether they be rectilinea I or not. 1n fact, slich an envelope, if
touched by the arcs of characteristics in the plastic range, is the clirve (C) of
Section I,along which the equation Ea (or Ep) is not verified since it is not a
charaeteristic. Bonneau's theorem then indicates that (C) canllol be lOllched by
the arcs of the characteristics within the plastic zone but only at its boundary,
or else the envelope is not tOllched by the arcs or the characteristics themselve.<;
but by their prolongations.
For problems ofplane deformation, the solution give~ curves (C) as boundaries
between the zones of type a (I) and type IJ (11). In this case the allowable stress
field is known in (I) (in limit equilibrium), and (C) is a characteristic or an
envelope of characteristics. Bonneau's theorem gives a particularly important
result in this last case, since it shows tha t if (C) is an envelope of characteristics
within the solid, an allowable prolongation of the stress-field does not exist.
Finally, in the case of a standard ma terial, (C) can also be a line or discontinuity
of the velocity, isolated and separating two rigid zones. In this case, 't =
11((J, X, y), the sign being determined by the sense of the velocity jump, and the
stress-fields are unknown in (I) and (II) . Bonneau's theorem specifics that if an
allowable stress-field can be associated with this mode of deformation, it
necessarily verifies equation Ea (resp. Ell) along the velocity discontinuity line.


, / r - - - - - - - - - - - --- -_ _


! _-"


Th c:; c rc:) ulls h(\ '1 (~ ; ill p()rlal~t ap p!ic:l:olls for the incomplete solutions of
problems of plane deformation for a standard material, as indicated below.

(I). From the viewpoint of applying the kinematic method and the use of the
particular properties of the incomplete solutions, it is unnecessary for the
equation Ea, (or E{I) to be satisfied either at the boundary of the deformed zones
in the solid or along the isolated slip lines.
I\n example is supplied by Ostrowska [4J which involves an envelope or
characteristics touched by those in the plastic zone.

(2). Ir, when constructing an incomplete solution, it is hoped to go further

than the simple ulilization of the kinematic method and thus havean incomplete
:wlution that can be completed to give the exact value of the limit loading, it is
obvious that equation Ell (or E{I) must be satisfied at the boundary or the de
formed zones in the solid and along the isolated slip lines; otherwise, the
;checking' of the rigid zones is certainly not possible.
Hill [3J, led by physical reasoning, took this condition into accollnt for the
construction of incomplete solutions in a study of the indentation of a block.
Assumptions were made concerning the development and localization or the
plastic zones.
The theorem will now be proved.
3. Lemma

Ifan a 1I0wa bl e stress-field ex ists in a sol id, thell it is cont in lIOllS i11 crossi I1g (C).

The allowable stress-field must satisfy the continuity of the stress components
and T in crossing (C).

Figure V. D. l

Thc principlc or this proor is duc to J. Mandel (privatc communication).



-- --

~;[re~s vector
T) at a point M or(C) is on the intrinsic curve or this point.
The loading function is assumed continuous with respect to x anc! y. Il
follows that at point M t,he intrinsic curve is identical on both sides of (C) to
that on (C). Now the only Mohr circle passing through T, representative of a
plastically admissible stress tensor at M, is the circle (r) which is tangential to
the intrinsic curve of point M in T.
11 may be deduced from lhis, since (f and Tare conlinuous across (C), thaI
(r) represenls the slress tensor at M, not only on (C) but also in the regions (I)
and (II) on both sides of(C), at points M. and M . Thererore, thcre is continuity
orlile slress tensor'ill crossillg (C),

In the Mohr diagram (F'igure V.B.I) the extremity T of the



The continuity of the stress-field in crossing (C) entails continuity of the

tangential derivatives of the stress along (C). I
4. Statement of the Differential RcI:ltions Necc..'isarily Verified Along (C)

Consider the case of Figure V.I3.2 where T is nega tive. 2 The relation (1) which
is salisfied along (C) is then

h((f, x, .1')




Figut'e Y.n.2

Conditions at M arc given by

11)" =-:- 11
T ",)"




and is necessarily conlinuous when crossing (C);

aT: ".

aT: "~ I'




and, according to the lemma, is necessarily continuous when crossing (C).

According to the compOItihility relations of I-Iadamard.

In the case in which (C) is OIn envelope of charOlcteristics, it is an envelope of the ex characteristics.



As the body forces are assumed to be continuous across (C), and since the
equilibrium equations must be satisfied in (1) and (II) so that the strcss-lield is
allowable, it follows that a-r x/8y and aO"Jay are continuous across (C) at poin t lvi.
For the expressi.on
g(x, y) = -r xy - I1(O"y, X, y)
Ihe derivalive with respecL to y is given by


oy - oy



8h 80"
ocry oy


=.-2y. __ ._.:....::J. _ _



According 10 the assumption of continuity for the 'derivatives of the loading

function, and the resulLs obtained above, oy/?Jy is necessarily continuous when
crossing (C). Now, in M,

= -r xy

y(x, y)

= --r - h(cr, x, y) = 0

"(cry, x, y)

According Lo the lemma of Section 3, 9 is a continuous function when crossing

(C). Also, if the loading function of the material is denoted by I, so that the stress
field is allowable in (1) and (II), it is necessary in both regions to have

I(a, x, y)



y) ~ 0


-r xy - h(O"y, x, y)


Therefore, the only possibility for

D~ {(-r.~y - h(cr y' x, y))}

10 be continllolls aL M whcn crossing (C)

is for iL to be zero: I

-;- {(-r xy - h(O"y, x, y))} = 0



According 10 Figure V.B.I,

Hence, by substiLution in (3),




tan (/) ----.l - oy

= 0 at


11 is convenient to express o/Dy[h(cr, x, y)] as a function or iJ/oy[R(p, x, y)J,

introduced in the Appendix A or Chapter IV:

-;- h(cr, x, y) = --:;- -;- R(p, x, y)
cos (P oy


Thlls lhe slress-field is nol only OIl lhe yield boundary on (C) and conlinuous across ii, bUI also,
on bolh sides ncar (C). il is necessarily langenliallo a field allhe yield boundary.




- - - -- - - --

Then (5) can be transformed by taking (6) into account and inlroducing the
derivatives with respect to x of the stresses by means of the equations of equi
librium. The components of the body forces in M following M . and M), are
given by X and Y. Hence,








-I- (J(X

I aR
+ ---cos (/) ay

Y tan 4))

= 0


Let M' be a poin ton (C) in the vicinity of M, set {J = - (a I + a 2)/2, and denole
by 0 the angle of the direction of greatest tension at 1'11' with flllx. Then, at M',

a.,. : p + R cos 20
{ r )' R sin 20
and hence, since


-a{J =
aa.. -_



ap (1


cos L )

...:.. SIl1 '/'

= ._- SII1


(/) Sill









2R cos 20 -


20 -ao -I- cos 20





and at a point M where 0 = n/4 -I- /2

aa .

at. =



ap. 2
- (I -I- SII1 eM ax

-" = -




4) cos cp -

2H cos (/) -- -



Sill (/) -


-I- cos



4) -



Thus (7) becomes


- -

R ao +
(p ax


- 2---



(X -I- Y tan (jJ) = P

Y tan

(M + -1- aR
-= 0
cos 4) ay


is the component of the body force for the lInit basis

M x and e{l is deli ned by

c a ' ell' where c a is tangential at I'll to

I oR
+ c/) - - - =
cos 4) ay

(c , e ) = a


using the notation of the appendix of Chapter IY. Thus, the following equation
is obtllined.


+ - 21<'-I

OR) dx

dO - ( (JP -I- -

= 0 along (C)


which is simply the equlltiol1 Ea' (See Appendix A of Chapter IY.)

It is obvious that if the case of + '! = h(a, x, y) had been considered along (C),
the equation E{I would have been obtained.



---- - --- - - -5. Remllrks

:i. 1

The reasoning given in Section 4 indicates the significance of the equation

I~ a. (or E/,). Along (C) lite extremil y of lhe slress r;eclor actill[J 011 (C) is 01/ lite
intrinsic curve al1d the equatiolls of equilibrium are satisfied 011 bOlh sides of(C),
witlt lhe yield criteriOI1 110t bein[J violaled.
This reasoning can be taken as a basis for a different introduction of the strcss
cil:lr:lctcrislics ill the pl:lstic :t.onc. At any point of the pl'lstic, the two
'marginal' faces on which 11'1 = h(a, x, y) (faces corresponding to the points of
cont:lct of the Mohr circle and the intrinsic curve) are to be considered. If (C)
is a curved envelope of the faces of a family, then one of the equations Ea. (or E{I)
III liSt bc s:ltisried along (C). This eqll:ltion is a differential relation between the
unkllown functions, and tllc'refore (C) is a characteristic or the stress equations.

Thc now rule or the material docs not appear in the above proofs. The con
sider:llions concern only the stresses, and express the possibility of the existence
oran allowable stress-field in the indicated conditions. It follows that the result
obtained is valid for non-standard materials, and also for elasto-plasticity.
M. !3ollllt:au (1947) Eqllilibre limite et rupture des milieux eOlltinlls.

111111. P(}IIl.~ cl

C/rflus.\'Ces, Scpt-O~l. 1947, pp. 609-653, Nov-Dcc. 1947, pp. 769-80 I.

.I. Ie Book~r (1970) t\ propcrly of limiting lines for a pcrb;tly plastic material.

Uiliv. Sydllcy Civ. Eng. Lab" Rcs . Rcpl. No. R 134, Mar~h 1970.

It 1'lill (1950) t\ theoretical investigation or the effect ofspeeimell size in the measure

ment of hardness. Phil. Ma{j., 41, pp. 745-753 .

.I. Ostrowska (1967) Solution or indentation problcm with ellvelope of slip lines.
/Jill. 11(:. 1'01. Sc:. sh';e Se. Tech., IS, No. 10, pp. 603-611.




Thc numbcrs rcfcr to thc chapter (Roman ligure~), and to thc paragraph within thc
chaptcr (Arabic figllfe~)

Active prc~sure --IV.24

Anisol ropy- - 1.2
A~:;oeiatcd (-now rule)-1.7
A~~ociation (Thcorcm of - ) V App . A
Axial ~ymmctry (uncontaincd plastic now
in - )

IV App. B

Dcviator (stres~--) ---1.2

()isconl illuily (sl r\.~ss -) -1 V . I (I
Di~cOlltilluity (velocity-)--I V .15
Di~locat ions-1.4
Dis~ipat ion- V .2
Domain or clasticity-I 1.3
DRUCKER (-'s postulatc)-1.7
DltUCKER and PRAGER (-'s yield
cri lcrion )--1 .2

BONN EA U (-'s thcorcm) V App. B
Boundary (clastic -)-11.3
Boundary (yicld-)-IU
Briltlc (matcrial, fracturc) 1.2

Elasto-plasticity- --II

Envclope (of characterislics)-V App . B

Ca vities (sphcrical and cylindrica\- )--11.5
Charactcrist ic lincs--I V_U
Clwractcristics (Mcthod 0"- -') I V.X
Charactcristics (Rclations along thc strcss
- ) IV.7
Charactcristics (Rcl;ltions along thc vclo
city-) IV . D
Coaxiality-I . IO
Completc solution-V.7
Consistcnt boundary conditions-111.5
Constitutive law-1.5
Containcd (- plastic now)--11.2
Convexity (of thc yicld function)-1.2
Corrcsponding sta tcs (thcorcm of thc-)
IV .23
COULOMB (-'s yicld critcrion)-1.2
COULOMB CS wcdgc)-V .9
Critcrion (yicld- )-1.1
Critical (\oad)-I 1.4
Curvc (intrinsic-)-1 .2

FELLENIUS (-'s Illcthod)-V.9

Finitc elemcnts (-for limit an:llysis)-V.9

Flow ( - rule)-1 .5



(-'s cqll:ltions)-IV.lJ

GVOZDEV (-'s thcorcm)-V.3


llAA R-- KA R MAN (-hypothesis)--IV

App. B

llardness (---test )--1 V.20

HENCKY (-'s nct)-IV . IO

I-I ENCK Y (- 's rclation)-I V.9

HENCKY (-'s thcorclll)-IV . IO

I-JILL (-'s principle)-I.G

Homogencous (-stress-ficld)-I V.l 0



~-<. ~~~------- ~-~~~~---~----------------~

Interrace (condition s
Irreversibility---I . I


tlle-)-V App.

Permissi blc (-Ioading)-V.3

Plane (--strain)--IV
Plastic (-Jeformation)-l.J
Plastic (-polenlial)-I.6
PRANDTL and REUSS (--eonslilutive

Kinematic approach--VA
K inema tic method-- V A
Killelllali<.: solulion--V.7
K0TTEI~ (-'s eClualions)-IV.7

RANKINE (-eClllilihrium)-IV.24
Rigid -plastic material-I I 1.1
I~ule (now-)--l.5


Limit analysis-V
Limit equilibrium-111.2
Limit load-Il.3
Lilllit (elastic--of a system)-I1.3
Loading proeess-I 1.3
Loading (- surrace)-l.1

MANDEL (-'s equations)-lV .7
MASSAU (-'s method)-IV.8
Minim\lll1 (--principle for the strain rates)
. MinimuI11 (--principle for the stresses)

SAINT-VENANT (-'s hypothesis)-1.5

(-stress field)-I V.I 0
Standard malerial-1.8
Static approach-V.3
Static method-V .3
Static solution-V .7
SLrain-hardening (or a system)-11.3
Strain softening-l 1.2
Superposition (Method of-)-IV .23


TERZAGHI (-'s formllIa)-IV .23
TRESCA (-'s yield criterion)-1.2


Non-homogeneous (Plane strain for-

material)- IV. App. A

Nl)l\-stan<lard (---system)--V App. A.3

Norlllality-- Ui

N<. N. I N y-IV .23

Parameters (loadillg-)-- lll.5

Passive pressure- l V.24

Path (loading-)-I1.2

Perfect plasLicity-l.l

Permanent (-deformation)- l.l

Permissible (-stress-lield)--V.2

Permissible (-velocityfield)--V.2


Uncontained (-plastic now)-11.2

Uniqueness (-or the stress field)-1l1.6
Uniqueness (-of the limit Ioadings)-V.3
Unloading.-1 .5

Visco-plasticity- I.5







Consulting Editors

T. W. Lambe
R. V. Whitman

Department of Civil Enyineering

Massachusells !nstitute of Technology

Applications of the Theory of Plasticity

in Soil Mechanics

J. Salel190n
Ecole Poly technique

Ecole Nationale de.\' Pont.\' el Chaus,'I(!es, Paris