This action might not be possible to undo. Are you sure you want to continue?
Computational Geomechanics
with special Reference to Earthquake Engineering
0 C Zienkiewicz, Institute for Numerical Methods in Engineering,
Swansea, Wales
A H C Chan, University of Birmingham, England
M Pastor, CEDEX* and ETS de Ingenieros de Caminos, Madrid, Spain B A Schrefler, University of Padua, Italy
T Shiomi, Takenaka Corporation, Japan
*
Centro de E s t ~ i ~ x f i & I & i d n / e Obras Publicas
Chichester New York
.
JOHN WILEY & SONS
Weinheim Brisbane Singapore Toronto
.
.
Copyright
1999 John Wiley & Sons Ltd, Baffins Lane, Chichester, West Sussex PO19 IUD, England National 01243 779777 International (+44) 1243 779777 email (for orders and customer service enquiries): esbooks@wiley.co.uk Visit our Home Page on http:l/www.wiley.co.uk or http:llwww.wiley.com
All Rights Reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, scanning or otherwise, except under the terms of the Copyright, Designs and Patents Act 1988 or under the terms of a licence issued by the Copyright Licensing Agency, 90 Tottenham Court Road, London, UK WIP 9HE, without the permission in writing of the publisher.
Other Wiley Editorial Offices
John Wiley & Sons, Inc., 605 Third Avenue, New York, NY 101580012, USA WILEYVCH Verlag GmbH, Pappelallee 3, D69469 Weinheim, Germany Jacaranda Wiley Ltd, 33 Park Road, Milton, Queensland 4064, Australia John Wiley & Sons (Asia) Pte Ltd. 2 Clementi Loop # 0201. Jin Xing Distripark, Singapore 129809 John Wiley & Sons (Canada) Ltd, 22 Worcester Road, Rexdale, Ontario M9W ILI, Canada
Library of Congress CataloginginPublication Data Computational geomechanics with special reference to earthquake engineering1 O.C. Zienkiewicz . . . [et al.]. p. cm. Includes bibliographical references and index. ISBN0471982857 1. Earthquake engineering 2. Mathematics. I. Zienkiewicz, O.C. TA705.C625 1998 624.1 ' 7 6 2 6 ~ 2 1 988795 CIP British Library Cataloguing in Publication Data
A catalogue record for this book is available from the British Library ISBN 0471982857 Typeset in 10/12.25pt Times from the author's disks by Pure Tech India Ltd, Pondicherry Printed and bound in Great Britain by Bookcraft (Bath) Ltd, Midsomer Norton, Somerset This book is printed on acidfree paper responsibly manufactured from sustainable forestry, in which at least two trees are planted for each one used for paper production.
Contents
Preface
1 Introduction and the Concept of Effective Stress
1.1 Preliminary Remarks 1.2 The Nature of Soils and Other Porous Media: Why a Full Deformation Analysis is the Only Viable Approach for Prediction 1.3 Concepts of Effective Stress in Saturated or Partially Saturated Media 1.3.1 A single fluid present in the poreshistorical note 1.3.2 An alternative approach to effective stress 1.3.3 Effective stress in the presence of two (or more) pore fluids. Partially saturated media References
2 Equations Governing the Dynamic, SoilPore Fluid, Interaction
2.1 General Remarks on the Presentation 2.2 Fully Saturated Behaviour With A Single Pore Fluid (Water) 2.2.1 Equilibrium and Mass Balance Relationship (u, w and p) 2.2.2 Simplified equation sets (up form) 2.2.3 Limits of validity of the various approximations , 2.3 Partially Saturated Behaviour with Air Pressure Neglected @ = 0) 2.3.1 Why is inclusion of semisaturation required in practical analysis? 2.3.2 The modification of equations necessary for partially saturated conditions 2.4 Partially Saturated Behaviour with Air Flow Considered (pa 0) 2.4.1 The governing equations including air flow 2.4.2 The governing equation 2.5 Alternative derivation of the governing equations of sections 2.12.4, based on the hybrid mixture theory 2.5.1 Kinematic equations 2.5.2 Microscopic balance equations 2.5.3 Macroscopic balance equations 2.5.4 Constitutive equations 2.5.5 General field equations 2.5.6 Nomenclature 2.6 Concluding Remarks References
>
vi
CONTENTS
3 Finite Element Discretization and Solution of the Governing Equations
3.1 The Procedure of Discretization by the Finite Element Method 3.2 up Discretization for a General Geomechanics Finite Element Code 3.2.1 Summary of the general governing equations 3.2.2 Discretization of the governing equation in space 3.2.3 Discretization in time 3.2.4 General applicability of transient solution (consolidation, static solution, drained uncoupled, undrained) Time step length The consolidation equation Static problemsundrained and fully drained behaviour 3.2.5 The Structure of the numerical equations illustrated by their Linear equivalent 3.2.6 Damping matrices 3.3 The uU Discretization and its Explicit Solution 3.3.1 The governing equation 3.3.2 Discretized equation and the explicit scheme 3.3.3 The structure of the numerical equations in linear equivalent 3.4 Theory: Tensorial Form of the Equations 3.5 Conclusions References
4 Constitutive RelationsPlasticity
4.1 Introduction 4.2 The general Framework of Plasticity 4.2.1 Phenomenological aspects 4.2.2 Generalized plasticity 4.2.3 Classical theory of plasticity 4.3 Critical State Models 4.3.1 Introduction 4.3.2 Critical state models for normally consolidated clay 4.3.3 Extension to sands 4.4 Advanced Models 4.4.1 Introduction 4.4.2 A generalized plasticity model for clays 4.4.3 A generalized plasticity model for sands 4.4.4 Anisotropy 4.5 Modified Densification Model 4.5.1 Densification model for cyclic mobility References
5 Examples for Static, Consolidation and Partially Saturated Dynamic Problems
5.1 Introduction 5.2 Static Problems 5.2.1 Example (a): Unconfined situationsmall constraint Embankment Footing 5.2.2 Example (b): Problems with medium (intermediate) constraint on deformation 5.2.3 Example (c): Strong constraintsundrained behaviour 5.2.4 Example (d): The effect of the K section of the yield criterion
CONTENTS
Isothermal Drainage of Water from a Vertical Column of Sand Modelling of Subsidence due to Pumping from a Phreatic Aquifer Air storage Modelling in an Aquifer Flexible Footing Resting on a Partially Saturated Soil Comparison of Consolidation and Dynamic Results Between Small strain and Finite Deformation Formulation 5.7.1 Consolidation of fully saturated soil column 5.7.2 Consolidation of fully and partially saturated soil column 5.7.3 Consolidation of twodimensional soil layer under fully and partially saturated conditions 5.7.4 Fully saturated soil column under earthquake loading 5.7.5 Elastoplastic largestrain behaviour of an initially saturated vertical slope under a gravitational loading and horizontal earthquake followed by a partially saturated consolidation phase 5.8 Conclusions References 5.3 5.4 5.5 5.6 5.7
vii
6 Validation of Prediction by Centrifuge
6.1 Introduction 6.2 Scaling Laws of Centrifuge Modelling 6.3 Centrifuge Test of a Dyke Similar to a Prototype Retaining Dyke in Venezuela 6.4 The VELACS Project 6.4.1 General analysing procedure 6.4.2 Description of the precise method of determination of each coefficient in the numerical model 6.4.3 Modelling of the laminar box 6.4.4 Parameters identified for the PastorZienkiewicz Mark 111 model 6.5 Comparison with the VELACS Centrifuge Experiment 6.5.1 Description of the models Model No. 1 Model No. 3 Model No. I I 6.5.2 Comparison of experiment and prediction 6.6 Centrifuge test of a Retaining Wall 6.7 Conclusions References
7 Prediction Applications and Back Analysis
7.1 Introduction 7.2 Example 1: Simulation of Port Island LiquefactionEffect of Multidimensional Loading 7.2.1 Introductory remarks 7.2.2 Multidirectional loading observed and its numerical modellingsimulation of liquefaction phenomena observed at Port Island Conditions and modelling Results of simulation Effects of multidirectional loading 7.3 Simulation of Liquefaction Behaviour During Niigita Earthquake to Illustrate the Effect of Initial (shear) Stress
viii
Influence of initial shear stress Significance of ISS component to the responses Theoretical considerations 7.4 Quay Wall Failure and a Countermeasure 7.4.1 Conditions and modelling Configuration Soil layers and properties Input Motion 7.4.2 Results and remarks 7.5 Lower San Fernando D a m Failure 7.6 Mechanism of Liquefaction Failure o n a n Earth D a m (the N Dam) 7.6.1 Objective of the analysis 7.6.2 Input motion 7.6.3 Conditions and modelling Soil properties Parameters for liquefaction Initial stress 7.6.4 Results of calculations 7.6.5 Remarks 7.7 Liquefaction Damage in the Niigata Earthquake of 1964 7.7.1 Results 7.8 Interaction Between Ordinary Soil and Improved Soil Layer 7.8.1 Input motions Earth pressure due to liquefaction 7.8.2 Safety for seismic loading External safety Internal safety 7.8.3 Remarks References 7.3.1
CONTENTS
8 Some Special Aspects of Analysis and Formulation: Radiation Boundaries, Adaptive Finite Element Requirement and Incompressible Behaviour
8.1 Introduction 8.2 Input for Earthquake Analysis and Radiation Boundary 8.2.1 Specified earthquake motion: absolute and relative displacements 8.2.2 The radiation boundary condition: formulation of a onedimensional problem 8.2.3 The radiation boundary condition: treatment of two dimensional problem 8.2.4 Earthquake input and the radiation boundary conditionconcluding remarks 8.3 Adaptive Refinement for Improved Accuracy and the Capture of Localized Phenomena 8.3.1 Introduction to adaptive refinement 8.3.2 Localization and strain softening: possible nonuniqueness of numerical solutions 8.4 Stabilization of Computation for Nearly Incompressible Behaviour with Mixed Interpolation 8.4.1 The problem of incompressible behaviour under undrained conditions 8.4.2 The velocity correction, stabilization process 8.4.3 Examples illustrating the effectiveness of the operator split procedure 8.4.4 The reason for the success of the stabilizing algorithm References
3.2 Subroutines for control and material data input 9.1 The top level routines 9.5.4 Major service subroutines 9.3 Description of major routines used in DIANASWANDYNE I1 9.CONTENTS 9 Computer Procedures for Static and Dynamic Saturated Porous Media finite element Analysis 9.2 Outline description of DIANASWANDYNE I1 9.5.2 Constitutive models available for general dissemination 9.3.3.3 Subroutines for mesh data input 9.1 Standard constitutive model interfacing subroutine CONSTI 9.5 Subroutines for the formation of element matrices and residual calculation 9.5.3 Other constitutive models implemented 9.3.1 Introduction 9.3.6 Systemdependent subroutines References Appendix 9A Implementing New Models into SM2D Author Index Subject Index .5 Constitutive model subroutines 9.4 Subroutines called by the main control routine for analysis 9.
often introduced with the linear approximations. Such simplifications are generally not useful and can lead to erroneous predictions. the basis of such computational approaches because a wider audience of practitioners and engineering students will require the knowledge which hitherto has only been available through scientific publications scattered throughout many journals and conferences. The multiple authorship not only ensures a speedy result. the phenomena of weakening and of 'liquefaction' in soil when subjected to repeated loading such as that which occurs in earthquakes. Here a validation of the twophase approach was available and close agreement between computation and experiment was found. These provide a useful set of benchmark predictions. it also introduces members of the research team which.Preface Although the concept of effective stress in soils is accepted by all soil mechanicians. In recent years. centrifuge experiments have permitted the study of some soil problems involving both statics and dynamics. practical predictions and engineering calculations are traditionally based on total stress approaches. have focused attention on the subject which has developed practical computer codes which are now available to both practitioners and researchers. in 1993. in the early seventies. When the senior author began. both students and colleagues. it became clear to him that a realistic prediction of the behaviour of soil masses could only be achieved if the total stress approaches were abandoned. the total stress calculation continues to be used by some engineers for earthquake response analysis. the application of numerical approaches to the field of soil mechanics in general and to soil dynamics in particular. A very important landmark was a workshop held at the University of California. during the last decade. can only be explained by considering this 'twophase' action and the quantitative analysis and prediction of real behaviour can only be achieved by sophisticated computation. The simple limit methods often applied in statics are no longer useful. Indeed. Since 1975 large number of research workers. have participated both at Swansea and elsewhere in laying the foundations of numerical predictions which were based largely on concepts introduced in the early forties by Biot. The essential model should consider the coupled interaction of the soil skeleton and of the pore fluid. However. It therefore seems necessary at the present time to present. Davis. in a single volume. which . The present book is an attempt to provide a rapid answer to this need.
the chapter is necessarily long as it starts from simple plasticity models and continues to the presentation of such topics as generalized plasticity. At this workshop a full vindication of the effective stress. necessary for an adequate description of the soil behaviour. critical state soil mechanics etc. we have endeavoured to present a unified approach and have used the same notation. A simplified version of SWANDYNE is outlined at the end of the book (Chapter 9) and an executable version can be obtained via the Internet using the URL at http://www. in this chapter we also introduce a simplified model of denszfication which. essential before numerical approximation. radiation damping and adaptive refinement are given..PREFACE xi reported results of the VELACS project (Verification of Liquefaction Analysis by Centrifuge Studies)sponsored by the National Science Foundation of USA.uk/ CivEng/swandyne/index. twophase approaches was clearly available and it is evident that these will be the basis of future engineering computations and prediction of behaviour for important soil problems. we address some rather specialized topics which help in the improvement of general programs but are not absolutely necessary. Chapter 8. . The first three chapters present the theory of porous media in the saturated and unsaturated states and thus establish general backbone to the problem of soil mechanics. During numerical studies it became clear that the geomaterialsoil. In the last chapter. Here. This phenomenon may be present at the outset of loading or may indeed develop during the dynamic process. deals with the very important matter of the quantitative description of soil behaviour which is necessary for realistic computations. Here special treatment of incompressibility. verification of computation by dynamic experiments in centrifuge (Chapter 6) and practical applications to earthquake engineering in Chapter 7. Such partial saturation is responsible for the presence of negative pressures which allow some 'apparent' cohesion to be developed in noncohesive soils. style and spirit throughout. Indeed.bham. Despite the large number of authors.htm. The book shows some examples of this validation and also indicates examples of practical application of the procedures described. The various solutions of static and dynamic situations shown in this book have been obtained by using the code named SWANDYNE which is available from the authors. We have therefore incorporated its presence in the treatment presented in this book and thus achieved wider applicability for the methods described. would often be present in a state of incomplete saturation when part of the void was filled with air. permits a realistic description of liquefaction and cyclic mobility phenomena consecutively with problems of applications to static or quasistatic problems (Chapter 5 ) .ac. when added to simple classical plasticity. Similarly the explicit derivative is also available. Chapter 4.
.
.
1 shows both a sketch indicating the extent of failure and a diagram indicating the crosssection of the encountered ground movement.2. theevident causeand the 'straw that broke thecamel's back' was the filling and the subsequent drawdown of the reservoir. However. Typical of this is the disastrous collapse of the mountain (Mount Toc) bounding the Vajont reservoir which occurred on October 9th 1963 in Italy (Miiller 1965). In theabovecollapse. On occasion he may have to apply his predictive knowledge to events in natural soil or rock outcrops.Introduction and the Concept of Effective Stress 1. displacing the water dynamically and causing an unprecedented death toll of some 4000 people from the neighbouring town of Longarone. or building foundations should be able to predict the safety of these against collapse or excessive deformation under the various loading conditions which are deemed possible. Figure 1. manmade conditions. Such static failures which occur.. The phenomenon proceeded essentially in a static (or quasistatic) manner until the last moment when the moving mass of soil acquired the speed of 'an express train' at which point it tumbled into the reservoir. subject perhaps to new. 1979. the overtopping of the dam would indeed have caused a major catastrophe with the flood hitting a densely populated area of Los Angeles. dams. dynamic effects such as those frequently caused by earthquakes are more spectacular and much more difficult to predict. fortunately at a much smaller scale. 1975). in many embankments and cuttings are subjects of typical concern to practising engineers. Perhaps the particular feature of this interac . Seed et al. This failure fortunately did not involve any loss of life as the level to which the dam 'slumped' still contained the reservoir. It is evident that the two examples quoted so far involved the interaction of pore water pressure and the soil skeleton.1 PRELIMINARY REMARKS The engineer designing such soil structures as embankments. We illustrate the dynamic problem by the near collapse of the Lower San Fernando dam near Los Angeles during the 1971 earthquake Figure 1. (Seed. Had this been but a few feet lower.
. However. failure of Mant Toc in 1963 (Oct. Such strength reduction phenomena are mainly evident in essentially noncohesive materials such as sand and slit. however. 1965) Plate 1 shows a photo of the slides (front page) tion. Clays in which negative. This is due to the 'weakening' of the soilfluid composite during the periodic motion such as that which is involved in an earthquake. thus reducing the interparticle forces in the solid phase of the soil and its strength. 9th): (a) hypothetical slip plane. What appears to have happened here is that during the motion the interstitial pore pressure increased. it is this rather than the overall acceleration forces which caused the collapse of the Lower San Fernando dam. capillary pressure provide an apparent cohesion are less liable to such strength reduction.2 INTRODUCTION AND THE CONCEPT OF EFFECTIVE STRESS Figure 1.1 The Vajont reservoir. escapes immediate attention. (b) downhill end of slide (Miiller.
.
once all the pore pressures are known. From this figure. finite element.2 THE NATURE OF SOILS AND OTHER POROUS MEDIA: WHY A FULL DEFORMATION ANALYSIS IS THE ONLY VIABLE APPROACH FOR PREDICTION For singlephase media such as those encountered in structural mechanics. The first has a granular structure of loose. singlephase analysis approaches. Provided that the dimension of interest and the so called 'infinitesimals' dx. It is clear here that the buildings behaved as if they were floating during the active part of the motion. are frequently used under static conditions. it is possible to predict the ultimate (failure) load of a structure by relatively simple calculations. . for problems of soil dynamics. it is possible to reduce the soil mechanics problem to that of the behaviour of a single phase. are large enough when compared to the size of the grains and the pores. Here we illustrate two apparently very different materials. quasistatic and dynamic phenomena which occur in soils and will indicate how a computer based. in which the pores of the solid phase are filled with one fluid. at least for static problems. it is evident that the approximation of a continuum behaviour holds. 1. etc. Similarly for soil mechanics problems such simple. One such case is that occurring under longterm load conditions in material of appreciable permeability when a steady state drainage pattern has been established and the pore pressures are independent of the material deformation and can be determined by uncoupled calculations. Then we can use again the simple. Using the concept of effective stress. However. but even here. However. generally uncemented.3 can be treated at all by the methods of continuum mechanics. analysis can be effective in predicting all these aspects quantitatively. particles in contact with each other. In this book. porous. the answer to the query concerning the possibility of continuum treatment is selfevident. The second is a solid matrix with pores which are interconnected by narrow passages. This behaviour is known as soil liquefaction and Plate 2 shows a photograph of some buildings in Niigata. pz etc. cannot be described by behaviour of a singlephase material. material on which both deformations and failure depend can thus only be determined once such pressures are known.4 INTRODUCTION AND THE CONCEPT OF EFFECTIVE S T R E S S This phenomenon is well documented and in some instances the strength can drop to near zero values with the soil then behaving almost like a fluid. the use of such simplified procedures is almost never admissible. limitload calculations. The strength of the solid. Indeed to some it may be an open question whether such porous materials as shown in Figure 1. Indeed on occasion the limit load procedures are again possible. The reason for this lies in the fact that the behaviour of soil or such a rocklike material as concrete. it is equally clear that the intergranular forces will be much affected by the pressures of the fluidp in single phase (or p l . if two or more fluids are present). dy. full justification of such procedures is not generally valid. which we shall discuss in detail in the next section. Japan taken after the 1964 earthquake. we shall discuss the nature and detailed behaviour of the various static.
by zero permeability being implied or by extreme speed of the loading phenomena. Although in early attempts to deal with earthquake analyses and to predict the damage and response. (b) a perforated solid with interconnecting voids Such drained behaviour. such undrained analyses were invariably used. which is frequently assumed for rapidly loaded soil. For this reason we believe that the only realistic type of analysis is of the type indicated in this book. however. 1993). In very finely grained materials such as silts or clays this situation may never be established even as an approximation. This was demonstrated in the same VELACS tests to which we shall frequently refer in later chapters. While the artifice of simple undrained behaviour is occasionally useful in static studies.3 Various idealized structures of fluid saturated porous solids: (a) a granular material. if all fluid motion is prevented.T H E NATURE OF SOILS 5 Solid / I Figure 1. adding generally a linearization of the total behaviour and an heuristic assumption linking the pressure development with cycles of loading. in general. Indeed. the behaviour predictions were poor. the pressures developed in the fluid will be linked in a unique manner to deformation of the solid material and a singlephase behaviour can again be specified. the complete solution of the problem of solid material deformation coupled to a transient fluid flow needs generally to be solved. We have not mentioned so far the notion of so called undrained behaviour. Here no shortcuts are possible and full coupled a n ~ l ~ v s e s equations which we shall introof duce in Chapter 2 become necessary. the infinite time required to reach this asymptotic behaviour. Indeed recent comparisons with centrifuge experiments confirmed the inability of such methods to predict either the pressure development or deformations (VELACS Arulanandan & Scott. in a general situation. theoretically.  . be linked again to the straining (or loading) history and this must always be taken into account. it is not applicable to dynamic phenomena such as those which occur in earthquakes as the pressures developed will. Thus. seldom occurs even in problems which we may be tempted to consider as static due to the slow movement of the pore fluid and.
e. (or dx. of course the dynamic forces are reproducible and scalable. though at considerable cost. So far no success has been achieved in modelling these and hence studies of structures with free (phreatic) water surface are excluded. at least qualitatively towards the end of the last century. 1913b and 1915) and indeed for other soil or rock structures. dx. This of course eliminates possible practical applications of the centrifuge for dams and embankments in what otherwise is a useful experimental procedure. 1895 and Fillunger. in Figure 1. such as air and water for instance. Unfortunately. gravity effects can be well modelled..3 CONCEPTS OF EFFECTIVE STRESS IN SATURATED OR PARTIALLY SA TURA TED MEDIA note 1. 1913a. similar concepts were used to define the behaviour of concrete in dams (Levy.1 A single fluid present in the poreshistorical The essential concepts defining the stresses which control strength and constitutive behaviour of a porous material with internal pore pressure of a fluid appear to have been defined. 1. dz in conventional notation). many factors conspire to deny in geomechanics a readily accessible model study. Boussinesq (1876) and Reynolds (1886) was here of considerable note for problems of soils. fill the pores.6 INTRODUCTION AND T H E CONCEPT OF EFFECTIVE S T R E S S At this point. i. In the context of the finite element computation we shall frequently use a vectorial notation for stresses. Later. As we shall see later under conditions when two fluids. writing . here.3. A typical civil engineer may well consider here the analogy with hydraulic models used to solve such problems as spillway flow patterns where the cost of a smallscale model is frequently small compared to equivalent calculations. dy. even here a barrier is reached which appears to be insurmountable.3 and include the total area of the porous skeleton. centrifuge models have been introduced and. If we thus define the total stress a by its components aq using indicia1 notation these are determined by summing the appropriate forces in the idirection on the projection. To remedy this defect. With suitable fluids substituting water it is indeed also possible to reproduce the seepage timescale and the centrifuge undoubtedly provides a powerful tool for modelling earthquake and consolidation problems in fully saturated materials. The work of Lye11 (1871). In all of these approaches the concept of division of the total stress between the part carried by the solid skeleton and the fluid pressure is introduced and the assumption made that the strength and deformation of the skeleton is its intrinsic property and not dependent on the fluid pressure. capillary effects occur and these are extremely important. Unfortunately. that of gravity though. The surfaces of cuts are shown. Scale models placed on shaking tables cannot adequately model the main force acting on the soil structure. perhaps it is useful to interject an observation about possible experimental approaches. or cuts. for two kinds of porous material structure. The question which could be addressed is whether a scale model study can be made relatively inexpensively in place of elaborate computation.
The above.3) in 1913 but despite conducting experiments in 1915 on the tensile strength of concrete subject to water pressure in the pores. The negative sign is introduced as it is a general convention to take tensile components of stress as positive. which gave the correct answers. Further. Fillunger introduced the concepts implicit in (1.3) that the strength of the material would be always influenced by the pressure p. argument leads to the following relation between total and effective stress with total stress or if vectorial notation is used we have a = a'.. It was the work of Terzaghi and Rendulic in 1934 and by Terzaghi in 1936 which finally modified the definition of effective stress to where n.m n p (1.. p acting. plausible. Now if the stress in the solid skeleton is defined as the effective stress d again over the whole crosssectional area then the hydrostatic stress due to the pore pressure.2 it would be possible to damage a specimen of a porous material (such as concrete for instance) by subjecting it to external and internal pressures simultaneously. it would appear from equation (1. only on the pore area should be where n is the porosity and Sji is the Kronecker delta. now called the effective areu coef$cient and is such that is . with values of porosity n with a magnitude of 0.10.4) where m is a vector written as The above arguments do not stand the test of experiment as it would appear that.CONCEPTS OF EFFECTIVE STRESS 7 This notation reduces the components to six rather than nine and has some computational merit. he was not willing to depart from the simple statements made above.
1. Here the work of Leliavsky (1947) McHenry (1948) Serafim (1954. to any porous material. only a very uniform and small volumetric strain will occur in the skeleton and the material will not suffer any damage provided that the grains of the solid are all made of an identical material. that such effects are insignificant. Thus in general the grains and hence the total material will be in a state of pure volumetric strain . 1964) made important contributions by experiments and arguments showing that it is more rational to take sections for determining the pore water effect through arbitrary surfaces with minimum contact points. (1996) addressed the same problem showing how an acrimonous debate between Fillunger and Terzaghi terminated in the tragic suicide of the former in 1937. the work of Biot (1941. The above requirement can be written in tensorial notation as requiring that the total stress increment is defined as or. Experiments performed on many soils and rocks and rocklike materials show. 1956a.2 An alternative approach to effective stress Let us now consider the effect of the simultaneous application of a total external hydrostatic stress and a pore pressure change. 1963) who found that interpretation of the various experiments was not always convincing. both equal to A p . using the vector notation In the above.7).3. 1956b. however. 1962) and Biot and Willis (1957) clarified many concepts in the interpretation of effective stress and indeed of the coupled fluid and solid interaction. However.8 INTRODUCTION AND THE CONCEPT OF EFFECTIVE STRESS Much further experimentation on such porous solids as concrete had to be performed before the above statement was generally accepted. The subject of effective stress is as of much interest to the senior author who directed his research to analysis of dams. while it is convenient to define stress components as positive in tension. viz Zienkiewicz (1947. can occur and that local grain damage may be suffered. 1955. It is evident that for the loading mentioned.6) and (1. This is simply because all parts of the porous medium solid component will be subjected to an identical compressive stress. Bishop (1959) and Skempton (1960) analysed the historical perspective and more recently de Boer (1996) and de Boer et (11. In the following section we shall present a somewhat different argument leading to equations (1. However. localized stresses. it appears possible that nonuniform. the negative sign is introduced since 'pressures' are generally defined as being positive in compression. if the microstructure of the porous medium is composed of different materials.
CONCEPTS OF EFFECTIVE S T R E S S 9 where K. Again. assuming that the material is isotropic. and (a) Internal pressure increment A. A block of porous. Frame before load Frame after load Negligible deformation (a) (b) Surface membrane AP Figure 1.4. spongelike rubber. the volumetric strain will be negligible as the solid components of the sponge rubber are virtually incompressible.4 A Porous Material subject to external hydrostatic pressure increases A. is the average material bulk modulus of the solid components of the skeleton. we shall have Those not familiar with soil mechanics may find the following hypothetical experiment illustrative. is immersed in a fluid to which an increase in pressure of A p is applied as shown in Figure 1.. Alternatively.. (b) Internal pressure increment of zero . If the pores are connected to the fluid. adopting a vectorial notation for strain in a manner involved in (1.1) where E is the vector defining the strains in the manner corresponding to that of stress increment definition.
= 1 have been simply added to the uniform volumetric compression.1 1) writing these as . However.1 la) and (1. if linear elastic behaviour is assumed.10 D is a tangent matrix of the solid skeleton implied by the constitutive relation with corresponding compliance coefficient matrix D' = C. we assume incremental properties. elastoplastic and elastoviscoplastic responses of the skeleton which.12). but generally will be defined by an appropriate nonlinear relationship of the type which we shall discuss in Chapter 4 and this behaviour can be established by fully drained @ = 0) tests. however. the principle of superposition requiring linear behaviour is not invoked and in this book we shall almost exclusively be concerned with nonlinear. The facts mentioned above were established by the very early experiments of Fillunger (1915) and it is surprising that so much discussion of "area coefficients" has since been necessary. irreversible.10 INTRODUCTION AND THE CONCEPT OF EFFECTIVE STRESS If. Although the effects of skeleton deformation due to the effective stress defined by (1. we note that the deforination relation of (1.1 1b).1 1) can always be rewritten incorporating the small compressive deformation of the particles as (1... the block is first encased in a membrane and the interior is allowed to drain freely. From the preceding discussion it is clear that if the material is subject to a simultaneous change of total stress Ao and of the total pore pressure Ap.6) with n. while the penultimate term is the strain due to the grain compression already mentioned viz Eq. AEO.= 1 is sufficient.. is simply the increment of an initial strain such as may be caused by temperature changes etc. the resulting strain can always be written incrementally in tensorial notation as or in vectoral notation with The last term in (1. on the other hand.. For assesment of strength of the saturated material the effective stress previously defined with n. then again a purely volumetric strain will be realised but now of a much larger magnitude. These of course could be matrices of constants. It is more logical at this step to replace the finite increment by an infinitesimal one and to invert the relations in (1. 1.
which is the tangential bulk modulus of an isotropic elastic material with X and being the Lame's constants. . In the above and the new form eliminates the need for separate determination of the volumetric strain component. Thus we can write 11 The reader should note that in (1.6. is defined. we note that.12) we have written the definition of the effective stress increment which can..3 !I !I  or we can write or simply Alternatively. in tensorial form.CONCEPTS OF EFFECTIVE STRESS 11 where a new 'effective' stress.. the same result is obtained as For isotropic materials. Noting that 6. be used in a nonincremental state as . of course. a".
The above definition corresponds to that of the effective stress used by Biot (1941) but is somewhat more simply derived. Pavtially saturated media The interstitial space. We shall.3. It should have been noted that in some materials such as rocks or concrete it is possible for the ratio K. that for water and S. but the discussion will be valid for any two fluids.3 Effective stress in the presence of two (or more) porefluids. may in a practical situation be filled with two or more fluids. This assumption is not totally correct in soil liquefaction or flow in soil shearing layer during localization. the meaning of a is no longer associated with an effective area. It is clear that. to be as large as 113 with a = 213 being a fairly common value for determination of deformation. We have also implicitly assumed that the fluid flow is such that it does not separate the contacts of the soil grains. that for air. (for example./K. . if both fluids fill the pores completely. oil and water and indeed the treatment described here is valid for any fluid conditions. we shall always have Clearly this relation will be valid for any other pair of fluids e. consider only two fluids with the degree of saturation by each fluid being defined by the proportion of the total pore volume n (porosity) occupied by each fluid. cr is a factor which becomes close to unity when the bulk modulus Ks of the grains is much larger than that of the whole material. In the context of soil behaviour discussed in this book the fluids will invariably be water and air respectively. We note that in the preceding discussion the only assumption made which can be questioned. of course recovering the common definition used by many in soil mechanics and introduced by von Terzaghi (1936).g. in this section. In such a case we can write. or the pores. In the above. material exposed to air is taken as under zero pressure).12 INTRODUCTION AND T H E CONCEPT OF EFFECTIVE S T R E S S assuming that all the stresses and pore pressure started from a zero initial state. is that of neglecting the local damage due to differing materials in the soil matrix. 1. Now. Thus we shall refer to only two saturation degrees S. however.
5 with and Occasionally the contact of one of the phases and the solid may disappear entirely as shown in Figure 1. pa pressure p = X.5 (a) and (b).19) where the coefficients xw and xa refer to water and air respectively and are such that The individual pressures pw and pa are again referring to water and air and their difference i. is dependent on the magnitude of surface tension or capillarity and on the degree of saturation ( p .17)) can thus be taken as I' = XwPw + XaPa (1. Depending on the nature of the material surface the contact surface may take on the shapes shown in Figure 1.5 Two fluids in pores of a granular solid (water and air). as capillary pressure).16) and (1. (b) both fluids in contact with solid surfaces (effective X. + . (a) air bubble not wetting solid surface (effective pressure p = p w .5a giving isolated air bubbles and making in this limit Figure 1.CONCEPTS OF EFFECTIVE STRESS 13 The two fluids may well present different areas of contact with the solid grains of the material in the manner illustrated in Figure 1.e. is often referred to therefore. The average pressure reducing the interstitial contact and relevant to the definition of effective stress found in the previous section (Equations (1. p.
1987). 1965. Narasimhan & Witherspoon 1978. and relative permeability. / X .14 INTRODUCTION AND THE CONCEPT O F EFFECTIVE STRESS Whatever the nature of contact. we shall find that a unique relationship between p.)/k. 1984.19. Alternatively. S. Certainly the arguments for thus extending the original concepts are less clear than is the case when only a single fluid is present.24). In either case the effect of p..16 and 1. In many cases occurring in practice. The concepts of dealing with the effects of multiple pore pressure by introducing an average pressure and using the standard definition of effective stress (1. 1977. can be easily neglected as the water pressure simply becomes negative from Equation (1. However.. Figure 1. = p.6 Typical relations between pore pressure head. the air pressure is close to zero (atmospheric datum) as the pores are interconnected. Indeed. the degree of saturation will similarly affect flow parameters such as the permeability k to which we shall make reference in the next chapter. 1975. as we shall see later Figure 1. 1980. 1.6 shows a typical relationship. Alonso et al. Neuman. We shall therefore use such a definition in the study of partially saturated media. negative air pressure occurs as cavitation starts and here the datum is the vapour pressure of water. . giving Many studies of such relationship are reported in the literature (Liakopoulos. Such negative pressures are responsible for the development of certain cohesion by the soil and are essential in the study of free surface conditions occurring in embankments. Note that relative permeability decreases very rapidly as saturation decreases . Bear et al. Saturation.17) were first introduced by Bishop (1959). Van Genuchten et al. the results obtained by this extension are quite accurate.(l) (Safai & Pinder 1979). .e. and the saturation Sw can be written i. k. 1979. h. Safai & Pinder. Lloret & Alonso. = k . (S.
(1980) Consolidation of unsaturated soils including swelling and collapse behaviour. Schiffman R. Biot M. 14821498. Eng. A. (1895) Quelques Considerations sur la construction des grandes barrages.. 112. McHenry D.. (Eds) (1993) Proceedings of the VELACS symposium. 28. A. Am. J. C o w ptes Rendus De L'Academie Des Sciences Serie IMathematique. (1962) Mechanics of deformation and acoustic propagation in porous media. W. Berkeley. 1180. 11. 12. USA. No. No. Fels Mecharzik. C. P. Osterr. R. 1821 85. (1871) Student's elements of geology. Boussinesq J. 179191. Eng. 2. London. Levy M.sient. Wochenschrift offentlichen Baudienst. Corapcioglu M.REFERENCES 15 REFERENCES Alonso E. H29. 30. 26. crllseitigem Wasserdruck. Proc. (1996) The origins of the theory of consolidation: the TerzaghiFillunger dispute. W. part Ilowfrequency range. E. (1965) Trczn. J. Fillunger P.flow through unsrrturated porous media. Soc. 4. London. Applicd Mec1z~1nic. No. Soc. 49. (1941) General theory of threedimensional consolidation. Soc. (1975) Galerkin approach to saturatedunsaturated flow in porous media in Finite Elernents in jluids. Bishop A. Biot M. Water Resour. . (1886) Experiments showing dilatancy. and Alonso E. 148212. Appl. 24. 20 1262. 39. L. 150167.. Trans. De Boer R. Dublin. Acoust. 2. A. Y. Teknisk Ukeblad. A. 2... Civil Engrs. and Hight D. Am. Phys. Wiley.. International Congress Large Dams. 168178. Bear J. Acoust. Biot M. P h j ~ .. Wochenschrift offentlichen Baudienst. Am. 532556. (1956b) Theory of propagation of elastic waves in a fluidsaturated porous solid. 10171034. Found. Dissertation. Biot M. Appl. (1915) Versuch iiher die Zugfestigkeit he. A. A. De Boer R. Arulanandan K. J.. Fillunger P. J. 3rd. 1751 86. (1978) Numerical model for saturatedunsaturated flow in deformable porous media 3. 14. Applications. Vol. A h . J. 155. and Witherspoon P. Stockholm. Narasimhan T. E. A. Fillunger P. Rotterdam. and Willis P. 443448. N. Water Resources Res.' Int. Gkotechnique. 859863. A. 59&601. J. savants h a n g e r s . 164. 46. Mech. Biot M. 2. (1996) Highlights in the historical development of the porous media theory. GPotechnique. and Scott R . D. Miiller L. Reynolds 0. Inst. Acad. 7 . 0sterr. (1984) Modeling of centrifugal filtration in unsaturated deformable porous media.. Proc. (1947) Experiments on effective area in gravity dams. University of California. 354363. Leliavsky S. Liakopoulos A. G. E.s Review. Phys. Appl. 33. Neuman S. M. and Gibson R. Baudienst. (1957) The elastic coefficients of the theory consolidation. 40. 9th Euro. Con/. Balkema.. and Balakrishna J. part11higher frequency range. (1 9 13b) Der Aujirieh in Tul~sperren. (1987) Special problems of soil: general report. Biot M. 449447. Lloret A. Appl. (1948) The effect of uplift pressure on the shearing strength of concreteR. (1965) The Rock slide in the Vajont Valley. (1876) Essai theorique sur I'equilibre d'elasticite des massif pulverulents.48. 0sterr. F. a property of granular material. Mem.. Soe. A. (1955) Theory of elasticity and consolidation for a porous anisotropic solid.. Belgique. (1959) The Principle of Effective Stress. 1. (1913a) Der Ayftrieh in Tcrlsperrcv~. 28. Soil Meel?. Lyell C. I. 444. No. Gens A. 567570. (1956a) Theory of propagation of elastic waves in a fluidsaturated porous solid.. Wochenschrift offentl. 288.
15. 1st ICSMFE. Concretes. 104108.. 29. Laboratorio Nacional de Engenheria Civil. von and Rendulic L. (1934) Die wirksame Flachenporositat des Betons. Archit Ver. M. L. Proc. Lee K. 215263. F. GCotechnique. 19.. Eng. and Saukin W. Congr. 3rd Annual Municipal Solid Waste Res. Geotech. (1954) A subpresseo nos barreyensPubl. Z. Idriss I. Ing. Zienkiewicz 0. (1975) Dynamic analysis of the slide in the Lower San Fernando dam during the earthquake of February 9. Large Dams. B. Vol. (1947) The stress distribution in gravity dams.C. Inst.EPA6001977026. 3.. and Pinder G. Eng. Seed H. Con$ Pore Pressures and Suction in Soils. 889911.244271. Zienkiewicz 0. W. J. Van Genuchten M. 101. 17. 8th.u. Div. Proc. Pinder G.16 INTRODUCTION AND T H E CONCEPT OF EFFECTIVE S T R E S S Safai N. Edinburgh. B. 95103. Int. P. Seed H. (1979) Vertical and horizontal land deformation in a desaturating porous medium. 1971. 1 . No. (1979) Consideration in the earthquake resistant design of earth and rockfill dams. (1960) Effective Stress in Soils. Proc. (1963) Stress analysis of hydraulic structures including pore pressures effects. 1925. 9. 5 4 5 6 . Water Power. 416. I. 112. ASCE. C. L. Serafim J. 55.. Civ. 86. Skempton A. Lisbon. Terzaghi K . (1977) Modeling of leachate and soil interactions in an aquifer . and Makdisi F. No. von (1936) The shearing resistance of saturated soils. L. . Symp.. T. and Rocks. Water Resources 2. F. Adv. M. (1964) The 'uplift area' in plain concrete in the elastic rangeC. 27. V. J. Ost. No. Serafim J.. Terzaghi K ..
which are essential for the study of earthquakes. SoilPore Fluid.4 dealing with simultaneous water and air flow in the pores. reduce to those governing the quasistatic situations of consolidating soils and indeed to purely static problems without modification. problem contains all the essential features of soil behaviour and the equations embrace and explain the vast majority of problems encountered in practice. most common. This feature will be used when discretization is introduced and computer codes are derived. The same discussion will show the domain of the validity of the assumptions of undrained and fully drained behaviour. This.1 GENERAL REMARKS ON THE PRESENTATION In this chapter we shall introduce the reader to the equations which govern both the static and dynamic phenomena in soils containing pore fluids. Interaction 2.3 dealing with partially saturated soil in which the air pressure is assumed constant and also finally in Section 2. Section 2. The limitations of the approximating simplification are discussed in Section 2.2 will deal with soil.Equations Governing the Dynamic. The notation used throughout this chapter will generally be of standard.3 in 3 dimensions . Thus: ui will be the displacement of the solid matrix with i = 1.2 in 2 dimensions or i = 1. We shall show here how the dynamic equations.2 by using a simple linearized example and deriving conclusions on the basis of an available analytical solution. saturated with a single fluid. tensorial form. We shall divide the presentation into three Sections. since a single code will be capable of dealing with most phenomena encountered in soil and rock mechanics. or indeed any other porous medium. In the same section we shall introduce a simplification which is valid for the treatment of most lowfrequency phenomenaand this simplified form will be used in the subsequent Section 2.
For completeness. are the permeabilities for air and water flow. are the relative degrees of saturation and k. The equations derived were subsequently recast in varying forms. These velocities are calculated on the basis of dividing the appropriate flow by the total crosssectional area of the solidfluid composite.11) of Chapter 1 when two fluids are present. or E refers to the strain components. or w and v to denote the velocities of water and air relative to the solid components. Indeed. (1990a) any nonlinear behaviour can be specified for the skeleton and therefore realistic models can be incorporated. E. and Zienkiewicz et ul. the mixture theory establishes of course identical equations but in the author's view. 1956b and 1962) and Biot & Willis (1957) but we believe it is slightly easier to follow as it explores the physical meaning of each term.and a refer to the appropriate . Similarly. with a and a"being the vectorial alternatives. de Boer and Kowalski (1983) found it necessary to develop a special plasticity theory for porous. and v. Further pa. total and effective stresses. we shall use w. 1956a. we shall include such mixture derivations of the equations in Section 2. This is a slightly different approach from that used in the earlier presentations of Biot (1941. SOILPORE FLUID.5 If correctly used. It seems that despite much sophistication of the various sets of coupled equations.18 EQUATIONS GOVERNING T H E DYNAMIC. Other symbols will be added and defined in the text as the need arises. . and P = XwPw f X & will stand for air and water pressure and the 'effective' pressure defined in the effective stress concept in Equation (1. 1955. Green (1969) and Bowen (1976)). Indeed we shall find that such models are essential if practical conclusions are to be drawn from this work. Later it became fashionable to derive the equations in the forms of so called mixture theories (see Green & Adkin (1960). introduces some arbitrariness in the selection of various parameters. S. The derivation of the equations in this chapter follow a physical approach which establishes clearly the interactions involved in the manner presented by Zienkiewicz and Shiomi (1984). p. linear elastic. Sa. (1990a and 1990b) etc. INTERACTION Alternatively. a. saturated solids. the form will also be used for the same quantity in vectorial notation. Here an important step forward was introduced by Morland (1972) who used extensively the concept of volume fractions. A full discussion of the development of the theory is given in the paper by de Boer (1996). In the equations of Zienkiewicz (1982). most authors limited their works to conventional. Zienkiewicz et (11. As mentioned in the previous chapter. Derski (1978) introduced a different derivation of coupled equations and Kowalski (1979) compared the various parameters occurring in Derski's equations with those of Biot's equations. behaviour of the solid. and k . Zienkiewicz (1982).. Similarly.
However. The latter is generally neglected in soil problems. be written in the general form using an incremental definition du" = D(de . However.deO) (2. . We thus define the strains as In the above. however. This effective stress is conveniently used as it can be directly established from the total strains developed.2. o = 1 will be assumed. For soil mechanics problems. for such materials as concrete or rock the value of o in the first expression can be as low as 0. it should be remembered here this stress definition was derived in the first chapter as a corollary of using the effective stress defined as below which is responsible for the major part of the deformation and certainly for failure.2 FULLY SATURATED BEHA VIOUR WITH A SINGLE PORE FL UZD ( WATER) 2. of the previous chapter which we repeat below.1 Equilibrium and mass balance relationship (u. In soils. I).5 but experiments on tensile strength show that the second definition of effective stress is there very much more closely applicable as shown by Leliavsky (1947).3b) The vectorial notation used here follows that corresponding to stress components given in (I. to which we will devote most of the examples.FULLY S A T U R A T E D BEHA VIOUR W I T H A SINGLE PORE FLUID ( W A T E R ) 19 2. D is the 'tangent matrix' and deOis the increment of the thermal or similar autogeneous strain and of the grain compression mp/3Ks. Serafim (1954) etc. w and p ) We recall first the effective stress and constitutive relationships as defined in equations (1.16). the difference between the two effective stresses is negligible as cu z 1. Constitutive relationships will still.
Thus If vectorial notation is used.) However. this definition needs to be modified and we must write where the last two terms account for simple rotation (via the definition in (2. as is often the case in the finite element analysis. SOILPORE FLUID... INTERACTION If large strains are encountered. The strain and rotation increments of the soil matrix can be determined in terms of displacement increments du. Thus in the derivations that follow we shall do sothough their inclusion presents no additional computational difficulties and they are included in the computer program. 1903b)Jaumann (1905) stress changes. as and The comma in the suffix denotes differentiation with respect to the appropriate coordinate specified. not summed. . j + du. We omit here the corresponding vectorial notation as this is not easy to implement.. the so called engineering strains are used in which (with the repeated index of du...20 EOUATIONS GOVERNING THE DYNAMIC.6b) ) of existing stress components and are known as the Zaremba (1903a. the shear strain increments will be written as dyij = 2 d ~ i= du. The large strain rotation components are small for small displacement computation and can frequently be neglected.
u?.e. If again we consider the same unit control volume as that assumed in deriving .8) T [u. We can now write the overall equilibrium or momentum balance relation for the soilfluid 'mixture' as where w. In derivations of the above equation we consider the solid skeleton and the fluid embraced by the usual control volume: dx . pf is the density of the fluid. ay  We shall usually write the process of strain computation using matrix notation as d~ = Sdu where u = (2. dt In about wi (or w) is the average (Darcy) velocity of the percolating water. The underlined terms in the above equation represent the fluid acceleration relative to the solid and the convective terms of this acceleration. where ps is the density of the solid particles and n is the porosity (i. dz.FULLY SATURATED BEHA VIOUR WITH A SINGLE PORE FLUID ( W A T E R ) 21 dy. b is the body force per unit mass (generally gravity) vector and p is the density of the total composite i. ax + ddu.u. dy . = ddu. = etc. dw. This acceleration is generally small and we shall frequently omit it.9) And for two dimensions the strain matrix is defined as: with corresponding changes for three dimensions (as shown in Zienkiewicz and Taylor 1989). the volume of pores in a unit volume of the soil). The second equilibrium equation ensures momentum balance of the fluid..e. Further.] (2.
the total volumetric strain. The final equation is one accounting for mass balance of the flow.12) represent again the convective fluid acceleration and are generally small./K. can be written as Note that the underlined terms in (2..n)dp/Ks and (iv) the change in volume of the solid phase due to a change in the intergranular effective contact stress (4. the permeability k is used with dimensions of [length]3[time]/[mass] which is different from the usual soil mechanics convention k' which has the dimension of velocity.5 (drii + Ks g) Here KT is the average bulk modulus of the solid skeleton and E.. throughout this book.22 EQUATIONS GOVERNING THE DYNAMIC.e. = . Their values are related by k = kl/p.e. Adding all the above contributions together with a source term and a secondorder term due to the change in fluid density in the process we can finally write the flow conservation equation .: Siideii= d~~~ mTdc = (ii) the additional volume stored by compression of void fluid due to fluid pressure increase: ndp/Kf (iii) the additional volume stored by the compression of grains by the fluid pressure increase: (1 . This storage is composed of several components given below in order of importance: (i) the increased volume due to a change in strain i. [length]/[time].~ the augmented storage in the pores of a unit volume of soil by occurring in time dt. Also note that. Here we balance the flow divergence w. assuming the Darcy seepage law. i. SOILPORE FLUID.Bqd~.. INTERACTION (2.11) (and we further assume that this moves with the solid phase) we can write In the above we consider only the balance of the fluid momentum and R represents the viscous drag forces which.= og + Sup): .g' where p[ and g' are the fluid density and gravitational acceleration at which the permeability is measured.
3) define the behaviour of the solid together with its pore pressure in both static and dynamic conditions. or u. again the boundary is divided into two parts r. The boundary condition imposed on these variables will complete the problem. (2.15b) as or in vectorial form where In (2. We shall omit them from further consideration here. Equations (2. (2) For the fluid phase. we specify the total traction t . The velocities of fluid flow wi or w The displacements of the solid matrix u. Summarising. on which the values of p are specified and r.13) and (2. where the normal outflow w. for the overall assembly.is given.16) the last two (underlined) terms are those corresponding to a change of density and rate of volume expansion of the solid in the case of thermal changes and are negligible in general. the displacement ui(u). we can thus write . with n. (a.16) together with appropriate constitutive relations specified in the manner of (2. being the ith G) component of the normal at the boundary and G is the appropriate vectorial equivalence) while for I?.FULLY SATURA TED BEHA VIOUR WITH A SINGLE PORE FLUID ( W A T E R ) 23 This can be rewritten using the definition of cu given in Chapter 1 (1.1 I). a zero value for the normal outward velocity on an impermeable boundary).. These boundary conditions are: (1) For the total momentum balance on the part of the boundary I?. is prescribed (for instance. The unknown variables in this system are: The pressure of fluid (water).. p = p.. ( t )(or in terms of the total stress avn.
SOILPORE FLUID. that some typical soil constants are implied in the formulation. with no net outflow. For instance. we have (neglect= ing the last two terms which are of second order). INTERACTION and Further and It is of interest to note.e..16) that for undrained behaviour when w. 0 i. as shown by Zienkiewicz (1982). we note from (2.. and If the pressure change dp is considered as a fraction of the mean total stress change mTda/3 or d0../3 we obtain the so called B soil parameter (Skempton (1954)) as Using the assumption that the material is isotropic so that .24 EOUATIONS GOVERNING THE DYNAMIC.
13) and (2. Lambe and Whitman. as will be seen from the next section. it is convenient to reduce the number of variables by neglecting the apparently small (underlined) terms of equations (2. 1925. of course.1 1) (2.16) together with the auxiliary definition system can of course be used directly in numerical solution as shown by Zienkiewicz and Shiomi (1984).2. This system is suitable for explicit time stepping computation as shown by Sandhu & Wilson (1969) and Ghaboussi & Wilson (1972) and more recently by Chan et al.16) using the definition (2.11) which now can be eliminated from the and (2.2 Simplified equation sets (up form) The governing equation set (2. These contain the variable w. B has. for unsaturated soils. (1.1 1)) sTu. a value approaching unity for soil but can be considerably lower for concrete or rock.(w) system. (w).FULLY SATURATED BEHAVIOUR WITH A SINGLE PORE FLUID ( W A T E R ) 25 where KT is (as defined in Eq.13).pii + pb = 0 (2. The first equation of the reduced system becomes (from (2. (1991). the value will be much lower (Terzaghi. We now have. in implicit computation. 2. 1969 and Craig. omitting density changes . However.14) and thus eliminating the variable w. Further. where large algebraic equation systems arise.10) the bulk modulus of the solid phase and p is once again Lame's constant.20b) The second equation is obtained by coupling (2. 1992).13) and (2.
13). shortduration. In eliminating the variable w. It is fortunate that the inaccuracies of the up version are pronounced only in highfrequency. The resulting system which is fully discussed in Zienkiewicz and Shiomi (1984) is not written down here as we shall derive this alternative form in Chapter 3 using the total displacement of water U = uR u as the variable. However. Equation (2. We now can rewrite (2.11) and (2. phenomena. another possibility exists for obtaining a reduced equation set without neglecting any terms provided that the fluid (i. Table 2. as the water displacement U then increases indefinitely.(w). We define where the division by the porosity n is introduced to approximate the true rather than the averaged fluid displacement. water in this case) is compressible. With such compressibility assumed.e. SOILPORE FLUID. + . Here a very small time increment can be used for the short time period considered (See Chapter 3).1 summarises the various forms of governing equations used. INTERACTION This reduced equation system is precisely the same as that used conventionally in the study of consolidation if the dynamic terms are omitted or even of static problems if the steady state is reached and all the time derivatives are reduced to zero. some loss of accuracy will be evident for problems in which highfrequency oscillations are important. However.(w) we have neglected several terms but have achieved an elimination of two or three variable sets depending on whether the two or threedimensional problem is considered. these are of little importance for earthquake analyses. As we shall show in the next section. since for such problems we can conveniently use explicit temporal integration. Thus the formulation conveniently merges with procedures used for such analyses.16) can be integrated in time provided that we introduce the water displacement u ? ( u R ) in place of the velocity w.26 EQUATIONS GOVERNING THE DYNAMIC.16) after integration with respect to time as and thus we can eliminate p from (2. It presents a very convenient basis for using a fully explicit temporal scheme of integration (see Chan et al. 1991) but it is not applicable for longterm studies leading to steady state conditions.
pf [w + wvTw] + pb = 0  u .n)pii .16) and of the approximations (2.n ) v ( v T u )+ a e n V ( v T u ) . (2.direction (XI= x) and then we have where D is called the onedimensional constrained modulus. (1980).(1 .21). An attempt to answer these problems was made by Zienkiewicz et al.13).11) (2. (2. to define the behaviour of the material and when the simplified equation system discussed in the previous section is applicable. 0 (representing undrained = conditions) was also considered.2.n) QV(VTu) n Q v ( V T u ) . also .13) and (2. of course..p approximation for dynamics of lower frequencies.16)) sTg pii .72) sTu+ a Q ( a .1 Comparative sets of coupled equations governing deformation and flow u .20).3 Limits of validity of the various approximations It is.21)) sTu.. important to know the degree of approximation involved in various differential equation systems. The basis was the consideration of a onedimensional set of linearized equations of the full systems (2. Exact for consolidation ((2.k' ( n u . Thus it is of interest to know under what circumstances undrained conditions can be assumed. (2.p i J + pb = 0 ( a . only convective terms neglected (3.20) and (2.pf U + pf b In all the above E a" = u + amTp and d u " = D ~ = DSdu + =0 2.w .nu) . We consider thus that the only physical variation is in the vertical.1 I).pii + pb = 0 u . The limiting case in which w.FULLY SATURATED BEHAVIOUR WITH A SINGLE PORE FLUID ( W A T E R ) 27 Table 2. For all these conditions the exact solution of the equation is possible. without introducing serious error. X I .p equations (exact) ((2.U. E is Young's modulus and u is Poisson's ratio of the linear elastic soil matrix.
In Figure 2. INTERACTION The differential equations are.~ l'' and a system of ordinary linear differential equations is obtained in the frequency domain which can be readily solved by standard procedure. The reader will note that the results are plotted against two nondimensional coefficients: .16): with Taking K. + cc For a periodic applied surface load a periodic solution arises after the dissipation of the initial transient in the form = jjell"' P = ~ e ~ etc. Thus at s=Lu=Ow=Oandatx=O~. 1980) we show a comparison of various numerical results obtained by the various approximations: (i) exact equation (Biot's.p equation approximation (labelled Z) (iii) the undrained assumption (w = 0) and (iv) the consolidation equation obtained by omitting all acceleration terms.1.=qp=O.11): In place of (2. labelled C. labelled B) (ii) the u .13): and in place of (2.1 (taken from Zienkiewicz et al. The boundary conditions imposed are as shown in Figure 2. SOILPORE FLUID.28 EOUATIONS GOVERNING THE DYNAMIC. in place of (2.
is the speed of sound. We note that. The second nondimensional parameter is defined as In the study. g is the gravitational acceleration. with the wave speed taken as we have T = = 0. In the above where L is a typical length such as the length of the onedimensional soil column under consideration.1s the parameter 112is therefore in the range 3. subject to an earthquake in which the important frequencies lie in the range Thus. < and when 111> lo2 the drainage is so free that fully drained condition can be safely assumed. T is the natural vibration period and T is the period of excitation. the problem of the earthquake response of a dam in which the typical length is characterized by the height L = 50 metres. for instance. the following values were assumed: = p f / p = 0. for instance. To apply this table in practical cases. Consider.FULLY SATURATED BEHA VIOUR WITH A SINGLE PORE FLUID ( W A T E R ) 29 where k' and k are the two definitions of permeability discussed earlier.333. and Figure 2. n(porosity) = 0.2 Summarizes the conclusions by indicating three zones in which various approximations are sufficiently accurate.9 < II~ 39 < 2 . fully undrained behaviour is applicable when II. some numerical values are necessary.333.
.1 The soil column variation of pore pressure with depth for various values of rl and . SOILPORE FLUID.30 EQUATIONS GOVERNING THE DYNAMIC.z (up approximation theory) B c (Consolidation "2 theory) (Solution (C) is independent of x ) Reproduced from Zienkiewicz (1980) by permission z.. of the Institution of Civil Engineers  ..(Biot theory) . INTERACTION Figure 2.
Reproduced from Zienkiewicz(1980) by permission of the Institution of Civil Engineers AI = k p ~ . moderate speed (w can be neglected) Zone 111 B # Z # C. with reasonable confidence: (i) assume fully undrained behaviour when I I l = 97k' < lop2 or the permeability k' < rnls. (This is a very low value inapplicable for most materials used in dam construction). P = pf / p .3.PARTIALLY SATURATED BEHAVIOUR WITH AIR PRESSURE NEGLECTED Drained (influence of z. (ii) We can assume up approximation as being valid when k' < m/s to reproduce the complete frequency range.Pkf /pin .1.5 s are still well modelled. B = Z # C. In the general case of . f = 2L/ V. n 0. . k . typically use the up formulation appropriately in what follows reserving the full form for explicit transients where shocks and very high frequency are involved. However. we have considered both the water pore pressure and the solid displacement as problem variables.2 Zones of sufficient accuracy for various approximations: Zone 1. P 0. when k' < lo' m/s periods of less then 0. Z /= 2~ 2 T / ~ f 2 ~ kp A2 = J L ~ / V .kinematic permeability. slow phenomena (w and u can be neglected) Zone 2. analysis. negligible) 31 Figure 2.1 Why is inclusion of partial saturation required in practical analysis? In the previous.33   and Ill is dependent on the permeability k with the range defined by According to Figure 2. We shall. Z = k=&lpg. therefore. V: = ( D + k f / n ) / p .kf / p l (speed of sound in water).33. fully saturated. Definition as in Figure 2. 2.3 PARTZALL Y SA TURA TED BEHA VZOUR WITH AIR PRESSURE NEGLECTED (pa = 0) 2. fast phenomena (w cannot be neglected only full Biot equation valid).2 we can. B = Z = C.
This is particularly true above the free water surface or the so called phreatic line. noting that generally we shall consider partial saturation only in the slower phenomena for which up approximation is permissible. a unique relationship exists between the degree of saturation Swand the pore pressures pw (see Figure 1. INTERACTION nonlinear nature which is characteristic for the problems of soil mechanics both the effective stresses and pressures will have to be determined incrementally as the solution process (or computation) progresses step by step. . This means that the history of straining (associated generally with shear strain) induces the solid matrix to contract (or the material to densify).2 The modification of equations necessary for partially saturated conditions The necessary modification of equations (2. However. leading finally to a decrease of contact stresses in the soil particles to nearzero values when complete liquefaction occurs. Indeed generally failure will occur prior to the liquefaction limit. Using this relation. This opening of voids will probably occur when zero pressure (or corresponding vapour pressure of water) is reached.32 EQUATIONS GOVERNING THE DYNAMIC. Note that both phenomena of densification and dilation will be familar to anybody taking a walk on a sandy beach after the tide has receded leaving the sand semisaturated.21) will be derived below. of course. Clearly liquefaction has occurred. Alternatively air will come out of solutionor indeed ingress from the free water surface if this is open to the atmosphere. and consequent capillary effects. Usually one is tempted to assume simply a zero pressure throughout that zone but for noncohesive materials this means almost instantaneous failure under any dynamic load. once voids open. The presence of negative pressure in the pores assures some cohesion (of the same kind which allows castles to be built on the beach provided that the sand is damp). Such densification usually will cause the pore pressure to increase. However if the pressure is not removed but reapplied several times the sand 'densgies' and becomes quickly almost fluid. This will imply development of negative pressures which may reach substantial magnitudes. Voids will therefore open up during the process in the fluid which is essentially incapable of sustaining tension.20) and (2.3. This cohesion is essential to assure the structural integrity of many embankments and dams. This obviously is the dilation effect. We have shown in Chapter 1 that. First note how when the foot is placed on the damp sand. SOILPORE FLUID.6). the material appears to dry in the vicinity of the applied pressure. In many soils we shall encounter a process of 'densification' implied in the constitutive soil behaviour.2 to deal with the problem of partial saturation without introducing any additional variables assuming that the air throughout is at constant (atmospheric) pressure. increase the strength of the soil and thus have a beneficial effect. Such negative pressures cannot exist in reality without the presence of separation surfaces in the fluid which is contained in the pores. The pressure will not then be vapour pressure but simply atmospheric. 2. which can be expressed by formula or simply a graph we can modify the equation used in Section 2. the reverse may occur where the soil 'dilation' during the deformation history is imposed. It is surprising how much one can learn by keeping one's eyes open! The presence of negative water pressures will.
we have neglected the relative acceleration of the fluid to the solid Equation (2. we must note that the effective stress definition is modified and the effective pressure now becomes (viz ch. isotropic. 1 sect. to be augmented by terms previously derived (and some additional ones). The mass balance will once again consider the divergence of fluid flow w.. and we note that: Such typical dependence is again shown in Figure 1.e. We note that its form remains unchanged but with the variable p being replaced by p.14).. permeability will be used here where I is the identity matrix.PARTIALLY SA TURA TED BEHA VIOUR WITH AIR PRESSURE NEGLECTED 33 Before proceeding.. will be considered.7 of Chapter 1.1). the water momentum equilibrium.21) will now appear in a modified form which we shall derive here. the conservation Equation (2. The correction is obviously small and its effect insignificant.7) we can write neglecting the weight of air. The value of k is. These are . However (2. Equation (2.20) remains unaltered in form whether or not the material is saturated but the overall density p is slightly different now.3) with the effective stress still defined by (2. defining the permeabilities. i. We thus have As before. however.16) has to be restructured though the reader will recognize similarities. 1. Finally. Equation (2. in general.8). only scalar. First. Thus in place of (2.3. dependent strongly on S. remains unchanged as However.
(neglecting density variation): .e. SOILPORE FLUID.. (ii) An additional volume stored by compression of the fluid due to fluid pressure increase: nSwdpw/Kf (iii) change of volume of the solid phase due to fluid pressure increase: (1 .17) when Sw= 1 and xW 1 i. on thermal expansion. when we have full saturation.21) gives.n)xwdpw/Ks (iv) change of volume of solid phase due to change of intergranular contact stress: KT/K. .) + (v) and a new term taking into account the change of saturation: ndSw Adding to the above.34 EOUATIONS GOVERNING THE D YNAMZC. the terms involving density changes. as in Section 2. INTERACTION (i) Increased pore volume due to change of strain assuming no change of saturation: Giideii = d ~ . This term is due to the changes of the degree of saturation and is simply: but here we introduce a new parameter Cs defined as The final elimination of w in a manner identical to that used when deriving (2.(~E.17). however. The above modification is mainly due to an additional term to those defining the increased storage in (2. place (? is different from that given in Equation (2. xwdpw/K.17) and we have in its = which of course must be identical with (2. the conservation equation now becomes: Now.2.
.33~) The small changes required here in the solution process are such that we shall find it useful to construct the computer program for the partially saturated form. on all the surfaces exposed to air. subsequently. net zero inflow) Clearly both conditions cannot be simultaneously satisfied and it is readily concluded that only the second is true above the area where the flow emerges. the reverse is true.3 A partially saturated dam. we still always assume that the parameters S.e. we mentioned several typical cases where pressure can become negative and hence saturation drops below unity. Previously. constant.swpfiij P + S. Figure 2. Only saturation (a) and pressure contours (b) are shown. we have apparently incompatible boundary conditions. Initial steadystate solution.3 we note that.PARTIALLY SATURATED BEHAVIOUR WITH AIR PRESSURE NEGLECTED 35 ( k j . with the fully saturated form being a special case. If we consider a typical dam or a water retaining embankment shown in Figure 2. and C. change slowly and hence we will compute these at the start of the time interval keeping them. In the timestepping computation.p . One frequently encountered example is that of the flow occuring in the capillary zone during steady state seepage.pfbj)). Of course when the flow leaves the free surface. The Phreatic line is the boundary of the fully saturated zone in (a) . These are: p. = 0 (i. . k . . Contour interval in (b) is 75 kPa.+aEii + Q* + SO = 0  (2. . The solution of the problem can of course be obtained from the general equations simply by neglecting all acceleration and fixing the solid displacements at zero (or constant) values.. = 0 and w.
2.4 we shall show how the effects of air movement can be incorporated into the analysis. recorded. Another example is given in Figure 2. 1977a and b. However. In the practical code used for earthquake analysis we shall use this partially saturated flow to calculate a wide range of soil mechanics phenomena. SOILPORE FLUID.. the procedures introduced are readily . a partially saturated zone with very low permeability must exist. E = 3000 kPa) Computation will easily show that negative pressures develop near the surface and that. Desai and Li 1983 etc.1 PARTIALLY SATURATED BEHA VIOUR WITH AIR FLOW CONSIDERED (pa > 0) The governing equations including air flow This part of the chapter is introduced for completenessthough the effects of the air pressure are insignificant in most problems. However.4.36 EQUATIONS GOVERNING THE DYNAMIC. Here a numerical solution of Zienkiewicz et a[. therefore.. .4 Test example of partiallysaturated flow experiment by Liakopoulos (1965).4. (c) data (linear elastic analysis. INTERACTION Figure 2. in some cases of consolidation and confined materials. The result of such a computation is shown in Figure 2. the air pressures play an important role and it is useful to have means for their prediction.. This contour is in fact the well known phreatic line and the partially saturated material procedure has indeed been used frequently purely as a numerical device for its determination (see Desai. (1990b) is given for a problem for which experimental data are available from Liakopolous (1965). for completeness in Section 2. (a) configuration of test (uniform inflow interrupted at t = 0) (b) pressures with . computed. Further.3 and indeed it will be found that very little flow occurs above the zeropressure contour.4 2.).
For completeness we repeat that equation here (now. a priori omitting the small convective terms) However. the simultaneous presence of water and oil is important in some areas of geomechanics and coupled problems are of importance in the treatment of oil reservoirs. The alternative approach of using the mixture theory in these problems was outlined by Li and Zienkiewicz (1992) and Schrefler (1995). be written in precisely the f same form as that for a single fluid phase (see (2. just as in Section 2. The procedures used in the analysis follow precisely the same lines as introduced here.1 1)).2 The governing equation The dynamics o the total mixture can.25) we have to write noting that For definition of effective stress we use again (2. For instance. just as in Equation (2. without equating the air pressure to zero i. In particular. Some simple considerations will allow the basic equations for the dynamic of the soil containing two pore fluids to be derived.e.3. Li et al. however.24) now. writing P = xwpw + (1 . noting that for water as in (2. 2. (1990) for flow of water and oil and Schrefler and Zhan (1993) for flow of water with air.27) and for the flow of air: .PARTIALLY SATURATED BEHA VIOUR WITH AIR FLOW CONSIDERED 37 applicable to other porefluid mixtures.xw)pa (2.36) For the flow of water and air we can write the Darcy equations separately.4. the treatment following the physical approach used in this chapter has been introduced by Simoni and Schrefler (1991). however.
6 of Chapter 1.30). pa K a + Pa + . aSW&+ Swpw K w + an Pw +xwpw+ nSw + nSw.13) can be rewritten in a similar manner using isotropy but omitting acceleration terms for simplicity.. p. in addition to the solid phase displacement u. SOILPORE FLUID. However. The approximate momentum conservation equations (see 2. w. while assuming isotropy.a p a nSa + nSa. we note that now (see (1.44) the air saturation can also be found.42b) Now.38 EQUATIONS GOVERNING T H E DYNAMIC. and Sw is unique and of the type shown in Figure 1.+Kw and for air an Pw PW KS xwpW+ nSw nSw. INTERACTION Here we introduced appropriate terms for coefficients of permeability for water and air.41~) (2. aSaii.+ i0= 0 ~ + Pa an Ks (2.+ S.(u).42~) VTv n an + crSami.+ S. now defines Sw and from the fact that Sw + S.+Swt).. Thus for water we have n .we have to consider the water presure pw and the air presure pa as independent variables. the mass balance equations for both water and air have to be written. pa + xada + nSa + nSaPa + so = 0 K a Ks  Pa (2. = 1 and xW+ xa = I (2.16)) and that the relation betweenp. . These are derived in a manner identical to that used for equation (2. We therefore have for water and for air Finally.41b) n VT w+crSwmi. A new variable v now defines the air velocity..So + + =0 n v..+ So = 0 Ks Pw (2.
Under assumption of local thermodynamic equilibrium. 2. All these theories lead to a similar form of the balance equations. This means that cu in equation (1.46) can be derived following Hassanizadeh and Gray (1990).e. it appears that where E is the time rate of change of the deformations. 1990).1982).1a) g ! = g.22. In the hybrid mixture theories. which results in Only in this case the two formulations given by Biot's theory and the mixture theory coincide. is in the constitutive equations. This was in particular shown by de Boer et al. Morland (1972)) start from the macromechanical level.16a) or (2. 11 + VP (2. the hybrid mixture theories and the classical Biot's theory. the level of interest for our computations. i. For instance. while the so called hybrid mixture theories (viz. Let's consider first the fully saturated case.1 that the governing equations can alternatively be derived using mixture theories. it follows that and the effective stress principle in the form of (2. 1980. The equations at macromechanical level are then obtained by spatial averaging procedures. Where the theories differ. k is a coefficient ad 8. the absolute temperatures of solid and water respectively.5 ALTERNA TIVE DERIVA TION OF THE GOVERNING EQUA TION (OF SECTION 2. The classical mixture theories (see Green (1969) and Bowen (1980. the pressure acting on the solid grains is introduced (Hassanizadeh and Gray. the concept of the solid pressure ps i. 1990) start from micromechanical level.45) has to be equal to one. Whitaker (1977) Hassanizadeh and Gray (1979a. (199 1) for the mixture theory. 197913. This is shown here in particular for the effective stress principle. Runesson (1978) shows that the principle of effective stress follows from the mixture theory under the assumption of incompressible grains.e.THEORY 39 We have now the complete equation system necessary for dealing with the flow of air and water (or any other two fluids) coupled with the solid phase deformation.48) holds also under nonequilibrium . ! r.. Further there exists a macroscopic thermodynamical approach to Biot's theory proposed by Coussy (1995). Equation (2. and 8. usually obtained from the entropy inequality. From the application of the second principle of thermodynamics.4) BASED ON THE HYBRID MIXTURE THEORY It has already been indicated in Section 2. because this was extensively discussed in Chapter 1.
see Equations (2. obtains where is the capillary pressure increment.20) (1.51) coincides with that of equation (2.22a. to be assumed that the solid grains remain incompressible. a = 1 this ofcourse limits the applicability. The first expression for partially saturated soils was developed by Bishop (1959) and may be written as follows. This equation has hence an incremental form and differs substantially from the previous ones. The formulations of the effective stress principle in finite form coincide if the Bishop parameter x = Sw and the solid grains are incompressible. but with Bishop's parameter equal to the water degree of saturation was derived by Lewis and Schrefler (1987) using volume averaging. 2. Coussy (1995). SOILPORE FLUID.e. Let us consider now the partially saturated case. There are again several conflicting expressions in literature. The same expression.19) where X.24 and 1.19) (1. Finally. using the ClausiusDuhem inequality. This assumption is now made and the governing . usually a function of the degree of saturation. the same assumption as with the mixture theory. is the Bishop parameter.49) if xW=Sw but this is of course not often the case. see (1. it has. Hassanizadeh and Gray (1990) find for the partially saturated case that which considering thermodynamic equilibrium conditions or nonequilibrium conditions but incompressible solid grains reduces to Taking into account that the solid pressure (2. It was concluded that in practical soil mechanics situations the resulting differences are small and appear usually after long lasting variations of the moisture content.36. i. INTERACTION conditions. Only several cycles of drying and wetting would produce significant differences. however. The practical implication of these different formulations for slow phenomena have been investigated in detail by Schrefler and Gawin (1996).40 EQUATIONS GOVERNING T H E DYNAMIC.b).
2. x. however.e.THEORY 41 equations are derived again. as has been done by Schrefler (1995) and Lewis and Schrefler (1998). 2. . In a Lagrangian or material description of motion. X" and of the current time t x = x. t ) given in its spatial description and referred to a moving particle of the 7r phase is If superscript a is used for phase. its inverse can be written and the Eulerian or spatial description of motion follows The material time derivative of any differentiable function f"(x. The state of motion of each phase is. Because of the nonsingularity of the Lagrangian relationship (2. For the full nonisothermal case the interested reader is referred to Lewis and Schrefler (1998). the position of each material point x" at time f is a function of its placement in a chosen reference configuration.. in the current configuration each spatial point x is simultaneously occupied by material points X" of all phases.1 Kinematic equations As indicated in Chapter 1. described independently.. 7r = 1. V" the left stretch tensor. i. using the hybrid mixture theory. a multiphase medium can be described as the superposition of all 7r phases. the Jacobian of this transformation must not equal zero and must be strictly positive.. since it is equal to the determinant of the deformation gradient tensor FT where U" is the right stretch tensor. . We first recall briefly the kinematics of the system. t To keep this mapping continuous and bijective at all times. the time derivative is taken as moving with the a .55). K . as throughout this book. and the skewsymmetric tensor R" gives the rigid body rotation. (x. Isothermal conditions are assumed to hold. . ) x.5.e. The differentiation with respect to the appropriate coordinates of the reference or actual configuration are respectively denoted by comma or slash. . i.
The balance equations have here been specialized for a deforming porous material.g the external supply of @ and G is the net production of Q. energy and entropy. The temperatures of each constituent at a generic point are hence equal. as here assumed. The constituents are assumed to be microscopically nonpolar. and for air qa = nS. momentum. SOILPORE FLUID. the microscopic situation of any .3 Macvoscopic balance equations For isothermal conditions..2 (Hassanizadeh and Gray (1980. In the averaging procedure the volume fractions q" appear which are identified as follows: for the solid phase qs = 1 . however.60) as outlined in Hassanizadeh and Gray (1979a. where flow of water and moist air (mixture of dry and vapour) is taking place (see Schrefler. g and G are given in Table 2. As throughout this book.q cP g 0 g g. i is the flux vector associated with a..i+h G 0 0 0 S P . The local thermodynamic equilibrium hypothesis is assumed to hold because the timescale of the modelled phenomena is substantially larger than the relaxation time required to reach equilibrium locally. At the interfaces to other constituents.r 0 t "I tmr .5. Table 2.42 EQUATIONS GOVERNING THE DYNAMIC.n: for water qw = nS. The values assumed by i.ir phase is first described by the classical equations of continuum mechanics.ir phase may be written as where i is the local value of the velocity field of the .ir phase in a fixed point in space. that the stress tensor is symmetric. 1990) and Schrefler (1995)). linear momentum and angular momentum are then obtained by systematically applying the averaging procedures to the microscopic balance equations (2. 1980).2 Micvoscopic balance equations In the hybrid mixture theories. 2. the constituents are assumed to be immiscible and chemically nonreacting. INTERACTION 2. the macroscopic balance equations for mass. The relevant thermodynamic properties @ are mass. while pore pressure is defined as positive compression for the fluids. For a thermodynamic property @ the conservation equation within the . 1995). Further. hence the angular momentum balance equation has been omitted. the material properties and thermodynamic quantities may present step discontinuities. All fluids are assumed to be in contact with the solid phase.5. stress is defined as positive tension for the solid phase.5r.2 Thermodynamic properties for the microscopic mass balance equations Quantity Mass Momentum Energy Entropy P i 1 i E X + 0. 1979b. This equation shows.
For the solid phase the linear momentum balance equation is hence The average angular momentum balance equation shows that for nonpolar media the partial stress tensor is symmetric ti: = t.] =0 where t" is the partial stress tensor.ir phase averaged over the total control volume. this equation reads where va is the mass averaged air velocity. and accounts for the exchange of momentum due to mechanical interaction with other phases..ir phase averaged over the part of the control volume (Representative Elementary Volume. . REV) occupied by the . The phase averaged density p. The linear momentum balance equation for the fluid phases is t.) + i. on the contrary.  a.) + p" Len(pi. + p"(b. p"a" the volume density of the inertia force. p"b" the external momentum supply due to gravity. at macroscopic level also and the sum of the coupling vectors of angular momentum between the phases vanishes. pke"(pr) is assumed to be different from zero only for fluid phases. The intrinsic phase averaged density p" is the density of the .ir from its mass averaged velocity. lost through evaporation and vw the mass averaged water velocity. For air..THEORY 43 The averaged macroscopic mass balance equations are given next.. The relationship between the two densities is given by For water we have where nSwpwew(p)= m is the quantity of water per unit time and volume. p"eT(pr) the sum of the momentum supply due to averaged mass supply and the intrinsic momentum supply due to a change of density and referred to the deviation i" of the velocity of constituent . is the density of the . For the solid phase this equation reads where u is the mass averaged solid phase velocity and p" is the intrinsic phase averaged density.rr phase.
SOILPORE FLUID. as already explained. Further. The equation of a perfect gas is hence valid where M . = pwSw + p a s a in the case of thermodynamic equilibrium or for incompressible solid grains. is the molar mass of constituent T . and in the solid phase is S r. separated from its vapour by a concave meniscus because of surface tension.' = and KT is defined by the following relation .68) gives the total stress a. is the fluid pressure.44 EOUA TIONS GOVERNING THE DYNAMIC. R the universal gas constant. and Q the common absolute temperature. = a .(1 . see Hassanizadeh and Gray (1980. It is assumed that R" is invertible. (2.68) with p. INTERACTION 2. It can be shown that the stress tensor in the fluid is where p. dry air and water vapour. its inverse being ( R " ) .67)..54). 1990). The momentum exchange term of the linear momentum balance equation for fluids has the form where v"5s the velocity of the r phase relative to the solid..pw.51). Most of them have been obtained from the entropy inequality. acting on a unit area of the volume fraction mixture This is the form of the effective stress principle employed in the following. written for air and water and of (2. The capillary pressure is defined as pc = pg . see Equation (2. Moist air in the pore system is assumed to be a perfect mixture of two ideal gases.5.4 Constitutive equations Constitutive models are here selected which are based on quantities currently measurable in laboratory or field experiments and which have been extensively validated. Dalton's law applies and yields the molar mass of moisture Water is usually present in the pores as a condensed liquid. with T = ga and r = g w respectively.n)pshii !I 1 11 (2. The sum of (2.
73) and (2.76) written for water and air. The effects of phase change are neglected and a vector identity for the divergence of the stress tensor is used. Finally.5 General field equations The macroscopic balance laws are now transformed and the constitutive equations introduced. By summing this momentum balance equation with equation (2. the following holds.67) and (2.67). to obtain the general field equations.69) assuming continuity of stress at the fluidsolid interfaces and by introducing the averaged density of the multiphase medium.5. k the intrinsic permeability and T the temperature above some datum. taking into account equations (2.65) the fluid acceleration is expressed. is the dynamic viscosity. In the case of more than one fluid flowing. where Kw is the bulk modulus of water. Equations (2. we obtain the linear momentum balance equation for the whole multiphase medium .74) are included. The linear momentum balance equation for the fluid phases is obtained first.THEOR Y 45 K$ = (p". and by taking into account the definition of total stress (2. the intrinsic permeability is modified as where k is the relative permeability. and introducing the relative fluid acceleration a"" Further. a function of the degree of saturation. equations (2. The terms are dependent on the gradient of the fluid velocity.59). 7 " ) T ) PT k. taking into account Equation (2.68) instead of (2. 2. For the ' " water density. yielding The linear momentum balance equation for the solid phase is obtained in a similar way. In Equation (2.72) are introduced. where p.
are introduced. For compressible grains. Further. see equation (2. For incompressible grains.75).89) and related remarks. after differentiation and dividing by pS. and kwii =!w J k. The macroscopic mass balance equation for the solid phase (2.krw and for air . as in Chapter 2. as assumed here. further simplifications are needed. is obtained as This equation is used in the subsequent mass balance equations to eliminate the material time derivative of the porosity.46 EQUATIONS GOVERNING THE DYNAMIC. yields the equilibrium equation (2.61). Then the derivatives are carried out. where the reference configuration is the last converged configuration of the solid phase. can be written for water where qWv? = W .63) is transformed as follows.4. Because of this we can neglect the convective terms in all the balance equations. in the linear momentum balance equation (2.79).78) the relative accelerations of the fluid phases with respect to the solid phase.0. SOILPORE FLUID. Neglecting. which are introduced next.1. the strain increments within each time step are small. the quantity of water lost through evaporation is neglected and the material time derivative of the porosity is expressed through Equation (2. An updated Lagrangian framework is used.28a) The linear momentum balance equation for fluids (2. yielding The mass balance equation for air is derived in a similar way To obtain the equations of Section 2. DSPS~r . INTERACTION The mass balance equations are derived next. First Equation (2.76) by omitting all acceleration terms. the material time derivative of the water density with respect to the moving solid phase and the relative velocity vWS. The mass balance equation for water (2.
35a) for incompressible grains except for the source term and the secondorder term due to the change in fluid density. This coincides with Equation (2. developing the divergence term of the relative velocity and neglecting the gradient of water density.75) Similarly. The mass balance equation for water is obtained from Equation (2. However. 2. dividing by pw.5 As this section does not follow the notations used in the use of the book. then the mass balance equations = are obtained in the same form as in Chapter 2 (with xW S. Finally. if for the solid phase the following constitutive relationship is used (see Lewis and Schrefler (1998)) where Ks is the bulk modulus of the grain material. this is not in agreement with what was assumed for the effective stress.).80).5. have been used.rr phase aTS acceleration relative to the solid. This yields where the first of Equations (2. This last one could be introduced in the constitutive relationship (2. Similar remarks as for the water mass balance equation apply. . the constitutive relationships for moist air. Equations (2. the mass balance equation for air becomes where again the first of equations (2.70) and (2.84) has been taken into account.71). taking into account the reference system chosen. In particular. we summarise below for purposes of nomenclature a" mass averaged acceleration of .86) has been taken into account and the gradient of the water density has been neglected.6 (Nomenclature) for Section 2.NOMENCLA TURE 47 where The phase densities appearing in Chapter 2 are intrinsic phase averaged densities as indicated above.
48 EQUATIONS GOVERNING THE DYNAMIC. macroscopic pressure of the . solid pressure pw pressure of liquid water p. capillary pressure p. microscopic stress tensor t" partial stress tensor 2" exchange of momentum due to mechanical interaction of the .rr phase with other phases u mass averaged solid phase velocity Un right stretch tensor vm velocity of the . SOILPORE FLUID. dry air pressure p.rr phase with respect to the solid phase vw mass averaged water velocity va mass averaged air velocity . molar mass of constituent K n porosity p. vapour pressure pa air pressure p.ir phase i local value of the velocity field R universal gas constant R constitutive tensor R" rigid body rotation tensor S intrinsic entropy source Sw degree of water saturation Sa degree of air saturation t current time t .ir phase KT permeability tensor Kw water bulk modulus m mass rate of water evaporation M . INTERACTION b external momentum supply material time derivative E specific intrinsic energy F" deformation gradient tensor f" differentiable function 6 fji fi aflaxi aflaxi g external momentum supply related to gravitational effects G net production of thermodynamic property P h intrinsic heat source i flux vector associated with thermodynamic property P k constitutive coefficient k intrinsic permeability tensor kr" relative permeability of .
Phys.CONCLUDING REMARKS v volume averaged water relative velocity w volume averaged air relative velocity V" left stretch tensor x" material point X" reference configuration E linear strain tensor @ entropy flux p increase of entropy 7" volume fraction of the T phase xw Bishop parameter X specific entropy p" dynamic viscosity 0. 168178.rr phase p. Phys. Biot M. part 11lowfrequency range. Biot M. In the next chapter. J. phase averaged density of the T phase p microscopic density a effective stress tensor ' 9 generic thermodynamic property or variable Super or subscripts ga = dry air gw = vapour a = air w = water s = solid 49 2. 12. No. 28. J. (1956b) Theory of propagation of elastic waves in a fluidsaturated porous solid. A.6 CONCLUDING REMARKS The equations derived in this chapter together with appropriately defined constitutive laws allow (almost) all geomechanical phenomena to be studied. absolute temperature of constituent T p averaged density of the multiphase medium p" intrinsic phase averaged density of the . 2. J. Soc. 179191. No. (1941) General theory of threedimensionalconsolidation. A. Acoust. Soc. (1956a) Theory of propagation of elastic waves in a fluidsaturated porous solid.. A. Am. Biot M . Appl. 2. 182185. Appl.. J. A. we shall discuss in some detail approximation by the finite element method leading to their solution. REFERENCES Biot M. Am.. . 26. Acoust.155164. (1955) Theory of elasticity and consolidation for a porous anisotropic solid. part Ilowfrequency range.. 28.
Kowalski S. SOILPORE FLUID. Wcrter Res.G. Acud. 22. J .. 0 . A. J. 191203. De Boer R. (1983) A Residual Flow Procedure and Application for Free Surface Flow in Porous Media. J.. 169186.. J. Eng. Appl. and Willis P. (1962) Mechanics of deformation and acoustic propagation in porous media. Polish Sci. Appl. Desai C. J. A.428438. De Boer R. Eng. (1969) On basic equations for mixtures. Phys. E.. J. Chichester. and Gray W. Int. 14821498. Craig R. J. S. (1979 a). Hassanizadeh M. Leipzig. Hassanizadeh M. Wiley & Sons. BUN. (1980) General conservation equations for multiphase systems: 3 Constitutive theory for porous media flow. and Adkin J. Bull.M. Nurn. Bishop A. 4.G. G. (1995) Mechanics of Porous Media. 24. and Gray W. Water Res. Desai C. Tech. J. UniversitaetGesamthochschule Essen. stress. Eng. Forschungsbericht aus dem Fachbereich Bauwesen. 47.. (1979 b) General conservation equations for multiphase systems: 2 Mass. Famiyesin 0 .. (1978) Equations of motion for a fluid saturated porous solid. Mech. 14151418. 2735. (1991) Porous Media. a Survey of Different Approaches. and Plischka J. 21. F. and Muir Wood D. momenta. 881887. 1. 201262. 131144. 947963. London. Tech. 11. De Boer R. (1980) Incompressible porous media models by use of the theory of mixtures. Sci. Adv. Oxford University Press. energy and entropy equations.M.. 18. 49. (1976) Theory of Mixtures in Continuum Physics. (1991) A Fully Explicit uw Scheme for Dynamic Soil and Pore Fluid Interaction. 1116. (1977b) Finite Element. 27. 2 5 4 0 . and Gray W. Mech. Rev. Math. C.W. L. Adv. 20. 3. New York. H. residual schemes for unconfined flow.. and Li G. Kowalski S. Academic Press. Water Res. Appl. 33. 13. London. (1992) Soil Mechanics (5th edn). Int. (1989) Poroese Medien. Sci. Green A. 54. 2.. J. Ehlers W. (1959) The principle of~ffective Bowen R. E. Acad. J. 11291148. Chapman & Hall. Green A.. 6 . Ehlers W. 98. Bowen R. J. (1905) Die Grundlagen der Bewegungslehre von einem modernen Standpunkte aus. 697735.. UniversitaetGesamthochschule Essen. 13431357. ASCE E M .. 39. 1113 Dec. General conservation equations for multiphase systems: 1 Averaging procedure. Hassanizadeh M. (1990) Mechanics and Thermodynamics of multiphase flow in porous media including interphase transport. (1957) The elastic coefficients of the theory consolidation. Derski W. 59&6O 1. 808 I. (1979) Comparison of Biot's equation of motion for a fluid saturated porous solid with those of Derski.. (1977a) DiscussionFinite element. (1960) Large Elastic Deformations and Nonlinear Continuum Mechanics. S.G. residual schemes for unconfined flow. Ghaboussi J. Quart. (1982) Compressible porous media models by use of theories of mixtures. and Kowalski S. Hassanizadeh M. Coussy 0 . Biot M. Mech. and Gray W. Int. Meth. Chan A.. Adv. Int. E. Meth. INTERACTION Biot M.. EM4. .. 26. Desai C. S.. and Wilson E. Jauman G.G. 2. (1983) A plasticity theory for fluid saturated porous solids. 10. 455461. Water Res. Adv.. APCOM Hong Kong. Adv.50 EQUA TIONS GOVERNING T H E DYNAMIC. 859863. Eng. Forschungsbericht aus dem Fachbereich Bauwesen. C. (1972) Variational formulation of dynamics of fluid saturated porous elastic solids. No. No. Water Res. Polish Sci. Teknisk Ukeblad. Int. Nurn. Eng. M. Bowen R. Appl. Sci. (1996) Highlights in the historical development of the porous media theory.
(1995) Finite Elements in Environmental Engineering: Coupled thermohydromechanical process in porous media involving pollutant transport. (SI Version). Terzaghi K.W. W. Li X. Chalmers University of Technology. Res.D.. M. 30. 154. Int. (1969) Soil Mechanics. Acad. undrained. 112. Crucovie. L. John Wiley & Sons. Chichester. and Whitman R. Reponse a M. J. Sandhu R. Water Resources Res. Zienkiewicz 0. 7196.Volume I : Baslc Formulation and Linear Problems (4th edn). Eng. Am.. K. Ph. 785814. (1989) The Finite Element Method . 444. Int.. Zienkiewicz 0. Whitaker S. (1991) A staggered finite element solution for water and gas flow in deforming porous media. C. Int. Appl. Lewis R. and Schrefler B. C.Eng. (1978) On nonlinear consolidation of soft clay. S. Cracovie.. No. Sci. (1954) The pore pressure coefficients A and B. 95. Skempton A. Nurn. Commun. J. 13. (1903b) Sur une generalisation de la theorie classique de la viscosite. and Zhan X. Num. and Xie Y. and Taylor R. A. in Geomech. Chichester. NO. Vienna. Trans. 2.REFERENCES 51 Lambe T. Archives of Computer Methods in Engineering. T. 45. 1. Leliavsky S.. ASCE EM. Natanson. Meth. Sci. Dissertation. Meth. Lewis R. and Bettess P.W. Academic Press. A. Schrefler B. and Schrefler B. and Wilson E.. Struct. Lisbon. (1993) A fully coupled model for waterflow and airflow in deformable porous media. Meth.. 20. 385395. (1947) Experiments on effective area in gravity dams. (1903a) Le priniple des mouvements relatifs et les equations de la mecanique physique. No. Zienkiewicz 0. 8. Liakopoulos A. Geophys. (1984) Dynamic Behaviour of saturated porous media: The generalized Biot formulation and its numerical solution... 641652. 890900. Zienkiewicz 0. Soc. Geom. Chang C. C. New York. D. von (1925) Erdbaumechanik auf bodenphysikalischer Grundlage.A. A. London.A. Anal. Num. Simoni L. Schrefler B. Zaremba S. W. Berkeley. 614621. University of California. Bull. Schrefler B. 30. 380403. and Schrefler B. McGrawHill Book Company. 2. Geotnech. Morland L. L. Civil Engrs. J . C. (1990) A numerical model for immiscible 2phase fluidflow in a porous medium and its time domain solution. Int. and Gawin D. Comp. Goeteborg. C. (1954) A subpresseo nos barreyens .S. L. J. (1998) The Finite Element Method in the Static and Dynamic Deformation and Consolidation of Porous Media.A.. and Shiomi T. 55. Acad. Li X. (1996) The effective stress principle: incremental or finite form?. Advances in Heat Transfer. Anal. 143147. (1987) The Finite Element Method in the Deformation rind Consolidation ofPorous Media. (1965) Transient flow through unsaturated porous media.. 213223. No. (1969) Finite element analysis of flow in saturated porous elastic media. Thesis. (1972) A simple constitutive theory for fluid saturated porous solids. Nurn. W.. 155167. (1982) Field equations for porous media under dynamic loads in Num. Meth. J .. Int.A. 11951212. C. USA. V. C. GPotechnique. Boston U. Reidel. (1992) Multiphase flow in deforming porousmedia and finiteelement solutions. Deuticke. Runesson K. Zienkiewicz 0 . New York. and Zienkiewicz 0. (1980) Drained. (1977). 77. Bull. Simultaneous heat mass and momentum transfer in porous media: a theory of drying. D. Wiley & Sons.Publ. 4. consolidating and dynamic behaviour assumptions in soils. 29. Serafim J. 6. 4. Gkotechnique. Laboratorio Nacional de Engenheria Civil. K. 21 1227. Zaremba S. 7. . Wiley & Sons. J.
Chan A. Proc. ...52 EQUATIONS GOVERNING THE DYNAMIC.A rational approach to quantitative solutions. C. Proc. A. Roy. (1990a) Static and Dynamic Behaviour of Geomaterials . 285309. H. Roy. and Bicanic N.. Part 11: Semisaturated problems. .. Paul D. Lond. INTERACTION Zienkiewicz 0. Lond. SOILPORE FLUID. A429. C. Xie Y. K. Schrefler B. Ledesma A. (1990b) Static and Dynamic behaviour of soils: a rational approach to quantitative solutions. Part IFully Saturated Problems. Soc... C . Soc.. 310323. and Shiomi T. Zienkiewicz 0. Pastor M. M. A429.
etc. d l d y .Finite Element Discretization and Solution of the Governing Equations 3. The dot notation implies time differentiation so that =& at a@  d2@ dl. which can be.and shape function Nk which are specified in spatial dimensions. z & etc. Here we shall use throughout the notation and methodology introduced by Zienkiewicz & Taylor (1989 and 1991) which is the most recent (fourth) edition of the first text for the finite element method published in 1967. In the application to the problems governed by the equations of the previous chapter we shall typically be solving partial differential equations which can be written as where A.1 THE PROCEDURE OF DISCRETIZATION BY THE FINITE ELEMENT METHOD The general procedures of the Finite Element discretization of equations are described in detail in various texts.. are 'discretized' or approximated by a finite set of parameters 6. B are matrices of constants and L is an operator involving spatial differentials such as d l d x . @ is a vector of dependent variables (say representing the displacements u and the pressure p). In all of the above. Thus we shall write . nonlinear. and frequently are. The finite element solution of the problem will always proceed in the following pattern: (i) The unknown functions Q.
C and P are matrices or vectors corresponding in size to the full set of numerical parameters 6k.1.(t) (3. In Section 3. and &. once again.4b) are usually the values of the unknown function at some discrete spatial points called nodes which remain as variables in time. i.e. Typical finite elements involving linear and quadratic interpolation are shown in Figure 3.54 FINITE ELEMENT DISCRETIZATION AND SOLUTION where the shape functions are specified in terms of the spatial coordinates i. if the parameters Giare time dependent. the parameters & represent simply the values of at specified pointscalled nodes and the shape functions are derived on a polynomial basis of interpolating between the nodal values for elements into which the space is assumed divided. equation (3. If time variation occurs. This can be. = 6.6) which is now an ordinary differential equation requires solution in the time domain. very suitable choice for the weighting function Wj is to A take them being the same as the shape function Ni: Indeed this choice is optimal for accuracy in so called selfadjoint equations as shown in the basic texts and is known as the Galerkin process.e. Usually. The present chapter will be divided into two sections. achieved by discretization in time and use of finite elements there although many alternative approximations (such as the use of finite differences or other integration schemes) are possible. (ii) Inserting the value of the approximating function 6 into the differential equations we obtain a residual which is not identically equal to zero but for which we can write a set of weighted residual equations in the form which on integration will always reduce to a form Where M .2 we shall consider solution of the approximation based on the up form in which the dependent variables are the displacement of the soil matrix and the +" .
A GENERAL GEOMECHANICS FINITE ELEMENT CODE 55 Figure 3.3. However. The timestep limitations of such explicit codes are severe and the code is therefore limited to relatively short time durations. Section 3. 3. On the other hand. Only two dimensions will be considered here and in the examples which follow. 1991) following the work described by Zienkiewicz and Shiomi (1984). (1990a).2 and is the basis of a code capable of solving all lowfrequency dynamic problems.1 and 2. 2.. The tensorial form of the equations can be found in Section 3.1 Some Typical two dimensional elements for linear and quadratic interpolations from nodal values pore pressure characterized by a single fluid.2. allows much longer periods to be studied. we shall allow incomplete saturation to exist assuming that the air pressure is zero. The formulation thus embraces all the features of the up approximation of sections 2. Indeed. we shall limit ourselves to the use of the condensed.1 UP DZSCRE TZZA TZON FOR A GENERAL GEOMECHANZCS FZNZTE ELEMENT CODE Summary of the general governing equations We will report here the basic governing equation derived in the previous chapter. The code based on this form is fully explicit and it is named GLADYS2E as it was developed in Glasgow (see Chan et al. the implicit form in Section 3. However.2. i.2 3.e.2. but extension to three dimensions is obvious. water.2.3 of this chapter is intended for the solution of dynamic problems where highfrequency effects predominate. . vectorial form of these which is convenient for finite element discretization.4.3. displacement vectors of the solid and of the pore fluid (water). consolidation problems and static drained or undrained problems of soil mechanics. with suitable accuracy control such codes can be used both for earthquake phenomena limited to hundreds of seconds or consolidation problem with a duration of hundreds of days. The variable here will be u and U i.e. The code based on the formulation contained in this part of the chapter is named SWANDYNE (indicating its Swansea origin) and its outline was presented in literature by Zienkiewicz et al.
The strain matrix S is defined in two dimensions as (see (2. the relative fluid acceleration terms are omitted as only the up form is being considered.11) and is here copied for completeness as In the above and in all following equations.24) with pa P = XwPw = 0.14) The effective stress a" is computed from an appropriate constitutive law generally defined as 'increments' by (2. EO .19)) generally taken as constant and a is the total stress with components The effective stress is defined as in (2.1) where a again is a constant usually taken for soils as and p the effective pressure defined by (2. (3. p the total density of the mixture (see (2.10)) { : z } = Sdu Here u is the displacement vector.56 FINITE ELEMENT DISCRETIZATION AND SOLUTION The overall equilibrium or momentum balance equation is given by (2.2) where D is the tangent matrix dependent on the state variables and history and corresponds to thermal and creep strains.
On occasion the approximation can be used...e.e.18) and (2. The effective stresses are determined at any stage by a sum of all previous increments and the value of p. above becomes where q. it has been included in the force term of the computer code SWANDYNEI1 (Chan. The above set defines the complete equation system for solution of the problem defined providing necessary boundary conditions have been specified as in (2. The total boundary I? is the union of its components i.A GENERAL GEOMECHANICS FINITE ELEMENT CODE 57 The main variables of the problem are thus u and p. and Assuming isotropic permeability. (effective area). Its inclusion in the equation will render the final equation system nonsymmetric (see Leung. (saturation) and X . The contribution of the solid acceleration is neglected in this equation.e. However. An additional equation is supplied by the mass conservation coupled with fluid momentum balance. 1995) although it is neglected in the left hand side of the final algebraic equation when the symmetric solution procedure is used.27) which can be written as with k = k ( S . . This is conveniently given by (2. ) . 1984) and the effect of this omission has been investigated in Chan (1988) who found it to be insignificant. determines the parameters S. is the influx i.19) i. having an opposite sign to the outflow w.
ordinary differential equation now becomes where . and prescribed boundary tractions i.58 FINITE ELEMENT DISCRETIZATION AND SOLUTION 3. To obtain the first equation discretized in space we premultiply (3. the natural boundary condition will be obtained by integrating the weighted equation by parts.12)) now defined to allow for effects of incomplete saturation as The discrete. automatically by a suitable prescription of the (nodal) parameters.e.18) are satisfied on ru and r.2.2 Discretization of the governing equation in space The spatial discretization involving the variables u and p is achieved by suitable shape (or basis) functions. equal in number of components to that of vector u contains all the effects of body forces.8) by (NU)= and integrate the first term by parts (see for details Zienkiewicz and Taylor. At this stage it is convenient to introduce the effective stress (see (3. writing We assume here that the expansion is such that the strong boundary conditions (3. 1989 or other texts) giving: where the matrix B is given as and the 'load vector'f ( I ) . As in most other finite element formulation.
15) can be written in discrete form: where of course D is evaluated from appropriate state and history parameters. and integrating by Finally we discretize equation (3. where C s . Sw.30~) i.e. This gives the ordinary differential equation where the various matrices are as defined below where Q* is defined as in (2.17) by premultiplying by ( N ~ ) ~ parts as necessary.A GENERAL GEOMECHANZCS FINITE ELEMENT CODE 59 is the MASS MATRIX of the system and is the coupling matrixlinking servation.. . and equation (3. Cw and k depend onp.23) and those describing the fluid con The computation of the effective stress proceeds incrementally as already indicated in the usual way and now (3.
The former is known as the SSpj .27) and (3. On the other hand. t.. 1995) was an extension to the original work of Newmark (1959) (see also Whitman 1953) and is called Betam method by Katona (1985) and renamed the Generalized Newmark (GNpj) method by Katona and Zienkiewicz (1985). which derivatives $ &. From the initial conditions. it is necessary to integrate the ordinary differential equations (3. acceleration in dynamical problems. (which can either be the displacement or the pore water pressure).27) and (3. . &. . families of singlestep methods evolved separately. exclusively. For the SSpj. at time station t. 1988).23). Both methods have similar or identical stability characteristics. or higher time derivatives are required. however. 1985a. and its . we have the known values of a. Before treating the ordinary differential equation system (3.6) by adding a forcing term: + +.28) in time by one of the many available schemes.23). . This was introduced by Zienkiewicz et al. The later method. On the other hand. all quantities in the GNpj method are defined at a discrete time station thus making transfer of such quantities between the two equations easier to handle.Single Step pth order scheme for jth order differential equation ( p 2 j ) . but distinct. 1985b). and t. Here we shall use the later (GNpj) method. they are inconvenient as most of them are not selfstarting and it is more difficult to incorporate restart facilities which are required frequently in practical analyses.60 FINITE ELEMENT DISCRETIZATION AND SOLUTION 3.+l. due to its simplicity. In all time stepping schemes we shall write a recurrence relation linking a known value 4. (3. . 1984b. 6. Two similar. with the values of are valid at time t. can be found and this solution is required if the initial conditions are different from zero. Wood. .+. (1980b. we can thus write: and From the first equation. (3. 1984) and extensively investigated by Wood (1984a. . 1990).g. Although there are many multistep methods available (see e. which was adopted in SWANDYNEI1 (Chan.. . A t and are the unknowns. .g. One is based on the finite element and weighted residual concept in the time domain and the other based on a generalization of the Newmark or finite difference approach. the singlestep methods handle each step separately and there is no particular change in the algorithm for such restart requirements. no initial condition e. the value of the acceleration at time t..3 Discretisation in time To complete the numerical solution.+. We assume that the above equation has to be satisfied at each discrete time i. we shall illustrate the time stepping scheme on the simple example of (3.2. The SSpj scheme has been used successfully in SWANDYNEI (Chan.28). .e.
A GENERAL GEOMECHANICS FINITE ELEMENT CODE 61 The link between the successive values is provided by a truncated series expansion taken in the simplest case as GN22 as Eq(3.36). in (3.23) and (3. the only unknown is the incremental value of the highest derivative and this can be readily solved for.23).27) and (3. we are considering here i.e. Returning to the set of ordinary differential equations.35) and (3.+l. we have: assuming that (3.28) and writing (3. and In the above equations. (3.34) is a secondorder differential equation j and the minimum order of the scheme required is then two: as (p > j) Alternatively. a higher order scheme can be chosen such as GN32 and we shall have: In this case. an extra set of equations is required to obtain the value of the highest time derivatives.28) at the time station t. Using GN22 for the displacement parameters u and G N l l for the pore pressure parameter p"'. This is provided by differentiating (3.27) is satisfied. we write: .
41) and (3. The parameters PI. If the damping matrix is nondiagonal. Dewoolkar (1996).. (1982) and Leung (1984).43) into equations (3. Inserting the relationships (3.6 or PI = 0. + p. and Ap. are as yet undetermined quantities. Usually some algorithmic (numerical) damping is introduced by using such values as P = 0. For P2 = 0 and PI = 0..+1 = u. The well known central difference scheme is recovered from (3. In practice. numerical oscillation may occur if no physical damping is present. remain as unknowns. However. and Ap.at and where Au. using the computer program SWANDYNE I1 in the modelling of a freestanding retaining wall.P2 = 0 and this form with an explicit u and implicit p scheme has been considered in detail by Zienkiewicz et al. the physical damping (viscous or hysterestic) is much more significant than the algorithmic damping introduced by the time stepping parameters and the use of either sets of parameters leads to similar results.605 2 P = 0. if the higher order accurate 'trapezoidal' scheme is chosen with P2 = PI = 112 and PI = 112.P2 and PI are usually chosen in the range of 0 to 1.62 u.41) if PI = 1/2.51. we shall have an explicit form if both the mass and damping matrices are diagonal.51 and PI = 0. in cases involving soil.42) yields a general nonlinear equation set in which only Au. therefore the second set was used and gave very good comparisons.au. However. reported that the first set of parameters led to excessive algorithmic damping as compared to the physical centrifuge results.+1 FINITE ELEMENT DISCRETIZATION AND SOLUTION = u. an explicit scheme can still be achieved with PI = 0 thus eliminating the contribution of the damping matrix. such schemes are only conditionally stable and for unconditional stability of the recurrence scheme we require P 2 > PI > 2 1  and 8 25 1 1 The optimal choice of these values is a matter of computational convenience. This set can be written as . + Au.6 and PI = 0. +&at u.515 2 PI = 0. the discussion of which can be found in literature.
+.A GENERAL GEOMECHANICS FINITE ELEMENT CODE 63 =0 wl ! j = QT+~.. for instance.43). . The equation will generally need to be solved by a convergent. The Jacobian matrix can be written as J= where which are the well known expressions for tangent stiffness matrix. Two points should be made here: (a) that in the linear case a single 'iteration' solves the problem exactly (b) that the matrix can be made symmetric by a simple scalar multiplication of the second row (providing KT is itself symmetric). Matthies and Strang (1979) and Zienkiewicz and Taylor (1991) . sap. The values of u.+l and p. iterative process using some form of Newton Raphson procedure typically written as I J { a san } l l = * ~ p.27) as the solution proceeds.LI~&A& ~ . + ~ p ~ A t A c ..}~+{sa&}~+~ AP.y ~ A ~ .5).44b) where @j1 F ~ J can be evaluated explicitly from the information available at and . In practice it is found that the use of various approximations of the matrix J is advantageous such as.F s + "1 j ! + + (3. The underlined term corresponds to the 'initial stress' matrix evaluated in the current configuration as a result of stress rotation defined in (2. at the time t. time t. the use of 'secant' updates (see for instance Crisfield (1979). n+ 1 where 1 is the iteration number and {~i. and In this A u i must be evaluated by integrating (3.+l are evaluated by equation (3.
dependent on the solution history. if stable. 1984). Here time is still real and we have omitted only the inertia effects (although with implicit schemes this apriori assumption is not necessary and inertia effects will simply appear as negligible without any substantial increase of computation). The iterative procedure allows the determination of the effect of terms neglected in the up approximation and hence an assessment of the accuracy. Such staggered procedures. (1980). . Thus for static or quasistatic problems it is merely necessary to put M = 0. Brezzi. We shall return to this problem in Chapter 8 where a modification is introduced allowing the same interpolations to be used for both u and p. 1971 and 1973. to be updated. i. and immediately the transient consolidation equation is available. useful when the permeability is sufficiently large. however. In a later chapter we shall discuss a possible amendment to the code permitting the use of identical up interpolation even in incompressible cases. 1989) and Figure 3. We shall return to such explicit processes in Section 3.44) can be amended to that of successive separate solution of the time equations for variables Au. The form of most of the elements used satisfies the necessary convergence criteria of the undrained limit (Zienkiewicz. The process of the timedomain solution of (3. and usually S = 0 resulting in a zeromatrix diagonal term in the jacobian matrix of Equation (3. 1974) convergence conditions or their equivalent (Zienkiewicz et al.19) if the BabuskaBrezzi (Babuska. Though the bilinear u and p quadrilateral does not. and indeed a wide choice is available to the user if the limiting (undrained) condition is never imposed. using an approximation for the remaining variable. only conditionally stable and is efficient only for phenomena of short duration.e. The matrix to be solved in such a limiting case is identical to that used frequently in the solution of problems of incompressible elasticity or fluid mechanics and in such studies places limitations on the approximating functions Nu and NP used in (3.3. respectively. f(*)= 0. In pure statics the time variable is still retained but is then purely an artificial variable allowing load incrementation. Until now we have not referred to any particular element form. It is. This explicit procedure was first used by Leung (1984) and Zienkiewicz et al.64 FINITE ELEMENT DISCRETIZATION AND SOLUTION A particularly economical computation form is given by choosing /32 = 0 and representing matrix M in a diagonal form. 1986) are to be satisfied. and Ap. it is. can be extremely economical as shown by Park and Felippa (1983) but the particular system of equations here presented needs stabilization.47). The incremental computation allows the various parameters. This was first achieved by Park (1983) and later a more effective form was introduced by Zienkiewicz et al. We note that all computations start from known values of u and p"' possibly obtained as the result of static computations by the same program in a manner which will be explained in the next section. Due to the presence of first derivatives in space in all the equations it is necessary to use 'Cocontinuous' interpolation functions (Zienkiewicz and Taylor. however. Special cases of solution are incorporated in the general solution scheme presented here without any modification and indeed without loss of computational efficiency. H .2 shows some elements incorporated in the formulation. (1988). In static or dynamic undrained analysis the permeability (and compressibility) matrices are set to zero.
biquadratic for u.A GENERAL GEOMECHANICS FINITE ELEMENT CODE 65 Quadratic for u Biquadratic for u a (ii) Bilinear forp (ii) Linear forp Linear for p D)0 Linear for u b Linear (with cubic bubble) for u Linear f o r p (ii) Figure 3. displacement (u) and pressure @) formulation (a) (i). (b) (i).. (ii).. x. 2 0 we regain the fully saturated behaviour described in Section 2... linear for p. linear (with cubic bubble) for u.. The reader will observe that when p. . the values of S. (ii). when p < 0..have to be determined from appropriate formulae or graphs.2 of the previous chapter. k.. bilinear for p: (c) (i) linear for u. Further we note that if p. 0 (full saturation) then we have > and the permeability remains at its saturated value However. when negative pressures are reached i.e. (ii). linear for Element (c) is not fully acceptable at incompressibleundrained limits. quadratic for u.2 Elements used for coupled analysis.. linear for p: (d)(i). . (ii). Thus for known Au increment the 0'' are evaluated by using an appropriate tangent matrix D and an appropriate stress integration scheme.
implicit.02 s which is the interval used usually in earthquake records. The length of the time step based on accuracy considerations was first discussed in Zienkiewicz et al. and later by Zienkiewicz and Xie (1991). Clearly for a scalar variable u the error term is given by the first omitted terms of the Taylor expansion i. undrained) Time step length As explained in the previous section. the remaining motion is caused by something resembling a consolidation process which has a slower response allowing longer time steps to be used.51) . the computation always proceeds in an incremental mannerand in the u .e. With unconditional stability of the implicit scheme. the only limitation on the length of the time step is the accuracy achievable. Zienkiewicz and Shiomi (1984). Indeed after the passage of the earthquake. (1984).2.43) and its comparison with a Taylor series expansion.66 FINITE ELEMENT DISCRETIZATION AND SOLUTION 3. static solution. In the examples that we shall give later we shall frequently use simply the time interval At = 0.46) with variables Au and Ap needs to be solved at each time step. The simplest process is that which considers the expansion for such a variable as u given by (3.e. (3. form and the equation system given by the Jacobian of (3.p form in general the explicit time stepping is not used as its limitation is very serious. a much longer length of time step could be adopted. drained uncoupled. Bergan and Mollener (1985). once the input motion has ceased and its record no longer has to be followed. However.4 General applicability of transient solution (consolidation. in scalar values Using an approximation of this third derivative shown below we have For a vector variable u we must consider its L2 norm i. llul12 = &% etc. earthquake problem short time steps will generally be used to follow the time characteristic of the input motion. Invariably the algorithm is applied here to the unconditionally stable. Clearly in the dynamic.
Static problemsundrained and fully drained behaviour Steady state. This can be deduced by rewriting the two.A GENERAL GEOMECHANICS FINITE ELEMENT CODE 67 and we can limit the error to This limit was reestablished later by Zienkiewicz and Xie (1991) who replace the leading coefficient of (3. the essence of the procedure is identical and this is given by establishing a priori some limits or tolerance which must not be exceeded. not to change the length of the time step by more than a factor of 2 or 112 otherwise unacceptable oscillations may arise. u but in general this suffices for quite a reasonable error control. 1998) the acceleration terms are generally omitted a priori. The consolidation equation \ In the standard treatment of consolidation equation (see for instance Lewis and Schrefler.28) omitting terms involving time derivatives. governing equations (3. The tolerance is conveniently chosen as some percentage 7 of the maximum value of norm llul12 recorded. The equations now become: . The procedure simply reduces the multiplier of the mass matrix M to a negligible value without influencing in any way the numerical stability.e. The time step can always be adjusted during the process of computation noting. and modifying the time steps accordingly. discrete. however.23) and (3. there is no disadvantage in writing the full dynamic formulation for solving such a problem. (static) conditions will only be reached under the extremes of undrained or fully drained behaviour. provided of course that an implicit integration scheme is used. Thus we write with some minimum specified. In the above we have considered only the error in one of the variables i.52) as a result of a more detailed analysis by Whatever the form of error estimator adopted. However.
Equation (3..57) can be solved once the history of the load applied has been specified. The solution so obtained is of course the well known.2). However. Indeed in this solution the negative pressure zones and hence the partially saturated regions can be readily determined following the procedures outlined in the previous chapter. However.p.56) can be solved independently of the first for the water pressures. = 0 coincides. drained.. which can only be done provided that some fluid compressibility is available giving S # 0.e. then (3. The existence of the two steady states is well known and what we have indicated here is a process by which the various matrices given in the original computer program can be used to obtain either of the steady state solutions. with totally impermeable behaviour H = 0 and f ( 2 ) =0 (3. determined as the first equation (3. If S = 0 i. interpolations are permissible (as shown in Figure 3..e.de") = D(Bdu  de") (3.. Solving (3. no compressibility is admitted we have the problem already discussed in the previous section in which only certain u .68 FINITE ELEMENT DISCRETIZA TION AND SOLUTION and with the effective stresses given by (3. can be eliminated directly and the solution concerns only the variable u.e.55) coupled with the appropriate constitutive law (3. With p.55).57) First we observe that the equations are uncoupled and that the second of these i.55) and the constitutive law are sufficient to obtain the unique undrained condition.28) we find that it becomes which on integration establishes a unique relationship between u and p.27) and are defined incrementally as dd' = D(de . The case of undrained behaviour is somewhat more complex.59) But on reexaming equation (3. We note that with k = 0 i.61) now has to be solved together with (3.61) for p.. if S # 0 p.. this does require an . (3.. behaviour.. which is not time dependent assuming that the initial condition of u = 0 and p.
an iteration scheme such as the Newton Raphson.+. it is possible to obtain such steady states by the code. 3. this process may be time consuming even if large time steps At are used. With a systematic change of the external loading.61) now become. in due course. Tangential Matrix or the Initial Matrix method can be adopted. QuasiNewton. converge with However..2. then the equation can be solved directly yielding the two unknowns u. for the undrained problem. Two types of undrained conditions exist: (a) when k = 0 throughout. and p. illustrated by their linear equivalent If complete saturation is assumed together with a linear form of the constitutive law we can write the effective stress simply as a" = DBu (3. (b) k # 0 but the complete boundary is impermeable..62) .5 The Structure of the numerical equations. Both cases can be computed with no difficulties. However.41) and (3. using the previous timestepping procedure. and if the material is nonlinear. Provided that the boundary conditions are consistent with the existence of drained and undrained steady state conditions the timestepping process will. If the material behaviour is linearly elastic. problems such as the loaddisplacement curve of a nonlinear soil and porefluid system can be traced.A GENERAL GEOMECHANICS FINITE ELEMENT CODE 69 alternative to the original computer program. A simpler procedure is to use the GNOO scheme with Equations (3.
29) The overall system can be written in the terms of the variable set [ii. This is one of the requirements of the patch test of Zienkiewicz et al. . in the absence of fluid compressibility.28) to the form given below and where p = p.70 FINITE ELEMENT DISCRETIZATION AND SOLUTION We can now reduce the governing up equations (3.. 1985) It is interesting to observe that in the steady state we have a matrix which.31) and (3. S and H are again symmetric matrices defined in (3. and K = 1. results in which only can have a unique solution when the number of i variables nu is greater i than the number of p variables n. (1986) and Zienkiewicz and Taylor (1989) and of the BabuskaBrezzi (Babuska. 1973 and Brezzi.30) and Q is as defined in (3. For undrained behaviour we can integrate the second equation when H = 0 and obtain an antisymmetric system which can be made symmetric by multiplying the second set equation by minus unity (Zienkiewicz and Taylor. as plT Once again the uncoupled nature of the problem under drained condition is evident (by dropping the time derivatives) giving in which p can be separately determined by solving the second equation.23) and (3.. Zienkiewicz et al.. (1985). B~DB~O is the well known elastic stiffness matrix which is always symmetric in form. 1974) condition.
23) to Indeed such damping matrices have a physical significance and are always introduced in earthquake analyses or similar problems of structural dynamics.3. (2. if the solutions of the problems are in the lowstrain range when the plastic hysteresis is small or when. the terms underlined in the above equations.3 3. it may be necessary to add system damping matrices of the form Cu to the dynamic equations of the solid phase.47).16) are repeated below as: omitting now only the convective acceleration. . Here for brevity the equations are now given only in tensorial notation. 1975 or 1993). M is the same mass matrix as given in (3.e.1 THE UUDZSCRETZZA TZON AND ITS EXPLICIT SOL UTZON The governing equation We shall now return to the original equations of Chapter 2 Section 2. With the lack of any special information about the nature of damping it is usual to assume the so called 'Rayleigh damping' in which where cr and p are coefficients determined by experience (see for instance Clough and Penzien. However. i.2.2.. purely elastic behaviour is assumed. to simplify the procedures. Thus (2. i.2.THE uU DISCRETIZATION AND ITS EXPLICIT SOLUTION 71 3.6 Damping matrices In general.1 l).e. In the above. 3.. when dynamic problems are encountered in soils (or other geomaterials) the damping introduced by the plastic behaviour of the material and the viscous effects of the fluid flow are sufficient to damp out any nonphysical or numerical oscillation.13) and (2. changing (3. density variation and source terms.2 without the introduction of the approximation used in Section 3.24) and K is some representative stiffness matrix of the form given in (3.
'J = 01.'. to approximate the average true displacement. (3. the manner suggested in (2. we shall use its total displacement defined as thus where the relative displacements are divided by the porosity n.22) of Chapter 2 giving in If instead of using the relative velocity w.72 FINITE ELEMENT DISCRETIZATION AND SOLUTION uR. equation (3.76). this system is inconvenient as second derivatives in time of both variables occur in each of the equations and thus completely diagonal matrices cannot be obtained by mass lumping. pui pf. A very simple modification can be made here as suggested by Zienkiewicz and Shiomi (1984). after some algebraic manipulations we have and .17)).n ijiR + pbi =0 (3. we obtain the following system rJ. The discretization of the above leads to the equation originally used by Ghaboussi and Wilson (1972).72~) time and find a direct expression for the pressure provided in that the compressibility. + a Q ( ' ~ ~ k+knu&).72a and b). is different from zero (see (2. Now in place of relative displacements of water.1 of Chapter 2.75) a. .73). . Thus we have Inserting this into (3. However.+ .. we introduce the relative displacement we can integrate. which leads to complete matrix decoupling of higher derivatives and which is therefore ideally suited for explicit schemes.$.. Starting from (3.  VP  (3. l / Q .76a) and The vector form of the above equation is presented in Table 2.
80a) with NUT the second (3.79b) by n and subtracting from (3. Weighting the first equation (3. thus leading to a convenient diagonal form in discretization. we have the following equation system: .THE uU DISCRETIZATION AND ITS EXPLICIT SOLUTION 73 This is conveniently rewritten as and Multiplying (3.80b) with N ~ 'on .2 Discvetized equation and the explicit scheme Approximating the two variables in terms of finite element interpolation. a in 3. and ~ integrating by parts over the whole domain.79b) is multiplied by n so that symmetry is preserved in the discretized equations.80) only ii.79a) we find that the first equation is now free from the acceleration of the fluid displacement and becomes The second equation (3. In the final equation system (3. occurs in the first equation and only the second. (now including the vector notation) and where Nu and N u are the appropriate shape functions. we have.3.
3 The structure of the numerical equations in linear equivalent (viz.3.74 FINITE ELEMENT DISCRETIZATION AND SOLUTION 3. we have D ~ i j k l= D ~ i j l k we have . 3.2.82) in turn: (i) Stiffness Matrix K where ni are the directions of the normal at the boundary.5) Considering each term in the equation system (3. Linearizing the constitutive equation we have and making use of the symmetry of shear strains.
THE UUDISCRETIZATION AND ITS EXPLICIT SOLUTION 75 (iii) Stiffness Matrix K2 Note that (iv) Solid Mass Matrix Ms (v) Solid Body Force (vi) Damping Matrix C2 (vii) Damping Matrix Cl .
76 FINITE ELEMENT DISCRETIZATION AND SOLUTION (vii) Stiffness Matrix K2 Transpose (viii) Stiffness Matrix K3 Noting again that (ix) Damping Matrix C3 (x) Damping Matrix C2 Transpose .
THE UU DISCRETIZATION AND ITS EXPLICIT SOLUTION 77 (xi) Fluid Mass Matrix My N. I ~. of course. now applied in terms of combined displacement variable: for which the whole problem can be written in a simple form of We recall here the form outlined in Section 3.1. nodal integration) allows a very cheap forward marching scheme. This is obtained by putting the following values for y in the original Newmark scheme or P2 for the Generalized Newmark scheme: for if P2 = 0 it disappears . for instance. we can use the generalized Newmark procedure. Although. implicit schemes can once again be used herethe explicit operation is of main interest because lumping (by. With full symmetry of the system.(npf = oi)dfl = .dfl6~. MflLULi = M~U (xii) Fluid Body Force Collecting all the terms and displaying in matrix form which was first presented by Shiomi (1983) and Zienkiewicz and Shiomi (1984): It is of interest to observe that the mass matrix of the system does not couple variables ti and U and here can be easilty diagonalized.np~N.
10) is scalar .. if the Newmark parameter p or the Generalized Newmark parameter PI is nonzero then the inclusion of the damping matrix may destroy this diagonal structure. the use of a small time step is a positive advantage as the nonlinearity may impose a large number of iterations in typical implicit schemes. then the damping matrix will assume a block diagonal structure of 4 x 4 for twodimensions and 6 x 6 for threedimensions if the variables u and U are numbered next to each other. is defined as: dy. (1991).78 FINITE ELEMENT DISCRETIZATION AND SOLUTION Although the mass matrices can be diagonalized rather simply. One of the major constraints on an explicit scheme is the limitation of the explicit time step. Noting that the engineering shear strain dy... However.e. 1991). if the crosscoupling terms are all zero.4 THEORY: TENSORZAL FORM OF THE EQUA TZONS The equation numbers given here correspond to the ones given earlier in the text. then the block diagonal structure will consist of 2 x 2 matrices whether it is twodimension or threedimension. for many nonlinear and transient problems. Generally. On the other hand. In any case. The following two cases can be identified. = 2 d ~ . Having dealt with the problem raised by the Rayleigh damping. kv # 0 for i # j. incrase in the critical time step length is possible and this an was shown in Chan et al. Therefore by a suitable reduction in the value of Kf. 1986). (see Chan et al. the results are not adversely affected until the value of Kfbecomes comparable to the bulk modulus of soil matrix K T . If anisotropic permeability is used with crosscoupling terms i. 3... Equation (3. Simon et al. for instance. hence only the diagonal contribution of the stiffness matrix is included in the solution matrix while the full matrix is retained in the right hand side during the calculation of the residuals. the offdiagonal term on the solution matrix can also be removed by further lumping (see. if Rayleigh damping with the full stiffness matrix is used then the diagonal structure will be destroyed. In the case when the fluid bulk modulus Kfis much bigger than the bulk modulus of the soil matrix KT the critical time step is found to be: where e is the minimum length between nodes and C1 is a constant depending on the type of element used (and such other factors as porosity). .
.12b) Equation (3.+ S.14) is scalar.13) is scalar. .w. and assuming isotropic permeability The summation range for the upper case indices will depend on the number of nodes with solid displacement and pore water pressure degreesoffreedom (dof) respectively. 79 v = (3..11b) !IP (5" IJ = 0. Applying Green's identity to the internal force term (first term on the lefthand side) . = n... + 1 1 (3.kii(p ..16) is scalar and n... Equation (3. on I? = I?. ~ 1 6 .(5.TENSORIAL FORM OF THE EQUATIONS (5 ..)= w... Equation (3..p/6.
+ 1. I Inserting the shape functions . k u p M .. + i0)dR = 0 .27b) 1.kd". N.. b . N.pN.20b and 3. = D i j ~ i(N.lJ o'llJ.dR]GLi = 1 NipbidR + 0 1 rr N.vpfb..  ax6.  N (kl(pw. .23b) 1 (~ do.. ..80 FINITE ELEMENT DISCRETIZATION AND SOLUTION Rearranging NZjordR +[ I. .iidT.22b) The definition of the B matrix in equation (3. N$(kgS. .).21) is not needed in tensorial form. + 1. + Sivpfbj) + a&.+ P . (3. ..i Niazi.$dR  ~ ~ . + N g f i dR = 0 Q* . < Q + + Neglecting source term and integrating by part the first part of the first term Ira N$nikvp. + NtidiiKk)  d&li) (3... 0. d' MKLsL..jdT. ~ f I0 r fi)  = (3.
W.. W. December. and Penzien J. Numtr. and Mollener (1985) An automatic timestepping algorithm for dynamic problems. Num. C.5 CONCLUSIONS In this chapter. 3. Math. 1. (1973) The finite element method with Lagrange Multipliers. H. McGrawHill. Ph. Chan A. I. Dissertation. Inc. 881887. 1113 Dec. Appl. No. Bergan P... Birmingham.CONCLUSIONS 81 equation (3.. Num. R. 49. 322333. G . McgrawHill Book Company. the governing equations introduced in Chapter 2 are discretized in space and time using various implicit and explicit algorithms. C. Clough R. Math. Anal. quasistatic and dynamic examples to illustrate the practical applications of the method and to validate and verify the schemes and constitutive models used. Famiyesin 0 . New York. 16. (1974) On the existence. They are now ready for implementation into computer codes. (1995) User Manual for DIANA SWANDYNE11. REFERENCES Babuska I. Bellman R. 179192. Chan A. H.. H. we shall show some applications for static. 0 . Mech. University College of Swansea.D. Babuska I. New York. uniqueness and approximation of saddle point problems arising from Lagrange multipliers. C. 299318. (1993) Dynamics of Structures (2nd edn). and Muir Wood D.33) is scalar. (1960) Introduction to Matrix Analysis (1st edn).. London. 0. . University of Birmingham. 20. (1975) Dynamics of Structures. (1971) Error Bounds for finite element methods. A. 8. (1991) A Fully Explicit uw Schemes for Dynamic Soil and Pore Fluid Interaction. Clough R. 129151. Brezzi F. Comp. Meth. APCOCM Hong Kong. McGrawHill. R2. Wales. School of Civil Engineering. and Penzien J. (1988) A unqied Finite Element Solution to Static and Dynamic Geomechanics problems. Chan A. In Chapters 57. Eng. R.
and Zienkiewicz 0. (1983) Stabilization of partitioned solution procedure for pore fluidsoil interaction analysis. 273384. (1973) Computational Methods in Ordinary Differential Equutions. Dahlquist G. (1996) A study of seismic effects on centiliverretaining ~valls with saturated backfill. Kungl. Interscience Publishers. A. University College of Swansea.82 FINITE ELEMENT DISCRETIZATION AND SOLUTION Crisfield M... Meth. L. Math. John Wiley & Sons. NUMETA 85. 18. (1956) Convergence and stability in the numerical integration of ordinary differential equations. A. R. translated and revised by Brenner J. Appl. Lambert J. 3353. Ghaboussi J. Gantmacher F. 14.for Transient Analysis. Chichester. Math. 13451 359. 30. A. University of Colorado. Meth. John Wiley & Sons. (1984) Earthquake response of saturated soils and liquefaction. BIT. Inc. Comp. Meth. Scnnd. and Felippa C. (1979) The solution of nonlinear Finite Element equations.. V. Num. Teknisku Hogskolans Handlingar. Park K. (1972) Variational formulation of dynamics of fluid saturated porous elastic solids. Meth. Chichester. Wales. M. 19. M. Comp. H. 2. J. 213225. Num. (1992) Multiphase flow in deforming porous media and finiteelement solutions.7 and Structures. C. Num. and Xie Y. C. L. K. Crisfield M. Eng. C. (1933) ~ b e die Bedingungen.D. unter welchen eine Gleichung nur Wurzeln mit r negativen realen Teilen besitzt. 1. and Schrefler B.. (1983) 'Partitioned analysis of coupled systems'. Meth. No. 11951212.. W. Ph. PhD Thesis. et al. (1991) Nonlinear Finite Elenzent Analysis of Solid. Park K. Eng. Chichester. A. Eng. Katona M. Crisfield M. Elsevier Science Publishers B. Boulder. (1978) On accuracy and unconditional stability of linear multistep methods for second order differential equations. 947963. (1959) A method of computation for structural dynamics. Chapter 3 in Computational Methods. ASCE E M . ASCE. Basel. Int. John Wiley and Sons Ltd. M. 16691673. Katona M. (1954b) The Theory of MatricesEnglish translation by Hirsch K. Chelsea Publishing Company. Lewis R. A. Int. 4. Mech. R. and Wilson E. Li X.. and Zienkiewicz 0. J. Newmark N. J. G. (1985) A general family of singlestep methods for numerical time integration of structural dynamic equations. (1990) A numerical model for immiscible twophase fluid flow in a porous medium and its time domain solution. Eng. Gantmacher F. Leung K. (1998) The Finite Element Method in the Static and Dynamic Deformation and Consolidation of Porous Media. Zienkiewicz 0. Mathematische Werke. unter welchen eine Gleichung nur Wurzeln mit r negativen realen Teilen besitzt. Proc. A. 20. (1985) A unified set of single step algorithms Part 3: The Betam method. Eng.. USA. 533545. Dissertation. C. Matthies H. 6794. 16131626. (1959) Stability and error bounds in the numerical integration of ordinary differential equations. Dewoolkar M. 21 1227. 21. Ann. 2. D. . Int. 1. Dahlquist G. Li X.C. Dahlquist G. 46. 2. 130. 1331 36. John Wiley & Sons. J. and Strang G. Chichester. 267278.. (1954a) Applications of the theory of matrices: Second part of ' A Theory o f Matrices'. Hurwitz A. EM4. 98. 8. a generalisation of the Newmark scheme. R. Struct. Hurwitz A. (1895) ~ b e die Bedingungen. Gantmacher F. Int. 45. Dept of Civil Engineering. No. Chelsea Publishing Company.. Num. (1997) Nonlinear Finite Element Analysis of Solids and Structures. (1954~) The Theory of'MatricesEnglish translation by Hirsch K.. A.. 1. (1979) A faster modified NewtonRaphson Iteration. G. 2.
C. J. (1982) Efficient dynamic solutions for single and coupled multipled field problems. J. Meth. S.. C. L. Int. Wood W. MIT.. Wood W. 10. (1984) A unified set of single step algorithms Part I: General formulation and applications. (1985a) Addendum to 'a unified set of single step algorithms. Wales. H. Hinton E. (1984) 'Coupled Problems and their Numerical Solution.REFERENCES 83 Paul D. New York. ICONMIG 5. Ph. (198413) A unified set of single step algorithms Part 2: Theory. Meth. 461482.' Chapter 1 in Numerical Methods in Coupled Systems. Anal. Houbolt. Zienkiewicz 0 . Eng. Num. 154. J. (1985) The coupled problems of soilpore fluidexternal fluid interaction: Basis for a general geomechanics code.D. Clarendon Press. Nunz. and Zhan X. 8. (1982) Liquefaction and permanent deformation under dynamic conditions Numerical solution and Constitutive relations.D. Engineers and Education. L. Int.. and Taylor R. and Chang C. 15291552. J. (1984a) A further look at Newmark. 2. L.. S. McGrawHill Book (UK) Ltd. 8. C. (1977) The Finite Element Method (3rd edn). Whitman R. Wood W. Zienkiewicz 0 . Wood W. Simon B. Zienkiewicz 0 . Wales. C. 20. Ph. Routh E. Schrefler B. (1955) Stability of a given state of motionthe advanced part of a treatise on the dynamics of a system of rigid bodies. Wood W.. No. 155167. Meth. W. University of Virginia Press. R. L. K. L. J.. K. Eng.Transient and Cyclic loads. Dissertation. Part 2: Theory'. Chapter 5 in Soil Mechanics. Wood W. Department of Mathematics. (1980a) Staggered. Hine N. of Swansea. University College of Swansea. Schrefler B. (1985b) A unified set of singlestep algorithms Part 4: Backward error analysis applied to the solution of the dynamic vibration equationNumerical Analysis Report 6185. Int. and Shiomi T. C. Archives of Computational Methods in Engineering. 20. Wood W. Dover. (1980b) An alternative singlestep algorithm for dynamic problems. L. A.. Zienkiewicz 0 .. 3140. Shiomi T. Meth. University of Reading. 20. (1990) Practical Timestepping Schemes. L. Routh E. Int. (1995) Finite Element in Environmental Engineering: Coupled thermohydromechanical process in porous media involving pollutant transport. J.  ... L. 7196.. C. C. 200202. 10091017. Zienkiewicz 0 . Zienkiewicz 0. John Wiley and Sons Ltd.. (1983) Nonlinear behaviour of soils in earthquakeC/Ph/73/83. (1930) Stability of a given state of motionthe advanced part of a treatise on the dynamics of a system of rigid bodies. London. The Earth. (1953) After Marcusion (1995): an example ofprofessional modesty. Geomech. Num. Dover. Num. A. Leung K. Earthquake Engineering & Structural Dynamics.. Num. 21. H. (1984) Dynamic Behaviour of saturated porous media: The generalized Biot formulation and its numerical solution.. etc. Time Marching Schemes in Dynamic Soil Analysis and Selective Explicit Extrapolation Algorithms in Proceedings of the Second Symposium on Innovative Numerical Analysis for the Engineering Sciences. Hinton E. V. Univ. and Taylor R. Dover. Zienkiewicz 0 . L.. Int. Eng. Water Resources Res. Chichester. and Taylor R. (1993) A fully coupled model for waterflow and airflow in deformable porous media. 1. Geomech. L. London. Routh E. Leung K.. T. Zienkiewicz 0 . timestepping formulae. J. and Paul D. Reading. 29.. C. J.. London. Chichester. Anal. Coll. Dissertation. Int. 17311 740. Wu J. 1165. Oxford. Eng. Zienkiewicz 0 . John Wiley. (1877) Stability of a given state of motionthe advanced part of a treatise on the dynamics of a system of rigid bodies (6th edn). C. (1986) Evaluation of uw and up FEM for the response of saturated porous media using Idimensional models. Num. J. 23032309.
and Xie Y. Xie Y. (1990a) Static and dynamic behaviour of geomaterials: A rational approach to quantitative solutions. (1985) Coupled Problems . Proc. Meth. Meth. 233239. Ledesma A. H. Zienkiewicz 0. (1989) The Finite Element Method .. M. (1986a) The Patch Test . McGrawHill Book Company. Pastor M.84 FINITE ELEMENT DISCRETIZATION AND SOLUTION Zienkiewicz 0.A condition for assessing F. Convergence. H. Eng. E. Int. H. C. and Bicanic N. C.Volume I : Basic C. Zienkiewicz 0. and Shiomi T. NO 5. Zienkiewicz 0. 285309. Int. Roy. Simo J. Part I: Fully saturated problems. Taylor R. C. C. J. C. K. A429. Num. Zienkiewicz 0. 10391055. Meth. Proc. Engrg.. J. C. K. London. C. J. Num. 3962. C. Eng. A429. (1988) Unconditionally stable staggered solution procedure for soil pore fluid interaction problems. Zienkiewicz 0. 1. Zienkiewicz 0. Num... Comm. In?. (1986b) The Patch Test for Mixed Formulation. International Journal for Earthquake and Structural Dynamics. 871887.. M. Soc.. Roy. 18731883. A. (1991) The Finite Element Method . . L. L. and Nakazawa S. and Taylor R.. C.Volume 2: Solid and Fluid Mechanics. C.. 22. Appl. L.. and Taylor R. Dynamics and Nonlinearity (4th edn).. and Taylor R. Meth. Paul D. Schrefler B.. Soc. M. Taylor R. C.. Zienkiewicz 0. 23. Num. 26... Zienkiewicz 0. Lond. 20.A simple timestepping procedure. C. Formulation and Linear Problems (4th edn).. Chan A. 3 10323. L. Lond. (1990b) Static and dynamic behaviour of soils: a rational approach to quantitative solutions. Qu S.. and Chan A. Part 11: Semisaturated problems. McGrawHill Book Company. London.. Paul D. L. (1991) A Simple error estimator for adaptative time stepping procedure in dynamic analysis. and Chan A.
models can be classified into two main groups: (i) Micromechanical or physical models. It can be said that history of plasticity began in 1773 with the work of Coulomb in soils. based on experimental results on punching and extrusion tests. plasticity based theories provide a consistent framework in which the behaviour can be accurately understood and predicted. nature of stress conditions.Constitutive Relations Plasticity 4. a specimen of soft clay subjected to a constant vertical stress shows a vertical deformation which increases monotonically with time. All materials present a response which depends on time to a greater or lesser degree. It was not until almost a century later. However. etc. and (ii) Macromechanical or phenomenological models. when Tresca. In fact. most of the geomaterials under normal engineering conditions present a mechanical behaviour which depends more on the level of stress. past history. Most of the models used in computer codes are of the second class. level of strain. based on the behaviour of grains or particles. For instance. applied later by Poncelet and Rankine to practical soil mechanics problems. the major part of the time dependence observed is generally connected with the pore water flow. St.1 INTRODUCTION The purpose of constitutive models is to capture some of the main features of the mechanical behaviour of solids under given conditions of temperature. Later. velocity of load application. proposed a yield criterion dependent on the maximum shear stress. pore pressure. direction of load increment and material structure than on time. For these. in 1864. Venant in 1870 introduced the concept of . Roughly speaking.
Along the plateau ABDE. which was generalized by Levy to threedimensional conditions. motivated by the development of both faster computers and numerical methods for boundary value problems. when von Mises and Huber. elastic and plastic parts of the strain increment was implied. The decomposition between elastic and plastic parts was introduced for plane stress conditions by Prandtl in 1924 and later.2). as plastic strain are present from very early stages of the test (Fig.2 THE GENERAL FRAMEWORK OF PLASTICITY Phenomenological aspects 4. If we unload at some point F and then reload again. became known as the LevyVon Mises equations. the observed behaviour in tension is schematized in Figure 4. introduce a general framework within which most models can be cast (section 4. Hardening plasticity was studied by Melan. If the specimen is subjected to an increasing strain. loading and unloading. and the material does not harden. consistency and uniqueness. first of all. the framework of what is known today as Classical Plasticity was established in 1949 by Drucker. from which plastic or irreversible strain upon unloading appears. The next significant step had to wait until the beginning of the new century. where it can be seen that the response is elastic and linear until a point A is reached. In the case of metals such as mild steel. the material is able to resist a higher load until new plastic strains develop (hardening). who introduced many of the concepts of modern plasticity such as loading surface. Reuss proposed a flow rule for the plastic component. proposed a new yield criterion which. If the specimen is unloaded. both loading and unloading follow the same path. The idea of plastic potential was suggested by von Mises in a work presented in 1928.2. without irreversible deformation. 4.1 The uniaxial behaviour of materials shows that irreversible strain develops in a way which depends on the type of material. the material behaviour is known as perfectly plastic. where the normal to the yield surface was used to provide the direction of plastic flow. Finally. the stress again increases. no distinction between total. in 1930 by Reuss for general conditions of stress. 4. who in 1938 generalized previously established concepts of plasticity to account for this effect. There exist today a great variety of models able to deal with most of the observed features of the mechanical behaviour of materials.2). The principal axes of stress and the increment of strain were assumed to be the same. the stress does not change until point E.1. Since then. with the flow equations derived by Levy.3 following with the description of advanced models capable of showing liquefaction phenomena in Section 4.4. neutral loading. The approach which will be followed here will. the stressstrain curve is different. much development has taken place. independently. . then deal with Classical Plasticity formulation in Section 4. In the case of soft soils such as saturated clays. a maximum load is reached from which the stress decreases until the material fails. The level of stress at which plastic strains appears does not change. There. Once a certain level of strain has been reached (E).86 CONSTITUTIVE RELATIONS  PLASTICITY isotropic flow rule. However.
However.2. and theories such as Damage Mechanics provide a suitable framework. some geomaterials such as concrete present degradation due to damage caused by the loading process to the structure of the material (Fig. boldface characters will be used for tensors. and lower case (such as a) for secondorder tensors a. Full understanding of this behaviour needs to take into account this process of degradation. 4.2 Generalized plasticity In the following. Loading and unloading shows clearly how the aparent elastic modulus of the material degrades as the test progresses.3)..2 Behaviour of soft clay Finally. uppercase (such as D) denoting fourthorder tensors Dijkl. . 4.THE GENERAL FRAMEWORK OF PLASTICITY 87 Stress t 0 Strain Figure 4.1 Behaviour of mild steel Stress I Strain 0 Figure 4. plastic models can be developed to reproduce the observed behaviour with an acceptable degree of accuracy.
devoted to the Introduction to Elastoplastic Constitutive Equations.3 Behaviour of materials with damage It is convenient to use a vectormatrix representation of tensorial magnitudes in numerical computations. The convention for products and its matrix equivalence is: (a) Double dot denotes contracted product in last two indexes (b) Tensor products are expressed by The behaviour of geomaterials depends on effective stresses as shown in Chapter 1. Basic theory If the response of the material does not depend on the velocity at which the stress varies the relationship between the increments of stress and strain can be written as . We shall here indicate the small alteration necessary to return to the matrix notation used in the previous chapters where a and D are vectors and matrices respectively. In Chapters 23 we have introduced this notation. which are denoted by a dash. we will not use the dash when referring to stress for the sake of simplicity. fourthorder tensors corresponding to matrices and secondorder tensors to vectors. in the first part of this chapter.88 CONSTITUTIVE RELATIONS  PLASTICITY 0 Strain Figure 4. However.
rateindependent constitutive laws. However. An inverse form is As the material response does not depend on time. Consequently. homogeneous. etc. in this simple onedimensional case. Xde = @(Ada) where X E %+ is a positive scalar (Darve 1990). the slope depends on the stress level. modification of material microstructure. the inverse of the slope at the point considered. different slopes are obtained in 'loading' and 'unloading'. This dependence is only on the direction.) Taking a closer look at point C. Before continuing. as C is a homogeneous function of degree zero on d u . This is a general relation embracing most nonlinear. is a homogeneous function of degree 1. some basic properties of C will be described. if we compare the slopes at points A . is a function of the increment of the stress tensor d u and variables describing the 'state' (or history) of the material. As can be seen. We will consider a uniaxial loadingunloadingreloading test schematized in Figure 4. and C depends on past history (stresses. Therefore. it is possible to write for loading .THE GENERAL FRAMEWORK OF PLASTICITY 89 where Q. strains. .4 where the constitutive tensor C is a scalar. being smaller at higher stresses. which can be written as from which the increments of stress and strain are related by where is a fourthorder tensor. for a given point. it can be seen that. which implies a dependence on the direction of stress increment. A2 and A3. of degree zero in d u . they are not the same.
90 CONSTITUTIVE RELATIONS  PLASTICITY / Strain Figure 4.d a . Leung and Pastor 1985. . Pastor Zienkiewicz and Leung 1985.10) This is the starting point of the Generalized Theory of Plasticity. 1982). or the hypoplastic laws of Dafalias (1986) or Kolymbas (1991). the simplest consists of defining in the stress space a normalized direction n for any given state of stress a such that all possible increments of stress are separated into two classes. Zienkiewicz and Chan 1990). among which it is worth mentioning the multilinear laws proposed by Darve and coworkers in Grenoble (Darve and Labanieh. 1985 and Zienkiewicz and Mroz. introduced by Zienkiewicz and Mroz (Mroz and Zienkiewicz. There are several alternatives to introduce the dependence on the direction of the stress increment. 1985) and later extended by Pastor and Zienkiewicz (Zienkiewicz. However. the total change of strain is not zero as This kind of constitutive law has been defined by Darve (1990) as incrementally nonlinear. Pastor.4 General stressstrain behaviour and for unloading We observe that if we consider an infinitesimal cycle with d a followed by . loading and unloading deL = CL : d a for n : d a > 0 (loading) dev = Cv : d a for n : d a < 0 (unloading) Neutral loading corresponds to the limit case for which (4.
the increments of strain using the expressions for loading and unloading are and It follows that material behaviour under neutral loading is reversible. As for such loading. and it can be very easily verified that for any infinitesimal cycle of stress (do. the accumulated strain is zero. This suggests that the strain increment can be decomposed into two parts where and We note that irreversible plastic deformations have been introduced without the need for specifying any yield or plastic potential surfaces. the tensor Ce characterizes elastic material behaviour. du) where d u corresponds to neutral loading conditions. and it can therefore be regarded as elastic. ~ ngU are arbitrary tensors of unit norm and H L j u two scalar functions and defined as loading and unloading plastic moduli. Indeed. nor hardening rules. It can be very easily verified that both laws predict the same strain increment under neutral loading where both expressions are valid and hence nonuniqueness is avoided.THE GENERAL FRAMEWORK OF PLASTICITY 91 Introduction of this direction discriminating between loading and unloading defines a set of surfaces which is equivalent to those used in Classical Plasticity as will be shown later. Continuity between loading and unloading states requires that constitutive tensors for loading and unloading are of the form and where n . but these surfaces need never be explicitly defined. All .
21) where we have taken into account that the product De : C' is the fourthorder identity tensor.e. nT. when HL is negative. definitions of loading and unloading have to be modified as follows: dcL = CL : d u for n : d u e > 0 (loading) dew = C u : d u for n : d u e < 0 (unloading) where du' is given by (4. First of all. a scalar X is introduced and the increment of strain is written as Both sides of above equation are now multiplied by n: D' n : D' : de from which we obtain = (n : D') : (C' : d u ) + (n : D') : dXngLlu (4. Substituting now . To account for softening behaviour of the material. Inversion of the constitutive tensor Implementation of a constitutive model into finite element codes requires on many occasions an inversion of the constitutive tensor in order to express the increment of stress as a function of the strain increment.du for. Should this not be the case. ngLlu and n. inversion would have to be carried out according to the procedure described below (Zienkiewicz and Mroz.. 1985).du' > 0 etc.17) We note here (and in what follows) that in matrix notation the product forms are written simply as deL = CL. i. This inversion can only be automatically performed when the plastic modulus H i s different from zero.92 CONSTITUTIVE RELATIONS  PLASTICITY that is necessary to specify are two scalar functions HLluand three directions.
. L / u + If we make use of the vectorial formulation to represent tensors. gives du which can be written as du = Del' : d~ = D' : d~  (n (D' : ngLIU) : Dr : d ~ ) HL/U n : D' : ~ . directions n and ngLlu and the elastic constitutive tensor.24) If we now multiply by D' both sides of d~ = Cr : d u we have. L .2.3 Classical theory of plasticity Formulation as a particular case of generalized plasticity theory Classical Plasticity Theory can be considered as a particular case of the Generalized Theory described above by a suitable choice of the plastic modulus. the above expression can be written as 4.dXD' : ~ . du = D' : d~ .THE GENERAL FRAMEWORK OF PLASTICITY 93 we obtain n : D' and : d~ = (HLIU n : D' + : ngLlu)dX (4. u + dXngLlu Substitution of the value of dX.
f depends also on the tensor variable a . Yield and failure surfaces Following experimental evidence. The plastic modulus is obtained through application of the so called 'consistency condition'. there is no plastic deformation. as in the case of kinematic and anisotropic hardening models. the requirement that during yield the stress point should always remain on the yield surface.e. or can be different in which case there is a nonassociative flow rule. with a discontinuity in the derivative of stressstrain curves.. A certain 'hardening law' has to be introduced. The loadingunloading direction is given by the normal to the surface where The direction of plastic flow is similarly derived from a plastic potential surface g(u) = 0 passing through the stress point considered. and. i. Therefore. the material behaviour predicted by Classical Plasticity models presents a sharp transition from the elastic to the elastoplastic regime. Both surfaces can coincide. In the interior of the yield surface. Here we will restrict the discussion to the isotropic case stated above. Sometimes. the plastic modulus is H = m. as will be discussed later. For all K . ) = 0. relating d~ to either incremental plastic work or to the increment of plastic strain. consequently.94 CONSTITUTIVE RELATIONS  PLASTICITY A yield surface is first introduced as where we have assumed that there is a set of scalar internal variables K accounting for the material state and characterizing the size (and shape) of the yield surface. for instance. and the flow rule is then said to be associative. plasticity theories postulate that irreversible or plastic strain appears whenever the stress reaches a surface f (av.
these surfaces can be different. (4. 02. of course. II is the first invariant of the stress tensor.38) JZ and J3 the second and third invariants of the deviatoric stress tensor s.K ) < 0. This is the reason why the yield surface is also known as the failure surface. a the three principal stresses. The scalar K usually characterizes the size of the surface. .THE GENERAL FRAMEWORK OF PLASTICITY 95 f (au. This is. material behaviour is elastic and K is constant. as in the case of materials with hardening. 3 . 11 = 01 + + 02 03 = u. the representation theorems of scalar functions of tensor variables allows a simpler expression for f which can be further simplified to where Y ( K ) is generally some measure of strength. If stress states in the interior of this surface. such as If the material is isotropic. and more complex descriptions are available. however. Care should be taken.. the material cannot sustain a higher stress and failure takes place. a simplification. and a .
The parameter ti in this case is associated with the (negative) plastic volumetric strain.43) and in Figure 4. consider the example given in Figure 4. To illustrate this. being the component in the direction of decreasing volumetric strain if plasticity is assumed associated.e.. where failure takes place. we unload at P2. (a) Yield surfaces (b) Stressstrain curve showing permanent strain upon unloading .5 where a specimen of soft clay is being loaded from an initial state PI to failure at P3. the second or deviatoric invariant and the first or mean. inside which behaviour of the material is elastic. i. as the initial stress is on the yield surface.P2 and P3 correspond to increasing values of ti as shown in Figure 4. softening and failure It is important to distinguish between the yield surface. It has to be noticed that plastic strain appears from the beginning of the test.( t i ) = 0. dti = d~. With each of these is associated an appropriate strain component n. (4. Deviatoric stress Deviatoric stress Axial strain (b) Figure 4.There.5 (b). yield surfaces are the ellipsesf .5 we show the yield surface in the space of two stress invariants. for instance.96 CONSTITUTIVE RELATIONS  PLASTICITY At this stage it is convenient to define also the Lode's angle 19often used instead of J 3 . hydrostatic stress invariant.. there will exist a permanent deformation even when the stress has come back to the original state. If. and the failure surface. with Hardening.5 Typical hardening behaviour of clay. Thus the three stages of loading P I .
Comparing the conditions at PI and Pz. Notice that slopes of the stressstrain curves contradict this definition.6. the elastic domain is bigger in the latter. Indeed. (a) Stress path (b) Stressstrain curve The process of increasing the size of the yield surface in this case is known as hardening. the size did not change and failure takes place as soon as the yield surface is reached.7 Softening behaviour. .6 Ideal Plasticity ( K = constant). and softening behaviour occurs. in the case shown in Figure 4.THE GENERAL FRAMEWORK OF PLASTICITY f 97 Deviatoric stress t Deviatoric stress Hydrostatic stress (a) ('J) Axial strain Figure 4. Some frequently used failure and yield criteria. and the material is harder in this sense. In another loading case. Pressure independent criteria : von MisesHuber yield criterion von Mises yield criterion assumes that plastic strain appears whenever the second invariant of the stress tensor reaches a critical value Y * . (a) Stress path (b) Stressstrain curve Deviatoric stress Deviatoric stress / Hydrostatic stress (a) Axial strain * (b) Figure 4. as the incremental response of the material is harder in the first case. Hardening is not a common feature of all materials.7. the size of the yield surface may decrease. as shown in Figure 4.
8 von Mises .8(b). Alternative expressions are (i) In the principal stress space (ii) In general stress conditions Taking into account that the condition J2 = constant.az.PLASTICITY where Y(K) is generally the tensile strength.Huber yield criterion. In the same figure we show a plane perpendicular to the hydrostatic axis. corresponds to stress states a . von 1 a 3 1 2 Mises criterion is represented in principal stress axes as a cylinder of radius = &!Y which is schematized in Figure 4. If the 2 3 value of limiting tensile stress is a y then we obtain a from which Figure 4. which is shown in Fig. which is referred to as the II plane.4. such that the distance to the hydrostatic axis a = a = a3 is constant. A simple method of determining the constant Y is to perform a tension test a = a = 0 and to determine the instant at which plastic strain develops.98 CONSTITUTIVE RELATIONS . (a) In the principal stress space (b) Section by l plane l .8(a). Its intersection with the von Mises cylinder is a circle.
a stress axes is and the expression of the criterion in principal which corresponds in the al. J2 and Lode's angle 8 Noting that .9).THE GENERAL FRAMEWORK OF PLASTICITY 99 Figure 4. . Tresca criterion The Tresca criterion. 3 In plane stress conditions. This condition. expressed in terms of the principal stresses reads (urnax . a axes to an ellipse with principal axes at 45" (Figure 2 4.50) Substituting now the maximum and minimum principal stresses by their values in terms of the invariants I .9 Von Mises criterion for plane stress conditions = 0.ffrnin) = Y (4. is based on the assumption that plastic straining of a material appears when the maximum shear strain reaches a critical value Y. proposed in 1864.
1 1. it has been used frequently in engineering practice. He assumed that failure occurs on a plane on which the shear stress T . Finally.100 CONSTITUTIVE RELATIONS  PLASTICITY Figure 4. tan q5 (4. Although it is not advisable to think of it as a yield surface.55) to describe the conditions under which failure takes place in soils. The section by the IIplane is a regular hexagon as can be seen in Figure 4. = 02 = a3 (Figure 4.  a.10 Tresca Yield criterion.10(b).. and the normal stress a.7 (compression negative) fulfill the above condition. . the plane stress condition a2 = 0 is represented by which are shown in Figure 4. (a) In principal stress axes (b) In the Il plane we can write finally When plotted in the space of principal stresses.10a). and most finite element codes include it. the Tresca yield criterion is a hexagonal prism. with its axis coincident with the hydrostatic axis a . Pressure dependent criteria : MohrCoulomb surface In 1773 Coulomb proposed the law T = C.
T H E GENERAL FRAMEWORK OF PLASTICITY 101 In terms of principal stresses or invariants.11 Tresca criterion for plane stress conditions Figure 4. we will write and which can be obtained from geommetrical considerations (Figure 4.12.12 MohrCoulomb law .) This results in Figure 4.
it is easy to obtain The MohrCoulomb criterion is represented in the space of principal stresses as a hexagonal pyramid. consists of a cone in which the axis is coincident with the hydrostatic axis.102 CONSTITUTIVE RELATIONS  PLASTICITY Figure 4.13. using the relationships between principal stresses and invariants. the friction angles corresponding to compression and extension are different. in fact. The section of this cone through the II plane is a circle and when plotted in the mean hydrostatic pressuredeviatoric stress plane. when plotted in the space of principal stresses.the relationship between the friction angle and a is .13 MohrCoulomb yield surface and From above. the intersection with it consists of two lines with identical slope (compression and extension). which has been depicted in Figure 4. Therefore. and. given the parameter a and a value of Lode's angle 19.if the intersections of the DruckerPrager cone and the MohrCoulomb surfaces are to coincide for a certain value of Lode's angle 8. The surface is written as and. DruckerPrager criterion The DruckerPrager criterion is an attempt to create a smooth approximation to the MohrCoulomb surface in the same manner as von Mises approximates Tresca.
THE GENERAL FRAMEWORK OF PLASTICITY
103
A similar relationship between cohesion and the parameter Y can easily be obtained as
These relationships have to be taken into account when trying to use the DruckerPrager criterion for plane strain conditions and what is known from experiments is cohesion and angle of friction. Under cylindrical triaxial conditions, i.e., uT = ( u l ,uz = u3)the angles of friction in compression and extension are different, and can be obtained using the above relationship with 0 = 7r/6 and 0 = 7r/6
ac =
2sin4 cos 2  1 2 sin4 sin
JS
from where
and
In a similar way,
The values of Y for compression and extension are
and
104
CONSTITUTIVE R E L A T I O N S PLASTICITY
Finally, it can be seen that, for a given value of a, the relationship between angles of friction in extension and compression is
It is left to the reader as an exercise to demonstrate that there is a value of sin& for which the friction angle in extension reaches 7r/2. It is interesting to mention that 'rounded' MohrCoulomb surfaces have been proposed in the past (Zienkiewicz and Pande, 1977), in which the slope M is assumed to vary as:
In this way, the yield surface is smooth. and coincides with the MohrCoulomb original surface in both triaxial compression and extension conditions.
Consistency condition for strain hardening materials
If we assume that the material hardening is of 'strain' type, there will exist a law relating the increments of K and E
Assuming that yielding occurs on the yield surface given by
and hence
This can be rewritten using (4.71) as
If. in the above expression, we substitute dd' it gives
THE GENERAL FRAMEWORK OF PLASTICITY
105
This expression can be further developed to
from which the plastic modulus is finally obtained as
Using the alternative vector notation, the above expression is written as
where d t i / d ~is a square matrix. ~) Local failure conditions or continuing deformation at a constant stress state can happen whenever H L = 0, which corresponds to:
for which either of the following conditions have to be fulfilled
d n dg (b).  = 0 with ~ E Pd u

dn # 0 ddl
Computational aspects
Most of the expressions given in the above sections simplify in the case of isotropic materials, as all necessary items can be defined in terms of stress invariants. The yield surface, for instance, can be expressed as
106
CONSTITUTIVE RELATIONS

PLASTICITY
or in terms of another alternative set of invariants. The constitutive tensor DePwhich appears in nonlinear finite element computations can be expressed, as it was shown above, as a function of directions n and n,, and the scalar H, all of them dependent on the invariants. However, what it is needed in computations is the general threedimensional form. Therefore, the expressions have to be transformed from the space of invariants to the general 3D space. In the case of the Classical Theory of Plasticity, the constitutive tensor DeP is written in vector notation as
where we have introduced
having dropped, for simplicity, subindexes L/ U referring to loading and unloading. This is precisely the Generalized Plasticity expression (4.28) with
n
=
af
du
and n,
=
ag
A simple way to obtain either gradient in terms of invariants I , , J2 and 0 is the following
where sin 38 =  2 5,312 Differentiating this expression, we arrive at
3d3
J3
THE GENERAL FRAMEWORK OF PLASTICITY
107
and, from here,
Taking into account now that
we have, finally,
which can be written in a more compact way as:
n = C~nl C m
where
+
+C m
and
It can be seen that the set of constants { C ; )depends on the yield criterion chosen, being independent of the vectors ni. Next, we will obtain the explicit form of nl, n2 and n3.
108
CONSTITUTIVE RELATIONS
nl

PLASTICITY
Vector
is given by
nl
or.
= =  
811 da,,  d(aikSki) 6 . 6 6. = / kr  r/ ! i au a ~ ! ~ doji
d {a\
aa,
+a,. + a,)
=1
from which, in vector notation,
To obtain vector nl we will use tensor notation
Taking into account that
we arrive at
and
THE GENERAL FRAMEWORK OF PLASTICITY
109
from which it follows that
Therefore,
Vector n3 is given by
where
1 J3 = j~ i j ~ j k ~ k i
After some algebra, the final expression for n3 is
Constants CI , C2, and C3 depend on the yield criterion chosen. In the case of the von Mises yield criterion, which can be written as f=J2y2=0 (4.106)
we find
3. Shear behaviour in axisymmetric triaxial tests: Drained and undrained tests Effects of density and confining pressure Liquefaction of loose sands Memory effects: overconsolidation. Liquefaction and cyclic mobility. there are a great variety of models. Isotropic consolidation. This has been made possible because of extensive work developed in laboratories throughout the world. Loading. = Sometimes. able to deal with situations ranging from simple monotonic stress paths to cyclic loading. 2. 4. rotation of principal stress axes and anisotropy. Most of the proposed benchmark tests deal with key aspects of granular and cohesive soil behaviour: 1. the initiatives of laboratories such as 3S in Grenoble. have to be mentioned. reloading and cyclic loading: Densification Pore pressure buildup. Here. No finite element code will provide results of better quality than that of the constitutive equation implemented in it. benchmark tests on some selected reference materials have been made available to constitutive modellers. Threedimensional effects 5. Today. 4. unloading and reloading. Memory effects.1 CRITICAL STA TE MODELS Intvoduction Constitutive modelling of soil behaviour is a keystone in the process of predicting the behaviour of a geostructure. Anisotropy: Material Induced by loading. 3.110 CONSTITUTIVE RELATIONS  PLASTICITY 0. as an important result. and. CERMES in Paris (both part of GRECO and GEO networks in France) and Case Western University in Cleveland (USA). Unloading.3 4. In the past years. an alternative expression using the 1D yield stress is used a and then C2 is a. coordination of effort between different groups has increased. .
This line was referred to as the Critical State Line and is one of the basic ingredients of Critical State Theory introduced by the Cambridge group (Roscoe. however. which made it necessary to introduce new concepts. in 1952. The last section was devoted to the introduction of elastoplastic constitutive models in the framework of Generalized Plasticity. the engineer needs to be acquainted with them. Several factors have contributed to it: 0 0 0 Industrial Finite Element codes do not implement constitutive models suitable to realistic geotechnical analysis. 1968). the associated behaviour is not valid. q) space where the residual states lie. To calibrate more advanced models. Drucker.3. Later. and a hardening law dependent on density. The simple models of Tresca and von Mises present severe limitations when applied to geomaterials in general and soils in particular. Schofield and Wroth.2 Cvitical state models fov novmally consolidated clays Hydrostatic loading: isotropic compression tests One of the basic features of soil behaviour is the importance of its density on its behaviour. this model is not able to describe plastic deformations inside the yield surface cone. p'.CRITICAL STATE MODELS 111 These progressively more sophisticated tests have helped to develop constitutive models of increasing complexity. and it was shown there how Classical Plasticity models can be considered as particular cases of the theory. as occur in common engineering situations. Both cohesive and granular soils exhibit changes of density caused by 0 0 a change in effective confining pressure p' and changes of arrangement of grains in the structure induced by shearing of the material. arriving not only at practical expressions describing volumetric hardening. Schofield and Wroth. but also at the concept of a line in the ( e . a dramatic gap between these recently developed constitutive models and those used in daytoday engineering practice. Moreover. extensive research on the basic properties of soils in triaxial conditions was carried out at Cambridge University (Henkel 1956. Gibson and Henkel introduced an elastoplastic model including two fundamental ingredients. Drucker and Prager proposed. 1958 and Roscoe and Burland. as it would predict large dilatancy at failure. paving the way to modern plasticity. in 1957. . and these ideas were further elaborated. There exists. an elasticperfectly plastic constitutive model with an associated flow rule which could be applied to limit analysis problems. 1958. The purpose of this Part is to describe classical elastoplastic models for soils. Special laboratory tests are frequently required to obtain material parameters which cannot be obtained by direct observation from raw data. At the same time. 4. together with their limitations. 1963. a closed yield surface which consisted of a cone and a circular cap. Roscoe. Schofield and Thurairajah. 1960 and Parry. Henkel. However. 1960).
in which the confining pressure is varied. where we have used the minus sign for consistency with the definition of p'. or the change of voids ratio e We will consider now a loadingunloadingreloading process 12345678.14 Hydrostatic compression stress path The simplest case in which the first mechanism occurs is hydrostatic loading of a soil specimen. If the initial state of stress is and the specimen is loaded according to the stress path will consist on a segment of straight line along the hydrostatic axis d.q ) (Figure 4. This cycle is followed by a loading branch 45. we have considered compressions as negative. (Figure 4. to p'. with a final pressure of p'.= d!= d'.112 CONSTITUTIVE RELATIONS  PLASTICITY Figure 4.15a) . The change of volume can be described either by the volumetric strain.14). and then the specimen is unloaded to pi and reloaded again to pi.. In above. or along the axis q = 0 if we are using the plane (p'. in which hydrostatic pressure p1 is increased from p'. The process is carried out slowly enough to prevent the development of intersticial pore pressure (drained conditions).
15 Hydrostatic compression test on a normally consolidated clay. both states are inside yield . no plastic deformation would have been produced. and. 1978) It is important to note that. if a MohrCoulomb or a DruckerPrager yield surface had been used. If such plastic deformation needs to be reproduced.16 illustrates this fact.15(b). (a) Experimental results (b) Idealized behaviour It can be seen that unloading and reloading branches differ.CRITICAL STATE MODELS 113 Figure 4. closed yield criteria should be used instead. the response of a soft cohesive soil can be idealized in the In($) e plot as a line of slope X (Points 1258)  or. and it can be related to the Plasticity Index PI by the empirical relation (Atkinson and Bransby. yield criteria which are open in the hydrostatic axis. this is a severe limitation of all finite element codes implementing. The parameter X depends on the type of soil. and can be sketched as shown in Figure 4. irreversible plastic deformation occurs from 1 to 2 and from 4 to 5. This behaviour is typical of soft clays. Figure 4. alternatively X dp' d ~ ? . If time effects can be neglected. However. = 1+ep1 This line is often referred to as the 'Normal Consolidation Line'. although volumetric strain developing in the branch 234 can be considered to be reversible. If the stress increases from point 1 to 2. and is one of the basic ingredients of modern plasticity models for soils. as unique options. therefore.
a plastic strain develops.= de.dp' 1 + e p'  During unloading from 2 to 3. = dp' ltey' K where K is a new constant characterizing the elastic volumetric response. Loading from 1 to 2 expands the yield surface. which can be assumed to harden as the soil densifies.. The observed volumetric strain can be decomposed into elastic and plastic parts according to d ~ . It can be related to the bulk modulus by K. no plastic strain is produced in the process. and subsequent reloading to 4 the behaviour will be purely elastic d ~ .16 Open and closed yield surfaces surfacefi. det = A + d ~ := . the yield surface expands and the soil continues hardening. and therefore. to the plastic volumetric strain is obtained as from where . The solution is to use yield surfaces intersecting the hydrostatic axis q = 0. . which will be denoted by p. .. = 1+e K Once the stress point reaches the yield surface again. A simple law relating the size of the yield surface.114 CONSTITUTIVE RELATIONS  PLASTICITY Figure 4.
and a vertical additional load (a. and 3 consist of a cell pressure a applied through a fluid. If a soil specimen which has been subjected in the past to a consolidation pressure of p: = p i is tested at a lower pressure pl. This soil is referred to as 'overconsolidated'. it is worth mentioning those problems related to accurate measurement of vertical loads. while soils at the normal consolidation line are called 'normally consolidated'. Here we will concentrate in shear behaviour of normally consolidated clays subjected to symmetric or cylindrical triaxial stress conditions which hereafter will be referred to as 'triaxial'. In addition to these. but can be generalized to more complex situations where the OCR will be the ratio of the measures of two stress states. usually water. or OCR.15(a) in the branch 345. a3)referred to as a 'deviator' applied with a ram. inhomogeneities caused by the development of narrow zones where the strain localizes. it will be possible to observe in the curve e . and measurements of axial. and the process 125 is referred to as 'isotropic consolidation'. Triaxial test So far. There are several problems like membrane penetration. . and it was mentioned that another mechanism causing densification was shear. It was seen that isotropic compression results in densification and hardening of soil. The triaxial test is commonly used in the laboratory to determine soil properties as the desired stress paths can be reproduced quite accurately. Both concepts can be easily understood in the framework of plasticity. The 'Overconsolidation Ratio'.lnp' a change of slope such as depicted in Figure 4. homogenization of pore pressures inside the specimen. Of course. as overconsolidated soils are characterized by the stress state being inside the yield surface at the initial state. radial and volumetric strain.CRITICAL STATE MODELS 115 The subscript 'c' refers to consolidation. is a parameter measuring the degree of overconsolidation. and friction with the upper and lower rigid caps. unless the soil is anisotropic. The stress conditions applied in a triaxial cell are sketched in Figure 4. known as 'shear bands'.17. changes in specimen crosssection. these definitions apply only to simple hydrostatic loading conditions. we have only considered stress paths where no shear strain is induced.
.17 Triaxial stress conditions Figure 4.) = pl(l. In the first case. and (ii) Consolidated undrained.a.116 CONSTITUTIVE RELATIONS .PLASTICITY Two main types of tests are currently used: (i) Consolidated drained.18 Consolidated drained stress path . a saturated soil specimen is brought to an initial state of hydrostatic stress I T u1= (ah. As pore pressures are zero (drained conditions). 1. and once the initial conditions are reached. The load is applied slowly to avoid pore pressure buildup.l) where the pore pressure is zero. Figure 4. the vertical load applied through the ram is increased. a. the total and effective stress coincide.18. . The stress path is depicted in Figure 4.
M.)I from which we have q =  (u'. alternatively. The slope of the stress path can easily be obtained as .q.. This fact occurs when applying compressionextension cycles. In above. The measures of strain are those workassociated top' and q Concerning Lode's angle. both stresses are negative. q) plane is shown in Figure 4. .19.. . as these invariants are given by and with = 1 (u'. the angle AOB is smaller than AOB'. a'. where it can be seen that. the soil will fail earlier in extension. or. and we have supposed that the absolute value of d. 8). is higher than that of a'. The last is very convenient. it can be seen that both the hydrostatic pressurep' and the deviatoric stress are increasing. Therefore. due to its positive slope. provided that (d.. if the failure surface is of MohrCoulomb type. it is kept constant during the test.CRITICAL STATE MODELS 117 In this figure. and for a given absolute value of the deviator q. in the spaces ( I .) does not change sign during the test. The stress path in the (p'. .. 2 Me. 19) or (p'.u i ) (4.J 2 . The stress path can be studied either in the space of principal stresses.127) which is precisely the stress induced by the vertical load applied through the ram.
ir Figure 4.20 Typical results of CD tests on normally consolidated clays .118 and CONSTITUTIVE RELATIONS  PLASTICITY from where. taking into account that d' is constant.19 Consolidated drained stress path in the p' q plane Figure 4.
measurements of pore pressure.. this is a simplification. It can be seen how the effective stress path bends towards the origin as a consequence of pore pressure increase caused by the tendency of soil to compact. During the test. Undrained behaviour of normally consolidated clays provides an interesting illustration of the shortcomings of the MohrCoulomb criterion when used as a yield surface. This test has been classically used to characterize soil behaviour under 'fast' loading. where pore pressures do not have time to dissipate. Figure 4. axial strain. The main features are the following: 0 There is a tendency of the soil to compact as the test proceeds. Failure takes place at a certain value of stress ratio q = M for tests performed at different confining pressures. the drainage valve is closed to prevent dissipation of pore pressure. vertical stress and cell pressure are taken to monitorize the stress path. Of course. It can be seen how the model overestimates soil strength because it .21 Consolidated undrained tests on normally consolidated clay The results obtained in compression tests on normally consolidated clays are similar to those depicted in Figure 4.22 compares the predicted behaviour of such a model with that observed in the laboratory. Again the failure takes place at a line of slope M. and a complete coupled analysis should be performed instead. where.CRITICAL STATE MODELS 119 Figure 4. in shortterm stability analysis.21 shows typical results obtained in drained consolidated clays.20. caused by the increase of p' and a rearrangement of soil particles. It has to be noticed that soil strength is lower in undrained than in drained conditions because of the generated pore pressures. which coincides with that obtained in drained tests. The test has to be carried out slowly enough for the pore pressure to be homogeneous through the specimen. after consolidation.. 0 0 A second type of triaxial test is the Consolidated Undrained (CU) test. Soil strength and compaction depend on confining pressure. Figure 4. and increase with it.
in which elastoplasticmodels for soilscould bedeveloped. The stress paths resulting either from consolidated drained and undrained tests lie on a unique state surface referred to as the 'Roscoe Surface'. This line was depicted in Figure 4.. There exists a line in the space (e. lnp') plane is parallel to the NCL. The projection of this line on to the (e. The basic ingredients of Critical State Soil Mechanics are the following: There exists a line in the (e. Schofield and Thurairajah.23). In p'.. no plastic strain is produced until the yield surface is reached.. shear deformation takes place without change of volume. and on the theoretical and experimental work of researchers from the University of Cambridge. depending on whether they lie in the space between both lines or not (Figure 4. who first introduced the ideas of volumetric hardening and a closed yield surface. This process will be endless. The interest of this line is that it provides a volumetric hardening rule which can be generalized to general stress conditions (Roscoe. Predicted  PLASTICITY Experiment p'o Figure 4.. At this line.15(b) (1258). In the MohrCoulomb model. q) where all residual states lie. if the flow rule is associative. 1960). the process is stopped by cavitation of the pore fluid.. This fact was found . and the stress path will turn to the right following the yield surface (BlC). which is referred to as the 'Normal Consolidation Line' (NCL). and divides initial states into 'wet' and 'dry'..120 CONSTITUTIVE RELATIONS .22 Predicted (MohrCoulomb) and observed behaviour in CU tests cannot predict the pore pressures caused by plastic volumetric strain which develop during the test. Gibson and Henkel(1957).. 1963). In reality. In addition to that. dilation and negative pore pressures will develop at failure. and the deviatoric stress will keep increasing continuously.. Critical state models It can be said that modern plasticity models for soils are based both on the pioneering work of Drucker. independently of the type of test and initial conditions (Parry. who provided the framework of Critical State Soil Mechanics.. In p') plane in which all stress paths in normally consolidated clays lie.
What is useful. as mentioned above. the yield surface is determined.24)...24 Constant water content lines as obtained from C D and CU tests (sketched) . is that it gives a hint of the kind of yield surface.. The first step. who plotted the water content contours obtained in drained tests and found that undrained tests paths followed these lines as well (Figure 4. during undrained paths the soil hardens as plastic volumetric strain is produced...23 Normal consolidation and critical state lines . simple elastoplastic models can be derived... Next. Roscoe. Schofield and Thurairajah (1963) assumed that incremental plastic work Figure 4. however. is to assume as a hardening rule where pb is a parameter characterizing the size of the yield surface. This fact is not directly applicable by elastoplastic models. From here. while the sum of the plastic and elastic increments of volumetric strain is kept constant. as these isolines are not yield surfaces corresponding to constant values of the hardening parameter.CRITICAL STATE MODELS 121 experimentally by Henkel (1960). In fact. from CD tests CU tests . w=ct..4 Figure 4..
along the surface we obtain .P/d&.P is obtained as where it is interesting to note that dilatancy is zero at the Critical State Line. The normal to the plastic potential surface is proportional to and 7 Therefore. dilatancy d. dilatancy is given by If we take into account that. = d&.122 CONSTITUTIVE RELATIONS  PLASTICITY was given by 6 WP = Mp'd&r from where Using the above expression.
25 Yield and Plastic potential surfaces of Cam clay Model Figure 4. This surface has been depicted in Figure 4. This model was further elaborated by Burland (1965).CRITICAL STATE MODELS 123 which can be integrated to obtain the plastic potential where p:.26 Yield surface of modified Cam clay model. . If the flow rule is assumed to be associated. who suggested an ellipse as a yield surface. is not directed along the axis. the normal will not be uniquely defined in threedimensional stress conditions. where it can be seen that the normal to the surface at p' = p'. the yield surface coincides with g. Therefore.25. although it can be assumed that the surface is rounded off in the proximity of the axis so that the normal is directed along it. is the abscissa at which the surface intersects the hydrostatic axis q = 0. The work dissipation was given by PC Figure 4.
who suggested that the moment at which the stress ratio (deviatoric stress in the plot) reaches the value at which residual conditions will take place later. it dilates. as occurs in the case of cohesive soils. In the first part of the test the sand contracts. Concerning the deviatoric stress. The behaviour of granular soils depends mainly on density.26 4. They depart from reality (i) when applied to overconsolidated soils. and then it softens.27. From the point of view of elastoplasticity. the behaviour is similar to that depicted in Figure 4. When sheared in triaxial conditions.124 CONSTITUTIVE RELATIONS  PLASTICITY from which dilatancy is obtained as with The yield surface can easily be obtained by integration of the above and is given by which is depicted in Figure 4. a critical state exists for sands. it can be assumed that dilatancy is always zero at the line 7 = Mg.3. and two extreme classes of behaviour can be identified. reaching a minimum void ratio e. it increases until a peak is reached. p Figure 4. Sands at very dense states can be prepared in the laboratory by vibration and not by compaction. Axial strain Axial strain Triaxial tests on dense and loose sand (schematized) . Therefore. and (ii) when applied to sands. This fact had been first established by Rowe (1962). and.27 * Dense . it stabilizes at residual conditions. Finally.3 Extension to sands The models described so far are able to predict with reasonable accuracy the behaviour of normally consolidated clays. The results sketched in the figure follow the ideas of Taylor (1948). the volumetric strain presents a peak. from there. where the plastic flow takes place at constant volume. as it is not possible to reproduce inelastic strain which develops inside the yield surface.
28 Behaviour of dense sand as predicted by critical state models either before reaching the critical state or at that point. 1975). The behaviour is elastic from 1 to 2. as the strength is underestimated. An important difficulty encountered is that the specimen is no longer homogeneous long before residual conditions. If such a process is reproduced with a critical state model. It can be seen there that the difference from the observed behaviour is important. and. as shown in Figure 4. . the results will be similar to those shown in Figure 4. 1978) or the Line of Phase Transformation' (Ishihara. Therefore. where the yield surface is reached. which behaves as a viscous fluid (Figure 4. As the soil dilates. very loose sands present liquefaction under undrained conditions. the best choice is to assume the sand as overconsolidated. therefore. It is important to note that the behaviour in the descending branch corresponds to increasing values of the stress ratio. Tatsuoka and Yasuda.29 (b). If such behaviour is modelled by a basic critical state model such as described in the preceding section. several investigators proposed a separate denomination of this line referring to it as the 'Characteristic State Line' (Habib and Luong. Under undrained conditions. and the stress path follows from 2 to 3. In fact. q) plane (Figure 4.28. it is not sound to assume this behaviour as softening. the results present more important discrepancies.CRITICAL STATE MODELS (4 A 125 2 3 1 ' t Axial strain Axial strain Figure 4.29(a)). Dense sands separate at the beginning of the test from the undrained stress path of an elastic material. the observed softening is rather of a structural than a material nature. where it stabilizes at critical state conditions. it softens. The phenomenon consists of a sudden drop of resistance of the soil. The separation from the vertical of the stress path shows that plasticity is present from the beginning of the test. which is a vertical segment on the @'. On the other side of the density spectrum.30). as the strain concentrates in the shear bands.
(a) Experimental results (b) Predicted Such behaviour cannot be described by the models presented above. and (iii) plastic deformations existing throughout the process. (ii) nonassociative plastic flow rules.21. If the material is assumed normally consolidated. . The three fundamental ingredients were: (i) hardening laws depending on deviatoric and volumetric plastic strain. the results will be similar to those shown in Figure 4.126 CONSTITUTIVE RELATIONS  PLASTICITY Axial strain (4 Axial strain (b) Figure 4.29 Undrained behaviour of dense sand in CU triaxial test. All these limitations motivated further research to extend the range of application of CS models. especially the dramatic loss of resistance.
. Failure will occur when . the Critical State Line at which can be crossed. as the plastic modulus is zero there otherwise. Therefore.30 Liquefaction of very loose sand The first is required to cross the line 77 = M. Deviatoric hardening was introduced by Nova (1977) and Wilde (1977).CRITICAL STATE MODELS 127 I Axial strain I Axial strain Figure 4. who assumed a hardening parameter of the type Y=E~+D< where (4.149) < is the accumulated deviatoric shear strain The size of the yield surface was made to depend on Y. without the plastic modulus is zero. and the plastic modulus in triaxial conditions was found to be proportional to ag + ap' Dag ay.
which results in a discontinuity of slope but with the desired result of coming back to CS conditions. and liquefactionlike behaviour can be modelled in the softening regime. Holubec and Sherbourne (1966 and 1967). and Pastor. .31 Plastic potential and yield surfaces for (a) loose sands (b) dense sands . The plastic potential and flow rules can be determined from experiment. proposed by Nova (1982).' Figure 4. Zienkiewicz. as shown in Nova and Wood (1978). the path will not return to the CSL. and failure will take place with dilation. Finally. This surface can be obtained from the dilatancy rule dg = (1 +a). assuming now that a single expression was valid for the full range of stress ratios where pi is the abscissa at which it intersects the p' axis. consists of making them drop D to zero. If a negative value of D is assumed. Nova (1982).128 CONSTITUTIVE RELATIONS  PLASTICITY which happens at a stress ratio higher than M. Zienkiewicz and Leung (1985) used as plastic potential a simplification of that proposed by Nova and Wood (1979). Once there. it is more sound to assume this process to be of the hardening type. Nova and Wood (1979). Pastor. as the stress ratio is continuously increasing. CSL n .. Zienkiewicz and Leung (1985). if D is kept constant. Humpheson and Lewis (1975). The second ingredient is a nonassociative flow rule. as suggested by Poorooshasb. then the plastic modulus becomes zero at a stress ratio lower than critical.. where surfaces were defined by different analytical expressions valid for different ranges of the stress ratio. (Mg  rl) CSL . Zienkiewicz and Leung (1985). Another possibility. As discussed above. a hardening law with saturation can be assumed to hold for D This law was suggested by Wilde (1977) and applied to a bounding surface model by Pastor.
depends on the relative density and indeed it can be assumed to be the same as D. they proposed curves belonging to the same family where M f # M. it can be shown that variation of pore pressure and densification in undrained conditions are related by the expression where n is the porosity. material remained elastic within the yield surface. Here. (1977) who developed densification laws in the context of the endochronic theory. They incorporated as basic ingredients a plastic potential and a yield surface. or phenomena occuring during cyclic loading. Figure 4.ADVANCED MODELS 129 As yield surfaces.. The immediate consequence is that these models are unable to reproduce either the behaviour of overconsolidated soils. as the latter is a direct consequence of the tendency of soil to compact.4 ADVANCED MODELS Zntvoduction 4. 1975). . 1982).4. The first consisted in developing new cathegories of models embracing classical plasticity as a particular case. it motivated an important effort of research. 1978 and Zienkiewicz et a/.1 So far. which proceeded along two main lines. the latter being allowed to expand or contract depending on whether the material was hardening or softening... we have discussed in the previous sections some classical plasticity models for soils which have proven to reproduce accurately enough the behaviour of soil under monotonic loading. both phenomena are related.31 (a) and (b) show plastic potential and yield surfaces for very loose and dense sands. These will be described next. in general. from which suitable densification laws were produced. such as pore pressure generation in fast processes or densification. An interesting fact reported by them is that the ratio M. and that of Zienkiewicz and coworkers at Swansea University. In fact. and the second was based on introducing the volumetric deformation which is produced by cyclic shearing of a soil as an 'autogenous volumetric strain'. we could mention the work of Bazant and Krizek (1976) and Cuellar et 01. where no plastic deformation can develop. who developed simple densification models and implemented them into coupled numerical models (Zienkiewicz er al. / M.. 4. As these are key aspects which reproduce failure of the soil caused by liquefaction or cyclic mobility phenomena (Martin et ul. Kf the bulk modulus of pore fluid and KT that of the soil skeleton. However. Indeed.
In the model proposed by Zienkiewicz et nl.. the accumulated deviatoric strain was quantified by a variable defined as < where eii is the deviatoric strain. We must. The densification is described by a law where A and B are constants and K a parameter defined as Here.. This interpretation has motivated the development of the so called densification models. however. y is a third parameter of the model. and B is given by where I@/ is the amplitude of the stress cycles (deviatoric) and a.159) where Dep accounted for the elastoplastic behaviour and d ~ for the densification " caused by cyclic loading. and (ii) liquefaction in the last cycle as it occurs in monotonic loading. The first successful theory was the multisurface .. (1978). The constitutive equation was written as da' = D. Kinematic hardening models The second approach consisted in extending the theory of plasticity beyond the limits imposed in the classical formulation.130 CONSTITUTIVE RELATIONS  PLASTICITY Densification models Phenomena such as liquefaction can be thought of being caused by: (i) an accumulation of pore pressure with the number of cycles. In general a nonassociative MohrCoulomb model with zero dilatancy is assumed here for the elastoplastic behaviour.. where simple elastoplastic behaviour of soil and the accumulation of pore pressure were taken into account by two different mechanisms.(d€  dq) (4.o the initial value of the mean effective stress at the beginning of the cyclic loading process. The 'densification' model differs from other elastoplastic models described in this text by a rather arbitrary separation of effects. remark that it is capable of modelling with good accuracy the liquefaction phenomena encountered in earthquakes and because of its simplicity it deserves wider use. For this reason we defer the discussion of its most recent forms to the end of this chapter.
A similar approach. and evolved to what is known today as 'Bounding Surface Theory' (Dafalias and Herrmann. for loading processes inside it. 1987. Memory of loading events was kept through the position and size of the surfaces at which stress reversal took place. This movement must comply with a rule which ensures that surfaces never intersect each other. the results coincide with those of classical plasticity. 1978. Prevost. Hirai. Since then. a field of hardening modulii can still be described by prescribing the variation between both surfaces. The number of surfaces allows us to keep track of loading events such as the maximum stress level reached. On the bounding surface. However.e. the price to pay in numerical computations is high. as observed by Sangrey.ADVANCED MODELS 131 kinematichardening model proposed by Mroz in 1967. Wang Dafalias and Shen 1990. 1982). Kaliakin and Dafalias 1989 and Bardet. Mroz and Norris (1982) showed an application of the model to the cyclic behaviour of normally consolidated and overconsolidated clays. kinematic or anisotropic hardening models have been shown to perform well in modelling liquefaction and other cyclic loading phenomena (Ghaboussi and Momen. the field of plastic modulii was made continuous in the whole domain enclosed by the outer surface. plastic strain develops according to classical plasticity theory. further developments and improvements have taken place (Mroz. Hirai. However. and simplified versions were sought. the 'subloading surface model' was proposed by Hashiguchi and Ueno (1977) and Hashiguchi. 1982. An elastic domain may also be postulated. such as may . 1987). An improvement of this 'multisurface model' was provided by Mroz.. such as memory of past events and plastic deformation during unloading. The elastic domain was assumed to shrink to a point. where a set of 'loading surfaces' within an outer 'boundary' surface was postulated. and the plastic modulus obtained through application of the consistency condition describing material hardening or softening properties. given by the normals to the bounding and plastic potential surfaces. 1989). In this way. the surfaces reached by the stress path translate until a new loading surface is attained. who introduced an infinite number of nested loading surfaces. 1986. In the case of loading processes beginning at the bounding surface. making the hardening modulus depend on the ratio of the sizes of active loading and outer or consolidation surfaces. Imamura and Ueno (1989). Dafalias. or points at which stress has reversed. et al. As the stress is increased from an initial state. i. Other elastoplastic. This model was independently proposed by Krieg (1975) and Dafalias and Popov (1975). Henkel and Esrig (1969). Large intensity loading events erase lower intensity events. corresponding to the volume enclosed by the inner surface. Bounding surface models and generalized plasticity If the number of surfaces is reduced to two. with directions n and n. Aubry. Norris and Zienkiewicz. 1977. the outer or consolidation and the inner or yield. Norris and Zienkiewicz (198 l). and found that the model was able to predict final states lying on an 'equilibrium line'. It can be seen that both the 'multisurface' and the 'infinite number of surfaces' models are able to reproduce most of the basic features of soils under cyclic loading. 1982.
its intersection with the bounding surface being taken as the image point. The main shortcoming of early BS models was their inability to reproduce plastic deformations which develop when unloading. in P were assumed to be those at Pss and the plastic modulus was interpolated according to a simple law where 6 is the distance from the origin to the stress point P.32).a line was drawn passing through the origin and point P. which was applied by the authors to reproduce the behaviour of both cohesive and frictional soils under monotonic and cyclic loading. and it was overcome within the more general framework of generalized plasticity (Pastor. A further step was given by Pastor. There. the difference is that bounding surface models are able to introduce plastic deformations by using some interpolation rules relating the stress point P(C in Figure 4. introducing a full generalized plasticity model in Pastor and Zienkiewicz (1986) and Pastor. Directions n and n. Here. but plastic deformations during unloading were introduced within the more general framework of generalized plasticity.132 CONSTITUTIVE RELATIONS  PLASTICITY I Bounding surface Figure 4. which is taken into account . the model was of bounding surface type for loading. Zienkiewicz and Chan (1990).32) to an image of it on the BS. Simple interpolation rules were proposed by Dafalias and Herrmann (1982). Hypoplasticity and incrementally nonlinear models One of the basic characteristics of the elastoplastic tensors Dep and Cep is their dependence on the direction of loading u = du/lldull. to obtain the image point PBS.32 Bounding surface interpolation occur in cyclic loading. y being a parameter of the model (Figure 4.32). 1985). and So the distance between the origin and the image point PBs. Leung and Pastor (1985). and by Zienkiewicz. PBs(Bin Figure 4. Zienkiewicz and Leung. Zienkiewicz and Chan.
as they are based on the assumption that the incremental nonlinearity in the constitutive tensor may be approximated by suitable interpolation laws once material behaviour along different stress paths has been established. such as dependence on u. Wang.2 and 3. in this sense.ADVANCED MODELS 133 in a simple way by introducing a direction n for each mechanism of deformation considered. 1990. 1986. Consequently. Alternatively. producing general expressions for the constitutive tensor. Hypoplastic models have been introduced also in Karlsruhe by Kolymbas and coworkers (Wu and Kolymbas. Among all the constitutive models of this kind it is worth mentioning those proposed by Darve and Labanieh (1982). (1991). the most promising framework in which this goal can be achieved. . (1986). and Kolymbas. If constitutive tensors along positive and negative directions of the principal directions are N+ and N. a simple incrementally nonlinear law could be given by where The model proved to reproduce well the behaviour of soils under both monotonic and cyclic loading (Darve. Dafalias and coworkers presented extensions of the bounding surface model within the framework of hypoplasticity (Dafalias. six values of C were required for the interpolation along a particular direction. Darve and Labanieh (1982) suggested that these paths could correspond to positive and negative directions along principal stress axes 1. Dafalias. Hypoplasticity is. They are referred to as incrementally non linear models. 1990). it is possible to provide general expressions for the constitutive tensors satisfying all necessary requirements. The constitutive tensor was assumed to be given by were u is the unit tensor along direction of loading increment. 1993). One of the first models which can be considered of this kind was introduced in Grenoble by Darve and Labanieh (1982). and since the early days of the theory it has been considerably improved (Desrues and Chambon 1993). Flavigny and Rojas 1985). Kolymbas and Wu. Dafalias and Shen.
d~ ' d ~ l L  d& dt. and de. To this end. experiments carried out by Balasubramanian and Chaudhry (1978) using constant p'lq stress paths suggest that dilatancy can be approximated by a straight line in the p'q plane (Figure 4.134 CONSTITUTIVE RELATIONS  PLASTICITY 4. q the effective confining pressure and a measure of deviatoric stress.4. it is possible to use values of dilatancy measured in laboratory tests d" . . are strain measures workconjugated top' and q SW = d : de = (p'q)..33). cr a material constant and 7) is the so called stress ratio defined as Direction n. To obtain a flow rule.q where e is the voids ratio and p'. direction n can be immediately obtained. where E. Dilatancy is therefore expressed as where M. We will begin by assuming that a residual critical state exists in the space e . is now given by . (2:) defined as If the ratio between the plastic increments of volumetric and deviatoric strain is assumed to be the same then the ratio of total (elastic plus plastic) increments observed in laboratory tests. is the slope of the critical state line in the p'q plane.p' .2 A Generalized plasticity model for clays Normally consolidated clays The simplest case of normally consolidated clay under virgin loading will be considered first.
5 I . 1985). Zienkiewicz and Leung (1985) according to test results reported by Frossard (1983). Concerning direction n.33 Dilatancy of soft Bangkok clay (after Balasubramanian and Chaudry) with This law can also be used to describe dilatancy of granular materials.O Figure 4. as it was suggested in Pastor. Therefore. we will have with . who performed experiments in three cohesive soils and found little discrepancies from plastic potential and yield surfaces (Atkinson and Richardson. we will assume that the flow rule is associative. following Atkinson and Richardson.ADVANCED MODELS 135 0.
= .. and Ho is a material constant. p'  Comparing now the above equation to the general expression for the plastic strain increment 1 d~~ = n. In p') plane. for which the increments of volumetric elastic and total strain are given by K dp' d ~ .183) for the stress path considered here. we will consider an isotropic consolidation test of a normally consolidated specimen. it can be concluded that the plastic modulus H L is given by The parameters X and K are the slopes of the normal consolidation and elastic unloading lines in the (e. decreasing as the later increases until reaching a value of zero at the critical line (7 = q/pl = M). . To generalize this expression of the plastic modulus to other conditions than isotropic compression paths. = l+epl from which the plastic volumetric strain increment is K ~ E P = (AI + e ) dp' . l+epl and x dp' de.136 CONSTITUTIVE R E L A T I O N S PLASTICITY and In what follows. we will make the assumption that plastic modulus depends on the mobilized stress ratio. To obtain the plastic modulus for virgin loading. we will drop the subindex 'g' referring to the plastic potential surface as it coincides with the yield surface.(n HL which particularizes to : da') (4.
ADVANCED MODELS 137 Therefore.. 1977) + where M. where f ( q ) is such that and A suitable form was proposed in Pastor. Below we define a smoothed version of Mohr's criterion widely used in practice (Zienkiewicz and Pande. Zienkiewicz and Chan (1990) where do = (1 a ) M and p can be taken as two for most clays. So far we have analyzed only the behaviour in the triaxial plane. is the slope of the critical state line obtained in standard compression triaxial tests and 0 Lode's angle Elastic constants are assumed to depend on p' according to the laws and K 1 dEt = dp' 1 ep' + . but above expressions can be immediately generalized to all threedimensional conditions by assuming that M depends on Lode's angle 0 according to a suitable law.
two elastic constants.10 0. p'Wa) 100. they were found from the constant p' tests reported. The parameter a controlling dilatancy can be found from dilatancy plots. The slope M of the critical state line on the (p' . the slope of the critical state line M.\yExperimants Predicted .34 Constant p' test on Bangkok clay (After Balasubramanian and Chaudhry) ::(': 100.138 CONSTITUTIVE RELATIONS  PLASTICITY where The model presented so far concerns normally consolidated clays under virgin loading. 0. 300. and it is given by Finaly.predicted Figure 4. In order to assess model performance. :sl I : . their predictions will be compared against a full set of tests carried out by Balasubramanian and Chaudhry in 1978 for soft Bangkok clay.. The proposed model has five parameters.q) plane is found from drained.35 Consolidated undrained tests on Bangkok clay (After Balasubramanian and Chaudhry) . 0 Experiments . 1r: .e. K and e as described above.20 Figure 4. undrained or constant p' tests. T o determine them the following procedure may be followed: The elastic constants can be easily determined from unloadingreloading tests. which appears in the plastic modulus. Here. 400. constant Ho can be found as a function of A. i. the constant a characterizing dilatancy and Ho. 200.
Figure 4. where .36) Overconsolidated clays The model described in the previous section can be extended to describe the behaviour of overconsolidated clays. The mobilized stress function proposed in Pastor. finally. First of all. the model is applied to simulate the consolidated drained behaviour of Bangkok clay (Figure 4. we present the results obtained for Consolidated undrained tests (Figure 4.34.36. 4. In the above. 1977).35).36 Consolidated drained tests on Bangkok clay (After Balasubramanian and Chaudhry 1978) The results obtained with the proposed model are shown in Figure 4. we will introduce a function accounting for memory of past history. To this end.ADVANCED MODELS 139 Figure 4. which will consist of storing the past event of maximum intensity. Zienkiewicz and Chan (1990) is which will be used in the plastic modulus where f (7) has been given in the previous section and CMAX is the maximum value previously reached by the mobilized stress function.34 shows the constant p' tests together with the experimental results.35 and 4. and. Next. we have introduced a deviatoric strain hardening function g(<) (Wilde.
37 Behaviour of normally consolidated Weald clay (after Henkel 1956) .39 the performance with the above assumption for cyclic tests with constant stress amplitude carried out by Taylor and Bacchus (1969). Table 4. Figures 4.140 CONSTITUTIVE RELATIONS  PLASTICITY and Therefore. For cyclic loading it is possible to obtain quite satisfactory results for clays using simple elastic unloading and thus avoid the introduction of additional parameters. It should be noted that for first or virgin loading of clays above expressions reduce to those previously proposed for normally consolidated clays.37 and 4. Bangkok clay Weald clay Taylor and Bacchus Figure 4. together with the model predictions.38 show the behaviour of normally and heavily consolidated Weald clay reported by Henkel (1956). Finally.1 gives the parameters used in the simulations described above. It is important to note that in overconsolidated clays the peak value of the stress ratio may be higher than M. We show in Figure 4. two additional parameters y and 00 are needed to extend the range of application of the model to overconsolidated clays. then decreasing to reach it as a residual state.
3 A generalized plasticity model for sands Monotonic loading Following experimental results reported by Frossard in drained triaxial tests. Dilatancy is zero at the line . dilatancy can be approximated by a linear function of the stress ratio 77 similar to that used in the preceding section for normally consolidated clays.39 Behaviour of clay under twoway straincontrolled triaxial loading (data from Taylor and Bacchus 1969) 4.38 Behaviour of overconsolidated Weald clay (OCR=24) 10 Number of cycles 20 Figure 4.ADVANCED MODELS 141 r z Axial strain Figure 4.4. using M. instead of M as we will assume that flow rule is nonassociated.
142
CONSTITUTIVE RELATIONS

PLASTICITY
which coincides with the projection of the critical state line on the plane (plq). This line has also been referred to as the 'characteristic state line' (Habib and Luong, 1978) or the 'line of phase transformation' (Ishihara, Tatsuoka and Yasuda, 1975) and plays an important role in modelling sand behaviour as will be shown later. It has to be noted that this line is not the critical state line, which will be reached at residual conditions. Whether the critical state line existed or not has been a matter of discussion during past years, due to the difficulty of obtaining homogeneous specimens at failure after shear bands have developed. However, recent experiments carried out at Grenoble by Desrues have shown that inside the shear band a critical void ratio is reached. During a test, this line can be crossed a first time, with the specimen still far from the residual state. If shearing continues, the stress path will finally approach the critical state line. Therefore, the condition 77 = M, represents two different states at which dilatancy is zero, the 'characteristic state' and the critical state. can The direction of plastic flow n , ~ be determined in the triaxial space by similar procedures used in cohesive soils, giving
with
So far, the behaviour of granular and cohesive soils coincides. However, use of nonassociative flow rules is necessary for modelling of unstable behaviour within the hardening region, and the direction n should be specified as different from n , ~ . We do this by writing
with
ADVANCED MODELS
143
where
Again, both My and M, depend on Lode's angle in the manner suggested in Zienkiewicz and Pande (1977). It has to be remarked that both directions have been defined without reference to any yield or plastic potential surfaces, though, of course, these can be established u posteriori In fact, it is possible to integrate above directions to obtain both plastic potential and yield surfaces
where the size of both surfaces is characterized by the integration constants pi, andpk. Both surfaces were depicted in Figures 4.31(a) and (b) for a mediumloose sand, together with experimental data obtained from accoustic emission (Tanimoto and Tanaka, 1986). Similar yield surfaces were proposed by Nova (1982). To derive a suitable expression for the plastic modulus HL. it is necessary to take into account several well established experimental facts: (i) residual conditions take place at the critical state line
(ii) failure does not necessarily occur when this line is first crossed. (iii) The frictional nature of material response requires the establishment of a boundary separating impossible states from those which are permissible. A convenient law was introduced in Pastor and Zienkiewicz (1986) in the form
where
together with
144
CONSTITUTIVE RELATIONS

PLASTICITY
limit the possible states, and where
are of a similar form to the expressions proposed for clays. To illustrate the predictive capability of the proposed model, we will consider next several sets of experiments reported in the literature (Castro, 1969; Taylor, 1948 and Saada and Bianchini, 1989) and which cover the basic features of granular soil behaviour under monotonic loading. (i) Very loose sands exhibit liquefaction under undrained shearing, as has been described in the preceding chapter. Considering the qualitative results shown there, it is important to remember that the material densifies during the whole process, which is shown by a continuous increase in pore water pressure, suggesting that the soil is hardening. This seems to contradict the fact that a peak exists, and the material can be thought of as softening. However, in a frictional material, strength has to be analyzed in terms of mobilized stress ratios rather than deviatoric stress, and no peak is presented by this parameter. This behaviour can be considered unstable in the sense of Drucker (1956, 1959)
having thus
If such a feature is to be modelled with a positive plastic modulus, the associated plasticity has to be abandoned, choosing
Figures 4.40(a), (b) and (c) show, respectively, the stress paths, deviatoric stress vs. axial strain and pore pressures obtained by Castro, together with the model predictions, which agree well with them. (ii) At the other end of the density range, peaks exist in deviatoric stress during drained shear of very dense sands, this effect developing progressively as density is increased. The factor H, is introduced in the expression giving a plastic modulus to account for: crossing of the Characteristic State line (7 = M,) without immediately producing failure, reproduction of softening residual conditions taking place at the critical state line.
ADVANCED MODELS
(a)
145
p'
392 KPa

200 400 600 Mean e r l t c b v c confinmg pressure p' (kPa )
1
2
3
Shear stram e, (36)
4
5
6
7
x
9
Figure 4.40 Undrained behaviour of Banding sand (data from Castro 1969) Computed results shown by solid line. (a) p'q plot (b) Deviatoric stress vs. shear strain plot (c) Pore pressure vs. shear strain plot
146
CONSTITUTIVE RELATIONS  PLASTICITY
0
j
10 I5 Axial strain (Of;)
20
0
5 10 IS 20 A x d strain (",I)
0
5
10 15 Axial strain ("h)
10
0
5
10 15 Axial strain ('14)
20
Figure 4.41 1948)
Drained behaviour of dense and loose sand (experimental data from Taylor,
To illustrate the role of the plastic modulus in the transition from softening to hardening regimes, let us consider a drained triaxial test (Figure 4.41). During a first part of the path, both H I ,and H, are positive and decrease in a monotonous way. At 17 = M,, i.e., when crossing the characteristic state line, H,, becomes zero, while H, is still positive. If the process continues, a moment arrives at rl, where
with
If the test is run under displacement control, the deviatoric stress does not change for an infinitesimal variation of the strain dp'
= dq = 0
ADVANCED MODELS
147
Meanwhile, H, has decreased, and consequently, the plastic modulus becomes negative. The soil has entered the softening regime, and from this moment the deviatoric stress will present a descending branch. The deviatoric strain hardening function H , will vanish as deformation progresses, reaching a final asymptotic value of zero at 17 = M,, this time at the critical state line. During the softening process,
and
It can be seen that there is no need on this occasion for nonassociativeness to ensure the existence of peaks as H is negative, and, in fact, very dense sands may exhibit the limiting associative behaviour with
The ratio M f / M g seems to be dependent on relative density, and in Pastor, Zienkiewicz and Leung (1985), a suitable relation was proposed as
where D, is the relative density. Figures. 4.41 and 4.42(a) and (b) show model predictions for dense and loose sand response in drained conditions (Taylor, 1948; Saada and Bianchini, 1989). Care should be taken when analyzing the results of tests in general, and in the case of dense sands in particular, as failure localizes along narrow zones referred to as shear bands. From the moment of their inception, the specimen is no longer homogeneous, and the experimental results correspond to a boundary value problem rather than a homogeneous body. However, we have to stress following facts: Even if the specimen is not homogeneous, softening must exist for the sample to exhibit a peak.
0
The overall response is governed by the ratio between the width of the shear band and the length of specimen. This effect is similar to what can be observed in numerical computations and which has been referred to as meshdependence. Experimental evidence seems to indicate the existence of a residual critical state. To obtain it, it is simpler to use loose rather than very dense specimens.
0
148
CONSTITUTIVE RELATIONS  PLASTICITY
I
I
I
"
r)
3 2 Axial strain (%)
1
2 50
Hollow cylinder Hostun sand conf.press.207kPA

&
i n i n
Test HH3 Hollow cylinder Hostun sand conf.press.207kPA
u
 Experiment
o
2 z 100

 Experiment
0
Prediction
'ti u
.m
150
Prediction
Axial strain (%)
a
8
0 0 0 "
Axial strain (%) (a)
Figure 4.42
Axial strain (%)
\
(b)
Drained behaviour of Hostun sand (Experiments from Saada and Bianchini 1989). (a) Compression test (b) Extension test
(iii) Undrained shearing of mediumloose to dense sands shows the intermediate characteristics to be discussed. Once the characteristic state line is reached, an upturn in the stress path is produced as the soil changes from contractive to dilative behaviour. If the material is isotropic, determination of the CSL position can easily be performed from a point at which the undrained stress path has a vertical tangent in (p'  q) space, as then,
and
dp' = 0
Figures 4.40(a), (b) and (c) show how relative density influences the undrained behaviour of sand, together with predictions of the proposed model.
ADVANCED MODELS 149 At this point. Threedimensional behaviour So far we have considered the triaxial response of soil (compression and extension). p' and q). the increment of plastic strain is given by . the soil response can be generalized to any path out of the triaxial plane. However. proposed relations have been made dependent not only on I. as no surfaces are involved and consistency conditions do not have to be fulfilled. Hence. but also on the third invariant or equivalently on Lode's angle 6. (iii) it is computationally efficient in FE codes. 1988) d a :n then. q and 19by a* and the Cartesian stress tensor by a . invariance of the contracted product u : n results in (Chan. (ii) it is very simple.226) n. as the stress point does not have to be brought back to the yield surface and tangent moduli are easily established. alternatively. a model has been produced such that: (i) it reproduces the most salient features of sand under monotonic shearing. and J2 (or. Zienkiewicz and Pastor. substituting in the above du* d a * =: da du results in da* n=n*du and similarly * du* = d a * : n* (4. da Finally. Denoting triaxial stress parameters p'. =n.
In this case. as illustrated in the following examples. However. in addition to this. Figure 4. and. a pure rotation of principal stress directions will not induce any response from the material. allows any specified stress or strain path to be followed. the effect of increasing q prevails over rotation of principal stress axes. 1989). the principal stress axis will change. as the model is defined in terms of stress and strain invariants. Figure 4. In Figure 4. Zienkiewicz and Chan. 1989) it is shown how the proposed model can reproduce this special path. Zienkiewicz and Chan.43 shows the predicted behaviour forp' and 0 constant tests against results obtained in the hollow cylinder device (Pastor. r = 0) the shear stress is increased leaving the o 2 specimen to drain freely. The model was calibrated with conventional triaxial compression and extension tests only. The hollow cylinder device is able to perform tests with rotation of principal stress axes by combining axial and radial stresses with a variable shear stress r. from a triaxial state (al.43 Constant b tests on Reid sand (experiments from Saada and Bianchini 1989) . q will increase while p' is kept constant. If.44 (Pastor. = 03.150 CONSTITUTIVE RELATIONS  PLASTICITY and the constitutive tensor CeP in Cartesian coordinates can be written as This of course.
.. Generalization of the proposed model for these phenomena can be made by considering several mechanisms. which is not always very accurate... Zienkiewicz and Chan ( 1 990) using models able to introduce strain and loadinduced anisotropy (Pastor..44 Shearing of a sand with rotation of principal stress axes (experimentsfrom Saada and Bianchini..ADVANCED MODELS 151 500  PREDICTED EXPERIMENTAL . or .45 depicts results obtained by Ishihara and Okada (1982) on undrained shearing of loose sands under reversal of stress. 1989) Experiments with pure rotation paths show a plastic volumetric strain under drained conditions and pore pressure generation under undrained loading. its importance increasing with it. from which unloading takes place. as under undrained conditions the volume is constant. it can be observed from experiments that higher pore pressures than those correspondent to elastic unloading appear. 1991 ). 1997). In fact. Instead of unloading along a vertical line the stress path turns towards the origin which indicates higher pore pressures than isotropic elastic. the response is characterized as isotropic and elastic in most classical plasticity models. and the new distribution of contacts makes the specimen anisotropic (Bahda. the volumetric elastic strain should be also zero and. Figure 4. as it was proposed in Pastor. Unloading and cyclic loading A first step towards modelling cyclic behaviour of sands is to understand what happens when unloading and reloading. p' should not change (a variation of p' causes a change in volumetric elastic strain). therefore. Two possible explanations are possible: (i) Either the material structure has changed after having crossed the characteristic state line.. This phenomenon depends on the stress ratio 7. Figure 4. Isotropic elastic unloading is characterized by zero volumetric plastic strain and. Concerning the former..
~ thus be provided by . we note that irreversible strains are of a contractive (densifying) nature. If we assume that plastic strains appear upon unloading.152 CONSTITUTIVE RELATIONS EXPERIMENTAL PREDICTED  PLASTICITY a 0.0 o 1.0 2. To determine the direction of plastic flow produced upon unloading.60 1 B I Mean confining pressure Mean confinin pressure P' ( W c m )  n w 1. a simple expression for the plastic modulus fulfilling these requirements was proposed in Pastor. 1982) (ii) Plastic deformations develop during unloading.0 AXIAL STRAIN (96) AXIAL STRAIN (%) Figure 4. can The direction n . Zienkiewicz and Chan (1990) 1 =H (5) 1 1 ' for >I and extends the range of the model so far proposed hierarchically.45 Undrained behaviour of loose sand under reversal of stress (experiments after lshihara and Okada. and that they are of a contractive nature.
46 shows the results obtained by Castro (1969) in his pioneering work. In the case of very loose sands. to take into account the history of past events.ADVANCED MODELS 153 where and Concerning reloading. Failure here is progressive since the stress path approaches the characteristic state line by its shift caused by pore pressure build up. this mechanism causes progressive pore pressure buildup leading to failure. Figure 4. and which are responsible of catastrophic failure of structures subjected to earthquakes. Both phenomena are largely caused by the overall tendency of medium and loose sands to densify when subjected to drained cyclic shearing. we will modify the plastic modulus introducing a discrete memory factor HDMas where C was defined above as and y is a new material constant. and strains produced during the next loading branch are of higher amplitude. as it was done in the previous section for clays.47 shows both the experimental results obtained by Tatsuoka on Fuji river sand.45 shows a prediction of this model extension for the experimental results of Ishihara and Okada. Figure 4. It is possible now to model cyclic phenomena as liquefaction and cyclic mobility which appear in loose and medium sands under cyclic loading. plastic modulus is given by Figure 4. liquefaction takes place following a series of cycles in which the stress path migrates towards lower confining pressures. If the load is applied fast enough or the permeability is relatively small. . Deformations during unloading cause the stress path to turn towards the origin. Denser sands do not exhibit liquefaction but cyclic mobility. Finally. it is necessary. Here.
5 u I I .47 Cyclic mobility of loose Niigata sand (predictions and experimental data from Tatsuoka.0 Axial stram (%) 0 Predicted 5 5  M 0 " Y. 2 2. u I p'1(kg I cm) .46 Liquefaction of loose banding sand under cyclic loading (predictions and experimental data from Castro.5 ij ? I 10.0 (c) 1 Experimental (4 OE$l Axial strain (%) 0 Exper~mental 10. Prcdicted (b) .5 Figure 4.154 Experimental (a) 2 CONSTITUTIVE RELATIONS Experimental  PLASTICITY (bi 1 0 pt (kg I cm) Predicted 5 Axial stram (%) Predicted 10 2 3 p' (kg 1 cm) 4 0 5 Axial stram (96) 10 Figure 4. 1969) (a) . p' (kg 1 cm) I 2 2. 5 0 g m o m Y. 1972) .
B.4. and the plastic modulus HL depend on the relative position of the stress C with respect to the point at which the load was reversed. To thisend. The direction of plastic flow is obtained again by defining a suitable dilatancy at C.4.. To obtain the values of H L .4. two elements are introduced: (i) a surface defining the maximum level of stress reached. d g which is interpolated from an initial value dgoto The initial value of the dilatancy at the reversal point d8 is given by .2 Model parameters used in simulations Fig.42 Fig.45 Fig.4. Xu and Peraire (1993) by improving the way in which the history ofpast events is taken into account. suitable interpolation rules are used. n is interpolated from n to n using a linear law.41(loose) Fig. defined on the same mobilized stress surface as B. Zienkiewicz.ADVANCED MODELS 155 Table 4.2 above gives the model parameters used in the preceding simulations. Zienkiewicz and Chan (1987) and Pastor.4.43 and 4.. and (ii) the point at which last reversal took place.4. n and n.44 Table 4. Directions n and n. In particular. The model can be further elaborated as shown in Pastor.46 Fig. D.47 Figs. and an image point.
unloading may be considered as a new loading process. The plastic modulus is interpolated between an initial value HUoand its final value at the image point on the mobilized stress surface H D . several improvements and modifications have been introduced. The initial value can be assumed to be infinite to decrease a possible accumulation of plastic strain under very low amplitude cycles. and it can be observed how the volumetric and deviatoric plastic strain produced decreases with the number of cycles. but of the past history as well.48 Interpolation rule where the constant C. For instance. the influence of sand densification under cyclic loading can be taken into account by introducing into the plastic modulus a factor H. < 1) varies with the density. In fact. Concerning the rule to obtain the image stress point D. It is important to remark that direction of plastic flow and unit vector n will not be functions of the stress state only. This interpolation law provides a smooth transition between unloading to reloading.156 CONSTITUTIVE RELATIONS  PLASTICITY Figure 4. C and D and which is 1 when C and D coincide. Finally.(O < C. . being close to zero for mediumloose sands. where f is an interpolation function depending on the relative position of the points B.49 shows the densification of a loose sand under cyclic loading. as depicted in Figure 4.48. It should be mentioned here that since the simple models we have described here were proposed. there are several alternative possibilities./ Figure 4. it can be obtained as the intersection of the straight line joining the reversal and the stress point with the mobilized stress surface.
4 Anisotropy Introductory remarks Anisotropy in geomaterials is caused either by the arrangement of particles such as occur in natural deposits in which the grains may have their major axes on the bedding planes. 4.ADVANCED MODELS 157 Stress ratio .4. 1997.0 2. in a consistent way.0 Figure 4. or by the spatial distribution of contacts and contact forces. it is found that the strength is higher when tested along the deposition direction. Bahda. 1994). the behaviour of sand under different conditions of confining pressure and relative density (Bahda. Pastor and Saitta. where research recently finished has succeeded to include state parameters describing. This effect can introduce important errors if not taken into account.0 0 Shear strain (%) 1 .49 Densification of a loose sand (predicted) particularly at CERMES (Paris) (Saitta. if the number of cycles to liquefaction is determined using a standard triaxial testing machine to evaluate the liquefaction potential of a natural sand . In the first case. 1997). For instance.1 .
a simple way to introduce anisotropic surfaces is to define a modified second invariant of deviatoric stress tensor where Aijkl is a fourthorder tensor characterizing material anisotropy. 1987. in any isotropic plasticity model (such as the CamClay. as proposed by Anandarajah and Dafalias (1986). Surfaces can be allowed to expand following isotropic hardening rules and to translate and rotate under anisotropic hardening laws. Hashiguchi. orientation and shape of isotropic yield. This effect can be introduced also by directly formulating the surfaces on the stress space (Banerjee and Youssif. 1993). Combination of kinematic and anisotropic hardening laws proposed by Mroz (1967). Multimechanism theories can also describe anisotropic behaviour. 1988 and di Prisco. Surfaces were made dependent on in addition to invariants of stress and strain tensors. All theories mentioned above are able to introduce the anisotropic response of geomaterials even when there is a single mechanism of deformation. loading or plastic potential surfaces. provided that they are not formulated in terms of the three stress invariants only. Ghaboussi and Momen. in such a way that those changes were dependent on tensors such as stress or strain and not only on their invariants. 1986) or by deriving them from modified anisotropic flow rules. Basically. Liang and Shaw.158 CONSTITUTIVE RELATIONS  PLASTICITY deposit. Baker and Desai (1984) suggested to include the effect of stress induced anisotropy via joint stress and stress invariants. Cambou and Lanier. the strength will be underestimated. Lanier and Nova. If J i is substituted by J. one finds yield surfaces which have been distorded and rotated. This method was initially suggested by Hill (1950) and extended by Nova and Sacchi (1982) and Nova (1986) to soils and soft rocks. Several theories have been proposed within the framework of plasticity to describe both initial and stressinduced anisotropy. the value obtained will be greater. for instance). have provided a suitable way to model anisotropic behaviour of soils (Mroz. 1982. 1980. If only initial or fabric anisotropy is to be considered. therefore. Multilaminate models as introduced by Pande and Sharma (1983) consider that deformation is caused by dilation and slip taking . Norris and Zienkiewicz. 1979. Hirai. Initial or fabric anisotropy could be reproduced as well by introducing initial movements and distortions on the surfaces. and. This approach is based on the representation theorem of scalar functions depending upon two symmetric secondorder tensors u and P in this case). anisotropy has been approached most of the times by changing the position. 1991.
Alternatively. Proposed approach It has been mentioned above that material structure or fabric has to be incorporated in the constitutive equations to account for both initial and induced anisotrpy. which can be incorporated into the constitutive equations. The structure tensor will have the form . it should be mentioned that material fabric plays a paramount role on geomaterials anisotropic response and it is in turn modified by the deformation process. The fabric may be approximated by a secondorder tensor. An interesting way has been recently proposed by Pietruszczak and Krucinski (1989). 1974). 1982 and Matsuoka. Pastor. Alternatively.ADVANCED MODELS 159 place at all possible contact planes within the material. including anisotropy effects. the class of symmetry will be defined by the set of operators Q which fulfill For instance. If Q is a rotation or reflection tensor. Lassoudiere and Meimon.o. Hujeux. The overall response is obtained by a process of numerical integration extended to sampling planes. if attention is focused only on planes normal to XY. and the second accounts for deviations of isotropy. it will be assumed that fabric can be described by a secondorder structure tensor A. transversely isotropic materials will be described by A invariant under Q given by Q = (a 0 0 cos0 sin0 s i n Oo :c ) where it has been assumed that the plane of isotropy is XY. YZ and ZX. 1983. which are made dependent on fabric tensor. This has been proposed by Darve and Labanieh (1982) and applied to complex stress paths. and can then be generalized to more general stress conditions. Zienkiewicz and Chan. the behaviour of the material can be assumed to be caused by superposition of responses to variations in 4.and 4 . 1990). Of all possible active planes only a reduced number of sampling planes is considered. which will determine its type of symmetry. First one corresponds to an isotropic hardening mechanism. and their responses are grouped together. Here. Finally. and consists on adding two components to obtain the increment of plastic strain. one finally arrives at a threemechanism model (Aubry. Both multilaminate and multimechanism models of the type described above can produce plastic strain under pure rotation of principal stress axes (Pande and Sharma.
(i) 6 @ 6 (ii) 6 @ A A t 3 6 (iii) 6 t3 A2 A' (iv) A t3 A (v) ABA' @ 6 A'BA (vi) A' @ A' where compact notation has been used. A . therefore.160 CONSTITUTIVE RELATIONS  PLASTICITY and it can be easily checked that the invariance relation A = Q~ . If the initial 'structure' of the material is described by A'. the tensor B referred to principal axes is given by where rows and columns include components of the tensor with their components ordered as and. If a transverse isotropic material is considered. B can be expressed as a combination of terms listed below. . Q is verified. according to where dA will depend on the plastic strain Now. Nova and Sacchi (1982) and Nova (1986). component 1133 is located at first row and third column. A will vary along the loading process. Following Cowin (1985). the structure tensor can be used to define a fourthorder anisotropy tensor B from which a modified second invariant Ji can be derived as suggested by Hill (1950).
and B'" on double tensorial products of S. However. and indeed some kinematic hardening models in which a back stress is introduced can be considered as particular cases of the theory outlined above. the advantage of being simpler to develop. only JS has been extended to account for anisotropy. can be introduced in a similar manner. It can be seen that the resulting anisotropy tensor depends only on five constants. can be generalized to anisotropic situations by substituting them by modified forms given above. geomaterial behaviour is also dependent on first and third invariants. Therefore. The first anisotropy tensor B' would be dependent on 6 and A. An interesting particular case is obtained when A is taken as Then. the result of substituting stress invariants by their modified forms can be viewed as introducing a rotation and a distortion of yield and plastic potential surfaces. the extended set of invariants is given by Finally. the constitutive law can be seen to be dependent on joint stressstrain invariants. New invariants fl and J. constitutive laws derived for isotropic materials in terms of I.ADVANCED MODELS 161 In the above. etc. A @ S 8 6 . If the three tensors introduced to produce the modified invariants are defined as: . and that the form proposed by Nova and Sacchi (1982) is a particular case of the above expression in which BI and B4 have been made one and zero respectively.'. as proposed by Baker and Desai (1984). and anisotropy should also be reflected on them. however. They present. by defining B' and B"' which are tensors of orders two and six. As mentioned above. So far.J: and J . A and such as S @ S @ S.
can be obtained. ~ . it will be a rotation of directions n and n . 1991). a modified form of the classical set of invariants (j'. To obtain the second invariant J:. In the case of generalized plasticity.162 CONSTITUTIVE RELATIONS  PLASTICITY it can be checked that this choice corresponds to a pure rotation of the yield and plastic potential surfaces. A possible solution consists of defining a new deviatoric stress t as from which and The modified invariants . g) Alternatively. Care should be taken when defining a new Lode's angle as the trace of the new modified deviatoric stress is not zero. Zienkiewicz. where surfaces are not introduced. Xu and Peraire (1993) introduced a rotation characterized by a new direction of the plastic potential and yield surface axes given by a unit tensor a given by: The first modified invariant was obtained as which is proportional to the projection of stress tensor along a. The model described in Pastor (1991) and Pastor. a modified deviatoric stress was defined as q. " (Pastor.
no such dependence was found for M. Therefore.ADVANCED MODELS 163 and e can now be introduced into the generalized plasticity model described in the preceding section. The tests performed by Yamada and Ishihara (1979) consisted of proportional. It will be shown next how to obtain a simple generalized plasticity model for anisotropic sands. Leung and Pastor. However. Miura and Toki (1984) analyzed the experiments of Yamada and Ishihara (1979) and concluded that soil anisotropy did not greatly affect the dilatancy behaviour of sand. If a specimen of such material is brought to failure.". is affected by soil anisotropy. Zienkiewicz and Chan. The samples exhibited a strong anisotropy as grains were arranged such that their long axes were horizontal. . different behaviour was observed along paths such as ZC and YC which have the same value of Lode's angle. A detailed description of both the testing procedures and the results obtained is given there. to describe the anisotropic behaviour of sand. J i and Jj or p'. confirming the linear relation between dilatancy and stress ratio proposed above for isotropic soils. This initial or inherent anistropy may be modified by subsequent strains developed as the material is loaded. It has been shown above how the material response can be described by providing suitable expressions for tensors n. depends only on Lode's angle 0 and q' is the stress ratio. two experimental facts will be recalled: If experimental data obtained on granular soils with initial anisotropy such as is given in Yamada and Ishihara (1979) are analyzed. 1985.~. grains will be reorganized as deformation increases. it can be found that the zero volumetric incremental strain surface may be described by the same simple relation proposed in the preceding section for isotropic sand where M. and the following relationship was proposed . 0 A similar analysis may be carried out to study how soil dilatancy. radial paths performed at constant p' and 0. and plastic modulus H L j u . simple models have been derived for isotropic materials (Zienkiewicz. 1985 and 1990) in terms of invariants I .. changing the initial structure. It was found that the parameter cu depended on 0. Pastor. Following this approach. q and 0. A generalized plasticity model for the anisotropic behaviour of sand Sand deposits exhibit anisotropic response caused by the alignment of sand grains on horizontal planes. which will be assumed to be independent of anisotropy. Concerning the plastic flow rule.
q. it should reflect material symmetries. while isotropic materials will be characterized by As the material is loaded. Therefore. Those depicted in Figure 4. General expressions for this functions have been suggested by Pietruszczak and Krucinski (1989). The modified invariants @'. 6 ) will be used. are taken as zero and one respectively. Cubic samples of sand were tested on a true triaxial apparatus along different paths. Here we will assume that dA may be expressed as where CClo and CClI two material parameters and E is the accumulated deviatoric are strain. deformation will produce rotation of grains and rearrangement of microstructure. It is interesting to notice how the anisotropic structure induced by the deposition process resulted in a higher overall stiffness for the specimen tested in com .. in the definitions of direction n and plastic modulus HLI('directly. Therefore. The tests were run on the Fuji River.50 correspond to compression and extension along the vertical. This resulted on a highly anisotropic structure which was modified by subsequent loading. and the modified invariants will be functions of both the stress and mixed stressstrain invariants. however. Figure 4. If CLlo and C. Microstructure tensor A was assumed to have an initial value A' corresponding to initial fabric. and no initial anisotropy exists. So far. A will coincide with EP. and dA will be a function of the increment of strain. with specimens constructed by pluviation of sand through water to simulate the natural deposition process.50 reproduces both the results obtained by Yamada and Ishihara (1979) and the model predictions (Pastor. and expressions giving MJH) and dilatancy holds for both isotropic and anisotropic materials. 1991). A will change. To show the performance of the proposed model. it has not been necessary to introduce the modified invariants described above..164 CONSTITUTIVE R E L A T I O N S PLASTICITY where subindex 'c' refers to values obtained for 6' = 30". Naturally deposited sands exhibiting transverse isotropy will have with axis X I coinciding with direction of deposition.
until the full liquefaction takes place or the shear stress reaches the phase transformation line. however.5. ... Predicted 4 (kglcm) Experiment Figure 4.1 MODIFIED DENSZFICA TZON MODEL Densification model for cyclic mobility Introduction The densification model is firstly introduced with the classical MohrCoulomb criteria for shear strength (Zienkiewicz rt ul.50 Anisotropic behaviour of Fuji River sand in triaxial compression and extension pression along the vertical. It cannot. That model is suited to simulate undrained tests in which the excess pore pressure is built up due to the constant cyclic load.MODIFIED DENSIFICATION M O D E L 165 3 . This can happen if the soil layer consists of very loose sand but the dense soil layer resists and transfers the .5 4. 1978). simulate the cyclic mobility behaviour after the stress point reaches the phase transformation line. The most important application is the soil layer problem. and how the situation was reversed when loaded in extension. the wave does not transfer upward and no response acceleration is observed. 4. After the stress reaches the failure line of the MohrCoulomb densification model in the middle of the soil layer.
Then (4. In order to simulate the cyclic mobility. C3 takes a constant value as follows. r is independent of the stress ratio until the reversal stress ratio is over the phase transfer line qR. parameter k defined by equation (4.Thereafter. Now the damage parameter ti has a linear relationship to the dilatancy.162) are generalized as introduced by Shiomi and Tsukuni (1998). Equations (4. K is a damage quantity that causes irreversible dilatancy. equation (4. The dependence of ti is included in the function h. Most experimental data refers to 0 and very few data are available for the relation between the dilatancy and rl.161) and (4.269) becomes very large when liquefaction takes place and causes numerical difficulty. Modification of the densification model. This is because the damage. The damage quantity is dependent on the stress ratio and the accumulated shear train <.P(~)IXK) . it should theoretically be 77.~ ) .3 Definition of function Before cyclic mobility r After cyclic mobility where a.166 CONSTITUTIVE RELATIONS  PLASTICITY upward wave propagation. K ) . The modification of the densification model for this kind of phenomena is introduced here.K ) . the above function can be modified further to avoid numerical difficulty.161) can be rewritten by the following functions if K ( K ) represents the limitation of the autogeneous volumetric strain at a large damage parameter k. But if the dilatancy depends on the stress ratio. Table 4. Therefore. The incremental formula can be as follows. The empirical formulation can be used for this purpose. In order to simulate a triaxial undrained test. h(71. where 71 is the stress ratio against the current mean stress and H is the stress ratio against the initial mean stress.162) can also be rewritten using the following functions. that is h ( 7 . b are constants. the function K ( K )is defined as and h ( q ) is replaced by h(r1. defined as follows 3 4q3 K ) =Y. The definition and simulation of a triaxial test is presented. Cyclic mobility can be included in the function h(r1).
.. table as { ( a l711.. B.. '11 lrll where 1 7 ~ tan& and q. the model parameters of the densification model.( a Zq2). = tan$. . Table 4. as shown in Table 4.1 0.. Before cyclic mobility (n= 1 ) After cyclic mobility (n=2) I .C~L(O) C ~ L ( O ) +I + B l K 1. (MPa) Model of stress path for dilation behaviour Definition of a ( q ) Function a ( q ) Before cyclic mobility After cyclic mobility For loading For unloading Defined by table (e. & is the phase transfer line and $f is the failure line.51 and is a function of the stress ratio q(= ~ / a : .g.( K K O ) Z  Loading Unloading cru I CZL C2Lt1+BI(~~U)  .1 q'? lrll 1 % 1 1 .3 0..6 Figure 4. ) ) . .5 (a) Definition of P ( K ) (a) for K < K.. 0. n = 2. y. 'n' is the index are for the strain region.. and n = 3 when K > k. and (b) for K 2 K.7) lrll 0. .51 Table 4.0 Czc + 1 +1B.. G The different function is used before and after the liquefaction as show in Table 4.5 0.5 a ( 7 ) is is the differential of the stress path shown in Figure 4.MODIFIED DENSIFICATION MODEL 167 where A.0 0. n = 1 before cyclic mobility.4 0. (a..4 o:.. after cyclic mobility. . )It is defined as an empirical function and described in the . = P(K) depends on the damage parameter.. Table 4.4.2 0.
An example of P ( K ) is shown in Figure 4.5) Table 4.53 shows the stress path of the experimental and numerical simulation. Strains become large after the cyclic mobility starts and loses . Tables 4.6 to 4. Figure 4.8 show the parameters of the model. The function P ( K ) reaches a certain value when the damage parameter ti becomes large. Both results show the cyclic mobility behaviour.168 CONSTITUTIVE RELATIONS  PLASTICITY Figure 4.52 Relationship between the damage parameter K and P(K)(CZI. The coefficients C7L and C2L' are coefficients for the loading process and the unloading process respectively. Figure 4.= 0. There are many coefficients but these coefficients can be determined from the undrained cyclic test data.52.54 shows the stressstrain curve.5 (b) Contu' for k > kc After cyclic mobility (n=3) Loading Unloading Here KO is the damage parameter when cyclic mobility takes place. Simulation of a triaxial test The following simulation of a triaxial test shows the capability of the modified densification model.
0.147) Calculation 0.02 0.50 90Mpa Table 4.0 CZL 0. yl AI B1 4.015 0. 25.0 x lo' 2.0 deg.016 A3 8.50 0. Table 4.01 5 ( = 0.06 0.004 Czo 4.147) g2 e 2 z: 2 rc m Vi 0.MODIFIED DENSIFICATION MODEL 169 stiffness at low shear stress.53 Stress path of simulation . This is well represented.7 Function 471) Table 4.05 u b G 2.55.8 Function C4L Experiment (q = 0.(MPa) 0.0 B3 100.04 0. Furthermore the enlargement of the strain after cyclic mobility takes place also shows good agreement.OMPa 41.60 0.00 0. as shown Figure 4.8 deg.0 n.0 y3 0.08 Effective mean stress o'.0 y2 A? B2 25.10 Figure 4.6 Parameters of modified densification model c 4 4 0.0 8.000 m 0.0 x lo7 1.
0 15 4 2 0 Shear strain y(%) 2 4 Figure 4.147) Calculation (q=0.r_ z P. n 0.54 Stressstrain curve of a triaxial test (experiment and simulation) 10 Number of cycles N 100 Figure 4.147) 0.55 Enlargement of shear strain due to accumulation of damage (experiment and simulation) .015 2 .170 CONSTITUTIVE RELATIONS  PLASTICITY Experiment (q=O.000 2 V) z 0.
GT9. 112. and Chaudhry. Mech. Z. 21. Studer. and Herrmann. N. Am. Numer. J. Ph. 21 1214. J. M. J. 12. M . Zienkiewicz. NO. Baiant. D . Models in Geomechanics.) (1990) Geonirrteriuls. Eng. and Lanier J. 4.35. (1978) Deformation and strength characteristics of soft Bangkok clay. and Desai. J. Bahda. Dafalias. An Introduction to Criticcrl State Soil Mechanics.for Grunulcw Noncohesive Soils. Comrnrm. R. Burland. 4. Bardet. C. Lassoudiere. (1984) Induced anisotropy during plastic straining. A. R. J. 103. Cambou. (1969) Liquefirction of'scmds. Anal. 12921 3 18. ENPCCERMES Paris. (1978) The Mechanics qfSoi1. R. 343. M~rth. 577580. R. J. Y. C. H. A. Atkinson.. Chan. Bianchini (Eds). R. Phys. Anal. G .. F. (Eds) Balkema. 0. Bahda. Num. N. Saada and G . Elsevier Applied Science. Series no. . Civ. Eng. (1986) Bounding surface plasticity 111: Application to anisotropic cohesive soils J. S. Krizek. London. L. 9 18. H.s. Hujeux. EM 102. H. (1773) MPm. Zienkiewicz (Eds). L. and Bransby. E. Materials. F. ASCE. in Soil MechanicsTransient and Cyclic L O N ~G . Meth. and Popov. N. Proc. Geotech. D. 966987.REFERENCES 171 REFERENCES Anandarajah. Num. Y. (1982) A double memory model with multiple mechanisms for cyclic soil behaviour. Pande and J. Darve. En& Div. (1985) The relationship between the elasticity tensor and the fabric tensor. J. Constitutive Equations and Modelling. F . (1982) Bounding Surface Formulation of Soil Plasticity. Computers and Geotechnics 6. 7 . F. (1986) A plasticity model for the mechanical behaviour of anisotropically consolidated clay. (1975) A model of nonlinearly hardening materials for complex loadings. (1977) Densification and hysteresis of sand under cyclic shear. (1988) Induced anisotropy in cohesionless soil: experiments and modelling.. McGraw Hill. Dungar. Castro. Mech. and Saitta. P. 10. F. Z. Baiant. Int. 173192. J. A. A. Geomechs.D. 443449. (1986) Bounding surface plasticity. Pietruszczak. Int. and Labanieh. Methods Geomech. and Meimon. Dafalias. F. C. J. and Dafalias. Pastor. (1997) A double hardening model based on generalized plasticity and state parameters for cyclic loading of sands. Atkinson. 112. P. P. Anal. Geotech. Methods Geon~ech. Y. Gkotechnique 15. Num. Int. GPotecliniqlre. 2913 1 1. Harvard Univ. Pande and S. (1989) Prediction of deformations of Hostun and Reid Bedford sands with a simple bounding surface plasticity model. Y. K. 521541.. and Yousif. 243275. and Richardson. (Ed. Thesis. Coulomb. (1988) Transformation of incremental plasticity relation from defining space to general cartesian stress space.. 8 1. (1997) Etude du comportement du sable au triaxial. J. 22538. Eng. I: Mathematical foundation and hypoplasticity. et Cuellar. 253282. 6. Num. and Krizek. 167185. S. C. F. Pande and 0. A. Synzp. (1985) Elasticity and normality in soil: experimental examinations. Div. 10. P Ch. Soc. 104. Appl. Int. Eng. B. 8. B. A. Acta Mech. and Silver. J. (1976) Endochronic constitutive law and liquefaction of sand. Dafalias. C. P. V. Darve. (1982) Incremental constitutive law for sands and clays: simulation of monotonic and cyclic tests. 1371 47. and Pastor. F.. S. B. Balkema. Cowin. Balasubramanian. Baker. Wiley. Rotterdam. G . A. S. PhD Thesis. ASCE. C. in Constitutive Equations . 11531 167. C. (1965) Correspondence on 'The yielding and dilatation of clay'. F. N . P. Y. R. Harvard Soil Mech. J. P. Mech. A. L. J. M. ASCE. Banerjee. 131148. ASCE.. Meth. Aubry. G .
Geotech. Kaliakin. 83124. pp. W. M. 14. 33. Hirai.. 641646. R. D. E. N. 4979. D . K. Levy.) in Modern Approaches to Plasticity.). Eng. Amicale des lngenieurs Anciens Eleves de I'Ecole Nationale des Ponts et Chaussees (Palaiseau. W. GPotechnique. M. Mech. 101106. 21 3224. Kolymbas. 2829 Sept. and Ueno. Engng. N. Liang. (1993) A mixed isotropic kinematic hardening constitutive law for sand in D. D. and Rojas. Nova. D. and Yasuda. C. Henkel. Clarendon Press. Computers cmd Geotechnics. R. C. (1977) Elastoplastic Constitutive laws of granular materials. J. (1871) J. D. Kolymbas (Ed. Ass. Kolymbas. Imamura. P. Gibson. 22. (1978) Sols pulvurulents sous chargement cyclique. Drucker.125. and Wu. K. Henkel. ASCE 122. K. E. D . Wiley. Y. R. Pande and 0. D. R. Int. Kolymbas (ed. N0. Tatsuoka. Balkema. Soil Mech. Huber. Proc. and Dafalias. and Henkel. (1989) Prediction of deformation behaviour of sands by the subloading surface model. (1960) The shear strength of saturated remoulded clay. (1983) Une equation d'ecoulement simple pour les materiaux granulaires. 11. 61. and Shaw. and Chambon. Krieg. T. Appl. J. Desrues. N. T. Merh. 91100. di Prisco. Soil Mecl~anicsTransientund Cyclic Loudk. 131148. Hashiguchi. ASCE. 117. (1991) Anisotropic Plasticity Model for Sands. Ghaboussi. (1950) The Mathemuticul Theory of Plusticity. Anal. L.sopismo technicxe. J. K. J. 2129. Appl. Found. J. Colorado. Drucker. J. pp. 369. J. Ishihara. (1982) Effects of large preshearing on cyclic behaviour of sand. 7382. Num.6. R. Trans. D . and Lanier. Nun?. Archive ofApplied Mechanics. J. S. A. S. L. M. Habib. J. D. (1952) Soil Mechanics and Plastic Analysis or Limit Design. A S M E . 313342 G . M. H. (1993) Introduction to Hypoplasticity. D. Hashiguchi. Quurt. Muth. Georechnique 6. 47. Pures et App. V. 13. pp. R. 9th. 157165. Murayama and A. D. (1991) An outline of hypoplasticity. Balkema. 338346. Saada and G . D . C. (Ed. Oxford.. Mech. 16. Elsevier. . and Prager. (1986) A class of incrementally nonlinear constitutive relations and applications to clays. (1989) Simplifications to the bounding surface model for cohesive soils. (1980) Constitutive equations of elastoplastic materials with elasticplastic transition. (1982) Modelling and analysis of cyclic behaviour of sands. Appl. J. Appl. 2. 1978). 109. and Momer. Materiuu. 81. K. Quurt. Boulder. Int. Schofield (Eds). Research Conference on Sheur Strmgth of' Cohesive Soils. in Modern Approuches to Plrsticity. (1975) A practical twosurface plasticity theory. Math. Soils und Foundations. F. Soils and Foundutions. J. M. 15. 10. Kolymbas. 3542. Bianchini (Eds).. Drucker. pp.. Zienkiewicz (Eds). (1904) C:~. Muth. (1993) A new rate type constitutive model for geomaterials: CloE.). (1956) On uniqueness in the theory of plasticity. (1957) Soil Mechanics and Workhardening Theories of Plasticity. 143151. F. Trans. Int. 1.172 CONSTITUTIVE RELATIONS  PLASTICITY Darve. A n d . Frossard. 533540.. Ishihara. Congr. JSSMFE. (1959) A definition of unstable inelastic material. 913933. pp. H.~ and Structures Sous Churgernent Cyclique. E42. P. and Luong. Flavigny. 22. C. Lemberg. (1987) An elastoplastic constitutive model for cyclic behaviour of sands. J. and Okada. in Constitutive Equntionsfbr Grunulur Noncohesive Soils. 266272. Mech. 2944. Mo~lernApprouches to Plusticity. H . 4366. Geomech. A S M E . and Ueno. Constitutive Equations of'Soils. Geomech. Drucker. Hashiguchi. (1975) Undrained deformation and liquefaction of sand under cyclic stress. F. Hill. 26. C. C. 309324. E. 1391 50. 503520. Metk. E. (1956) The effect of overconsolidation on the behaviour of clays during shear. Appl. S. Elsevier.
Anal. Nova. 0. R. 0 . R. Pande. Zienkiewicz (Eds) pp. Norris. 7. Zienkiewicz. Mech. (1991) Modelling of Anisotropic Soil Behaviour. (eds). Mroz Z. V. (1979) Application of an anisotropic hardening model in the analysis of elasticplastic deformation of soils. N . 415450 C. (1983) Multilaminate model for clays: a numerical evaluation of the influence of principal stress axes. 4357. L.REFERENCES 173 Martin. Middleton (Eds). and Toki. (1982) A model of the stressstrain relationship of orthotropic geological media. Matsuoka. Num. Int. Nova. 1. Saada and G. Geomech. 445458.. Jackson and Son Publ. N. pp. Computers and Geotechnics. Theor. . 203221. E. J. R. Mech. F. Parry. D. Pande and 0. C. G. H. Geomech. and Norris. Pastor. 397418. A. Mroz. Soils and Foundutions. G. Nova. Norris. 31. D. G. M. Proc. A. John Wiley and Sons. A. Appl. 927949. Z. 343. 134. Ch. (1978) An anisotropic hardening model for soils and its application to cyclic loading. R. Camp. (1986) An extended cam clay model for soft anisotropic rocks. Z. and Chan. R . Pastor. G. 18. Z. in Soil MechanicsTransient and Cyclic Loads. Meth. 451469. Con$ Numerical Models in Geomechanics. pp. C. M. (1938) IngenieurArchiv. 163175. 7786. (1982) A Constitutive Model for Soil under Monotonic and Cyclic Loading. C. GT 5. Num. J. and Geotech. Nova R . (1982) Elastoplastic and viscoplastic constitutive models for soils with applications to cyclic loading. C. Meth. Desai and R. Soils and Foundutions. 0 . 0 . 2. W. (1978) an experimental program to define yield function for sand. Eng. and Zienkiewicz 0. 'I 7~+21'J ' U ~ W V q am' SVCIJ. M. Gbotechnique. in Soil MechanicsTrunsient und Cyclic L o ~ d sG. 2. C. M. K. Z. and Sacchi G. Num. and Zienkiewicz.pp. Eng. C. Arch.. Nova. (1987) A Generalized Plasticity Continuous Loading model for Geomaterials. M. C . S. C. (1977) On the hardening of soils. Mater. Mech. John Wiley and Sons. H. H. C. A. 2nd. 31st March4 April. and Zienkiewicz. B. and Chan. 42338. Civ. Anal. (1975) Fundamentals of liquefaction under cyclic loading. G. R. (1985) Uniform formulation of constitutive equations for clay and sand. J. M. 11. 0 . V. S. Bianchini (Eds). pp. V. Solids. M. and Van Impe..ximation. Pastor. Proc. (1974) Stressstrain relationships of sands based on the mobilized plane. Anal. Int. Mroz. A. Zienkiewicz. (1960) Triaxial compression and extension tests on remoulded saturated clay. Z. Stos. Int. in Constitutive Equations for Granular NonCohesive Soils.. 535549. Melan. (1989) Generalized plasticity model for threedimensional sand behaviour. Gallaher (Eds). 3. H. in Numerical Methods in Engineering: Theory and Appro. Mech. 22. V. Soils and Foundations. (1979) A constitutive model for sand. W. No2. (1981) An anisotropic critical state model for soils subjected to cyclic loading. pp. 116. Pande. Am. 0. Martinus Nijhoff Publ. 3. A. Phys. N. 15. Mroz. and Wood. Meth. 4761. and Zienkiewicz. Int. J. 255278.C. and Wood. Mroz. Ghent (Belgium). Geonzech. 29. N. 29. D. 14. C. and Seed. G . (1984) Elastoplastic stressstrain relationship for loose sands with anisotropic fabric under threedimensional stress conditions. 10. S. N. R. 374. 1661 80.. 0. G. (1986) A generalized plasticity hierarchical model for sand under monotonic and cyclic loading. Geotechnique. Giotechnique. and Zienkiewicz. Norris. Balkema. H. Nova.C. Finn. Pande and J. Miura. H. Mroz.. Zienkiewicz. A. J. Pastor. Pande a n d 0. and Sharma. 6988. 6. (1967) On the description of anisotropic workhardening. 173208. Soc. Zicr~kicwiiz<Eds(. 9. 24. 131150. 7.
Poorooshasb. L. C. (eds. W. Holubec. Num. 9. Delft. 8. C. J. 7th. D. W. D. Wang. A. Int. Zienkiewicz. Y. (1968) On the generalized stressstrain behaviour of wet clay. pp. I. A. 21 1240. Geotnech.al. H. Holubec. Saitta. 13. Arch. Klasse. Y. Meth. Int. J. (1963) Yielding of clays in states wetter than critical.. Tresca.. 70. F. B. N. 799809. Paris. Elsevier. Geotech. 241252. B. GPotechnique. application a la liquefaction des sables et a la sollicitation d'expansion de cavite. Roscoe. (1977) Mathematical modelling of monotonic and cyclic undrained clay behaviour. I. 473. Leckie (eds. Math. Anal. Pastor. W. 4.. (eds) (1989). Int. P. Pastor. (1967) Yielding and flow of sand in triaxial compression (Parts I1 and III). H. Int. C. (1864) Cornptes Rendus Accrd. (1966) Yielding and flow of sand in triaxial compression (Parts I1 and HI). C. M. and Wroth. Rowe. I. and Wroth. (1989) Description of anisotropic response of clays using tensorial measure of structural disorder. de SaintVenant. 26.ASCE. (1993) Modelling of sand behaviour: cyclic loading. K. I.. 1. Mech. 116. (1969) Dynamic cyclic strain tests on a clay. (1958) On the yielding of soils. F. P. A.phys. 505510. A.. Shiomi. Taylor. 0 . 899916. Saada.). A. anisotropy and localization. 0. S. Cambridge University Press. Schofield. Geotech. J.. Num. J. and Tsukuni. M. N. Mech. pp. Prandtl L. N. J. and Bianchini. and Leung. in Modern Approaches to Plasticity D. D . Soils and Foundations. (1958) Critical State Soil Mechanics. 9831 001. math. J. Mech. 754. and Chan. Wiley. Int. K. H. Heyman and F. Geotech. and Esrig. 195216. K. Meth. skewsymmetric tensors and vectors. (1971) On isotropic functions of symmetric tensors. A. P. H. Thimus et. Proc.. Soc. 59. 151190. R. S c i .). (1870) Cot7lptes Rendus Acad Sci. von Mises. Balkema. H. Appl.for Granular NonCohesive Soils. 376397. and Found. Stos. A. Meth.174 CONSTITUTIVE RELATIONS  PLASTICITY Pastor. ENPCCERMES Paris. A. 29. Xu GuangDou and Peraire. pp. and Shen. P. 3. H. P. GPotechnique. Poorooshasb. A. 535610. in Engineering Plasticity. McGrawHill. 500527. (1990) Bounding surface hypoplasticity model for sand. (1962) The stressdilatancy relation for static equilibrium of an assembly of particles in contact. Ang. 6980. Num. 582. H. J. 2253. Roscoe. Kolymbas (Ed. Eng. H. J. K. R. C. A. (1990) Generalized Plasticity and the modelling of soil behaviour. 469492. C. Roy. Zienkiewicz. Mecl~. Prevost. 1st Int. Sangrey. N. M. Zienkiewicz. K. Mexico. Henkel. Pietruszczak. Smith. B. Schofield. and Sherbourne. Geomech. (1948) Fundammtuls of'Soil Mechunics. B. 401409. J. and Krucinski. A. Wilde. M. Paris. 237249. C. Poromechanics. 477498. G . Tanimoto. H. G . Z. Can. (1924) Proc. (1977) Twoinvariant dependent model of granular media. Sci. Taylor. J. (1930) Zeits. Muter. (1969) The effective stress response of a saturated clay soil to repeated loading. Dafalias. T . Eng. 14. Constitutive Equations. Cong. Anul. D. 10. and Bacchus. A269. J. Eng. 266. 8. J. . Mech. J. K. and Tanaka. Cot$ on Soil Mech. S. Roscoe. (1986) Yielding of soils as determined from acoustic emission. pp. 179190. 6. (1998) Applicntion of Root's dynamic equation to seismic liquification problem. 9. Proc. Schofield.) Balkema. I1 Nonassociative models for sands. (1985) Simple model for transient soil loading in earthquake analysis. S. N. 43. 0 . and Burland. and Thurairajah.. Anal. Can. Reuss. Can. P h D Thesis. A. and Sherbourne. pp. (1913) Gottinger Nachrichten. Geomech. (1994) Modelisaton elastoplastique d u comportement mecanique des sols.
A t d Metl~. C. C. 2. And. Etg. and Mroz Z.~ Tvunsimt und Cyclic Lourls. N. (1985) Generalized Plasticity formulation and application to Geomechanics. Zienkiewicz. (1978) Nonlinear seismic response and liquefaction. Desai. Hinton. Int. 0. C. Zienkiewicz. Y. and Chang. Zienkiewicz (Eds). pp. C. D. E. Gudehus (Ed. W.REFERENCES 175 Wineman. Muter. pp. Nutn. T. 0. pp. J. Leung. H. Mech. 381404. S. C. Soils und Foundutions. (1975) Associated and nonassociated viscoplasticity and plasticity in soil mechanics. G . Num. Humpheson C. (1985) Simple model for transient soil loading in earthquake analysis. (1990) Numerical testing of the stability criterion for hypoplastic constitutive equations. C. T. C. N. Metlzods in Geomech. pp. Wiley. (1964) Material symmetry restrictions on constitutive equations. and Lewis. Int. W. A. H. H. Geotnerh. 453476. Ch. 25. 0. 17. 0. C. 655680 John Wiley and Sons. R. and Pastor. Zienkiewicz. A n d .Wu and Kolymbas. Chang.. Chapter 5. 2. J. C. K. 0. 19. 671689. C. in Soil Mpchunic. K. Pande and 0. K . Mech. Muter. G.). 1 Basic model and its application. and Pipkin.  . Arch. 184214. 33. 0. G . 71104. (1982) Liquefaction and permanent deformation under dynaniic conditions Numerical solution and constitutive relations. Zienkiewicz. 195201. M. John Wiley and Sons. and Hinton E. and Gallagher. (1977) Some useful forms of isotropic yield surfaces for soil and rock mechanics.. and Ishihara. Zienkiewicz.C. 179190. G&technique. Yamada. A. and Pande. Rat. Zienkiewicz. (Eds). 7994. R. 9.. Mrcli. (1979) Anisotropic deformation characteristics of sand under threedimensional stress conditions. S. in Finite Eletnmts in Geomechanics. 29. Leung.
.
1 INTRODUCTION In this chapter we deal with the solution of static and quasi. An excessive amount of the latter is a measure of failure generally or loss of servicability where the stress distribution or its magnitude of pore pressures are indicative of the state of the material stressed. This is a definition which is often accepted. on occasion a finite displacement can be specified as failure by the engineer and knowledge of displacements is important even if these are not excessive. Less computational effort is involved if the acceleration terms are neglected and the G N l l scheme is used for both the skeleton displacement and pore pressure. which answers the question posed by the engineers and which provides the determination of failure. It is clear (as indicated earlier) (see Equation (3. Failure is sometimes associated with continuing displacement without a load increment. it is not . Since finite deformation and partial saturated examples are only given in this chapter. It is the dejormution and displucement which are observable and must be determined. However. In the latter category the phenomena associated with consolidation are typical and both fully and partially saturated cases will be presented. Consolidation and Partially Saturated Dynamic Problems 5.66)) that the problems of consolidation can be directly solved by the code based on our formulation. For these the knowledge of the constitutive relation discussed in the previous chapter is of paramount importance. examples involving hypothetical dynamic behaviour using finite deformation and under partially saturated condition respectively are also given.Examples for Static. is to be used at all times. In both classes of problems we are concerned with the deformation and movement of the soil or of its associated foundation. as with slow motion the dynamic effects become automatically negligible.static problems in which dynamic (inertial) effects are negligible. However. In the first three chapters of this book we have formulated the dynamic problem and its solution with the time dependence being retained in the final discretized equations. but the simplest constitutive law.
Particular attention is paid to the effect of the plastic dilatancy effect.7. Sections 5. It is followed by the comparison between the consolidation of a fully saturated and a partially saturated soil column in Section 5. the analysis of an isothermal drainage of water from a vertical column of sand is compared with experimental results in Section 5.3. in Section 5. followed by a consolidation phase. such as that involved in the stability of embankments and some foundation problems. Lastly in Section 5. In this section we shall endeavor to show the reader that numerical solutions.7.4. The effect of air flow is considered in Section 5. It is followed by the modelling of subsidence due to pumping from a phreatic aquifer in Section 5. .4 we have shown how the static problem can be dealt with by the dynamic code without any loss of efficiency when an appropriate time stepping scheme.4. A twodimensional consolidation example is introduced in Section 5.e.3 for a soil layer consolidating under fully saturated and partially saturated conditions. In Section 3. the elastoplastic largestrain behaviour of an initially saturated vertical slope subjected to gravitational load and horizontal earthquake acceleration.1. we shall begin with some typical static problems.6 by the modelling of a flexible footing resting on a partially saturated soil. The comparison of the small strain and finite deformation formulation under earthquake loading is given for a onedimensional soil column example in Section 5. In this chapter. is employed.7.5. In the next section. is given.7. a comparison of consolidation and dynamic results for various fully saturated and partially saturated problems using both small strain and finite deformation formulation. Since fully saturated behaviour is but a special case in our formulation.7.2 some typical static problems using a nonassociative MohrCoulomb material model. 5. are necessary and provide otherwise unavailable information.7.2 STA TIC PROBLEMS In this section we shall present some typical problems of static analysis both of the fully drained and totally undrained kind (the latter usually involving fairly rapid load application during which seepage can be considered negligible). Then we return to the theme of bearing capacity in Section 5.178 CONSOLIDATION AND P A R T I A L L Y SATURATED D Y N A M I C PROBLEMS obvious that static problems can be directly dealt with by the general program though it has been customary to create special programs for such analyses. though costly by comparison with the simple limit methods which are widely and often successfully used in geomechanics.2. First.35.6 are devoted to various problems concerning consolidation under partially saturated conditions. the GNOO scheme. the failure loads obtained by numerical computation are not very dependent on the plastic flow rules and again match well the predictions by conventional (limitbased) analysis. is given. The examples chosen will show that: (a) For a relatively unconstrained situation. The comparison begins with the consolidation of a fully saturated soil column in Section 5.5 which is the numerical modelling of air storage in an aquifer.2. we shall introduce first in Section 5. i. respectively. Lastly.
The superscript' implies that these quantities are effective stress parameters. the difference in the collapse load between the assumptions of associative and nonassociative behaviour is very small. as effective parameters are always used. The computation was carried out by Zienkiewicz et a1 (1975) from which Figure 5. The value of the cohesion at the collapse point is identified as that obtained by slip circle analysis (despite the coarse mesh employed). Embankment The first example is that of an undrained soil in an embankment in which the angle is steeper than that of the internal friction and in which cohesion has been added. 5. Here no exact closed . (c) Drained and undrained behaviour can be studied effectively by a single specification of material properties using the effectivestress concept.2. refers to the normal tensile effective stress. shows the collapse of a footing exerting a uniform load on the soil stratum. Further. also taken from the same reference as the previous one. the effective stress will be the same as the total stress n. Footing This example.STATIC PROBLEMS 179 (b) For situations in which nonhomogeneous material distribution or geometry provides an appreciable amount of confinement. there is no confusion if the superscript' is dropped. and finally: (d) The solution can be quite sensitive to the form of the yield section assumed in the Tplane In all the examples we shall specify the material properties by a yield surface of MohrCoulomb type with a straightline tangent and in which 'yield' is assumed to occur when the shear stress T exceeds a given cohesion c' and a friction angle o': where a'. limitbased methods do not predict the failure loads satisfactorly. In this problem the collapse is achieved by progressively reducing the cohesion. For the dry problem.1 is taken. In this section.1 Example (a): Unconfined situationsmall constvaint In this set of examples we shall study the importance of associative and nonassociative flow rules and will find that here the choice of the flow rule is of almost negligible importance as far as failure loads are concerned.
0 r 5 10 15 Cohesion : k ~ / m ' Collapse cohesion by slip circle .0 .s ~ oconverged t Material propenles E=Zx 1 0 J k ~ l r n L v=0. 0  8 I I II 6. 0  .180 CONSOLIDATION AND PARTIALLY SATURATED DYNAMIC PROBLEMS Cohesion : k ~ / m ' 20 25 30 I 5 8.@=20) m * .0 K   1 b .8 .5 .b=0.213.0 I y I Nonassociative flow rule (7.25 @=2O y20 k ~ l r n ' .
is just not true as we shall illustrate in the following sections with examples in which some degree of constraint exists. however.STATIC PROBLEMS 181 Figure 5. The two previous examples suggest that numerical analysis adds little to the solution of the problem from the point of view of practical geomechanics. Reproduced from Zienkiewicz (1975) by permission of the Institution of Civil Engineers form solution exists. The second example illustrates however a case where differences between associated and nonassociated rules becomes appreciable. but for collapse.54 while numerical solutions show little influence on the flow rule employed giving 1. the main addition appears to be only the increased cost of analysis and little additional information has been gained.165 for both associative and nonassociative flows. Different degrees of dilatancy now affect the load appreciably.3.2. Indeed. 5. . progressive decrease of cohesion) (a) Associative 0 = 4 = 20" (b) Nonassociative 0 = 0 4 = 20".2.09 to 1.4. Here the different plastic displacements of each layer are such as to cause a collapsed load for which simple geomechanic methods is not sufficient. a mechanism suggested by Prandtl (1920) and Terzaghi (1943) gives loads which are compared with the numerical ones in Figure 5. This. In the table attached to the figure we show that a limit approach gives safety factors ranging from 1. Once again very close results were obtained for the limit loads by both associated and nonassociated plasticity models. This case is that of an Axisymetric triaxial sample loaded between two rough platens as shown in Figure 5.1 Embankment deformation flow patterns and maximum effective shear strain contours (gravity constant.2 Example (b): Problems with medium (intermediate) constraint on deformation The first example here is that of a heterogeneous embankment illustrated in Figure 5.
resulting in almost no overall change in volume during deformation. Prescribed load 4 Material properties Plane strain 32 quadratic elements 0 20 Applied footing pressure (Iblin') 40 60 80 100 120 140 160 180 Terzaghi Nanassociated flow' rule ( q j = . In such conditions the total volumetric strain is controlled by the compressibility of the pore fluid and this is generally small.'\ 0") :C MohrCoulomb material = \wNot Prandtl !on\ I JP"(~P'~ I0 lblih c$ = 20" Critical State 4 model I I Figure 5.3 Example ( c ) : Strong constraintsundrained behaviour Probably the highest degree of constraint on the deformation of the soil skeleton is the behaviour occuring during undrained deformation.182 CONSOLIDATION AND PARTIALLY SATURATED DYNAMIC PROBLEMS 5.2 (a) Strip load on a foundation of a weightless c4 material. ideal plasticity with associated and nonassociated (nondilatant) flow rules and strainhardening plasticity.2. Mesh and loaddeformation behaviour .
.e. = 0. are compensated by equal and opposite changes of pore pressure Ap and the mean effective stress remains unchanged i. any changes of mean total stress Aa. Reproduced from Zienkiewicz (1974) by permission of John Wiley & Sons Limited Such behaviour has two consequences: (i) If the compressive modulus (Bulk modulus) of the skeleton is small compared with that of the fluid.STATIC PROBLEMS 183 (ii) Figure 5.2(b) Relative plastic velocities at collapse (drained behaviour) (i) Associated MohrCoulomb (dilatant) (ii) Nonassociated MohrCoulomb (zero dilatancy) (iii) Strain dependent critical state model. Ad.
Reproduced from Zienkiewicz (1975) by permission of the Institution of Civil Engineers Thus if the material is initially at the point of yielding as shown in Point A of Figure 5.184 CONSOLIDATION AND PARTIALLY SATURATED DYNAMIC PROBLEMS O r i g i n : & r Granular fill material ) = 35 l Granular material 0 10 20 0 20 40 60 Organic silt 80 100 .165 Reference WhitmanBailey (1967) WhitmanBailey (1967) WhitmanBailey (1967) Sarma (1973) Sarma (1973) Be11 (1968) Figure 5.26 1.25 lbln) ) = 33 = ) = 20 = 125 1wrt2 y=llolb/f? 120 140 Method WhitmanBailey Bishop Fellenius Sarma MorgensternPrice Bell Associated MohrCoulomb Nonassociated MohrCoulomb Safety factor 1.33 1. (ii) If the plastic yielding is such that deviatoric strain changes of the skeleton must be accompanied by volumetric strain changes (as for instance required by associative laws) then the only way to achieve overall straining without volume change is to compress (or expand) the fluid. ..241.542 1.165 1. With the volume expansion required by associated .557 1. .49 1.3 Layered embankment problem (a) geometry and material properties (b) finite element mesh (53 quadratic isoparameric elements) (c) relative shear strain rate contours at collapse (d) table of safety factors using various methods.5 any deviatoric stress changes will occur without changes of the mean stress and the material will behave like a von Mises (or Tresca) solid with respect to total stress changes providing no volumetric stress occurs during such straining.09 1.
deviatoric stress plane. It will be observed that for a nonassociative behaviour with no dilatancy in failure. the failure loads are almost . Figure 5. Reproduced from Zienkiewicz (1978) by permission of John Wiley & Sons Limited Figure 5. Reproduced from Zienkiewicz (1974) by permission of John Wiley & Sons Limited flow rule.STATIC PROBLEMS .0 4.Prescribed load 4 1 85 Material properties 0 1.0 6.0 5. or for a critical state model where failure occurs with no volume changes.0 C = 10 m / m 2 @=45 Displacement v:m x 10 4 Figure 5.4 Axisymmetric sample between rough platens. the material will develop negative pore pressures during plastic straining and will gain strength continuously.6 shows the same footing problem as that in Figure 5.0 3.0 2. Changes of total stress from point A.2 but now solved (on the same mesh) introducing the undrained assumption.5 The MohrCoulomb trace in the mean stress . Effect of degree of dilatancy 9 = 0 (zero dilatancy) and 9 = $J = 45" (fully associated flow rule).0 7.
5. we have assumed that the full MohrCoulomb relation is employed which requires the yield (andlor flow) surface giving. Reproduced from Zienkiewicz (1974) by permission of John Wiley & Sons Limited identical with those of total stress analysis with Tresca assumption. expanding material allows no overall failure. However the nonassociative.186 CONSOLIDATION AND PARTIALLY SATURATED DYNAMIC PROBLEMS Applied footing pressure (lblin')  Pressure (lblinL) IV up to = 3500 Iblin but no collapse load found I Elastic ideal plastic (Mohr Coulomb) associative (dilatant) I1 Elastic ideal plastic (Mohr Coulomb) non associative (zerodilatancy) 111 Elastic strain dependent plastic (criticalstate) IV Elastic ideal plastic total analysis of non Mises Figure 5.5 for the material at different response.7.2. difficulties arise at corners whlch need special treatment (see Zienkiewicz and Taylor. the Tresca yield surface rather than the better known and simpler von Mises one. In Figure 5. with this type of definition. However. .4 Example (d): The effect of the T section of the yield criterion In all of the previous examples. we show the 7r plane section of the MohrCoulomb surface for 4 = 20" and various approximations of it. in the absence of friction.6 Load deformation characteristics (undrained conditions) for plane footing. It must be remarked that in the present case the failure loads are governed entirely by the cohesion existing in the material as otherwise the strength would be simply zero when gravity were absent giving an entirely different starting point (A) in Figure 5. 1991) and many simplifications have been suggested in the literature.
 om press ion cone sZienkiewiczPande ompromise envelope Extension cone MohrCoulomb Figure 5.3 c = 10 lbIin.2 f = 20" 32 Parabolic elements used Soil condition . Reproduced from Zienkiewicz (1978) by permission of John Wiley & Sons Limited .7 The /r plane section of the MohrCoulomb surface with 4 = 20" and various smooth approximations.STATIC PROBLEMS 187 1" .LoEompression cone Compramise cone E = 30000 Iblin2 n = 0.drained d Figure 5.2).8 Loaddeformation curves for ideal associated plasticity for various forms of the MohrCoulomb approximation (solution on the same mesh as in Figure 5. Reproduced from Zienkiewicz (1978) by permission of John Wiley & Sons Limited 1 / i / .
In the experiment. Furthermore. however.e.difference in the collapse load!! The simple set of examples presented was computed directly before the advent of the dynamic code. recomputed several typical cases using the method suggested in this chapter and have found an identical failure load without extra effort.3 ISOTHERMAL DRAINAGE OF WATER FROM A VERTICAL COLUMN OF SAND It is very difficult to choose some appropriate tests to validate the twophase flow partially saturated model because of the lack of any analytical solutions for this type of coupled problem where deformations of the solid skeleton are studied. Some difference in displacement response is obtained due to the difference in plasticity formulation (viscoplasticity formulation had been used in these examples) and a nonlinear iterative scheme has also been used. but only for onephase flow. 1995). the same interpolation is used for pressures and displacement fields. At the start of the experiment i. Here (Gawin and Schrefler (1996)) attention is paid to the effects of twophase flow. Another is to replace the surface by a circle as is done in the well known DruckerPrager surface (1952) The effect of such an approximation is surprising as shown in Figure 5. (1990b). In this example.7594. isoparametric Lagrangian elements are used. As discussed in Chapter 4 many possibilities arise. as well as by Schefler and Zhan (1993). linear elastic material behaviour is assumed. the water supply was stopped and the tensiometer readings were recorded. by a seperate set of experiments. There are also very few documented laboratory experiments available. as compared to the previous solution (Gawin et al. It is of interest to note that the full MohrCoulomb and the extension cone in the DruckerPrager give close representation.2 is used as a test bed. During the numerical simulation. t = 0. One is to use 'smoothing' and avoid corners in manner described by Zienkiewicz and Pande (1977) and Gudehus (1978).8 where the same foundation problem as that of Figure 5. The same example was solved in Gawin et al. (1995). at t < 0.e. a column of one metre high Perspex was packed with Del Monte sand and calibrated to measure the moisture tension at several locations along the column. The porosity n = 29. One of these is the experimental work conducted by Liakopoulos (1965). Zienkiewicz et al. Liakopoulos' saturationcapillary pressure and relative permeability of watercapillary pressure relationships were approximated using the following equations: . until uniform flow conditions were established. and the hydraulic properties of Del Monte sand had been measured by Liakopoulos and reported in his PhD dissertation (l965). Schrefler and Simoni (1988). This test was also used by Narasimhan and Witherspoon (1978). 5. although the smoothing appears to cause about 20'%. on the isothermal drainage of water from a vertical column of sand. water was continuously added from the top and was allowed to drain freely at the bottom through a filter. together with the saturatedunsaturated flow of mass transfer. We have.188 CONSOLIDATION AND PARTIALLY SATURATED DYNAMIC PROBLEMS which appear feasible. Before the start of the experiment i. to check their numerical models.
For the bottom surface. p . = 0 for r > 150s. = OPa. For the purpose of numerical modelling. p .. besides uniform flow conditions i. which corresponds to bubbling pressure of the analysed sand (Liakopoulos. similar to Gawin er al. static equilibrium was also assumed. was assumed to change linearly from the initial value to zero for t < 150s and the base is fixed in both displacement directions. S = 0. = OPa (dashdot curves) are compared with the results for onephase flow (fine curves) in Figure 5. At the beginning.e.3MPa. eight and ninenoded isoparametric finite elements of equal size and different time steps in the time domain with (At = IOs. For the latter case.. means atmospheric pressure. The profiles of vertical displacements. is the relative permeability for the water phase. where p. is given in the unit of Pa and k.4 and Biot's constant cr = I. The boundary conditions were as follows: For the lateral surface.. the horizontal displacement and the fluid outflow is zero. unit vertical gradient of the potential andp. Is or 0. so the Young Modulus of the soil was assumed to be E = 1. Poisson's ratio v = 0..I S O T H E R M A L DRAINAGE OF W A T E R 189 where p.0001. For the top surface.5s) gave practically the same results. the switching between saturated and unsaturated solution was performed at p. . Liakopoulos did not measure the mechanical properties of the soil. The calculations were performed for onephase flow with gas pressure fixed at the atmospheric pressure in the partly saturated zone. = 2kPa (solid curves) andp.e. =pa. as well as for twophase flow. into 10 and 20 four. the Young's modulus value has been manually 'fitted' for the case. as well as at p.10. while waterp. 2 0. The resulting profiles of water pressure for the twophase flow with switching at pc2kPa shown in Figure 5. 1965). the column was divided. in turn. = 2kPa i. (1995). in order to analyse the effect of the 'switching' value on the solution obtained.9 as solid curves with different symbols at different time stations and for onephase flow in dashdot curves are compared with the experimental results of Liakopoulos (1965) in fine lines. The solution from onephase flow showed better agreement and this is because in Schrefler and Simoni (1988)... water saturation and capillary pressure for twophase flow with switching pressure atp. . is the equivalent value of saturation with the additional lower limit of k. = 0 at the top surface.998. The relative permeability of the gas phase was assumed to follow the relationship proposed by Brooks and Corey (1966): where S. pa = pa. Schrefler and Zhan (1993) and Schrefler and Simoni (1988).
= OPa in the upper part of the sand column. The discernable differences are caused by the different assumptions about gas flow i.10. . The differences between the one and twophase flow solutions are more appreciable for saturation of water and capillary pressure as shown in Figures 5. The onephase flow solution tends to the twophase flow solution with switching at p.998 in the first case. the gradients of saturation and capillary pressure are higher for the twophase solution in the upper zone where gas flow occurs. This example shows that the modelling of the transition from fully saturated to partially saturated condition and vice versa is very sensitive to the procedure adopted. = 2kPa indicated by solid curves and onephase flow solution indicated by dashdot curves) with experimental results of Liakopoulos (1965) in fine lines. This is qualitatively in accordance with the solution obtained by Schrefler and Zhan (1993). In the lower zone. nevertheless the qualitative similarity of the gas pressure profiles in the zone. where gas flow occurs. is obvious. There is a small difference in the pressure amplitudes. but higher up there is a qualitative change in the soil behaviour caused by the presence of gas under pressure (see also Figure 5. no gas flow for zones where degree of saturation is greater than 0.13 while at pc = OPa. while in the middle region it forms some characteristic constantvalue zones corresponding to switching values of capillary pressure or saturation.13). The solution for twophase flow with switching atp.1 1 and 5. although their final values are similar. Reproduced from Gawin and Schrefler (1996) by permission of MCB University Press Ltd There are some noticable differences between one and twophase flow solutions for vertical displacements as shown in Figure 5. where no gas flow occurs. = 2kPa are given as solid curves in Figure 5.12. The profiles of gas pressure for twophase flow with switching at p. as dashdot curves. the differences are small. In general.9 Comparison of the numerical solutions for water pressure (twophase flow with switching at p.190 CONSOLIDATION AND PARTIALLY SATURATED DYNAMIC PROBLEMS Figure 5.e. = 2kPa is similar to the onephase flow solution at the bottom of the column.
= OPa (dashdot curves) with onephase flow solution (fine curves) made for degree of saturation of water.ISOTHERMAL DRAINAGE OF WATER Vertical displacement (m) 191 0.8 1 Height (m) Figure 5.2 0. = 2kPa (solid curves) and p. = OPa (dashdot curves) with onephase flow solution (fine curves) made for vertical displacements. Reproduced from Gawin and Schrefler (1996) by permission of MCB University Press Ltd .11 Comparison of twophase flow solutions with switching at pc = 2kPa (solid curves) and at p.4 0.10 Comparison of the twophase flow solutions with switching at p.0045 5 0 0.6 0. Reproduced from Gawin and Schrefler (1996) by permission of MCB University Press Ltd Saturation () Figure 5.
6 0. = 0 Pa (dashdot curves) with onephase flow solution (fine curves) made for capillary pressure).192 CONSOLIDATION AND PARTIALLY SATURATED DYNAMIC PROBLEMS Capillary pressure (Pa) lo O O 0 0 0 0. = 0 Pa (dashdot curves). Reproduced from Gawin and Schrefler (1996) by permission of MCB University Press Ltd Figure 5.12 Comparison of the twophase flow solution switching at p.2 Height (m) 0.4 0. = 2kPa (solid curves) and at p.13 Comparison of the gas pressure profiles (the twophase flow solution) for switching at p.8 1 Figure 5. Reproduced from Gawin and Schrefler (1996) by permission of MCB University Press Ltd . = 2kPa (solid curves) and p.
For the inner lateral surface (radius 0.5m from the bottom. The relationships between capillary pressure.2. The process is assumed to be isothermal.. The boundary conditions were as follows: For the bottom surface.e. An aquifer of 10m depth sited on an impervious layer was subjected to pumpage of 20m3/h within a height of 2.3m). an outflow rate of 1. full saturation of water with unit vertical gradient of potential was assumed. E = 22MPa. At the beginning a fully saturated state with unit vertical gradient of the potential was assumed. Reproduced from Gawin and Schrefler (1995) by permission of John Wiley & Sons Limited .4 MODELLING OF SUBSIDENCE DUE TO PUMPING FROM A PHREA TIC AQUIFER This example (Gawin et a1 (1995)) deals with subsidence of saturatedunsaturated land due to pumping from an axisymmetric aquifer. as well as a mechanical equilibrium state. For the outer lateral surface (radius 100.~ m / s and n = 0. which was solved previously by Safai and Pinder (1979) and then by Meroi (1993). v=0.5m from the bottom is assumed and the horizontal displacement at this surface is assumed to be zero. satura 0 10 20 30 40 50 60 Distance (m) 70 80 90 100 Figure 5.. the pressure of the air is fixed at atmospheric pressure p. k' = 2 x 1 0 .3m). At the top surface.179kgmp'sp' within the height of 2.MODELLING OF SUBSIDENCE 193 5.14 Resulting profiles of saturation (full curves) compared with solution of Safai and Pinder (1979) in broken curves. the vertical displacement is fixed.1. The mechanical and hydraulic properties of the soil were assumed to be the same as in Safai and Pinder (1979) and Meroi (1993) i..
15 Resulting saturation and vertical displacement profiles (full curves) compared with solution of Meroi (1993) in broken curves. A threebythree integration scheme was applied. Temporal discretization was performed with an initial step size of 1 min for the first 10 hours. For numerical purposes.4) were again applied in the partially saturated zone. Reproduced from Gawin and Schrefler (1995) by permission of John Wiley & Sons Ltd tion and relative permeability of water followed those proposed by Safai and Pinder (1979). 10 min for the next 20 hours and then 1 hour for the rest of the 28 daysthe total required time of analysis.194 CONSOLIDATION AND PARTIALLY SA TURATED D YNAMZC PROBLEMS Distance (m) Distance (m) Figure 5. the aquifer was simulated by 50 eightnoded isoparametric elements (five in height and ten in radial direction) using the same mesh as in Meroi (1993). .3) and (5. The relative permeability of gas proposed by Brooks and Corey (1966) as given in (5.
only one phase flow is considered with the gas pressure fvted at atmospheric pressure in the partially saturated zone. Initial conditions are no displacement and a constant water pressure of 5. 30h and 28 days are compared in Figure 5. with a rigid soil skeleton. q.066 M N / or 50 ~ ~ Patm.p.44 x 10p4kg/s/m.2 and initial aquifer pressure of 5. The profiles of water saturation on the upper surface and vertical displacements for time stations of 10 min.AIR STORAGE MODELLING 195 The resulting profiles of water saturated on the top surface of the aquifer are compared in Figure 5. The water density for standard conditions is 100kg/m3 and the air density..f. u h 7 u.15 with the results of Meroi (1993) in broken curves. In Meroi (1993). The porous medium system was assumed initially fully saturated with a reference permeability of 5 x lop"m2.3mNs/m2 and 2 4 p ~ s / m 2respect. The final values at the centre matched very well. = 0 top surface. ively. = 0.22kg/m3.14 with the results of Safai and Pinder (1979) in broken curves. The water and air viscosities are selected as 0. = 2. under isothermal conditions at 149°C (or 300°F). 3h. with Young's modulus E = 692kN/m2. 5. q. 10h. The following relationships between the relative permeabilities of water and air. the capillary pressure and water saturation. uh = 0 bottom surface. = 0. = O. Meiri and Karadi (1982) simulated this problem by a onedimensional finite element model.5 AIR STORAGE MODELLING IN AN AQUIFER With the same geometry as in the column drainage test example of Section 5. The ~ all boundary conditions are as follows: lateral surface. are employed in the simulation: 2lli k. q. air storage in an aquifer is simulated by the partially saturated model (Schrefler and Zhan (1993)). q. = 0. = ST . In this simulation the skeleton is considered deformable. 1.4. proposed by Brooks and Corey (1966).. It has to be mentioned that the model of Safai and Pinder (1979) neglects the fluid accumulation due to changes in the degree of saturation. q. which explains at least part of the differences in the early stage of the pumping phase. showing relatively good agreement. a porosity of 0.066 M N / at ~ points. which could explain the difference in time transident behaviours.(air injection). = p.
8 49.196 CONSOLIDATION AND PARTIALLY SATURATED DYNAMIC PROBLEMS Hrs 00 10 20 t30 *40 m +.9 50. air pressure and saturation of air storage modelling in an aquifer.2 Vertical displacement (m) Pressure of water (atm) Pressure of gas (atm) Saturation of water (Dotted lines.0 50. water pressure.2 4.70 0.1 0. results from Meiri & Karadi) Figure 5. Reproduced with permission from Schrefler and Zhan (1 993) American Geophysical Union .2 49.0 0. I 50 60 +80 0.16 Profiles of vertical displacement.1 50.
The vertical displacement. 3. In order to point out the influence of the solid skeleton deformation. air flow and solid deformation. X is the pore size distribution index and p b is the bubbling pressure.16 for different time instances.17 Profile of saturation of air storage modelling in an aquifer with double Young's modulus for Figure 5. have their importance. Water and air pressure have a minimum at the interface of the two fluids. and 1 . the deformable skeleton model has greater desaturations. correX ~.6 0. The saturations are compared with the result reported in Meiri and Karadi (1982). indicated with dashed lines. The column was first contracting and then finally expaning vertically with continuing air inflow from the top. = (Sw Swc)/(l. air pressure. .8 0.AIR STORAGE MODELLING 197 I I I I I I 0. a second simulation was performed with a double Young's modulus.0.7 0. sponding to sand with an intrinsic permeability of 5 x 10I3m2.e. The new saturation profiles are shown in Figure 5. Reproduced with permission from Schrefler and Zhan (1993) American Geophysical Union (c where S.5 0.2. While the propfile of saturation after 80 hours of pumping is roughly the same with both Young's moduli. In general.4 0.16.17. and saturation versus depth of the column are shown in Figure 5. i. 6 8 k ~ / m respectively.0 I Saturation Figure 5. water pressure. i. The values for Swc. The effects of a deformable skeleton assumption can be seen clearly..9 1. and p b are given as 0. Swc the irreducible is is saturation.Swc) the effective saturation. at the onset of air flow the desaturation is greater in the rigid skeleton model while later on.e.. These simulations confirmed that the model proposed can reproduce this quasistatic case and that the main features of the model. the profiles of shorter time spans are different. the situation is reversed.
in Figures 5.18."~American Geophysical Union .22. S.. capillary pressure increasing versus top surface such that on the top surface S. B and C of Figure 5.6 FLEXIBLE FOOTING RESTING O N A PARTIALL Y SA TURA TED SOIL In this example (Schrefler and Zhan (1993)).. = 1 from S ... Since no iterations were performed between the two programs during a time step. (1987). a flexible footing resting on a partially saturated soil is modelled. From a computational point of view.21 and 5. using two separate computer programs.. The deformation after 100000s is shown in Figure 5. q.18 Discretized crosssection for flexible footing resting on a partially saturated soil.r = 101.198 CONSOLIDATION AND PARTIALLY SATURATED DYNAMIC PROBLEMS 5. Furthermore. Reproduced with permission from Schrefler and Zhan (1993) ci.) and u h = u.19. Hence the results of the two approaches are not expected to be the same. for comparison. i. together with a modified effectivestress principle and the fully coupled system of equations is solved as such. one for the twophase flow field and one for deformation analysis. u h = 0. for t 2 0. S.. = 405.(p.75. This problem was first solved by Lloret et 01.325 kPa). respectively.3 kPa (4 p. The discretized crosssection (19 x 12 m) is shown in Figure 5. (1987) is based on state surfaces while here the capillary pressure relationship by Brooks and Corey (1966) is used. p. = S. this represents some sort of matrix partitioning and was applied to the original fully coupled system of equations. = 0.20 and water pressure and gas pressure contour lines. the approach used in Lloret rt LEI. Finally. Figure 5.. = 0. acting downward): all other conditions are as for t < 0. the top surface outside the footing. The boundary conditions are as follow: for t < 0. bottom surface. = 0. = 0. The wetting process is here simulated for a time span of 10's. = 0. the change in the vertical displacement with time of three surface points A. it is not clear how far the coupling was preserved through this procedure. the results for fully saturated conditions are also indicated. lateral surface. The contour lines of the water degree of saturation at the same time is shown in Figure 5.23 and. q.18 is shown in Figure 5.e. (pa = p w =p. = 1 and the top surface under the footing f.
Reproduced with permission from Schrefler 0 and Zhan (1993) ( American Geophysical Union First the soil is compacting. and with increasing time the crosssection swells.19 Deformation shape at t = 100000s. .FLEXIBLE FOOTING RESTING 199 Figure 5.20 Water saturation at r = 100000s. with much larger vertical displacements below the footing. The constitutive model used in Schrefler and Zhan (1993) required some refinement in order to simulate collapse upon first wetting and rewetting of the soil. in the presence of a mean compressive total stress. Reproduced with permission from Schrefler and Zhan (1993) 0 American Geophysical Union Figure 5. as is shown have happened a t point A.
22 Air pressure (atmospheres) at t = 100000s.200 CONSOLIDATION AND PARTIALLY SATURATED D YNAMZC PROBLEMS Figure 5.21 Water pressure (atmospheres) at t = 100000s. Reproduced with permission from Schrefler and Zhan (1993) 0American Geophysical Union . Reproduced with permission from Schrefler and Zhan (1993) American Geophysical Union /c~ Figure 5.
it would be of interest to compare the small strain formulation . Reproduced with permission from Schrefler and Zhan (1993) 0American Geophysical Union 5.23 Change in vertical displacement with time at three different points as indicated in Figure 5. Although the finite deformation solution has not been described in this book.COMPARISON OF CONSOLIDATION AND D Y N A M I C RESULTS 201 I o4 I o5 I o6 108 I o7 Time (seconds) 1 Figure 5.7 COMPARISON OF CONSOLIDATION AND DYNAMIC RESULTS BETWEEN SMALL STRAIN AND FINITE DEFORMA TION FORMULA TION A largedeformation model with a number of simplifications has been developed by Meroi et ul.18. (1995).
The problem. different load levels were considered. At high load levels. 5. Meijer. It is assumed here. the theoretical relationship between the applied load and the final displacement is logarithmic for this example and it is represented in Figure 5. in accordance with Prevost (1982). Bathe et al. 1984. These studies have been reported in Meroi (1993) and they showed good agreement with the corresponding results in the literature. Lewis and Schrefler. 1976) were first solved and then consolidation solutions of fully saturated soils were obtained (Carter et al. A backward difference integrator was used because of its superior efficiency in consolidation analyses reported in Prevost (1982). and subjected to a step load applied at the top level. 1993. fully saturated by water.3.1 Consolidation of fully saturated soil column This example is concerned with the consolidation of a one. reaching a maximum intensity for the uniformly distributed load equal to the Young's modulus of the ground. The results obtained under this assumption are represented by curve b. 1976. Prevost 1981 and 1984). an elastic modulus of lGPa and a zero Poisson ratio are adopted. The computational results (curve a) are in very good agreement with the theoretical behaviour (curve c). E. The boundary conditions for the displacement field are such that: all nodes are horizontally constrained and the bottom level is fixed with no vertical movement. The largedeformation model has been extensively validated with respect to solutions reported in the literature. a nearzero void ratio may occur computationally. 1979. a specific permeability of 0. 1988. Kim et al. All the examples contained comparison with small deformation developed in Chapter 2 and 3. (1995) to test the new features offered by the largedeformation approach. With the finitedeformation formulation. 1982 and 1984) results of onedimensional elastic consolidation.01 mls. 1987. 1976. For comparison with Prevost's (1981. which is onedimensional since each vertical section can be considered as a plane of symmetry. infinitely extended in the horizontal direction. and at this stage the soil behaviour should be described by the elastic characteristics of the compacted grain itself: a relationship between the elastic modulus of soil and its void ratio has been given in Monte and Krizek (1976).24 by the dotdash curve. with drainage allowed only through the top surface.202 CONSOLIDATION AND PARTIALLY SATURATED DYNAMIC PROBLEMS introduced in Chapters 2 and 3 and to investigate the effect of finite deformation. Atmospheric pressure is assumed at the top level and impermeable boundaries are imposed at the lateral and bottom surfaces. The dashed curve (curve d) represents the results of linear analysis while the dotted curve (curve e) the solution obtained by Prevost (1982). for the sake of simplicity. that the soil elastic modulus becomes ten times as large as the initial one when the porosity approaches zero. For the purpose of validation.dimensional ten metre deep ground. Heyliger and Reddy. a number of onephase static and dynamic problems (Bathe and Ozdemir. .7. Monte and Krizek. An initial porosity of 0. due to the assumption of constant elastic modulus. Five examples have been presented in Meroi et ul. is modelled as a saturated soil column under plane strain condition. 1975. Shantaram et al. All of them are reproduced in this section.
25. A onedimensional column of 7 metres high was modelled in accordance with Kim et al. In the present analysis.6 times the elastic modulus of the soil matrix. the time factor (Lewis and Schrefler. A further improvement to the finite deformation model consists of the introduction of the dependence of the absolute permeability on the void ratio. curve d: linear analysis response.0.24 Vertical settlement versus load level in onedimensional elastic consolidation problem with fully saturated conditions: curves a and b. the maximum vertical displacement versus the normalized time T. b and c refer to load levels equal to 0.5E.7. Prevost solution. shows the influence on computational results of such a relationship. curve c. theoretical solution for the finite deformation regime. permeability is assumed to be a linear function of the void ratio. Reproduced from Meroi (1995) by permission of John Wiley & Sons Limited In Figure 5.t. where c.2.2 Consolidation of fully and partially saturated soil column This example refers to the work by Advani et ul. 5. Figure 5. varying from the initial value to zero when porosity becomes zero. is the coefficient of consolidation and c. computational results with the finite deformation approach. Consolidation with constant and variable permeability is described. curve e. (1993). respectively. respectively.4 and 0. by curves a and c for the load level equal to 0.t/h2 = c. for constant and variable elastic behaviour. The significant increase in time (in logarithmic scale on the figure) necessary to reach full consolidation can clearly be seen. in which time is given in the logarithmic scale. ( T .26. = c. (1993) using plane strain .3E and by curves b and d for a load level of 0. respectively. 1987)) for different load levels: curves a. (1993) and Kim et at.COMPARISON OF CONSOLIDATION AND D YN'4MIC RESULTS 203 Figure 5. but partial saturation conditions were also considered.
204
CONSOLIDATION AND PARTIALLY SATURATED DYNAMIC PROBLEMS
elements. The material properties are: porosity of 0.5, dynamic permeability of 4 x 101 lm4/Ns, a n elastic modulus of 6GPa and a Poisson ratio of 0.4. The resulting time factor c, = 1.049 x lo'. The uniformly distributed load applied in the vertical direction reaches a limiting value of 1OGPa. In Figure 5.27, the vertical settlement, normalized with respect to the corresponding values of Terzaghi's theory, are plotted against normalized time for the case of fully saturated conditions for small deformation (curve a) and large deformations (curve b). The results are in substantial agreement
Figure 5.25 Vertical settlement versus normalized time in the onedimensional consolidation problem with constant permeability and elastic modulus under fully saturated conditions and finitedeformation assumptions. Curves a, b and c refer to load levels equal to 0.2, 0.4 and 0.6E. Reproduced from Meroi (1995) by permission of John Wiley & Sons Ltd
Figure 5.26 Vertical settlement vs. normalized time in the onedimensional elastic consolidation problem with fully saturated conditions. Curves a and c refer to a load level of 0.3E while curves b and d to 0.5E with c and d having variable permeability. Reproduced from Meroi (1995) by permission of John Wiley & Sons Ltd
COMPARISON OF CONSOLIDATION AND DYNAMIC RESULTS
205
with the ones presented in Kim et al. (1993).The large deformation analysis was also performed for the case of permeability linearly dependent on the void ratio (curve c). In the same figure, the large deformation results for three different homogeneous partially saturated initial conditions were also plotted. These partially saturated initial conditions were imposed by assigning the initial capillary pressure distribution corresponding to the degree of saturation via the capillary pressure relationship in Figure 5.28. While the small deformation analysis with air pressure equal to the atmospheric pressure gives the final results in one time step (curve d), in the large deformation analysis a real consolidation process has taken place.
Figure 5.27 For the model drawn, vertical settlement, normalized with respect to the corresponding values of Terzaghi's theory, vs. normalized time: fully (curve a, smalldeformation analysis; curves b and c, largedeformation analyses with constant and variable permeability, respectively) and partially saturated conditions (curves d for small deformation analysis). Reproduced from Meroi (1995) by permission of John Wiley & Sons Limited
Figure 5.28 Saturation and relative permeability vs. hydraulic head. Reproduced from Meroi (1995) by permission of John Wiley & Sons Limited
206
CONSOLIDATION AND PARTIALLY SATURATED DYNAMIC PROBLEMS
Figure 5.29 Time transient water pressure. normalized with respect to applied load. for the fully saturated case at the top. Curve a, smalldeformation approach; curves b and c, largedeformation cases with constant and variable permeability, respectively. Reproduced from Meroi (1995) by permission of John Wiley & Sons Limited
The dissipation of water pressure for the fully saturated case in the top element is shown in Figure 5.29 for small deformation as curve a, and for large deformation with constant and variable permeability as curves b and c
5.7.3
Consolidation of twodimensional soil layer under fully andpartially saturated conditions
This example is also taken from Kim et ul. (1993) and consists of a twodimensional, plane strain analysis of an 8m deep soil layer loaded by a uniform pressure of 6GPa by a 16m wide foundation (=2b). Because of symmetry, only half of the model is considered, which extends for 48m in the horizontal direction. All the material data are the same as in the previous section, apart from the Poisson ratio which is assumed to be zero in accordance with Kim et al. (1993), the resulting time factor is c, = 0.375 x The boundary conditions for displacements are taken as: the bottom level is fixed, while laterally only vertical displacements are allowed. Both rigid and flexible footings are considered. In order to model the rigid footing, a multipoint constraint technique, i.e. tiednodes technique, is employed. Pressure is assumed to be zero at the top surface including the underside of the foundation in accordance with Kim et al. (1993). Since a net flow is allowed through the foundation, it is assumed to be permeable. Besides the fully saturated case, a partially saturated model is also investigated with an initial saturation of 92%).The initial condition is obtained by imposing a suction in accordance with the relationship given in Figure 5.28. In the case of a rigid foundation under partially saturated conditions, the foundation is now considered as impervious.
COMPARISON OF CONSOLIDATION AND D Y N A M I C RESULTS
207
Displacements of the top node at the centre line, normalized according to
w, = Ew/2bq are shown in Figure 5.30 for the rigid foundation and in Figure 5.31
for the flexible one. In both figures, curves a refer to smalldeformation analysis introduced in Chapters 2 and 3, for the fully saturated case, and in perfect agreement with results proposed by Kim et 01. (1993) and with the analytical solution of Booker (1974). Curves c refer to largedeformation analysis. In the case of a rigid foundation, the result is also given by Kim et 01. (1993) and is in good agreement with the one reported here. Consolidation with the linear dependence of the permeability with the void ratio is indicated by curve d. Curves e describe the behaviour in the case of an initially uniform partial saturation of 92% and large deformations. For permeable flexible foundation, the behaviour is almost time independent, while for a rigid impermeable foundation, some consolidation takes place and the time for transient behaviour is longer than the fully saturated case, because the foundation impermeability forced the water not only to follow a longer path, but also one with a smaller relative permeability. Figure 5.32 represents the deformed mesh in the case of a flexible foundation. giving consolidation patterns at T, equal to 0.01, 0.1 and 0.55. For the same dimensionless times, Figures 5.33 and 5.34 give the results for a rigid foundation with full and partial saturation, respectively. The comparison of the different deformed shapes allows one to appreciate the influence of the different fluid pressures. In particular, a swelling close to the foundation during the first stage of the analysis can be observed under fully saturated conditions.
Figure 5.30 Model description and normalized settlement of the top node at the centre line versus normalized time for the rigid footing. Curves a and b, smalldeformation regime for fully and partially saturated initial corlditions, respectively; curves c and d, finite deformation analysis from initial fully saturated conditions with constant and variable permeability, respectively; curve e, finite deformation result from initial partially saturated conditions. Reproduced from Meroi (1995) by permission of John Wiley & Sons Ltd
208
CONSOLIDATION AND PARTIALLY SATURATED DYNAMIC PROBLEMS
Figure 5.31 Normalized settlement of the top node at the centre line for the flexible footing case versus normalized time. Curves a and b, smalldeformation analysis for fully and partially saturated initial conditions; curves c and d, finite deformation analysis from initial fully saturated conditions with constant and variable permeability, respectively; curve e, finite deformation results from initial partially saturated conditions. Reproduced from Meroi (1995) by permission of John Wiley & Sons Ltd
Figure 5.32 Deformed mesh for flexible footing, the fully saturated case: the consolidation pattern is given at dimensionless time, T, equal to 0.01, 0.1 and 0.55. Reproduced from Meroi (1995) by permission of John Wiley & Sons Ltd
COMPARISON OF CONSOLIDATION AND DYNAMIC RESULTS
209
Figure 5.33 Deformed mesh for rigid footing, the fully saturated case: the consolidation pattern is given at dimensionless time, T, equal to 0.01, 0.1 and 0.55. Reproduced from Meroi (1995) by permission of John Wiley & Sons Limited
Figure 5.34 Deformed mesh for rigid footing, initial partial saturation of 92%: the consolidation pattern is given at dimensionless time, T, equal to 0.01, 0.1 and 0.55. Reproduced from Meroi (1995) by permission of John Wiley & Sons Limited
5.7.4
Fully saturated soil column under earthquake loading
This example is used to test the code's capability of modelling large deformation with plasticity. This example of liquefaction performed by Zienkiewicz et al. (1990b) and Xie (1990) was used for this testing purpose and the PastorZienkiewicz (1986) model as described in Chapter 4 is adopted for the sand. It can be seen that the results were not significantly affected by the use of geometric nonlinearity because the fluid pressure can rise even without the large deformation during a cycle of earthquake
210
CONSOLIDATION AND PARTIALLY SATURATED DYNAMIC PROBLEMS
loading. The first 10 seconds of the NS component of the ElCentro 1940 earthquake is taken as the horizontal base acceleration input during the consolidating phase of the sandy soil under a uniformly applied load of 600kPa. The geometry and mechanical characteristics of the model are the same as those given by Zienkiewicz et 01. (1990). and, in particular, a n initial elastic modulus of 4.5 MPa is assumed. Figure 5.35 shows the increase in pore pressure, both for large deformation (upper of the twin curves) and for the small deformation approach during the first 15 seconds for
Figure 5.35 Pore pressure versus time in the generation phase at three reference points of the drawn model, both for largedeformation (upper one of the twin curves) and for the smalldeformation approach. Reproduced from Meroi (1995) by permission of John Wiley & Sons Limited
Figure 5.36 Horizontal displacements versus time at points A and D for both largedeformation (upper of the twin curves) and for the smalldeformation approach. Reproduced from Meroi (1995) by permission of John Wiley & Sons Limited
The fully saturated domain after excavation is modelled by linear triangular finite elements.7.36. Pore pressure and vertical displacements are shown in Figures 5.38 respectively. an undrained analysis is carried out with horizontal constraints at the vertical slope nodes. Reproduced from Meroi (1995) by permission of John Wiley & Sons Limited . both for the large (the one with the smallest final value of the twin curves) and for the smalldeformation approach. Figure 5. For any pair of curves. 5. To impose initial. the above mentioned horizontal constraints are released and the soil is allowed to desaturate. the one with smaller final value belongs to the largedeformation approach.37 Pore pressure versus time in the dissipation phase for the given points. 0 0 1000 2000 3000 4000 5000 6000 t (s) 0 Figure 5.39 shows the final pressure distribution plotted over the deformed mesh. 1990) subjected to gravitational loading during the first 25 seconds of ElCentro NS component of horizontal acceleration and during the following consolidating phase. Then the seismic excitation is applied at the bottom nodes. The horizontal displacements obtained using the two different approaches at points A and Dare given in Figure 5. with the piezometric level at the top nodes of the domain. It can be noted that the amount of vertical displacement induced by the earthquakes is not significant in comparison to the one induced by the surface applied load.5 Elastoplastic largestrain behaviour of an initially saturated vertical slope under a gravitational loading and horizontal earthquake followed by a partially saturated consolidation phase This last example considers the elastoplastic largestrain behaviour of a 9.15m vertical slope (Chen and Mizuno.37 and 5.COMPARISON OF CONSOLIDATION AND D Y N A M I C RESULTS 211 three reference points on the finiteelement model. fully saturated conditions.
(1990).212 CONSOLIDATION AND PARTIALLY SATURATED DYNAMIC PROBLEMS Figure 5.39 Final pressure distribution over the deformed mesh.0 Figure 5. in accordanced with Zienkiewicz rt al. Reproduced from Meroi (1995) by permission of John Wiley & Sons Limited PRESSURE TIME: 2570. both for the large (the one with smallest final value of the twin curves) and for the smalldeformation approach. horizontal displacements are fixed and a hydrostatic pressure distribution with atmospheric value at the two corresponding top nodes is assigned. Reproduced from Meroi (1995) by permission of John Wiley & Sons Limited At the left and right vertical sides of the domain. The material characteristics of soilsaturation relationships included.38 Vertical displacement versus time in the consolidation phase at the given points. are assumed with reference to the PastorZienkiewicz (1986) model of the clay core of the San Fernando dam. .
.
41 Vertical (v) and horizontal (h) displacements versus time during the seismic and consolidation phases of the top point of the slope.214 CONSOLIDATION AND PARTIALLY SATURATED DYNAMIC PROBLEMS 5 10 15 20 25 time (s) 100 1000 time (s) Figure 5. Reproduced from Meroi (1995) by permission of John Wiley & Sons Limited .
. Booker J. 10531065. However. (1979) The analysis of finite elastoplastic consolidation. hlt. Kim C. 7. 92. Booker J. H . (1995) Coupled heat.Meth. Lee T. Heyliger P. T. 36. 10. Solids & Structures. A. 8192. and Ozdemir H. H.. Advani S. 9. C. and Wilson E. Int. Eng. (1952) Soil Mechanics and Plastic Analysis or Limit Design.. Lee J. while in Figure 5. J. (1990) Nonlinecrr Ancrlysis in Soil MechcinicsTlzeory and Implement~ztion. Math. 147160. No. Bathe K . H. Meth. P. 20. Chichester. Int.40. University of California. Meth. 161179. J. Gawin D. Ramm E. Brooks R.CONCLUSIONS 215 The saturation distributions are represented over the corresponding deformed configurations in the area close to the slope at different times in Figure 5. (1993) Hygrothermomechanical evaluation of porous media under finite defromation: part I1 model validations and field simulations. 13. Liakopoulos A. it would be useful for practical purposes for the formulation to be validated using model experiments and this will be introduced in the next chapter. S. and Lee J. Chen W. 353386. S. Int. (1996) Thermohydromechanical analysis of partially saturated porous materials. (1988) On a mixed finite element model for large deformation analysis of elastic solids. Int. Struct.. 3. Num. 107129. Eng. N u r ~ Meth. R. and Schrefler B.. and Schrefler B.. Cornput. (1974) The consolidation of a finite layer subject to surface loading. Lewis R. R. Int. Berkeley. Coinp. A. 131145. J. H. water and gas flow in deformable porous media. (1987) The Finite Elenzent Method in the Drfbrmation crnd Consolidution of Porous Mrdiu.. 10.. and Corey A. 5. Anal. and Prager W. Lee T. W. The results obtained from static and consolidation analysis are highly satisfactory and compared well with available analytical and experimental solutions. Amsterdam. and Reddy J. Nurn. Eng. Drucker D. Gawin D. (1976) Elasticplastic large deformation static and dynamic analysis. W. 969987. L. J. Elsevier. 6168. and Small J. 36. and Mizuno E. REFERENCES Advani S. Int. (1993) Hygrothermomechanical evaluation of porous media under finite deformation. A. Eng. . F.. and Kim C. No.. Appl. and Schrefler B. N. Nonlinenr Mrch. 113143. S.. Nurn. Baggio P.. J.8 CONCLUSIONS The formulation developed in Chapter 2 and discretized in Chapter 3 is used to analyse various onedimensional and twodimensional problems with fully saturated and partially saturated soil respectively. Part IFinite element formulations. J. J. IR2. N. (1975) Finite element formulations for largedeformation dynamic analysis. W. (1965) Transient flow through unsaturated porous media. ASCE IR. Fluids. John Wiley & Sons. Dissertation..  . J. 6. Bathe K. Carter J. Geomech. J. 157165. C.. (1966) Properties of porous media affecting fluid flow. USA. C. 23. S. R. Eng. Quart.41 the time history of vertical and horizontal displacements of the top node of the slope is reported. Nzm~. D.
.339351. N u t ~ Anrrl. J. E.. M. Int. Dissertation. M. A 6. and Lewis R. Zienkiewicz 0. 19.~. Univ. A. H. N. 53 1548. (1981) Consolidation of anelastic porous media. A. 30. and Pande G . of'EnginrrrPrevost J. and Witherspoon P. in Mec~l~crnics ing Materirr1. Mech.. Gens A. G e m ~ ~ r h . C. Coll. Chapter 26.. 671689.. Rotterdam. and Krizek R. Netherlands. 1. Ph. 1. Shantaram D. Inc. and Alonso E. bit. Int.. J. H. W. heat flow in deformable porous media. 4. (1975) Associated and nonassociated viscoplasticity and plasticity in soil mechanics. GPotechnique. Ledesma A. GPotechniyue. A429. No.fCr M(rthernntik und Mechnrlik. Pastor M. Meroi E. 29. A n d . C. (1990) Finite element solution and adaptive analysis for static and dynamic problems of saturatedunsaturated porous media. Meiri D.216 CONSOLIDATION A N D P A R T I A L L Y S A T U R A T E D D Y N A M I C P R O B L E M S Lloret A. large deformation effects and fluid interaction.D. (1986) A generalised plasticity hierarchical model for sand under monotonic and cyclic loading. NZUII. numerical study of deformation and failure of rock masses. Schrefler B. (1991) The Fitiite E/(wer~t McthorlVolutne 2 Solid rrnti Fluid M e c I ~ ~ n i cDynuniic~~ n Nunline(rritj' (4th erln). Balkema. J. airflow and Heut Fluid Flolc. McGrewHill Book Con~pany. A. (1990) Static and Dynamic behaviour of soils: a rational approach to quantitative solutions. Wales.s. (1976) Dynamic transient behaviour of twoand threedimensional structures including plasticity. Schrefler B. . F. 14. L.Soc. Proc. A. (1982) Sin~ulationof air storage aquifer by finite element model.. Wtrter Resourc. 310323. 1 . 8. Zienkiewicz 0.. Zhan X.D.. Applications. 81106. of Swansea.laminate model of rocks a . M. Soil Terzaghi K. Wuter Rrsources Rex. Anrrl. Xie Y. Vol. 13 11 50. (1995) A coupled model for water flow. Lond. J.. A S C E E M . and Pinder G. Schrefler B. L. Prevost J.  . 10171034. (1988) A Unified approach to the analysis of saturated unsaturated elastoplastic porous media in Nurnericcrl Met1iod. J. Safai N. A. 318.. and Simoni L. and Zienkiewicz 0 . Rotterdam. n d . Appl. 107. Ghent.. Batlle F. (1993) Comportamento non lineare per geometria di mezzi porosi parzialmente saturi. 219247. Geotnecli. Errrtl~qunkeEngitic~ering& Structurcil Djworvics. C. Humpheson C.. 1. (1921) ~ b e die Eindringungsfestigkeit plastisher Baustoffe und die Festigkeit von r Schneiden. Merh. Int. R. . Universiti di Padova.. 169186. (1984) Comparison of finite and infinitesimal strain consolidation by numerical experiments. New York. (1979) Vertical and horizontal land deformation in a desaturating porous medium. C. A h . (1977) Timedependent multi.. and Bicanic N. Int. Nurn. and Zienkiewicz 0 . Xie Y. (1943) Tl~c~oreticcil Mecl~tmics. Nuni. Conzp. Roj*. 5. 155167. Wiley. (1993) A fully coupled model for waterflow and airflow in deformable porous media. (1984) Nonlinear transient phenomena in soil media. Nurw Metl~. No. A.John Wiley and Sons. Part 11: Semisaturated problems. NUMOG II. Geoniecll. N. H. 2. Schrefler B.. Balkema. 26. Prevost J. 25. Zeitschr~ft Ange~c~rmclte . . M. (1976) Onedimensional mathematical model for largestrain consolidation. Thesis. Innsbruck. 1925. Monte J. Narasimhan T. and Taylor R. Zienkiewicz 0 . A. April. C. 4955 10. Ph. 1520. and Zhan X. Meroi E. and Karadi G .es. Schrefler B. A. 53 1547. (1982) Nonlinear transient phenomena in saturated porous media. A . Istituto di Scienza e Tecnica delle Costruzioni.. Y. Owen D. 561578. (1987) Flow and deformation analysis of partially saturated soils in Grouncln~rter ejfkcts in Geotrc~linicul Engineering. and Zienkiewicz 0 . Eng. J. (1995) Large strain static and dynamic semisaturated soil behaviour. C. Meijer K. Wuter Resources R e x . L.s in Geon~echcinicr. Geornech. (1978) Numerical model for saturatedunsaturated flow in deformable porous media 3. and Simoni L. Zienkiewicz 0 . Prandtl L. C. Chichester. J. d London.
To test fully the possibilities offered by the formulation and the models presented we should seek examples where: (1) the repeated loading generates substantial pore pressures and possibly liquefaction. Further. we presented several examples of the application of the full formulation for various static and consolidation problems. the departure from linear. In the next chapter. Even if an earthquake of the desired magnitude with the desired effect did happen. we shall show such reconstructions known frequently as backanalysis. Class C predictions and therefore mistrusted by some as of course the soil parameters could be adjusted to achieve the already known measured results. This effectively tested the limit behaviour of various constitutive models and also the interaction of slow drainage with deformation during the consolidation process. particularly if some idea of the input motion is available through measurements within reasonable proximity. elastic. The problem did not. in which the computation precedes the actual event. or indeed ever. These in the soil mechanics problems are. its precise detail of input would not be available before it happened. stresses are present and at least partial inelastic permanent deformation results. and in the second case. stretch predictive capacity: in the first case of limit behaviour it gave answers which were quite well known in general. however.Validation of Prediction by Centrifuge 6. nearfailure. of course. the study of earthquake response presents the greatest challenge here but it is hardly possible to measure it on a site where both conditions 1 and 2 above will occur within a reasonable time span. For this reason it is desirable to attempt scale model tests of earthquake events for which both Class A and Class B predictions are possible (The last one being computed . so beloved by soil mechanicians. What is often possible is to reconstruct catatrophic events. it would be almost impossible to achieve the so called 'Class A' prediction. behaviour during consolidation was small. (2) the problem is such that nonlinear.1 INTRODUCTION In the previous chapter (Chapter 5 ) . Clearly.
The scale model in the centrifuge is usually rather small and thus the whole frame of the test has to be rotated at high speed.4 represents results of a very major study undertaken in USA under the name of VELACS (VERification of Liquefaction Analysis by Centrifuge Studies . the basic theory of centrifuge modelling in geotechnical applications is introduced with particular attention to the use of an alternative fluid in order to achieve dynamic compatibility of diffusion and inertial behaviour. In this example. etc. it is less successful for soil mechanics problems. It is necessary. This study was funded by the National Science Foundation. .218 VALIDATION OF PREDICTION BY CENTRIFUGE simultaneously with the measurement on the model). The shaking table has been used with great success in modelling the dynamic effect of structures but. producing a fairly uniform field of up to 100g in the model area. we did not use scaling as the centrifuge itself and the artificial earthquake itself.2. During this study. For this reason.Arulanandan and Scott. both Class A and B type predictions will be possible. For this particular case. the centrifuge are kept constant and the linear din~ensions was invented and this device permits a very considerable increase in gravitational acceleration.5) where we describe a somewhat similar test done on an embankment wall at Colorado by Dewoolkar. very many alternative situations were investigated and it will be seen later that excellent comparisons were obtained. The reason here is that for typical soil problems gravity is the most important external force and this is obviously not modelled correctly in a scaled model in which densities reduced. 1993). In Section 6. we shall draw our comparisons in the following sections entirely from the centrifuge and here. USA and involved some twenty laboratories in various parts of the world performing numerical predictions which later were to be compared to several centrifuge studies in the USA and Cambridge Geotechnical centrifuge.. unfortunately. we recommend a study of various publications of Professor Schofield and others concerning the physics of centrifuge modelling. Two possible scale models of environmental conditions exist: (a) the shaking table. Only in one or two cases were the results obtained later and these are specially marked as Class B. the only results which are available will be presented by backanalysis in the following chapter. Section 6. In both cases. In the first section describing the centrifuge test. the same remarks apply. all the predictions are of Class A type as results had to be presented to the organisers before the centrifuge tests were attempted. to remark that to date in no case has it been possible to perform centrifuge models with a free fluid surface and such structures as dams. UK. 1987) performed on the Cambridge geotechnical centrifuge. cannot be modelled because of the restriction on the partially saturated conditions. we shall concentrate on a model of a dyke (Venter 1985. For this reason. It is precisely this substitution of pore pressure which rendered centrifuge testing unsuitable for the modelling of prototype dynamic events under partially saturated conditions and here lies one of the limitations of the procedure. For the interest of the reader. Here comparisons of computations are done simultaneously with models and perhaps this section should be classified under Class B verifications. retaining embankments with different levels of water at different sides. were considered to be the 'prototype'. In this section and indeed in the later section (6. and (b) the centrifuge. however. as we mentioned.
the linear length is scaled by: where the superscript M denotes the model scale and P denotes the prototype it intends to model. As we consider in the numerical analysis that the centrifuge experiment is a prototype itself. one would conclude that time is also scaled by N times: t The maintanence of the same stress level is important for soil behaviour as the stress: strain behaviour of soils is highly stress level and strainhistory dependent. If we assume that the density of the mixture and fluid. We can write the mixture equilibrium equation (see (2. This is included to aid the readers in the interpretation of the centrifuge results and their comparison with the numerical results. this would require the acceleration to be scaled by 1/N times with: and Comparing with the linear dimension in (6. issues concerning farfield boundary conditions and particle size will not be dealt with. Schofield and Zeng.1). 1995). 1980 and 1981.11) neglecting only convective terms) for model and prototype respectively. Furthermore. Lee & Schofield. Steedman and Zeng. the following publications can be referred to (Schofield. it is the purpose of this section to explain the concept of dynamic compatibility in centrifuge modelling which led to the use of a different pore fluid from water in saturated centrifuge model tests. 1992.SCALING L A W S OF CENTRIFUGE MODELLING 219 6. For readers interested in the centrifuge modelling of dynamic events. . a brief description and derivation of the centrifuge scaling laws for models are described.2 SCALING LA WS OF CENTRIFUGE MODELLING In this section. Assuming an N (typically between 50 and 200) scale model is introduced in the centrifuge. together with the stress statet are maintained the same in both the model and the prototype. 1988.
5 and 6. For instance. its compressibility should not be too different from that of water.14) again neglecting only the convective terms) is considered: Considering the equation for the model and prototype separately: and As the velocity is required to be the same in the model and the prototype if the scaling of the displacement. in the Cambridge test of the dyke.16)) which implies that pore pressure is the same in the model and the prototype if the compressibilities and the void ratio remain the same. a dynamic event will happen N times as fast as its corresponding prototype. however. 6.4. a problem arises when the fluid flow equation (see (2. 1985) because water has a viscosity of 1 cs.6) we require the permeability to be scaled by I/N: This cannot be readily achieved if the same solid material. So far. The scaling relationship is then applied for the fluid mass conservation equation (see (2. Also for any source term its rate must also be the same. there is no problem in the scaling. pore fluid and porosity are used.13) and (2. This reduces the permeability by N times . the frequency and wavelengths of the wave within the pores should still be within the range of laminar Darcy's flow. acceleration and time is as described in Equations (6. for most practical earthquake events. in the model. If any substitute fluid is used. silicon oil with the same density as water is used.220 VALIDATION OF PREDICTION BY CENTRIFUGE Therefore. 6. One of the solutions to this problem is therefore to use a different fluid.1. though the bulk modulus of water does vary because of the amount of air dissolved in it. However. The viscosity is chosen to be N centistokes or cs (Dow Corning Limited. thus pushing up the frequency of the dynamic events by N times.
1986) on the LeightonBuzzard sand used in the centrifuge experiment. 4noded for soil displacement (u) and 4noded for pore pressure (p). The Finite Element idealization and the boundary conditions are given in Figure 6.2. in order to avoid problems in the interpretation of the centrifuge tests.2.1 are the measurement devices.2.The computer code used was SWANDYNEI (Chan. This is done by using silicon oil (Dow Corning Limited. 1994. Also shown in Figure 6.e. Simply. If the behaviour of soil is controlled mainly by the stress state and its strain history then the centrifuge model is able to predict the generic behaviour of the prototype under earthquake conditions. Also shown in the figure are the positions of 10 pore pressure transducers. The comparison is done for as many points as possible so that an overall picture.3 CENTRIFUGE TEST OF A DYKE SIMILAR TO A PROTOTYPE RETAINING DYKE IN VENEZUELA The test represents a fully dynamic analysis with transient behaviour. which include 11 PPTs . CulliganHensley and Savvidou. The first study is performed with 44 element.1. the lefthand side one is a measured value from experiment.pore pressure transducers. However. soil parameters for the Pastor . water was chosen as the pore fluid and consolidation was therefore found to be applied more rapidly than the corresponding prototype. Other substitute fluids used include Metolose (Dewoolkar. it would be difficult at this stage to extend such tests to dynamic events. while the right hand side is the computed value. the use of another liquid will in all probability lead to a different value for surface tension as well as different drying and wetting characteristics. there is now a possibility that the bulk modulus of the fluid and its damping characteristics are different from those of water.Zienkiewicz Mark111 model are identified.1. 1996). 7 ACCs accelerometers and 3 TSTstotal stress transducers. 1988) using SSpj time stepping scheme. The physical model is a centrifuge experiment performed by Venter (1985) on the Cambridge Geotechnical Centrifuge. 4 accelerometers and the LVDT presented for displacement comparison purposes. The test was done at 78g and the material data for the Finite Element analysis are given in Table 6. However. 1991a and b) and the scaling of the capillary was reported to be modelled correctly (Hellawell.Linear Voltage Displacement Transducer. 1991 and 1993.CENTRIFUGE TEST OF A DYKE 22 1 and the above relation is retained. Although the use of the same fluid in experimental modelling of pollutant transport in a semisaturated environment has been reported (Cooke. 1 LVDT . The layout of the experiment is given in Figure 6. The principle of the centrifuge has been briefly explained in Section 6. together with appropriate mechanisms can be obtained.  . An oil reservoir is created behind the dyke to provide seepage conditions. Cooke and Mitchell. With the triaxial test results (Venter. 6. the centrifuge reproduces similar stresses and strain history as experienced by the prototype on the scaled model. i. They are listed in Table 6. Silicon oil is used so that the diffusion equation and the dynamic equation can have the same timescale under the centrifuge condition. The model is built in a strong box with a dyke lying on a sand bed. 1985) of viscosity of 80 centistokes (80 times the viscosity of the water). for semisaturated material. In the VELACS exercise. For each pair of graphs presented in this section. 1995).
Mg Mf 14 Predicted * * emin= 0 . Keso KesoIP.1% Properties of the Rigid block: Density = 2 0 0 0 k ~ m Mass per unit width: 18.222 Table 6.0.00015 sec (data point spacing of the measurements) Total number of time steps: 1024 0's for the SSpj scheme = 0.5 Bilinear interpolation for both u and p .1 x 104ms' Permeability of the sand drain gacceleration at lg = 9..81ms* gacceleration for the centrifuge test (78g) = 765. = 1.80 Permeability of the sand bulk = 2.444444 Porosity (n) Initial void ratio (e) = 0.1 x 106ms' = 2.18rn~~) Gauss Point for all cases (GaussLegendre) = 2 x 2 No incremental strain subdivision is performed Initial stress method is used for nonlinear iterations Convergence criteria for the nonlinear iterations: Residual force norm versus current external force norm (for each phase) 5 0.2kg per metre ~ Moment of Inertia per unit width: 0. 6 5 ( D ~ 100%) em.025(DR = 0%) = * As suggested in original paper (Pastor and Zienkiewicz 1986) Table 6.1 VALIDATION O F PREDICTION BY CENTRIFUGE Soil model data  Test Adjusted Po(kPa) Kevo KevoIP.alpha = 1.2 Material data for finite element analysis = 1908kgm' Bulk density (average soilpore fluid) = 980kgm' Density of the pore fluid = 1.5. O O OO = 0.092 x 109Pa Bulk modulus of the pore fluid Biot .069503 kgm2 per metre Size of time step = 0.
 I Sections through model KVV03 showing dimensions a n d transducer locations Water level / n=nuh k = 2 . Figure 6. Concrete dyke 3.1  . 0 1 1 overflow 5 Oil sea 6. 1 x 10/ k = 2 . Coarse sand drains 223 117 Baw wlth prcscr~bed earthquake movement R i g d boundry 7 Model d ~ m m r l o n rnm In Pore prcssure measurement @ PPT2626 x Acceleratmn measurement PPT2628 x(69) PPTZR4h ~(90) .CENTRIFUGE TEST OF A DYKE I Sand nuxturc 2. Shaking at bottom 0 Measured Location Computed location Computed deformation shape at the end of the earthquake (10 magnification) Figure 6.2 Finite element idealization . Retaming wall 4. I x lo Impervious u=v=O ii .
02 0.0000 0.0006 .08 0.0002 0.0.OOlO 0.12 0.0012 0.06 Figure 6. 0.3(a) Comparison with centrifuge results.0b 0.14 1 0.04 0.0004 g E .00 0.12 0.0008 0.06 0.0014  I .0.14 (ii) 0. 0.0002 .02 0.02 0.224 VALIDATION OF PREDICTION B Y CENTRIFUGE I I A .04 0. DOF: 2 OF NODE: 26 .06 0.14 (iii) 0.10 0.08 Seconds Input motion 0.0.04 0.00 0.2 5 .08 0 10 Seconds Measured 0.10 0. (44. Gamdm=2) (tip) input motion (iii) vertical displacement of the dyke (ii) & .12 Computed Vertical d~splacement the dyke of 0.
00 0. so does G.3(c) gives the comparison of pore pressure transducer D. no parametric study has been performed i. Also shown in Fig. I and J are slightly worse but the overall prediction on the pore pressure rise is very good.3(b) Comparison with centrifuge results: (44. H. E predicts lower pore pressure rise and F is quite satisfactory. The comparison is excellent.12 0.e. F. although oscillation is more pronounced.3(a)6. the predicted trend is still correct. The pore pressure transducer A gives good agreement too as shown in Figure 6. This oscillation found in the excess pore pressure could be due to the proximity of the rigid boundary condition either at the bottom of the container or near the underside of the dyke. the rising time and the shape of the rise are also predicted. The agreement of C is remarkable although the size of oscillation is larger.CENTRIFUGE TEST OF A DYKE 225 The results are given in Figures 6. More oscillation is seen in B.06 0.10 0.3(a) is the vertical displacement at the crest of the dyke. The mean value is almost the same as the measured value.3(b) although the value is slightly lower.00 0.3(h).loo00 20000 Q 20000 30000 30000 I 0 0. C. a prediction is stated as one of Class B1. Let us consider the transducers in two groups: (1) Farfield where the influence of the structure is less at locations A.i2 0 .14 Seconds (ii) Mrusr~red Cornputrd DOF I OF NODE: 14 Figure 6.04 0.08 0.08 0. K 0. the result is very good.02 0.04 0. nevertheless. G and H. EXCESS PORE PRESSURE AT POINT A DEVICE TYPE 6 P R E S U K t 1KANSDUCER D E V I C E h U M B b R 2851 DOb I OFNODE 198 EXCESS PORE PRESSURE AT POINT B DFVICF TYPE h PRESSURE TRANSDUCER DFVICF N U M B E R : 2846 40000 40000 30000 30000 6 20000 h\~~.06 0. Gamdm=2) (top) excess pore pressure at point A (bottom) excess pore pressure at point B .1'0 0. however. The input motion was taken from the accelerometer ACC1244 attached to the box. 6.02' 0. E. Besides predicting correctly the final displacement.~~+~+d@+ ~0000 10000 g loo00 0 5 01 0000 . The reader is reminded that this was the first set of soil parameters obtained directly from the soil model tester. The other graph in Figure 6.
D. for instance. I and J. This may be due to the . the dominant frequency is more akin to the input frequency (approximately 120Hz). Initially. When the same analysis is repeated using the fully explicit uw formulation (Chan et al. Similar behaviour has been found in the proximity of a retaining wall in (Dewoolkar. the value on the higher level (L and M) departs from the experimental value. 1996). As the average relative fluid acceleration is neglected. all the accelerometers show good agreement with experimental results. One possible reason for this oscillation is the proximity of the impermeable solid boundary and the fact the average relative fluid acceleration has been neglected in the up formulation.226 VALIDA TION OF PREDICTION B Y CENTRIFUGE EXCESS PORE PRESSURE AT POINT C 80000 70000 1 Devlce type pressure tranducer device numher 2848 Xi1000 Doll l of node: 98 70000) 700001 De\ I C type: ~ pressure tranducer deb ~ c number 2338 e Doll l of node 138 Figure 6. Even if there is substantial oscillation such as location H. point IPPT2628 (Figure 6. It can be observed that comparatively less oscillations are found at locations which are farfield.4) when compared with Figure 6.significantly less oscillation is observed in. The predictions of N and 0 are reasonable. the oscillations are much more pronounced and the frequency is more akin to the frequency of the structure and of a higher frequency nature (approximately 240Hz). 1995). As the soil weakens. (bottom) excess pore pressure at point D (2) Nearstructure where the influence of the structure is more pronounced at locations B. any volume change near an impermeable boundary will behave in an undrained manner and lead to a large rate of change in pore water pressure via the fluid continuity equation.3(c) Comparison with centrifuge results: (top) excess pore pressure at point C. However.3(f). for the nearstructure locations.
06 0 08 0.3(d) Comparison with centrifuge results: (top) excess pore pressure at point E.10 0 12 0.14 Sccondc . (bottom) excess pore pressure at point H .3(e) Comparison with centrifuge results: (top) excess pore pressure at point G. (bottom) excess pore pressure at point F 0 !I0 0 O? 0 01 0 00 (ii) Aleom~d 0 ox 0 10 Seconds 0 1? 0 11 Figure 6.hlemreed (11.CENTRIFUGE TEST OF A DYKE 100000 227 100000 IDewce type pressure tranducer dewce number 2847 j Dof I of node 140 I Device lype pressure rranducer dewce number 2855 I I of node: 196 10000 0 I I I I I I I 0 00 0 02 0 04 0. Figure 6.
3(g) Comparison with centrifuge results: (top) acceleration at point L.10 012 014 (ii) SECONDS .Deblce type: pressure tranducer d e v m number 30000 < zoo00 40000 2626 30000 of node.&!LUSW~ Figure 6.3(f) Comparison with centrifuge results: (top) excess pore pressure at point I.ifcnnemd 10000 I I I r y y I ' l 000 008 010 Sccmds 01: 011 O(i0 002 001 000 008 010 I 012 I 011 (ii) iornpu~d Figure 6.14 SECONDS (i) 'y 300 zoo 1 Debice type: I acceleronleter device nunihcr.06 0.loo00 0 10000 5 2 loo00 0   g I I I I I I I Oil! 001 000 .228 40000 VALIDATION OF PREDICTION BY CENTRIFUGE Device type pressure tranducer d e v m number: 2628 40000 1 Dof I of node: 17 . (bottom) acceleration at point M .23 $ 20000 : .06 0 08 0.10 0 I? 0.04 wed 0. I39 0'00 0 02 0.1928 300 300 7 Dof I of node. 1583 3004 Duf I o f node: 142 002 004 0. (bottom) excess pore pressure at point J Dcwcc type: I accelcromctcr dcvlcc numbcr.08 0.
0 Oh .oO { .14 ii) Wwwwd (b. 0.10 012 0. 103 Figure 6.3(h) Comparison with centrifuge results: (top) acceleration at point N.02 .0 T m e (seconds) ib) Figure 6.11 (102 (104 006 008 SCCONDS 0.melcromare d c \ m numbrr. 0 .0 . (b) with 96 elements .100 5 zoo 300 0.16 0. 1938 Dof I of node. 0O X .00 300 O.mrrrd Dc\lcu tvne I .12 Excess prcssurc (kPa) 60.04 . (bottom) acceleration at point 0 Excess plessure (kPe) 20..04 0.0 40.00 0. I I I 1 0. O 12 .08 Time (seconds) 0. 0. 0.CENTRIFUGE TEST OF A DYKE 300 200 229 p I00 .4 Numerical results of excess pore pressure at point I (PPT2628) using fully explicit GLADYS2E with 33 element: (a) with 44 elements. 0 10 .
the other side loose soil layers in laminar box soil layers in rigid box Model No. Massachusetts Institute of Technology. As the soil weakens. 1994).2Sloping Model No. and Rensselaer Polytechnic Institute). 1994) and these were not made available to the 'centrifuge experiment participants'. California Institute of Technology. Despite a number of shortcomings. Cambridge University. the mean effective stress is not reduced as much as in the physical experiment. 6.3A 0 layered loose sand in laminar box loose sand layer in laminar box sand layer one side dense.. Nevertheless.230 VALIDA TION OF PREDICTION BY CENTRIFUGE amount of shear wave energy being transmitted from the bottom to the top.3) when apparently the centrifuge tests were first commenced by seven universities (University of California. It is instructive for the readers to note that. USA.4 T H E VELACS PROJECT Although many verification exercises have been performed by the authors (Chan. these results represent a Class B1 prediction and a set of more convincing Class A predictions are given in the next and subsequent sections. due to the oscillatory nature of the pore water pressure. 1988) and (Zienkiewicz et NI.5 taken from Arulanandan and Scott (1993)) were selected for the verification: Model No. 1992 (Table 6. less shear wave should be transmitted. This double blind policy was introduced to minimize possible 'cheating' and thus enhance the credibility of the results. The numerical predictions were kept in sealed condition by a third party (Thompson and Lambe. Such 'Class A' predictions were submitted by twenty 'predicting participants' by 30th September.4aStratified Model No. However. all numerical predictors used computer codes based directly or indirectly on the Biot theory and approximation form introduced in this book (Smith. However. 1993) funded by the National Science Foundation. A numerical prediction of several postulated tests was requested from 'predicting participants' before the experiments are performed and results obtained for 'centrifuge experiment participants'. 1990) and others using reported centrifuge resultsa more systematic study became recently possible through the project VELACS (VErification of Liquefaction Analysis by Centrifuge StudiesArulanandan and Scott. 1 which represents a level soil layer. Boulder. the results represent an excellent comparison with the experimental results accounting for possible experimental errors. Princeton University.4bStratified . except for the simpliest model No. Nine centrifugal models (see Figure 6.1Horizontally Model No. University of Colorado. Davis. Therefore the shear modulus is not reduced adequately leading to the excessive transfer of shear wave energy. some of the specified centrifuge tests could not be carried outand additional computations ('Class B') mostly because of the difference in the prescribed and actual input earthquake motion were requestedwithout however supplying other experimental results.
THE VELACS PROJECT 231 I Model No. Figure 6. 4 l Laminar box I n Laminar box n * II ~ o d e No. t n 3r Rig~d Box . 1993) Project .5 Centrifuge model configurations for class A: PredictionsVELACS (Arulanandan and Scott. Davis and The California Institute of Technolog?. 1 l Rigid Box n 7. I l Internurional Conference on the Verification of N~inlerical Procedure for the Annlvsis ($Soil Liquefaction Problems 17 . t ~ o d eNo.6m silt 1 ~ o d eNo. . 7 Laminar box 1 Laminar box A Sand: Dr = 40% E n d Dr = 40% 2" 1 I( ~4 1 Model No.20 October 1993 Orpnized by The University of Californaia. 1 Sand: Dr = 60% 1 Rigid Box I 1 Model No. ' n ~ o d eNo. 4 l Rigid Box 1 Sand: Dr = 60%] 1 Model No..
232 VALIDATION OF PREDICTION BY CENTRIFUGE m Q x x X X X X X X X PI X m m 0 < x x x x d m Q x x x x x X X X X G x x x x x x  .
The appropriate boundary conditions are then applied at the boundaries of the model. 1993ah). together with the senior author. Time required for the subsequent dynamic and consolidation analysis are also taken into consideration so that a reasonable mesh is chosen.THE VELACS PROJECT 0 233 Model No. 2.1 General analysing procedure This analysis procedure is applicable to all the predictions performed by the authors: The information about the centrifuge model is gone through in detail and key data noted. A selection of these will be presented in the subsequent sections in this chapter with a brief description of the experimental setup and conlparisons with numerical results on the same scale. A finite element mesh is generated using a preprocessor. presented four predictions for model Nos 1. The SWANSEA group led by the senior author of this book.4.12A submerged embankment in rigid box submerged sand embankment with slit core gravity quaywall with sand backfill structure embedded in stratified soil layers Most of the tests have been performed at more than one centrifuge for centrifuge validation purpose and considerable scatter of results can be found between the centrifuge results.3. But before going into the detail of the prediction..4a.3). presented eight predictions for model Nos. . Lastly. 1 . Zienkiewicz. . 1993ag) and another of the authors.4.. 1.11A Model No. Most of the predictive results can be classified as good or excellent. Chan. Tied nodes are used to model the laminar box behaviour (see Section 6. SWANDYNEI1 (implicit up with full saturation) and MuDIAN (implicit uU with fullsaturation.1994) were also involved in the process of overviewing numerical predictions. 1993ad). 4a. 1994) and Shiomi (Sture ct (I/. is prescribed at the fluid nodes concerned and the pressure on the solid phase is also applied.7A Model No. 6. 3. to be introduced in Chapter 7) based on their work in University College of Swansea. The hydrostatic pressure..6A Model No. Three of the authors led three different groups of predictions using computer codes SWANDYNE4 (implicit up with partial saturation). 2. together with one of the authors. Chan (Chan rt ul. 6. Shiomi. which is assumed to be constant throughout the analysis. Pastor. the following sections are devoted to the analysing procedure using SWANDYNE11.. The second author. 11 and 12 (Shiomi et (I/. presented seven predictions for model Nos. 7 and 11 (Zienkiewicz et d . 4b. 7 and 11 (Chan et ul.2. 4b.
4. Rayleigh damping of (minimum) 5'%. pore pressure and displacement are used in the program. a gradual change in time step is used to avoid numerical instability. The dynamic analyses were performed using a Generalized Newmark (Katona and Zienkiewicz. 1985) scheme with nonlinear iterations using initial linear elastic tangential global matrix. The choice of the time step depends on the number of the stations in the earthquake input and the frequency of the input. If it is not.234 VALIDATION OF PREDICTION BY CENTRIFUGE The appropriate permeability and gravitational acceleration are then included. In order to avoid tensile stress and high stress ratio. (xii) The consolidation then follows the dynamic analysis. a new static initial analysis is performed with modified parameters to obtain equilibrium. permeability and other geometric properties were kept constant during the analyses. (xiii) The results are first plotted using a simple post. iteration schemes etc. (viii) A noearthquake dynamic run is then performed to check if the initial stress state is in the correct equilibrium condition. until a reasonable and numerically stable result is obtained. The parameters used are described in Section 6.processing program to check its validity. If the result does not seem reasonable. Usually a larger time step is used for the consolidation analysis. (xiv) Various plots are then performed for the final report. Mohr Coulomb elastoperfectly plastic model is used for the initial analysis with a reduced frictional angle of 25. The time step used is usually equal to a simple multiple of the earthquake spacing. Then a nonlinear analysis is performed for the earthquake stage with the supplied horizontal and vertical earthquake with proper scaling..e.4 is assumed. Since total quantities e. The models are modelled at the model scale so the appropriate acceleration level is the centrifugal acceleration imposed. .g. When the initial stress state is acceptable. A KO value of 0.4. The constitutive model used is the PastorZienkiewicz (1986) mark111 model. A static analysis was performed to determine the initial stress state of the model. postprocessing is required to obtain the excess pore water pressure and relative displacement required by the specification. The full dynamic equation is used for the consolidation stage of the analysis with the appropriate mass matrix. The void ratio i. (xi) The earthquake phase of the analysis is then plotted to check for any anomaly. the dynamic analysis is repeated with another set of numerical parameters. (vii) The output of the static analysis is considered carefully to check if the initial pore pressure distribution is reasonable and also if the stress state is acceptable. a linear elastic analysis is performed to note the basic behaviour of the finite element mesh.is applied at 100Hz which is the dominant frequency in the earthquakelike motion input.
2 Description of the precise method of determination of each coefficient in the numerical model The determination of each coefficient of the PastorZienkiewicz mark I11 model follows the procedure outlined in Section 5. Mg corresponds to the maximum slope obtained by this method. this value is usually taken as 0. In this exercise.4.THE VELACS PROJECT 235 (xv) Other postprocessing e. In the following sections. the way to identify each of the parameters required by the model is illustrated. One undrained monotonic and cyclic test is taken from each of the loose sands (Dr = 40U%). respectively. As drained monotonic. 6. it can be obtained by matching the critical stress ratio that the behaviour of the soil changes from contractive to dilative behaviour in the case of dense sand. In this section. In this exercise. Mf (dimensionless): can be determinated by matching the shape of the stress path in the q versusp' plot in an undrained triaxial test. the critical stress ratio is used. The tests should be done with samples having relative density around the intended relative density. Alternatively.45 and it is also used in this exercise. each parameter will be discussed in turn: Mg (dimensionless): can be estimated from the graph plotting stress ratio versus the shear strain or axial strain. excess pore water pressure ratio. a g (dimensionless): can be obtained from the slope of the graph between the dilatancy and stress ratio over the Mg graph. M can serve as a good starting point for the evaluation of its value. (1992a and 1992b). the undrained test is more relevant. During earthquake and other rapid loading. undrained monotonic and undrained cyclic tests are the most widely available tests in common engineering applications. dense sand (Dr = 60%) and silt experimental data sets. of (dimensionless): is usually taken to be the same as a g so that the loading locus and plastic potential are having the same shape. Mg can also be obtained from the drained test using the intercept of dilatancy versus stress ratio plot. These results are produced using a soil model subroutine for DIANASWANDYNE I1 interfaced with a soil model testing program SM2D and the experimental results are also plotted on the same graph for the monotonic test. the stress ratio plot was used. they are chosen in the parametric determination process.5 of (Chan. The comparison of the constitutive model and the physical undrained triaxial tests has been given in Chan er al. The ~ value of D R . 1988) and is being reproduced in this section. spectral analysis and response spectrum are calculated for reporting purposes.g. KevOc (dimensions of stress): represents the value of the bulk modulus at the mean effective stressplO. It can be obtained by matching the initial slope of the . Mg is approximately equal to the maximum value of stress ratio that the test reaches. However. It can also be estimated from the q versus p' plot with a tangent drawn from the origin to the residual stress path in an undrained triaxial test.
236 VALIDATION OF PREDICTION BY CENTRIFUGE mean effective stress p' or pore pressure versus axial strain plot in an undrained test. (x) Huo (dimensionless): is determined by matching the initial slope of the first unloading curve. this was done so that the end point in the predicted curves stayed close to the experimental data. The interface between structure and soil is assumed to be perfectly bonded. this was done so that the end point in the predicted curves stayed close to the experimental data.y (dimensionless): is determined by matching the rate of change of the ~ ~ slope of the first unloading curve or by matching the number of cycles in a series of loading and unloading. The second method is used in this exercise. (vii) $o (dimensionless): is usually taken as 4.4..3 Modelling of the laminar box The way that the boundary conditions were incorporated in the numerical method has been given in the relevant prediction papers. (viii) fiI (dimensionless): is usually taken as 0. (vi) KesOc (dimensions of stress): represents the value of three times the shear modulus at the mean effective stress p'O. Only rigid block and four noded linear isoparametric elements for both the displacement (u) and pore pressure (p) are used in the analyses. No change of boundary condition is made during the analyses. The second method is used in this exercise. (xiii) pb (dimensions of stress): is the initial mean effective stress of the undrained triaxial test. (ix) Ho (dimensionless): is determined by fitting the curves in p' or q versus the axial strain plot. The laminar box (Hushmand et al. In the VELACS exercise. Its value can be adjusted so that a better match of the curve of pore pressure versus axial strain can be obtained. (xi) . (xii) (dimensionless): is determined by matching the slope of the first reloading curve or by matching the number of cycles in a series of loading and unloading. 1988) is treated with tied node facility. In the VELACS exercise.2 and this value is taken here. It can be found by matching the shape of the q versus p' plot for undrained tests also. Its value can be adjusted so that a better match of the curve of q versus axial strain can be obtained.2 and this value is taken here. The horizontal and vertical nodal displacements at the two ends of the soil are restrained to have the same value. It can be obtained by matching the initial slope of deviatoric stress q versus axial strain plot in an undrained test. . 6.
2 pk = 4kPa. The parameters obtained are as follows: Mg = 1. The parameters obtained are as follows: Mg = 1.45 KevOc = 770kPa KesOc = 1155kPa.1 were carried out by three universities.15. = 2. Stadler et a1 (1993).COMPARISON WITH THE VELACS 237 6. 6.30.2.5. 10m (in prototype scale) thick. pl = 0.6(a). KevOc = 400kPa.2 p:.is given below: (9 Loose sand: (Dr = 40%) Experiment CIUC4051 was used. was mainly subjected to horizontal base motion. KesOc = 1520kPa. RPI is the primary experimenter and UC Davis and the University of Colorado conducted duplicate tests.45. a f = a g = 0.1 Description of the models Model No. Model No.4 Parameters identified for the PastorZienkiewicz Mark I11 model There are quite a number of parameters in the PastorZienkiewicz mark 111 model which require definition. in a laminar box. P1 = 0. Ho = 900. in a laminar box was subjected to base shaking.2.4. = 2. = 4kPa HO = 600 HUO= 4000 kPa Y H = 2yDM= 0. The test was instrumented as shown in . Ishihara 1994) A watersaturated uniform layer of loose sand (Dr=40%). the elastic modulus is proportional to the mean effective stress Po = 4.32. ~ (ii) Dense sand: (Dr = 60%) Experiment CIUC6012 was used. af = ag= 0.5 COMPARISON WITH THE VELACS CENTRIFUGE EXPERIMENT 6.03 af = cug = 0. 1992) which provided the standard soil model test results for the numerical predictors.15 Mf = 1. The parameters obtained are as follows: Mg = 1. KesOc = 2600kPa.50. Mf = 1.2 PI = 0. Bo = 4. The test was instrumented as shown in Figure 6. Po = 4. the elastic modulus is proportional to the mean effective stress. ?DM = 4 (iii) Silt: Experiment CIUCBS13 was used. Huo = 40000kPa. Ho = 750. YH. The permeability of silt is also calculated for this level of mean effective stress.. Mf = 0.. 3 (Scott et a1 1993) A watersaturated layer of sand deposited at l l m (in prototype scale) thick.2 pk = 4kPa. TH. KevOc = 2000kPa. Three CIUC (Isotropically Consolidated followed by Undrained Compression test) experimental results starting from 40kPa were chosen to identify the parameters. Centrifuge test resuts for model No. the elastic modulus is proportional to the mean effective stress. The experimental results were taken from (Arulmoli et al.yDM= 8. Huo = 100000kPa.45. The set used in the Class A prediction for (Chan et a1 1993a . 1 Taboada and Dobry (1993). The 40kPa ones were chosen because they are close to the mean effective stress value at the middle of the centrifuge model.
7(a). laboratory data were available only for 40% and 60% relative 1 I Instruments for model No.6 0.238 VALIDATION OF PREDICTION BY CENTRIFUGE Figure 6.0 k E  2 0.6 0.2 0.4 0. Although the model contained sand deposited at both 40% and 70% relative densities.4 5 5 0.E mesh RPI Test.6 2 4 6 8 12 14 16 18 20 0 2 4 6 (Second) T~me Expcrirnental and predicted horizontal ~ c e l c r a t i o n H 3 A 10 8 10 12 (Second) Time 14 16 18 20 (Second) Time (Second) Time .2 1 ID1 (Sccond) Time 11 0 (Second) Tune 0. I M LVOTI LVOTS Concentrated masses (aluminum alloy rectangular rings) F.2 0 4 0 6 0 3 0.2 5 0.2 00 5 0.4 0.
4 .2 0.s 0.2 0. the material parameters for relative density 60% is used.6 0.0 0.COMPARISON WITH T H E VELACS 239 densities.g E + C 0.2 0. 0.6 0 2 4 6 0 y U (Second) Time 8 10 12 14 16 18 20 (Second) Time Expennental and Predicted horizontal acceleration AH5 (KNM:) (KNM') 2 1003 Z 80'2 a 60E 40203 0 0 RPITest 2 P5 2 100 $ 802 60 Prediction P5 k g X I 4020O < 2 1 I 8 r/] 0 10 20 30 40 50 60 70 80 90 '00 l b 20 30 40 50 60 70 80 90 100 (Second) (Second) Ti]% Ti'"e Experimentel and predicted excess porepressure (KNM~I (KNM') 2 1100 2 100 2 80 Prediction RPITest 2 802 60 60 2 P6 P6 a 40 40 m 1 W l l 5 % m 0 k g X  20 O W W 0 (KNM') 0 10 20 30 40 50 60 70 80 90 100 10 20 30 40 50 60 70 80 90 100 (Second) (Second) Time Time Experimental and predicted excess porepressure P6 (KNM') RPITest 2 P7 m I I I I I I I I I 2 100 80 I 2 60  i L Prediction g 3 20 0 X I W . The predictors were expected to infer properties of sand at 70%)relative density based on the properties at 40'% and 60% relative densities.2 2 + 0. The centrifuge results for this model were carried out by CIT as primary experimenter and UC Davis and RPI as duplicate experimenters.0 0. In our case.4 c . I I I I l I 1 I 0 10 20 30 40 50 60 70 80 90 I00 (Second) Time g 20 O 0 10 20 30 40 50 60 70 80 90 100 (Second) Time Experimental and prcdicted excess porepressure P7 .4 414 noded elements g 0.4 0.
The centrifuge test of this model was carried out by Cambridge University only and was instrumented as shown in Figure 6. 0. 11 (Madabhushi and Ferg 1994) A soil and water retaining wall was subjected to base shaking. predictions of the SWANDYNE program compared well with experiment. No repeat test was conducted. 1: instrumentation.2 Comparison of experiment and prediction Overall. 6. finite element mesh and comparison of experimental and predicted results . The use of fully coupled equations for the soilpore fluid interaction and of a simple soil model based on the generalized plasticity for the soil skeleton. both introduced in detail in this volume. This rigorous analysis procedure promises to be a reliable tool for practical problems in engineering.8(a).6 VELACS Centrifuge Model No. the following problems have to be addressed. However. A surcharge was added on the backfill of the retaining wall.240 VALIDATION OF PREDICTION BY CENTRIFUGE Model No.2 I I I I I I I I I I I 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 0 1 0 0 0 l O P 3 0 4 1 3 W W 7 C I ~ W l 0 0 (second) (second) Time Experiment$ and predicted vertid displaccmmt LVDTl Figure 6.1 0.5. The relative density of the sand in this model was approximately 60%. The close agreement of the predicted pore pressures and displacements with those measured in centrifuge tests affirms the reliability of the computation procedure in the SWANDYNE program when used with carefully calibrated material properties and model parameters. RPITcst 1 Prediction p C. forms a consistent and powerful prediction procedure.
it is obvious that the experimental results of the surface acceleration is not necessarily reliable after soil liquefaction. 3 (KNIM') (KNIM 2 F. This can be improved by using 4 4 noded elements. We also have recognized that the numerical result of the surface acceleration obtained by using mixed 84 noded elements. E. probably gives a much higher peak value of acceleration. finite element mesh and comparison of experimental and predicted results .COMPARISON WITH T H E VELACS 241 (a) Although the predicted horizontal acceleration on the surface of the soil layer is much higher than the experimental results. 3: instrumentation. mesh (second) TIME (second) TlME " h (KNIM ) 2 0 50 100 150 200 250 (second) Tme Prediction M 20 Y 0 0 50 100 I50 (second) Tme 200 250 Experimental and ~redicted excess porepressure P8 (loose sand) Experimental and predicted excess porepressure P9 (dense sand) Figure 6. The prediction also showed that the peak acceleration on the surface decreased Instruments for model No.7 VELACS Centrifuge Model No. There is no large difference in the acceleration at the middle points between prediction and experiment.
) VELACS Centrifuge Model No.242 .f " !lTI E k20 CITTest i: 3 0 0 50 100 I50 200 250 g 3 0 S O I00 I50 200 250 (KNlM ) (second) Time (second) Ttme (KN M ) lKN. 3: instrumentation. finite element mesh and comparison of experimental and predicted results .h4) (second) Time IKNIM ) ":TI ti 20 9 0 0 50 100 I50 200 250 1 (second) T~me . VALIDATION OF PREDICTION BY CENTRIFUGE &. .. 200 250 150 200 250 w " 0 50 I00 I50 (second) Time Expenmental and predicted excess porepressure PB (loose sand) (sccond) Tme Ex~nmentaland predtctcd cxcess pareprcssurc P9 (dense sand) 0 50 100 150 200 250 0 50 100 (m) (second) Time (m) (second) Time (second) Time Prediction 4.3 1 5 44 0 50 100 150 200 250 0 50 100 I50 200 250 (second) Time Experimental and predicted vertical displacement 1 5 (loose sand) (second) Time Exper~mentaland pred~ctedvert~caldisplacement L6 (dense sand) Figure 6.7 (cont.
Similar phenomena can be found in the centrifuge tests and earthquake history. 11: instrumentation.2 0.0 0 5 10 LVDT 1 15 20 25 30 Prsdiction 35 40 Figure 6.8 VELACS Centrifuge Model No.5 0.0. finite element mesh and comparison of experimental and predicted results .COMPARISON WITH T H E VELACS 243 gradually when liquefaction occurred.3 2 0.0.. In model No.7. 1. the differences were not large.0) I (1.Uoundaty after sarthquate t s s ~ c 0..VDTI) horizontal n 0. The experimental 0 (m) 2 4 6 8 10 I Initial boundary . however. (b) The predicted time histories of porepressure agreed closely with those recorded in the centrifuge tests..1 2. .4 localion  (25.1 0. the experimental results showed slightly faster pressure dissipation.
The predicted values of surface settlement in model No. although the prediction of surface settlement in the loose sand agreed closely with that measured in the centrifuge tests. 1 were lower than those recorded in the centrifuge test. (c) The prediction of displacement showed similarities to those measured during centrifuge experiment. (b) Finite element mesh with slip element. in heavy line (Dewoolkar. In model No.75" AC4 ELEVATION Figure 6. 3 by RPI and CIT gave obviously different peak values of porepressure and made comparison very difficult.9 (a)Schematic model configuration of test M M D l (Dewoolkar 1996). 3. 1996) .244 VALIDATION OF PREDICTION BY CENTRIFUGE results of model No. The predicted postearthquake consolidation was consistent with the values measured in the centrifuge test. the numerical analysis 0. The horizontal displacement of the retaining wall in model No. but the difference was not too marked. 11 was well predicted.
1996). .1 0 0. Numerical .1 1.9 (b). Although the numerical modelling was performed at MMDI (SWANDYNE 11) Mesh2 MMDI (SWANDYNE 11) Mesh2 0.10. 1996). 6. It is of interest to carry out further research into this phenomenon. a consistent stress state cannot be found as the initial condition for the dynamic analyses. More research is also needed on the modelling and calibration of data for the behaviour of dense and and silt under cyclic loading. 1996) . horizontal accelerations within the soil and dynamic bending strains in the wall can be found in Figure 6.5 0.4 Deflection (inch) (a) 0.6 10 0 10 20 Total lateral earth pressure (psi) ('J) .6 CENTRIFUGE TEST OF A RETAINING WALL The schematic model configuration of test MMDl in Figure 6. (b) Total lateral earth pressure on the wall (Dewoolkar. the program SWANDYNE introduced by the authors was used.9 (a) has been taken from (Dewoolkar. The experimental results are striffer than the theoretical and numerical results because of the stiffening of the retaining wall by the weld at the root of it. Without this layer of slip element. Selected results for the horizontal accelerations at the tip of the wall. A layer of slip element can be found behind the retaining wall in Figure 6.12. despite a later repeat of the experiments which showed that different results could also be obtained (Dobry.3 0. The same is true for the horizontal displacement at the tip of the wall and longterm excess pore pressure trace in Figure 6.1 0. The mechanism involves a failed dense sand wedge moving into the liquefied loose sand due to the lack of support. Almost all numerical predictors failed to achieve this qualitative difference. : Theoretical + : Experimental Figure 6.CENTRIFUGE T E S T OF A RETAINING W A L L 245 did not predict more settlement in dense sand than in the loose sand which was recorded in all experiments for this model. Good agreement with theoretical and experimental results has been found for the static analysis in Figure 6. 1996).10 (a) Static deflection of the wall (Dewoolkar. In all comparisons. All of them showed good agreement with experimental results.2 0.
3 0.1 0... . .5 Figure 6.. 1996): (a) Horizontal wall accelerations at the top. .) . ... .2 0.. ... .4 0..4. . .. .AC9 . ' . :..... . Only the amount of damping has been varied to investigate the effect of different levels of damping.. ....... .  4 0 1 .4 0. . ..:...5 Time (s) MMDI SWANDYNE 11 . .. .. . . . :.. .2 0. .. . ... .. .. . all the material parameters used are derived from the VELACS exercise as given in Section 6. 40 0 1 0. .. . .. (b) Horizontal soil accelerations. . . I (b) MMDI (DI) (model scale) : AC8 .2 03 T~me (s) 04 05 0" 0" c o 0. .:. . the up formulation derived in Chapter 3 has been used to compare with various centrifuge tests performed on the Cambridge Geotechnical Centrifuge ... .... .4 0.. .3 Time (s) 0. 6.4.. .....3 Time (s) 0. 0 01 02 03 04 05 0 0.11 Comparison with centrifuge test MMDl (Dewoolkar.2 0.. .1 0.5 T~me (s) (c) Time (s) MMDI SWANDYNE I1 MMDI (DI) (model scale) 2 500 g . (c) Dynamic bending strains of the wall different stages of the centrifuge.. . .. . ....7 CONCL USZONS In this chapter.1 0. .. .. ..500 0 Ol 0. . .. .246 (a) VALIDATION OF PREDICTION BY CENTRIFUGE MMDI (DI) (model scale) MMDI SWANDYNE 11 . .
and Fruth L.1 0.3 Time (s) 0.5 Time (s) Figure 6. (eds. F. Balkema. Rotterdam. and Scott R. H. Irvine.2 0.12 Comparison with centrifuge test MMDI (Dewoolkar. 1924.D. Rotterdam. UC Davis. project No.5 4 0 05 1 15 2 2 5 Tune (s) 3 35 4 0 0. Wales. Arulanandan K . U C Davis. Ph. and Zeng X. C. A.for the Ana1~vsi. Muraleetharan M. (1988) A unified finite element solution to static and dynamic geomechanics problems. (eds. 900562. Earth Technology Corporation. Rotterdam.5 1 1.) Arulanandan K . qf and Scott R.s Soil Liquefimion Problems. 1720 Oct. (1992) VELACS Laboratory testing program . In the next chapter.. we are going to show numerical predictions of practical examples and the use of the numerical procedure in design. ArulmoIi K. REFERENCES Arulanandan K. and Zeng X. 1 in Verification o f Numerical Procerlures. Balkema. S. 5 1 3 3. Dissertation. Chan A. California. 295300.REFERENCES (a) 247 MMDI (DI) (model scale) MMDI SWANDYNE I I  0.2 1 0 0. 1720 Oct. A MMDl (Dl) (model scale) 2 I0 2 2 2 a 5 a 3 0 0 . F. 1996): (a) Dynamic wall deflections.4 0.r ?fSoil LiqutIfaction Problems.) Arulanandan K . M. 1 in Verificution o f Numerical Proceclures. (1993) Proceedings of VELACS symposium. A. A. Arulanandan K. Eds. A. . (b) Longterm excess pore pressure and the VELACS project.5 2 2. A. Balkema.Soil data report. and Scott R. .. .5 T ~ m (s) e (a) MMDI SWANDYNE I I PP6 (b) = 15 a . Very good and excellent agreements have been obtained thus validating the formulation and the computer code for various types of analysis under saturated condition. A. (1993a) Experimental results of Model No. H. University College of Swansea..for the Ana1ysi. F. ( I 993a) Experimental results of Model No.
Famiyesin 0. (1995) User Manual for GLADYS2E. December.for the unulysis of soil liquefuction yrobletns. University of Birmingham. C. (1992b) Report No. and Muir Wood D . Glasgow University. Glasgow University. C. H. C. 0 . 3 in Ver~ficcirionof Nutnericul Procedures . Balkema. H. and Scott R. Chan A. Chan A. Balkema.) Arulanandan K. Cooke B. F. A.ul Procedures . and Scott R. (eds. (1994) Overview of the Numerical Predictions for VELACS Model No. Famiyesin 0 . A. Kingston. Chan A. Rotterdam. 7. A. Glasgow. and Scott R. Rotterdam.) Arulanandan K.. 90993 1. Famiyesin 0 . and Muir Wood D. Famiyesin 0 . .. Cooke B. U C Davis. and Mitchell R. and Muir Wood D. A. Queen's University. 1720 Oct. 196263. and Muir Wood D. 28. Balkema. 489510. H. 0 .. Chan A. Cunudiun Geotecl~nical Journal. 0 . U C Davis. H. London. F... (1991a) Evuluution of' Contcrn~incrntTrunsport in Purtiully Srrtur n r d Soil in Centrifuge '91. Cooke B. (ed..General Description. C. A. and Muir Wood D. (eds. and Scott R ..) Arulanandan K..uI blenu. U C Davis. U C Davis.) Arulanandan K .s.. Balkema. and Mitchell R . (eds. Famiyesin 0 . C. C. H.^ . and Ito K. 71 1720. CEGE92230: Numerical Simulation Report for the VELACS Project . Glasgow. Famiyesin 0 . (1993) Physical modelling of contaminant transport in the unsaturated zone in Geotecl~nicul Munugement of Waste and Contaminution. and Scott R. H. 1. 1720 Oct. 3 in Verifi'cntion ofArun~ericcrl Proceduresfor tile Anrrl~vsis oJ'Soil Liquejkction Problems. H. Siddharthan R. and Scott R. Rotterdam. Canada.. A. 4a in Vmficution of' N~in~eric. H. 829833. F. In Ver~ficntion of Nun~c~riccrl Procedures . the of (eds. J.. 1. 1720 Oct. Cooke B.. (19938) Numerical Prediction for Model No. U C Davis. Famiyesin 0. R. 4b in Ver~ficution Nurnericnl Proceclures f i ~ rtlw Ancrlysis q f Soil Liquclfirction Proof' blenis.) Arulanandan K. 0 . 0.248 VALIDATION OF PREDICTION B Y CENTRIFUGE Chan A.is @Soil LiquLlfirction Problems. Chapter 8. 87108. UC Davis. and Scott R. Rotterdam.) Arulanandan K. A. (1992a) Report No. Famiyesin 0 . Chan A. A. U C Davis. F. Chan A . 1. F. (eds. F.for the Anrrl~~sis Soil of Liquefnction Prohlenis. A.. Balkema.. 0 . 343362. CEGE92231: Numerical Simulation Report for the VELACS ProjectClass A prediction of RPI model. and Muir Wood D. 623630. H. A. Blackie Academic & Professional. (1993a) Numerical Prediction for Model No. N.. School of Civil Engineering. 1. C. 1720 Oct.for the Analj. Ph. Birmingham. 1. Chan A. 1. (1993b) Numerical Prediction for Model No.s Soil Liqwfirction Proof' No. Rotterdam. Dissertation. Department of Civil Engineering. and Scott R. 1. Balkema. 1720 Oct. 835850. (1991) Centrifuge modelling of flow and contaiminant transport through partially saturated soils.. 0. (eds. No.fi. A. J. 1720 Oct. H. Ontario. Famiyesin 0. 0 .. (1993d) Numerical Prediction for Model procedure.r Anulysi.. (19930 Numerical Prediction for Model No. 0. 0 . Rotterdam.r. F.. (eds. ( 1 9 9 3 ~Numerical Prediction for Model ) No.D. R. C. Balkema. in Verification of Nun~eric. A. C.fbr the Ancr1j~si. Balkema. Chan A.~ Soil Liqu+ction Problems. 6. (1993e) Numerical Prediction for Model No. (1995) Environmental geomechanics and transport process in Geotechniccrl Centrifuge Technology. 3 in firificution of Nun~ericrilProcec/~rrc~. and Muir Wood D. Rotterdam.s~forthe Anulysis if Soil Liqurfirction Problems. C. J. Rotterdam. 1720 Oct. (1991b) Physical Modelling of dissolved contaiminant transport in an unsaturated sand. and Muir Wood D. 1720 Oct. Chan A. Famiyesin 0 . and Muir Wood D . and Muir Wood D ..) Arulanandan K. CulliganHensley P.) Taylor R. Balkema. Department of Civil Engineering. A. C.. 1 in Verification qf Numericrrl Procetlure. Rotterdam. 11.) Arulanandan K . and Savvidou C. U C Davis. H. Chan A. 14431456. A. (eds. F.
1. and Zienkiewicz 0 . Madabhushi S.. and Scott R.. UC Davis.fur the Anrilj~sis of Soil LicpyfucNo. 15931 606. 2. in Verifi'crrtion of' N~mic~icnl tion problem. ( 1 9 9 3 ~Numerical Prediction for Model No. (1994) An overview of numerical procedures used in the VELACS project. Ph. 2.. E. No. G. Rolla. Balkema. in Vcvific~tion c~f'Numc~ricc11 Proc~er/ures. C... 1720 Oct. A. C. Dow Corning Data Sheet.) Arulanandan K . U C Davis.. Balkema. Hushmand B. N. 1720 Oct... (eds. Shigeno Y. lshihara K . (1993b) Numerical Prediction for Model No. Cambridge. (eds. (1993a) Numerical Prediction for Model No. Shiomi T. in Verifierition qfNunirricu1 Procer/ure. Int. and Crouse C. No. F. 1. in Verificution qf'Nuniericr11Procerlure.. MO.. Cambridge University Engineering Department. April. Schofield A. Ghent. Smith I.fi. and Schofield A. C. 2..) Arulanandan K. 12. Shigeno Y.furtlie Anrrlysis of'Soil Liqurfuction Problenls. Dissertation. UC Davis. Rotterdam. 977986. (1988) Centrifuge Liquefaction tests in a laminar 38. 4558. M. (eds. 3.nriniic. 391394.. 1. Itit. 227268. M.. (1996) PhD Thesis. (1994) Modelling transport processes in soil due to hydraulic. 30.fOr Anulj. Schofield A. Rotterdam. F. ) Shiomi T. USA. 253262. 435462. A. and Zienkiewicz 0 .. density and electrical gradients. and Zeng X. and Zienkiewicz 0.) Arulanandan K . F.(eds. Conf: on Recent Advclncm iri Geotechnicrrl Engineering r~ndSoil Dj. and Scott R. (1985) A unified set of single step algorithms Part 3: The Betam method. (1981) Dynamic and Earthquake geotechnical centrifuge modelling. (1994) An analysis of the seismic behaviour of quay walls in Veri/icution of'Numericc11 Procedurc~s.. 1720 Oct. U C Davis. box. (eds.CUEDIDsoilslTR24.s. (1994) Review of the predictions for model 1 in the VELACS program in Verificution of Nurnc~riculProcedures. Hushmand B. University of Cambridge. (eds. a generalisation of the Newmark scheme. Scott R . . 1. U C Davis. Dobry (1996) Private communication Dow Corning Limited (1985) Dow Corning 200 Fluid in Information about Silicone Fluids Bulletin: 22069D01. 2. F. Merh. 108 11 100. 13531368. A. Shigeno Y. (1980) Cambridge Geotechnical Centrifuge Operations . 3 primary test description and test results in Verificrition qf Nuniericul Prucmiurrs jbr rlie Anc11y. (eds.otec/iriiq~w. 13211338.) ji. J. 3. F. P. Shigeno Y. University of Colorado. 1720 Oct. F. Num. Schofield A. Scott R . Balkema. CI. Rotterdam. 10671074. 131150. 38.~ the Anulysis o f Soil Licpefirction Problems. Eng. F. 1720 Oct. (1988) Centrifuge modelling of sand embankments and islands in earthquakes. 13451 359. and Scott R.20th Rankine Lecture. and Scott R. (1986) A generalised plasticity hierarchical model for sand under monotonic and cyclic loading. and Scott R. and Scott R.) Arulanandan K . A. C. A. 1720 Oct. 1720 Oct. A. and Zienkiewicz 0 . (3993d) Numerical Prediction for Model Proc~edurcs.. (1992) Design and performance of an equivalentshearbeam (ESB) container for earthquake centrifuge modelling . Hellawell E. F. Shiomi T.sis qf' Soil Liqu</action Prohlerns. Pastor M.) Arulanandan K . and Scott R. No. G .s. I I. 213219. GPotechniqzre. UC Davis. C.) rlie Arulanandan K . NUMOC II. 21.r Arulanandan K . C. and Scott R.~. UC Davis. Shiomi T.REFERENCES 249 Dewoolkar M .) Arulanandan K .. 1. U C Davis.5. Proc. Boulder.sis qf Soil Liquqfirctiori Problen~s. N. F. Lee F. N. Katona M.. and Zienkiewicz 0 . I . University of MissouriRolla. and Rashidi H. England. and Zeng X. F. GPotechiiyue.fOrthe Aticil~vsisq f Soil Liqwfirction Problems..^.. B.r the Anulysis qf Soil Liquefirctiori Probleriis. H. N. and Zienkiewicz 0.D. 1720 Oct. (eds. in the Verification of' N~nericuIProceO'ure. (1993) Model No.fl)r Anrrlysis qf Soil Liquejuctioti Problrrils.
4a. London Chapter 7. 1.. Huang M. and Pastor M. (1985) KVV03 data report: Revised data report of a centrifuge model test and two triaxial tests. F. (1994) VELACS: Overview of numerical predictions for model No. (eds. 583591.) Arulanandan K. (1987) Modelling the response of sand to cyclic loads. 2. (1986) Triaxial data report: Report on seven triaxial tests.. Zienkiewicz 0. Cambridge University Engineering Department. Zienkiewicz 0 .. 1720 Oct.. 3. H. 16351646. in Verrfication of Numerical Procedures for the Analysis qf' Soil Liquefuction Problen~s. Balkema. and Pastor M. (eds.(eds. Sture S. (eds. Huang M. 1. Tecl~nology. (1994) Project VELACS management and coordination in Verif~cation Numerical Procedures .. and Scott R . Balkema. and Pastor M.. Venter K. and Pastor M.(eds.fbr the Andysis of Soil Liquefirction Problems.) Arulanandan K. K. C.. (1993f) Numerical Prediction for Model No.) Arulanandan K . M.250 VALIDATION OF PREDICTION BY CENTRIFUGE Stadler A. Schofield and Associates. in Veriji'cation c?f'NumericcrlProceclure.fbr rlze Ancrlysis qf' Soil Liqugfuction Problen~s. 1. Taboada V. Venter K. Zienkiewicz 0 . (1993a) Numerical Prediction for Model No.. in Verifi'cation of Numericcrl Procedures for the Anal~vsiso j Soil Liquejuction Problems. Zienkiewicz 0.A rational approach to quantitative solutions. F. C. A... A. A.. A. ( 1 9 9 3 ~Numerical Prediction for Model No. Arulanandan K. (ed. and Scott R . and Zeng X. Part I . A429.. in Verif~cation Numerical Proceclures. F. Shiomi T. 1720 Oct. (1993b) Numerical Prediction for Model No. in Verrfication of Numericrrl Procedures jbr the Anu1. (1990) Static and Dynamic Behaviour of Geomaterials . (1994) Review of prediction 'A' on model I1 in Ver.. 1. Rotterdam. in Veriji'cation ofNumerica1 Procedures. 777782. 1.) Arulanandan K. England. Zienkiewicz 0 . 259276. Zienkiewicz 0 . Towhata I. 1720 Oct. and Iai S. KO HonY.) Arulanandan K .. Rotterdam..423434. Andrew N. 731736. and Dobry R. C.. C.s Soil Liquefuction Problems.. UC Davis. Huang M. C. F. Balkema. 1720 Oct. and Scott R. Schofield and Associates. K. 1720 Oct.for the Analysis qfSoi1 Liquefirction Problems. F. Rotterdam.) of Arulanandan K. V. Huang M. Zienkiewicz 0 . 1 at RPI in Verification of Nurnericd Procedures.. 1720 Oct. and Scott R.) Arulanandan K. and Scott R. and Scott R.fi'crrtionqfNun~ericcr1 Proceduresfor the Anabsis o f s o i l Liyuefuction Problems (eds. Lond.) Arulanandan K . C. (1993) Experimental results of Model No. UC Davis. Y. UC Davis.Fully Saturated Problems. Andrew N. 1720 Oct. (eds. U C Davis. F.for the Analysis qf'Soil Liyuefuction Problems. T. 2.) Arulanandan K . Huang M.. and Pastor M. Rotterdam.. V. 4b. and Shiomi T. Soc.) of' Arulanandan K. Blackie Academic & Professional.. and Scott R. A. Thompson P. U C Davis. C.for the Ancrlysis qj'Soil Liqwfuction Problems. 6. and Sture S. 1720 Oct. (eds. Huang M. U C Davis. 2544. F. C. Cambridge. 12.. in ) Verification of Numerical Procedures. 1. 873880.... A. and Scott R. (eds.). U C Davis. and Pastor M.. F. U C Davis.fbr the Analysis qf'Soil Liquefuction Problems. Venter K. and Scott R.vsis of' Soil Liquejuction Proble~ns. U C Davis.. C.fbr the Anulysis of'Soil Liyuefuction Problems. 168195. 1 in Ver~fi'cutionqf Numerical Procedures for the Ancr1~~si. Rotterdam.) Arulanandan K. Dissertation. 1720 Oct. Chan A. 1. F. F. (1993) Experimental results of Model No. . C. F..) Steedman R. 1. U C Davis. Cambridge. (1993e) Numerical Prediction for Model No. 285309. Ph. 1. 1. A. N. Pastor M. (eds. (1995) Dynamics in Geotechnicirl Cmtr~f'uge Taylor R. A. A.. and Pastor M. Huang M. (19938) Numerical Prediction for Model No. Proc. (eds. in Verification of Numerical Procedures. Law H. and Lambe P.. and Scott R . qf' (eds. (1993d) Numerical Prediction for Model No.. V. and Scott R. 1720 Oct. F. U C Davis. 1720 Oct.. Roy.D. Paul D. S. 675680.. 318. Zienkiewicz 0. England. U C Davis. Balkema. 12671280. 7. A. 2.. Balkema. and Scott R. 1720 Oct.s. 16071612..
UC Davis. F. 11.. and Scott R.REFERENCES 251 Zienkiewicz 0 . C. (eds.) Arulanandan K . 1720 Oct. Huang M. in Verification of Numerical Proceduresfor the Analysis of Soil Liquefaction Problems. (1993h) Numerical Prediction for Model No... 9971006. . and Pastor M. 1.
.
Dam failure I: Lower San Fernando Dam. For example. It is important that soil parameters are determined from the site investigation data including laboratory test data.axial test data. Soil layer liquefaction problems 11: simulation of liquefaction behaviour during Niigata earthquake (1964). Quay wall failure in Kobe City (1995).1 INTRODUCTION In this chapter we will introduce case histories of the twophase liquefaction analysis (effective stress dynamic analysis) for real engineering problems. For example. to illustrate the effect of pore water migration on stability of dams (1971). the cohesion and friction angle can be determined directly from the drained tri. to illustrate the effect of initial shear stress. Seven examples are introduced. dilatancy . Soil structure interaction problem: building tilted due to Niigata earthquake (1964). to illustrate the effect of multidirectional loading (1995). design The most difficult part of liquefaction analysis is the determination of the soil parameters. But parameters of liquefaction are not so easy for simple constitutive models. Effect of counter measures using the deep soil mixing methoda prediction.Prediction Applications and Back Analysis 7. Soil layer liquefaction problems I: simulation of recorded data at Kobe Port Island. Dam failure 11: investigation of liquefaction failure mechanism for an earth dam at Hokkaido (1993).
As the original densification is not capable of representing cyclic mobility phenomena. most analyses are done using only one or twodimensional models. is included to illustrate the effect of pore water migration on the stability of dams. For this reason. The lower San Fernando dam example.1 Zntroductory remarks Although a real earthquake inevitably consists of multidirectional components. the HyogokenNanbu Earthquake of 1995. a modification for the original model is presented in Section 4. This contrasts with the Niigata Kawagishicho problem in example (2). and the earth dam at Hokkaido is included to show an investigation of the liquefaction failure mechanism using a numerical method.e. In this chapter. difference of failure line from phase transformation line.2 EXAMPLE 1: SIMULA TZON OF PORT ISLAND LIQUEFACTION . The quay wall was heavily damaged during the same earthquake as in example (1) i. Both show a similar failure mechanism but they are included here as they showed different failure patterns. Here. Example (6) is included for its interest in threedimensional analysis. Thus the input motion at the site can be estimated from the measurements in the vicinity. dynamic earth pressure acted on the embedded structure when liquefaction occurred and this also presents important problems. the densification model mentioned in Chapter 4 is a useful model for design engineers.EFFECT OF MULTIDIMENSIONAL LOADING 7. two dams are analysed and reported in examples (4) a rockfill dam and (5) an earthfill dam. yield surface. The aim of this . Example (7) is included as a case of dynamic earth pressure action and shows how safety can be increased. In this chapter the densification model is used in most of the examples. The foundation behind the quay wall was not damaged because of the counter measure installed. and so on. typical of liquefaction. The example is taken from an apartment block heavily damaged during the Niigata Earthquake of 1964.254 PREDICTION APPLICATIONS AND BACK A N A L Y S I S parameters are dependent on the shape of the potential surface. The example is also a very rare case in which the input motion is available at the deep base support layer and this was recorded at the damage site. (4). Dams are another type of structure for which liquefaction is important. Examples (1)(7) are all back analyses of liquefaction induced by various earthquakes. In most of these cases.2. 7. Example (3) is a quaywall problem. Example (1) is not only typical of soil layer problems but is the only case in which accelerations were recorded at four positions in depth during the HyogokenNanbu Earthquake in 1995. only records of permanent deformations are observed and no measurements are available for acceleration or pore water pressures.5 of Chapter 4. where the input motion was estimated from the data recorded at a location over 100 km distant. hardening parameters of shear behaviour. The earthquake heavily damaged Kobe City and over five thousand people died.
there has to be an influence of MDL on the liquefaction induced by earthquakes since negative dilatancy. though not simply the arithmetic sum. under a given number of cyclic loading. The accumulated shear strain is the sum. e. attributed the settlement on the shaking table test to the liquefaction strength. 1980). e. 3'%. except for the process of pore water pressure buildup and other details of response. has multidirectional movement and initial static stresses with a shear component will always be found in such soil structures as dams. which causes liquefaction. Their conclusion is that the settlements caused by combined horizontal motion are almost equivalent to the sum of settlements caused by components acting separately. Settlement of a thin dry sand layer was studied by Pyke (1973) for the case of a sand layer shaken on the shaking table in one or two horizontal directions. They found that the cyclic stress ratio dropped approximately 25%35'%. Most studies of liquefaction analysis in the past have considered only one horizontal component of the earthquake. This liquefaction strength is also influenced by the loading irregularity (Ishihara and Nagase. It was found that the unidirectional loading along the principal axis of the earthquake orbit agreed well with results from the horizontally multidirectional loading for the maximum response acceleration. and even less attention has been paid to the initial stress condition. It was also found that the effect of vertical loading is not significant. 15 or 20) has been studied at the end of the 1970's (Seed et al. causing 3% strain in a direction depending upon the pattern of the two components for loading. From the numerical analysis point of view. MDL and initial stress obviously play important roles in the geotechnical numerical analysis if material nonlinearity occurs.g. dykes and in natural ground. 1988).g. 1978 and Ishihara and Yamazaki. Both are important for precise numerical prediction and studies of these are reported in the individual sections. and this will be discussed in Section 7. Circular and random motions were applied for the twocomponent tests. Firstly MDL will be discussed and followed by ISS in the next section. Ghaboussi and Dikmen (1981) first . 1988). vertical accelerations superimposed on the horizontal accelerations could cause a marked increase in the settlements. Twophase dynamic equations used in the examples are those that have been derived in Chapters 2 and 3.. depends on the accumulated shear strain. The effect of MDL has been studied experimentally by several researchers.EFFECT OF MULTIDIMENSIONA L LOADING 255 section is to investigate the importance of multidirectional loading (MDL) in engineering practice. The case studies are made on the liquefaction events that occurred during the HyogokenNanbu Earthquake of 1995. and Ishihara and Yamazaki obtained directly the liquefaction strength with a twodirectional simple shear test apparatus under undrained conditions. The effect of MDL on the liquefaction strength (stress ratio to induce a certain shear strain. Seed et ul. the effect of initial (static) shear stress (ISS) was. inevitably. While vertical accelerations of less than lg cause no settlement if acting alone. however. while very few studies have been made on the behaviour of the ground subjected to multidirectional earthquake loading. In reality an earthquake.3. of the six components of the strain (three deviatoric strains and three shear strains). The volumetric strain due to consolidation following liquefaction also differed due to MDL (Nagase and Ishihara.
subjected to the El Centro Earthquake was solved as a case study. Even when the amplitude of the acceleration is increased to the resultant peak accelerations from the two directions.8m.3 times larger than the earthquake and of a stronger direction between northsouth (NS) or eastwest (EW). (1995) suggested the use of input motions that are 1.Om) was 314gal (3.2). In this model. The direction. 1963).14msp2)for the NS and 288gal for the EW direction. Fukutake et al. together with other variations such as soil properties and input motions. Twotofour very large amplitudes can be seen from the figures. was about 20 degrees from north to west. A hypothetical.8m during zero to five seconds. Several large amplitudes were clearly seen. The diagram at the bottom right of Figure 7.256 PREDICTION APPLICATIONS AND BACK A N A L Y S I S studied the effect of MDL on a soil layer problem with their proposed numerical method using the fully coupled Biot's equation with uU formulation (Zienkiewicz and Shiomi. the results were different from the results of the twodimensional analysis. Therefore there is a need to survey the threedimensional behaviour of the level ground subjected to MDL. horizontally layered ground.1 shows the orbit at GL83.8m. liquefaction took place along most of the seaside of Kobe City. where the liquefaction phenomenon has been observed throughout the island and settlement was about 20 cm (estimated from the relative gap between buildings supported by piles and the ground surface after the earthquake).1 shows the orbit of the records at Port Island for the HyogokenNanbu earthquake in 1995. the decrease of effective stress is a function of both components of the horizontal shear strains. They are about half of the value record at GL83. 7.2. This section investigates the effects of MDL on a real site. 1984) for the dynamic equation and a nonlinear material model. Sand boiling and flushing water due to liquefaction occurred in many places including Port Island where an array of seismometers was set at four depths (KansaiKyogikai 1995).. The diagram at the bottom left of Figure 7. although this conclusion is considered premature. As an alternative. The time history of the direction was the NorthSouth direction (Figure 7. At first. and earthquake motions for the studies were recorded during the HyogokenNanbu earthquake on January 17th in 1995.1 shows the orbit of the NS (northsouth)UD (up down) motion.2 Multidirectional loading observed and its numerical modellingsimulation of liquefaction phenomena observed at Povt Island During the HyogokenNanbu Earthquake. Figure 7. Twodimensional analysis showed a marked difference in the buildup behaviour of pore water pressure and some differences in the surface velocity spectra. when the maximum amplitude occurred. the UD component was not considered significant. . At GL83. This direction is considered as the principal axis of the earthquake components. back analysis of the observed data is explained and then the parameter study for M D L is reported. The maximum acceleration at the surface (GL O. and the effective stresspath approach for dilatancy (Ishihara et al. The site is a typical soil 'column' on the Port Island in Kobe City. The material model was based on the hyperbolic stressstrain relation for shear (Kondner and Zelasko. 1975).
81~1 800 observed G.EFFECT OF MULTIDIMENSIONAL LOADING obscrved G. The record at GL83.8m . . (gal) 800 observed 800 G. The other properties are calculated using the data shown in the table and the soil properties at a similar site.2 shows the material properties of the ground layer at Port Island. (gal) 800 800 800 400 0 400 EW Acc. Table 7.1 Orbit of an observed earthquake record (after KansaiKyogikai 1995) Conditions and modelling The Effect of MDL was studied by simulating the above records using a column of finite elements. The column of soil used in the numerical modelling is shown in Figure 7. 800 400 400 EW Ace. 83. 83. For example. (gal) 800 800 800 Figure 7.L. 1983).L. Case 2 was studied to investigate the influence of vertical input motion on the liquefaction phenomena. (gal) 0 800 800 400 0 400 EW Ace.8m w 9 in Z 400 u m 0 m " Y n 3 400  M 0  400 1100 400 0 400 NS Ace.L. 0. Case 1 simulates the observed record by incorporating all three directions of the earthquake motion. Cases 35 are for the comparison between two and threedimensional modelling.L. Four cases are studied.3.0m observed 257 G. friction angle and liquefaction strength were calculated through the N value of standard penetration tests (Tokimatsu and Yoshimi. 83. Case studies performed are shown in Table 7.1.8m was introduced as the input motion.
... EW and U D NS and EW only NS and UD only EW and U D only Principal axis and U D only 3 D analysis 3 D analysis 2 D analysis 2 D analysis 2 D analysis Results of simulation Figure 7.. ....0 5 10 15 Time (Sec) 800 20 25 30 Principal direction 0 800 0 .0m was fully liquefied at about 6 seconds and kept liquefied.4 2  0 ' . These agreed well initially. The observed acceleration shows a very long period wave....... while the calculated acceleration has a component of higher frequencies. EW and principal direction Table 7..1 Case 1 2 3 4 5 Cases studied Analysis type Input motion NS.5 shows the time history of the excess pore pressure in Cases 1 (NS+EW+UD) and 2 (NS+EW).. The layer between GL 12.2 Acceleration of NS.i 5 10 15 Time (Sec) 20 25 30 Figure 7. Figure 7.. however..4 shows the time history of response accelerations in the NS direction overlaying the observed accelerations.. But the builtup speed ...258 PREDICTION APPLICATIONS A N D BACK ANALYSIS g s U J Time (Sec) 800 ..6m and 14. some difference in the response is found at the surface after 5 seconds......
In order to investigate the effect of vertical input motion. excess pore pressure response without UD motion (bold . at a deeper level of GL 15.0 to 61. a tendency for liquefaction at a deeper layer was found in this large impacttype earthquake.12. Nevertheless.6 to .0 79.4m.12. Therefore.14.0 5.0 to 79.0 19.0 to 27.8 to 50.0m.5. pore pressure continued to build up and almost reached full liquefaction.8 32. (mls) 259 Average Nvalue Density (kNim3) Figure 7.3 Analytical model of the excess pore pressure above the layer was slowed down after 6 seconds. Case (2) (NS+EW only) was then analysed.0 and deeper Soil type Manmade fill Manmade fill Manmade fill Manmade fill Silty c!ay Layers of gravelly sand and silt Silty clay Diluvium sand V.0 to 32.8 to 19. On the other hand.6 .8 16.0 to 5. (mls) V.0 61. This is because the input motion transferred from the bottom to the surface was significantly reduced by the sudden loss of the strength at the layer between GL .EFFECT O F MULTIDIMENSIONAL LOADING Table 7.0 to 12. contrary to popular belief.16. The excess pore pressure response using the updown component (UD) has a very high frequency shown by the thin solid line in Figure 7. Depth (m) 0.6m and .0 27.0 50.2 Soil layer and material properties Layer no.
260 600 PREDICTION APPLICATIONS AND BACK ANALYSIS . The response obtained with unidirectional EW direction loading showed a different response while the response obtained with unidirectional NS direction loading was close to the one obtained with multidirectional loading... 8 m 0 5 10 Time(s) 15 G.simulation o b s e r v e d 0 5 10 Time(s) 15 20 Figure 7. 8 m . 1 6 . . 8 3 . In order to find out the effect of the direction of input motion... . This implied that vertical input motion induced only compressible wave in the water but did not affect the response of the soil skeleton... Next the results of unidirectional loading (Figure 7.L.simulation G. .4 Response acceleration of the NS direction solid line) passes through nearly the mean value of the excess pore pressure obtained with UD motion.. 32..simulation 0 5 10 Time(s) 15 20 G. 0 m 600 20 observed ..L. 0 . Therefore the result of Case 1 is found to be very similar to Case 2 when the high frequency is filtered out..simulation G.7) and multidirectional loading (Figure 7.5) were compared. response with unidirectional loading ..6 and Figure 7.L.8m 20 . 0 5 10 Time(s) observed 15 . The buildup of pore water pressure could have been dominated by the NS component of the earthquake.L. This is because the NS direction was very close to the principal direction of the earthquake motion..
4m 1 6.15.9.D I M E N S I O N A L LOADING 261 0 5 10 I5 Time (See) 20 25 GL12.EFFECT OF M U L T t . In this case. This tendency was found the same as the results of Ghaboussi and Dikmen (1981).5 Excess pore pressure ratio (NS + EW + UD) in the principal direction was then calculated and shown in Figure 7.8m Time (Sec) Figure 7. Maximum vertical acceleration was hardly influenced by the horizontal excitation.L.8.0m 30 Time (See) G. the results were very similar to those obtained using the NS component alone. Results of the NS direction for Case 1 (NS EW UD) and Case 3 (NS + UD) showed similar behaviour. Response of Case 5 with input motion in the principal axis was similar to the response of Case 1 (NS + EW) at the same (principal) direction. + + .611114. Effects of Multidirectional loading Maximum accelerations for uniand multidirectional loading are plotted in Figure 7.
9.6m 1 4 . But it is not . and also in static problems. they could be modelled implicitly by adjusting the shear strength of the material if only shear resistance is important. Only a few studies have been done for in situ KO (Hatanaka and Uchida. 0 m 30 0 5 10 15 Time (Sec) 20 25 30 Figure 7. at rest) or static analysis due to selfweight.3 SIMULA TION OF LIQUEFACTION BEHA VIOUR DURING NZIGATA EARTHQUAKE T O ILLUSTRATE THE EFFECT OF INITIAL (SHEAR) STRESS In addition to MDL.0m I I .L. 1995).12. ISS due to selfweight is usually assumed through KO (earth pressure coefficient. However.Om 30 0 5 10 15 Time (Sec) 20 25 G.L. but the real value could not be obtained by the current measurement.6 Excess pore pressure ratio (NS) 7. Consequently there are very few reports in numerical analyses to investigate the effect of ISS. No reliable means for measuring initial stress in situ is sufficiently well developed at the current time.262 PREDICTION APPLlCATIONS AND BACK A N A L Y S I S 0 5 I0 15 Time (Sec) 20 25 G. ISS also induces significant effects to the results of the liquefaction analysis.
.L.T H E EFFECT OF INITIAL ( S H E A R ) S T R E S S 263 0 5 10 15 Time (Sec) 20 25 30 . 44  1.  0 5 10 15 Time (Sec) 20 25 G. 'with ISS' and 'without ISS' conditions.4m 1 6 .6m 1 4 . This assumption was made to simulate the ordinary calculation procedure for the liquefaction safety factor.7 Excess pore pressure ratio (EW) straightforward. This means that the vertical stress changes to maintain the mean stress. 8 m 30 0 5 10 15 Time (Sec) 20 25 30 Figure 7. However. there are some analysis codes which do not detect any significance for this problem.O in the case of the condition 'without ISS' and 0. The mean stresses were kept the same at the same depth since the liquefaction strength has long been assumed to be the same for the same mean stress in the soil column problem (Yoshimi.12. A layered soil ground was analysed with two different initial conditions.e.0 .   :   . The analysis of these examples was very significant.15. 0 m 30 . The same results were obtained for most of the constitutive models in which criteria depend on the maximum shear stress in threedimensional space. since not only shear strength but also the dilatancy characteristic is affected by ISS.5 for the case 'with ISS'.L. The earth pressure at rest KOis 1. i. This difference in response is obviously caused by the different constitutive equation used in different codes. 1991).0 44  0 5 10 15 Time (Sec) 20 25 G.
The difference in the results is mainly caused by the initial shear stress (ISS) component of the initial stress.L. We are going to comment on this using an example of a onedimensional layer problem and show how big the difference is. However different choice could give completely different the results.0 5 10 15 Timc (Sec) 20 25 G. .L. 1991) and (Shiomi and Shigeno.011111 .8111 30 0 5 10 15 Time (Sec) 20 25 30 Figure 7.3.264 PREDICTION APPLICATIONS AND BACK A N A L Y S I S 0 1. However.L. it is very difficult to evaluate the initial stress since no one knows how the ground was made and no measurement is available even for the ratio of the lateral earth pressure to the vertical stress. 1993).4m 16.6m 1 4 .Om 30 r 1 0 5 10 15 Time (Sec) 20 25 G.15.1 Influence of initial shear stress Initial stress is calculated or n~odelleddifferently depending upon the analysis codes and/or analysts (Shiomi.9.12. Therefore. several calculations should be made to find the appropriate initial stress which is chosen according to the engineer's experience. rr (11.8 Excess pore pressure ratio (principal direction) 7. 0 m 30 0 5 10 15 Time (Sec) 20 25 G.
KO = 0.e. Acc. The response near the surface of 'with ISS'.j A 0 200 400 600 800 0 Max.10 shows the maximum response acceleration and displacement.j NS A. m. i . The pore water pressure of 'with ISS' built up quickly and higher. d. (gal) RO O (b) Figure 7.T H E EFFECT OF INITIAL ( S H E A R ) S T R E S S ( N S + EW + UD) N S (observed) 265 +UD (NS + EW + UD) ( p r i n c i p a l axis + UD) UD (observed) NS 4 EW (NS + EW + UD) EW (observed) . Both cases reach full liquefaction. The existence of ISS places the stress closer to the failure line so the material with ISS becomes weaker and this is the main reason for the smaller acceleration and larger displacement in the case with ISS. is larger for acceleration and smaller for displacement than those of 'without ISS'. but the excess pore water pressure is about 10%less in the case of 'without ISS'. i. Acc.o. (gal) (a) 200 400 600 Max. This tendency was almost the same as the laboratory element tests .9 Maximum response acceleration (a) horizontal component. The difference is about 301% acceleration and 20% for displacement. This means that the stiffness of the soil material for the case 'with ISS' is reduced more than the case 'without ISS'.EW.5.i.11 shows the time history of the liquefaction ratio. (b) vertical component Significance of ISS component to the responses Response acceleration The influence of ISS is significant. Figure 7. for Excess pore water pressure Figure 7.
. .  0 40 80 120 Max. 1 2 Max. disp. (cm) 3 KO = 0.266 PREDICTION APPLICATIONS AND BACK ANALYSIS [I.10 Profile of maximum response acceleration and displacement 0 2 4 6 8 10 Time (s) Figure 7.5 K O = 1...:. acc...0 Figure 7. (gal) 0 . ....11 Time history of pore pressure ratio .. .
An external force due to an earthquake produces a large horizontal shear stress and the principal stress direction is rotated. There is no substantial evidence. which uses the flow rule. Theoretical considerations ISS can therefore be classified into two types. at about 8 seconds.5 seconds. That means that ISS is not involved in the formulation. It should be noticed that ground layer analyses frequently neglect ISS.' The constitutive models developed as an extension of a onedimensional shear soil column model often ignored ISS (deviatoric stress due to the difference of vertical stress and horizontal stress) since most of the models use the hyperbolic stressstrain model for the shear behaviour and their stresses always start at zero.1 1 may give a clue as to why the liquefaction strength curve gives the same value as for the same mean stress. The maximum shear stress acts for the direction of approximately 45 degrees. A soilcolumntype dynamic effective analysis can obviously be classified into Type I. deviatoric shear stress (vertical stress minus horizontal stress) is relatively large. that ISS should affect liquefaction differently. The criteria for the final liquefaction is either 5% shear strain in deformation or pore pressure reaching 95% of the initial vertical stress. as indicated by the experiments.0 at almost the same time. the soil beneath a structure has almost no shear stress in the horizontal direction. Type 11. The liquefaction strength curves were determined by the number of cycles taken for the samples to reach final liquefaction. a slope such as that of a dam has ISS close to the horizontal direction. 7. ISS is seen in the case where an external force is applied in a perpendicular direction to ISS. produce different results for the conditions 'with ISS' and 'without ISS. Constitutive models based on a typical elastoplastic theory. however.12). The liquefaction ratio is 75% for the 'with ISS' and 40%)for the 'without ISS' at 3. however. For example.OUA Y W A L L FAILURE AND A COUNTERMEASURE 267 by Vaid and Finn (1979) and Hyodo et ul. Type I. This proved the effect of the . The pore water pressure ratios reached 1. Figure 7. It was found that the existence of ISS creates a slower buildup of pore water pressure. For example. These two types of ISS might work differently. ISS is the case where an external force is applied to the direction parallel to the maximum shear direction of ISS. the incremental shear stress has mainly the effect of rotating the principal stress with the increment in the equivalent stress being small. the upper soil layers with ISS were weakened more than in the case without ISS. In the example problem.4 QUA Y WALL FAILURE AND A COUNTERMEASURE Although quay walls surrounding a Hotel at Kobe were greatly damaged due to liquefaction during the HyogokenNanbu Earthquake of 1995 (Figure 7. (1988) except for the final pore water pressure. the foundations of the building had no damage. In this case the incremental equivalent shear stress is equal to the external incremental shear stress. In this case.
The countermeasure was lattice walls. .268 PREDICTION APPLXCATIONS A N D B A C K A N A L Y S I S Figure 7. The building was built on a pier shaped ground.12 Damaged quay wall along the building (after Suzuki ct rrl. The latticeshaped stiffened ground walls (LSSGW) were built surrounding the piles.4. At the design stage. prediction analysis for the structural design stage was conducted to confirm the effects of the LSSGW. 1995). which are made of the stiffened ground by mixing cement to reduce the shear movement of the ground (Suzuki et al. 7. (1995)) countermeasure for the foundations..1 Conditions and modelling Configuration Twodimensional analysis was conducted.13. The numerical model of the foundation and building is shown in Figure 7.
The foundation was surrounded by quay walls. J . Soil layers and properties The zone where liquefaction was anticipated was from G L (ground level) 0.13 Numerical model of foundation and building which projected into the sea.I . The layers from G L . The building was considered as elastic and MohrCoulomb criteria was used for the LSSGW. LSSGW and ground with piles were modelled into separate groups. The piles were supported at the layer G L 32.7 m. Its soil properties (shown in . L m 8 5 ' 8 ' ~ 2 ~ 7 6 ' 8 2 ' fT8.OUA Y WALL FAILURE AND A COUNTERMEASURE 269 a 10 3m 13 4m  34m 15 8m  26 1m 32 7m (a) Sectmn plle .4m.13.1 I .0 m to . The constitutive model used here was the MC (MohrCoulomb) Densification model. J .13.5'6 ( b ) Pldne Figure 7.13. Deformation due to the earthquake in the shorter section of the building was anticipated larger than that in the longer section so that the numerical study was made in the shorter section.7 m were soft blue clay or silty soil.4m to 32.I. The foundation of the building was made using a repeated pattern of piles and LSSGW shown in Figure 7. J . These two groups were connected at a corner of the lattice.
02100 0.1997 r A B 1.0 5.OO 1. The liquefaction strength R I 5 is the only data available for the liquefaction parameter. R l s or RZ0 is the shear stress ratio against the confining pressure when liquefaction takes place after 15 or 20 cycles of the undrain triaxial test.02100 0.3 Soil properties Layer No.4 are for the densification model and are determined to satisfy R 1 5 . This is largely because safety against liquefaction is normally judged by the liquefaction strength.00 34. Depth (m) PREDICTION APPLICATIONS AND BACK ANALYSIS Porosity Friction angle (deg) Shear modulus (MW 34. O O 5. Parameters r.1574 0.1534 0.02000 0.0 5. 2 3 4 5 RIS 0.14 shows an example of a fitted curve for the liquefaction strength.Figure 7.0 0.00 79. O O 1. A and B in Table 7. log N Figure 7.270 Table 7.14 Liquefaction strength of laboratory test and calculation .1495 0.00 38. O O 1.00 92.01200 Table 7.00 141.4 Parameters of liquefaction ( R l s and densification model) Layer No.00 59.3) were obtained by a site investigation conducted during the design stage.0 5.00 Table 7.
7. Figure 7. was used. EW component. However.17).16 and 7. This result matches the investigation after the earthquake.15 Predicted earthquake for Meriken Oriental Hotel Input motion Input motions for design were E l Centro of 1940. there was no damage in the foundation zone. The figure for Line 4 shows the result for the quay wall.2 Results and remarks Figure 7.18 shows the profile of the pore water pressure at the end of the earthquake. The pore water pressure within the LSSGW was less than 30'% of the vertical component of the initial effective stress shown by the dotted line. The figure for Line 1 shows the results for the centre of the model. Large pore water pressure built up behind the quay wall and caused the failure of the quay wall during the HyogokenNanbu Earthquake (Figures 7. It is necessary for discontinuity to be avoided between the layers of LSSGW and the undersoil layer. The pore water pressure behind the quay wall reached the full liquefaction state. That was the intention of reinforcement of the soil foundation. NS component and some other records of large earthquakes which occurred in Japan. Taft of 1952. The latter pore water pressure buildup can be easily anticipated because stiffness of LSSGW became much greater than the support layer below the LSSGW so that a large shear strain is induced around there. . An artificial earthquake generated from a design spectrum is also often used. Relatively large pore water pressure built up around the toe of the LSSGW but did not cause any damage. Here an earthquake predicted at the site from an earthquake recorded at the Kobe Ocean Weather Station during the HyogokenNanbu Earthquake of 1995.4.16 shows the permanent deformation as the quay walls of both sides tilted and the back fill was extensively damaged.O U A Y W A L L FAILURE AND A COUNTERMEASURE 271 Time (s) Figure 7.
i. These small strains were due to the stiff LSSGW and could produce no liquefaction. Therefore.272 PREDICTION APPLICATIONS A N D BACK ANALYSIS Figure 7.19 shows the maximum shear stresses which occurred in the soil layers and in the LSSGW itself.17 Contour of pore water pressure Figure 7. The strains of the outside of the LSSGW.e. These stresses indicated that most of the shear forces acting in the ground were shared by the LSSGW. soil strains of the inner zone of the LSSGW were 0.16 Permanent deformation of quay wall Figure 7.19% at most of those which took place at the bottom of the LSSGW zone. backfill of the quay .
18 Pore water pressure and initial effective stress (see Figure 7. larger.19 Shear stress of lattice shaped stiffened ground wall and ground wall. so that the liquefaction took place and then more.l > ( x I 0 kPa) <Line 2 > ( x 10 kPa) <Lme 3 > ( x I0 kPa) Figure 7.13 for lines 1 to 4) Figure 7.Q U A Y W A L L FAILURE AND A COUNTERMEASURE 0 2 4 E 6 8 273 5 g I0 12 1 4 JA ' ? W Pore Pressure 5 5 0 5 10 15 20 25 Pore Pressure 0 5 10 15 20 Pore Pressure Pore Pressure <Line 4 > ( x I OkPa) <Line . strains were induced. . were very large (more than 10%) since the quay would be able to move freely towards the sea.
(1975) and Seed (1979) is remarkable in attempting to explain why the failure occurred apparently some 6090 seconds after the start of the earthquake. The actual collapsed dam and a 'reconstructed' crosssection are shown in Figure 7. which was recorded to last some 14 seconds. resulting from an earthquake. 1979) . Here. is typical of what can occur in a poorly consolidated soil structure affected by shaking.20 Failure and reconstruction of original conditions of the Lower San Fernando dam (Seed. the effect of cohesion resulting from negative pore pressure and the influence of such parameters as permeability and relative density on the dynamic response of the dam. are illustrated. the reconstruction of the event by Seed et ul.5 LOWER SAN FERNANDO DAM FAILURE The failure of the lower San Fernando earth dam in 1971 with nearly catastrophic consequences. the details are presented. 'migrates' in the post earthquake period to regions closer to the 'heel' of the dam where it triggers the failure.274 PREDICTION APPLICATIONS AND BACK ANALYSIS 7.20 following Seed (1979). Crosssection through embankment after earthquake Reconstructed crosssection Figure 7. Zienkiewicz and Xie (1991) have reported some results of numerical simulation of the failure. Although full comparative measurements are not available. The hypothesis made here was that the important pressure buildup occurring as a result of cyclic loading which manifested itself first in the central portions of the dam. In particular.
21 Idealization of San Fernando dam for analysis: (a) material zones (see Table 7. Figure 7. indicating clearly the 'phreatic' line and the suction pressures developing above. First an initial.5): (b) displacement discretization and boundary conditions.22 shows such an initial steadystate solution for the saturation and the pore pressure distribution. and (c) pore pressure discretization and boundary conditions We show in Figure 7. finite element meshes and boundary conditions used in the present computations. Figure 7.22 Initial steadystate solution: (a) Pressure (kPa).21 the material idealization.LOWER S A N FERNANDO D A M FAILURE 275 Base excitation Rigid boundary )le boundary Impermeable boundary Figure 7. and (b) Saturation contours . elastic static analysis is carried out by the full program considering a semisaturated condition and assuming the gravity and external water pressure to be applied without dynamic effects.
22(b) but the saturation in the semisaturated zones will be close to 1.0 2020.5 PREDICTION APPLICATIONS AND BACK A N A L Y S I S Material properties used in the Lower San Fernando dam analysis K f (Pa) 2 2 2 2 v Material zone p. Indeed. The displacements at some characteristic points and the development and decay of excess pore pressures are shown in Figure 7.2857 x 10' 0. Since p.0. 1990) are summarized in Table 7. in the otherwise cohensionless granular soil since the effective stress is defined as When we reduce the parameter b of Van Genuchten's formula by a factor of 100. a full nonlinear dynamic computation is carried out for the period of the earthquake and continued for a further time of 200 seconds.2857 x 10' 0.375 0.375 lop3 lo' lo' lo' x 109.0 2020. This indeed was conjectured by Seed (1979).24 and Figure 7.pw (or the S.0 2020. The material properties assumed to describe the various zones of the dam using the constitutive model described in Chapter 4 and also found in Pastor and Zienkiewicz (1986) and Pastor er al.(Pa) 1 2a 2b 3 2090.0 980.2857 x lo9 0. the pore pressure distribution of static solution will be almost the same as in Figure 7.25.23 shows the displaced form of the dam at various times.0 980. The suction pressures which developed above the phreatic line give a substantial cohesion there.0 10'' 10" lo1' lo1' Porosity k (mis) 0.2287 Starting with the above computed effective stress and pressure distribution. it is seen that the cohesion is of the value Swp. an almost immediate local failure develops in the dry material upon shaking.) . It is noted that deformations are increasing for a considerable period after the end of the earthquake. (1988. without such cohesion. This undoubtedly is aided by the redistribution of pore pressures.5.276 Table 7. Figure 7. Near the upstream surface the pore pressures continue to rise well after the passage of the earthquake. 980. preliminary computation indicates that. curve in the following equation.(kg/m3) pf ( k g / m 3 ) K. is assumed as zero. It is also noted that the pattern of deformation is very similar to that which occurred in the actual case showing large movements near the upstream base and indicating the motion along the failure plane.0 980.375 0. The amount of cohesion depends on the S.375 0.h. .
00700 2.0500 5.23 Deformed shapes of the dam at various times: ( i ) 15 s (end of earthquake): (ii) 30 s. O O 0.0 1.0667 0. The dotted lines in Figure 7. results in stronger cohesion in the upper part of the dam.5 2. If the permeability of the dam material is sufficiently high.24a are the results of computation occurred because of the now increased cohesion in the upper regions of the dam.90 After Van Genuchten rt rrl. (iii) 90 s: (iv) 200 s.0 4.0842 0. Table 7.6 Coefficients of saturation function h .01740 0. it may be impossible for an earthquake to cause any buildup of pore pressures in the embankment.0 0.0689 0.LOWER S A N FERNANDO DAM FAILURE 277 (iii) 90s Figure 7. (1977) The higher value of S. since the .?(cm') 1 u(cmI) h c1 For sand' For San Fernando I 0.
(d) at point I (see Figure 7. (b) at point E. .24 Horizontal (left) and vertical (right) displacements: (a) at the crest (dashed line represents the result of computation with increased 'cohesion').278 PREDICTION APPLICATIONS AND BACK ANALYSIS Figure 7. (c) at point H.21a).
Figure 7.21a) pore pressure can dissipate by drainage as rapidly as the earthquake can generate them by shaking.25 Excess pore pressure at points A to H (see Figure 7.26 shows the results which indicate a rapid dissipation of pore pressures and much reduced permanent deformations. .L O W E R S A N FERNANDO D A M FAILURE 279 Figure 7.
In this case significantly larger displacements are recorded at the early stages of the earthquake shaking as shown in Figure 7. is considered equivalent to the relative density Dr [Pastor et a1 19851.906 for material zones 1. . (c) horizontal displacement on the crest.21a) In an additional analysis the relative density of the dam material is assumed lower.604 and 0.0.280 PREDICTION APPLICATIONS AND BACK A N A L Y S I S Figure 7. (b) deformed shape of the dam at 200 s.453. excess pore pressure at point D (see Figure 7.24.5 fixed. 2b and 3. respectively. With M. Mf values are now reduced to 1.0.27. values in Table 7. (e) excess pore pressure at point A. (d) ~~ertical displacement on the crest.26 Results of analysis with increased permeabilities: (a) deformed shape of the dam after 15 s. which implies in the present constitutive model that the ratio Mf/M. Za.
Any trial to find the failure mechanism by conventional engineering methods such as the sliding analysis. (ii) 10s. the failure mechanism of an old irrigation dam was studied. refer to the dam as the N dam for simplicity. (iii) 15s. no clear slipline was found in the section of the dam.. which was made after the failure. were not successful. Sand boils due to its liquefaction were found at the foot of the dyke. The dam section could then be examined easily. showing deformed shapes at: (i) 5s.27 Results of analysis with softer materials. It was decided that the dam had to be decommissioned and a section of the dam was removed soon after the damage was found. (iv) 200s. The liquefaction analysis (Shiomi et al. Although the top of the dam had large cracks. 1993). We shall. due to liquefaction..6. 7. The analysis explained reasonably well the observation. . 1996) was then conducted using the dynamic effective stress approach using MuDIAN (Multiphase Dynamic Interaction Analysis: Shiomi et al. The dam collapsed completely during the Nihonkai Nanseibu Earthquake of 1993.6 MECHANISM OF LIQUEFACTION FAILURE ON AN EARTH DAM ( T H E N DAM) 7. however.MECHANISM OF LIQUEFACTION FAILURE 5s 28 1 (ii) 10s (iii) Figure 7.1 Objective of the analysis In this section.
due to the sloping of the dam side. There we found that we could not alter the initial shear stress itself with respect to the shear failure criterion. As the actual maximum acceleration could not accurately be determined. In the dam analysis. its influence on the dilatancy then came into question. The dam quickly liquefied and failed using those parameters. At first a static analysis was conducted for the selfweight of the dam. Then liquefaction analysis was conducted with the densification model. so the record at a dam nearby was used for input motion. Also in this study. which was proposed by Zienkiewicz et a1 (1978). 0.6. a reasonable simulation result was obtained by controlling the influence of the initial shear stress condition. three different values of the maximum acceleration for the input motion to the base of the analysis model were attempted and they were 134ga1. Therefore. The liquefaction parameters were estimated from a liquefaction strength curve. because of the presence of the initial shear stress. for its liquefaction properties.29 show the response spectra and time history of the recorded earthquake respectively. this relation was further investigated. the failure pattern of the earth dam is evaluated. Finally. This can neither be neglected nor reduced.28 Response spectra of input motion . soil parameters and initial stress conditions. 7. To find the reason for this we surveyed the influence of mesh. However. The N Dam is located about 70km from the epicentre so the maximum acceleration could have been as large as 200250ga1. which was only one test data set.1 1 Time (Sec) Figure 7.28 and Figure 7.282 PREDICTION APPLICATIONS AND BACK A N A L Y S I S The modelling of the constitutive relationship and the initial stress condition plays an important role in the analysis and is reported here in detail. the treatment of the initial shear stress condition is a difficult issue. Figure 7.2 Input motion There was no earthquake acceleration record at the N Dam. 200gal and 250gal. according to the attenuation curve of distance from the epicentre. The dam was about 130km from the epicentre.
5Hz for the vertical mode. The first natural frequency was 2.30(a)) was in the centre of the dam and 'shielded' the water flow. The configuration and structure of the dam are shown in Figure 7.MECHANISM OF LIQUEFACTION FAILURE 283 MAX.8 (23.3 Conditions and modelling The Ndam is an old earth dam built in the 1920's. 134.0m. The height of the dam from the bottom of the reservoir was about 10.29 Time history of input motion 7. Cohesion c and friction angle 4 were obtained from the drained tests. Therefore the twophase region (soil skeleton + water) was assumed only at the upstreamside of the core. Densities y Random IV (silty sand) ' (a) Sedimentation Effective stress analysis Figure 7. Soil properties The soil parameters used in the analysis are shown in Table 7. The various values of the shear moduli G were evaluated through the N values of the standard penetration test.6.30.211gal) 1501 0 ' ' ' ' ' 13 Time (Sec) ' ' ' ' ' 26 I Figure 7. (b) FEM mesh and boundary condition .3 Hz for the horizontal mode and the second was 3.30 Configuration of the dam: (a) distribution of soil properties. The core of the dam (dark area in Figure 7.7.
31 shows the liquefaction strength curve used for this analysis. (1944)). The liquefaction strength used here was obtained from the Nvalue and laboratory test. Table 7. The parameters of the densification models are listed in Table 7.7.) Number of cycles N Figure 7. The other marks shows the liquefaction strength calculated from the Nvalue of the standard penetration test.. Figure 7. Permeability was determined through 20% particle size D20 by Creager's approach (Creager et al. Parameters for liquefaction Material parameters for liquefaction are determined from the liquefaction strength curve as in the previous example.284 PREDICTION APPLICATIONS AND BACK ANALYSIS were obtained by the physical tests.7 Soil properties Name Core I Core I1 Random I Random I1 Random 111 Random IV Random V Sand Sand & gravel Sand & gravel Sandy rock I Sandy rock I1 Sandy rock 111 NValue y(kN/m3) G (MPa) Porosity Permeability c(kPa) (mJs) 4 (deg. The solid line shows the liquefaction strength calculated by the assumed soil parameter for the densification.31 Liquefaction strength . The solid dot shows the data obtained by the undrained triaxial test.
Figure 7.162) to the following equation in this analysis.007 0. since both influence the buildup of excess pore water pressure.2998 A 0. the initial shear stress causes a rapid buildup of pore water pressure.6. the case where the input motion was 200gal showed good agreement to the failure pattern observed.3 0 0.4 Results of calculations Among the three case studies.28 0. 7. RIzo 0.32 shows the initial maximum shear stress.004 B 1 r.78 1 1 8 6 3 Initial stress Initial stress was calculated for selfweight assuming that Poisson's ratio was 113.4 4.M E C H A N I S M OF LIQUEFACTION FAILURE 285 Table 7.17 2 50.6 32.2239 0. 3. The dilatancy of the sand may not depend on the initial shear stress if the material pass a long period after the construction of the dam.65 86.18 59.8 Model's parameter for densification model Soil type Random I1 Random IV Gravel Nvalue F C ('%I) Djtr (mm) Average u. which does not agree with ordinary engineering experience. So the actual magnitude that hit the Figure 7.01 1 0.32 Contour of maximum shear stress . With the densification model. the initial shear stress was ignored in the densification model modifying (4. Therefore. The initial mean effective stress and shear stress are important.7 20.1692 0. where Oo is the initial shear stress ratio.2 29.
34 shows the progress of the deformation of the dam in the case of 200gal.9117 At 10 seconds At 20 seconds At 26 seconds Figure 7. The largest depth of the crack C1 was 1. The cracks C2 and C3 are shown on the surface of the upper stream. At about 10 seconds.1 1.34 Transition of the deformation of the dam .2m. SCALE :1 .33 shows the damage to the dam.33 Crack observed after the earthquake dam could be said to be about 200gal. Figure 7. This deformation might have caused the observed cracks C2 and C3 DISP.286 PREDICTION APPLICATIONS AND BACK ANALYSIS Figure 7. Figure 7. the surface of the upper stream showed movement toward the toe of the dam. where three cracks were observed.Om1.
33. The side boundaries of the excitation direction are modelled by the periodic boundary.5 Remavks The liquefaction analysis made clear the failure mechanism of the dam. 7. The objective of this analysis is to evaluate the threedimensional effect on liquefaction analysis. Until that time. The largest tilt was about 60 degrees. This might have caused crack C 1.36. so that a direct foundation could be used. . 7.LIQUEFACTION D A M A G E IN T H E NIIGA T A EARTHQUAKE OF 1964 287 Figure 7. But a sand foundation lost its strength due to liquefaction and then the apartment tilted. This mechanism was explained by a rather simple densification model with a simple modification.35. Since sand region has liquefied and dyke sank under selfweight no significant slip line developed. The apartment was a wall structure but the fivestory building is modelled by beam elements adapting the natural frequency equivalent.35 Contour of excess pore water pressure at 26 seconds at the surface of the upper stream as shown in Figure 7.6. C1. The side boundary perpendicular to the excitation and the bottom boundary are modelled by a viscous damper. The constitutive model used here is that of the Pastor and Zienkiewicz generalized plasticity model. Excess pore water pressure was very high at 26 seconds as shown in Figure 7.84m due to the liquefaction near the dam core. Half of the problem is modelled considering the symmetric shape. This caused the overall land slide and the largest crack. Then the top part of the upper stream began to settle 0. it was believed that a sand layer was stronger than clay. During the first 20 seconds the deformation of the upper stream surface became enlarged continuously.7 LIQUEFACTION DAMAGE IN THE NZZGA TA EARTHQUAKE OF 1964 An apartment built directly on a sand foundation was tilted substantially due to the Niigata Earthquake in 1964. The analytical model is shown in Figure 7.
Figure 7.37.41(b) shows the bird'seye view of the deformation. Table 7. The stress contour under the building is higher due to the overburden load. The excess pore pressure beneath the building is large due to the overburden load.288 PREDICTION APPLICATIONS AND BACK ANALYSIS Figure 7.40 shows the initial stress of ground under and far from the building.41(a) shows the deformation due to earthquake.42 by comparing the results at sections A and B. Figure 7. The Nvalue is the number of loading cycles for the strain amplitude. Layers 5 and 6 show permanent horizontal deformation for both sections A and B and the ground surface settled. The liquefaction strength calculated by the soil parameters is shown in Figure 7. Figure 7.1 Results Figure 7. This can be seen in Figure 7. Parts of the ground element are eliminated to show the inside.39 shows the result of the undrained triaxial test by changing the applied stress ratio. All layers have the possibility of liquefaction but layers 79 are stronger than the upper layers.36 Analytical model The input motion used is the recorded data at AkitaKencho for the Niigata Earthquake of 1964.7. . 5% is calculated. 7.38. The time history is shown in Figure 7.9 shows the soillayer model and material parameters.
Test.4 A .37 Recorded data at AkitaKencho for the Niigata Earthquake of 1964 0 Lab. 10 Number of cycles N Figure 7. Layer 3. Test.9 Lab.6 Sirnulation."" 0 1 2 3 4 5 Time (Sec) 6 7 8 9 10 Figure 7.8. Test.8. Layer 5.LIOUEFACTION DAMAGE IN T H E NIIGA T A EARTHQUAKE OF 1964 289 .38 Liquefaction strength . Layer 7.9 Simulation. Layer 3.4 Simulation. Layer 5. Layer 7.6 Lab.
4 (GL2.8.) .(~P. 0 20 40 60 Mean stress (kPa) N=20 80 100 0 20 40 60 Mean stress (kPa) N= 5 80 100 ( I ) Layer 3..I4.) Cohesion (kPa) Q (deg.0 .Om) Mean stress (Wa) Mean strcss (kPa) Mean stress (kPa) N= 5 Mean stress (Wa) (3) Layer 7.05.9 (GL8.39 Stress path of element simulation Table 7.9 Material parameters Depth (m) porosity Poisson's ratio G (kPa) Permeability (rnls) v.290 PREDICTION APPLICATIONS A N D BACK ANALYSIS .Orn) N=5 Figure 7.
0 5.LIQUEFACTION DAMAGE IN THE NIIGATA EARTHQUAKE OF 1964 291 For the PZ model Depth (m) GL2.35 0.014.0 1.0 GL8.8 5.40 Initial stress condition predicted by static analysis .58 tr.0 1.11 0.8 2.02 0.0 8.50 0.5 2.7 2.1 1 1.6 3.35 0. M.0062 2.6 5.54 0.5 7" YDM 0.7 2.12 0.6 3. 8 " 1.7 (a) Section A (b) Section A Figure 7.0 GL5.46 0.6 5.62 1. ru.7 6.0 3.11 1.35 0. Ho Hue 2.0  MI 0.39 0.
41 Deformation of the ground and building . SCALE : 5.47E02 Figure 7.47E02m (a) Section A DISP. SCALE : H 5.292 PREDICTION APPLICATIONS AND BACK A N A L Y S I S DISP.
The area where power stations are built is often by the sea due to the water cooling requirements and the earth from these areas is inevitably made up of filled sand. In order to do this.) 7. the major mechanism should be extracted requiring investigation of the stress for both the inside and outside of the improved soil (Kishion et al.. This method improves the strength of the soil in the ground by mixing it with cement and makes the stiffness usually over 10 times greater than that of the original soil layers. This reinforcing method uses a deepsoil mixing method. the static prediction method is required to ensure seismic safety of the stiffened ground. According to common engineering practice.8 INTERACTION BET WEEN ORDINARY SOIL AND IMPROVED SOIL LA YER It is very important to prevent liquefaction for key facilities in the industry such as thermal power plants and in order to estimate correctly their damage if subjected to a major earthquake. Thus the foundation itself will not liquefy but should also be able to share the horizontal seismic force. 1998). .41 (corlt.INTERACTION BETWEEN ORDINARY SOIL 293 Figure 7. This example demonstrates how the dynamic effective analysis can be used to evaluate the effectiveness of the newly developed soil reinforcing method for foundations. The summary is reported here. which is transferred from the bottom of the superstructures.
43 numerical models for a crosssection of a power plant is shown.0 70. 00 20.00 1.0 &Pa) Figure 7. The dark .0 30.0 40.0 83. The building is about 320m long and 65m wide and modelled as a simple shear beam with stiffness equivalent to the whole structure like some form of superelement.42 Vertical stress immediately after the earthquake stops (Pore Pressure) In Figure 7.0 50.294 PREDICTION APPLICATIONS A N D BACK ANALYSIS ( a ) Section A ( b )Section B Section A Gl B c E F 0 C H 0.
Short dir. the details of the project will not be finalized. In this analysis. by laboratory tests.6 7.INTERACTION BETWEEN ORDINARY SOIL 295 Figure 7.0 223.45 8. 1 2 24.54 4 13. The soil property is determined by soil sampling at a real site as shown in Table 7. Height (m) Long dir.1 11.0(13. all necessary soil tests mentioned above have been completed but in most cases.0 16.0(7. The liquefaction analysis is conducted for a probable major earthquake artificially generated for the Tokyo Bay area.54 5 13.5(24. 1944).0 165. (b) lumped mass at node (a) Element No. Table 7.34 3 24. only limited data will be available for the prediction analyses. the shear velocities ( V s )by the wave velocity investigation at the field.10. The Nvalue is obtained by the standard penetration test.5 42.43 Configuration of the numerical model and mesh (shallow layer model) hatched zone is the improved soil which is supported by a strong support layer with shear wave velocity of over 400mIs in which condition the soil layer will not liquefy. No. 9. The surrounding area is the sandy soil layer where liquefaction is anticipated if a major earthquake strikes.8 29.1(34. 34.1 1.2 5 7. since they will be carried out at a very early stage of construction.4 92. 40.9 2680.8 7 8 5. occasionally they need to extrapolate.5 46.0 108. unit x 10Jm'jm (b) Node.82 0. 25.53 (sectional area.10 Equivalent beam model of a building: (a) stiffness.1 187. 34. Height (m) Long dir.1 4 18. drained cohesion and friction angles ( c and 4) and liquefaction strength. Engineers will be forced to interpolate between the limited amount of data. 44.4 6 0. At this stage. 529. Density and porosity are obtained by the physical property test.1 57.0 153. The maximum velocity is 50 mls.51 3 18.0(29. The material properties of the building are summarized in Table 7.0 1910.56 8.0 241.0 88. permeability from particle size DZo(diameter at which 20'51 of the soil is finer) according to Creager's experimental data (Creager et al.4 3. 1 2 29. (unit: Mglm) .0(18.5 74. Short dir. 363.3 6 7. 17.
40 2.80 11 11 7 3 3 3 27 .90 13.70 4.30 6.80 8.60 19.80 1 1.05 14.Table 7.40 5.11 Soil properties of DMS problem Soil type depth (m) Nvalue density Share ( ~ ~ / m ~ ) wave velocity (mlsec) Poisson's Friction ratio Angle (degree) Cohesion (kPa) Permeability (cmisec) Rl2O ud/2uo1 Soil particle density Porosity Fine sand SFs Silty sand Finesand Fine sand Silty clay Fine sand SFs As1 As2 Dcl Dsl 1.
which shows the limit stress ratio against the number of cycles of loading (Figure 7.44).1 Input motions The input motion used in the study was an artificial earthquake determined in 1992 as a design earthquake for the Tokyo Bay area. this curve does not give the information about how fast the excess pore pressure is built up. This earthquake can be used as input Figure 7.44 Liquefaction strength of soil layers .INTERACTION BETWEEN O R D I N A R Y SOIL 297 Determination of liquefaction strength The direct soil properties for liquefaction behaviour obtained in ordinary engineering practice are the liquefaction strength curve.8. 7. However. Therefore the parameters of the constitutive model are not determined only by the liquefaction strength curve but also by the engineer's experience.
Earth pressure due to liquefaction Seismic forces acting on the side of the improved soil ground is necessary for static safety analysis as well as seismic intensity.64 (17. and artificial earthquakes recommended by Japan Architectural Centre.72sec) I Time (Sec) (b) Acceleration spectra Figure 7. EW. A flat spectrum between 0.46 Pressure at the side boundary of the improved soil ground: (a) maximum distribution.2 and 0. = 300500 mls).45 Rinkai 1992 artificial earthquake motion at a depth of bearing layer such as the Edogawa Layer (V. an artificial earthquake. 1952. Among them the pressure acting to the Earth pressure (kPa) (a) Figure 7.6 Hz at maximum is assumed. (b) comparison with Westergaard's . To determine the seismic intensity in the Tokyo Bay area. Figure 7. NS.298 400 1 PREDICTION APPLICATIONS AND BACK ANALYSIS Max : 3 10. Taft. several earthquakes were used such as El Centro.45 shows the time history of the Rinkai 1992. 1940.
U a 0 m 0 3 . (3) an overturning of the improved soil ground. 7.99 0. The results obtained by the effective stress FE analysis agreed well with Westergaard's formula for dynamic water pressure. (2) a subgrade reaction at the bearing layer. A guideline for depth limitation was obtained from this numerical research. there will be a limitation for deeper foundations since the deviatonic stress becomes higher as the layer becomes deeper. (1) the possibility of the improved soil ground slipping at the boundary surface to the support soil layer.89 0.01 1.d w B E Heavier Deep layer Heavier Shallow Lighter layer Heavier Deep layer Heavier Shallow Lighter layer Heavier Deep layer Heavier Shallow Lighter layer Heavier Deep layer Heavier Shallow Lighter layer Heavier Deep layer Heavier Shallow Lighter layer Heavier Deep layer Heavier 0. The safety factor is calculated against the strength of the support soil layer.94      176 205 24 1 253 0.2 Safety for seismic loading External safety Evaluations for the external safety of the improved soil ground were carried out for.12. Table 7.89 0. The maximum total pressure from both sides is shown in Figure 7.75  x 5 2 V] x . Therefore.89 0.56 .57 1. Using the strength of the improved soil ground.99 221 239 290 646 247 310 330 310 81 78 92 186 147 105 1 16 142 Design 224 269 326 688 1. which has a value only for cohesion and zero for the friction angle according to the current design rule.INTERACTION BETWEEN ORDINARY SOIL 299 side boundary of the improved soil ground was not well known in the case when the surrounding soil layers were liquefied.10 224 269 326 688  Analysis Design 0.91 2.46 (a) shows the pressure of the skeleton and the water at each level from left to right.12 Results of FE analysis and simple calculation (design procedure) Check items Ground kind Subgrade reaction for ISG (kPa) Slip [safety factor] Maximum contact pressure (kPa) Subgrade reaction (kPa) Horizontal shera force (kPa) Vertical shera force (kPa) Shallow layer Building type Lighter Side Long Short Short Short Long Short Short Short Long Short Short Short Long Short Short Short Long Short Short Short Long Short Short Short Analysis 221 239 293 513 1.90 0.5 as shown in Table 7.98 0.83 0. t: M 3 2? V) 0 . a liquefaction analysis was made.51 0. Safety factors for the subgrade reaction were about three times larger than the compression. Safety factors for the slip failure of the examples were over 1. Therefore.44 2.69 1.43 1.48 0.8.46(b). Figure 7.
Internal safety In order to check the safety of the improved soil ground. Shear stresses at the side area of the improved soil ground are very low due to liquefaction.12. Figure 7.35m to G L ..48. Contour lines for the maximum value at a time of the shear stress (T. The maximum stress naturally occurs at the bottom edge of the building or of the improved soil ground as shown in Figure 7.48 is almost horizontal.. Distributions of the maximum value of the shear stress along the assumed horizontal slip line for the depth from GL6. The shear stress close to the foundation has sharp stress concentration at the edge of the building and low stress at the centre of the building.300 PREDICTION APPLICATIONS AND BACK ANALYSIS 3 0 2 0 1 0 0 10 Horizontal distance from centre (m) 20 30 40 Figure 7. The possibility of overturning was very small since the side ground supported the improved soil. are shown in Figure 7.48 Contour of the shear stress . There.47 Stress distribution of the vertical component but further research is necessary for the limitation taking into account the dependency of the confining pressure for the strength of the improved soil ground.49. the maximum stresses at the local point were checked against the strength for normal stress.) under the building shown in Figure 7.. very little moment was seen.0m. This means that the failure line can be developed along the line.
) calculated by the liquefaction analysis. From this study it is concluded that the failure mechanism for the safety check is adequate for horizontal section but overestimated the vertical one. The analyses concluded that the following mechanisms should be checked and are sufficient for .8. (kPa) Figure 7. the accumulated shear stress per unit length along the vertical or horizontal section was compared with the shear strength of the material..3 Remarks The safety of the improved soil ground by the deepsoil mixing method for very large structures was examined by dynamic effective stress FE analyses.. In the horizontal direction. Figure 7. As in the ordinary slipline analysis approach. the result of the calculation was much lower than the material strength. In the vertical section. 7.INTERACTION BETWEEN O R D I N A R Y SOIL 301 Horizontal distance from centre (m) Figure 7. The failure line may be horizontal near the bottom and/or vertical near the edge. The footing type of failure was not triggered by the seismic force in this problem. the values were comparable and less than the failure strength.50 Shear stress distribution along the vertical line Various types of possible shear failure mechanisms were also checked.50 (a) shows the distribution along the horizontal section for the maximum value at the time of the average shear stress. They were both examined using the shear stress (r.49 Shear stress along the horizontal line Distance from left end of improved soil ground (ni) Shear stress r.
7 . Endochronic constitutive law for liquefaction of sand. Dynamic behaviour of ground and earth structure Numerical method and problem specification. Proc. Part 1: design procedure and its verification by effective stress analysis. First In t. and Tatsuoka F. and Zienkiewicz O. Shiomi T. Soil and Foundations.N. on hehaviour of ground and earth structure during earthquake held by Japanese Societj. of Soil and Foundation Engineering in Tokyo. (1989). Proc. Jackson and Son. (ed. Fukutake K. in Numerical Methods in Geon~eclzr~nics. Div. geomechanics  . 5063..D.S. and Hinds J. Creager W.P.C. 201212. and Zelasko J. and Ishihara K. Symp.^ and Foundutions 28 1 Mar 1988. Hyodo M. Ishihara K. Vol. 199204. Yasufuku and Fuji (1988).J. Tokyo. Steel and Timber Dams. (in Japanese) Ishihara K. Geotechnical Engineering Division. Kansai . A. Proc.. and Ohtsuki A. Applicability of 2D analysis and merit of 3D analysis in liquefaction phenomena. J.. Liquefaction analysis for multidirectional shaking. Conference on Eurthquake Geotechnical Engineering. Narikawa M. Kansai . Ishihara K.. REFERENCES Bazant Z. G.. (3) maximum contact pressure at the bottom of the improved soil ground..A. and Uchida A. Hierarchical Model for Sand under Monotonic and Cyclic Lor~ding. (1988). Justin J. 1995 Kishino Y. Kondner R. 15(1). The results of the analyses also helped to determine the seismic intensity and other external forces such as the earth pressure to the side boundary of the improved soil ground. 197. London. A Generalised Plasticitv. of Mech. Nagase H. Soil Dynamics and Earthquake Engineering. Rockfill. On Threedimensionul Evaluation o f Ground Failure. (2) slip failure at the bottom.. and Krizek R. (1963) A hyperbolic stressstrain formulation for sands. 4 Oct 1988. Cyclic simple shear tests on saturated sand in multidirectional loading.U. Swaboda) 169180. 2"d Pan American Conference on Soil Mechanics and Foundations Engineering. 488. Ghaboussi J.F. 6th US National Conference on Earthquake Engineering No. soil. Masuda A. 4559. Pastor M. G. and Fujikawa S. Proc.. I. Proc.(eds. 605627. Ishihara K. 289324.P. (1944) Engineering for dams. (1995). (1981). 131150. on the use of' laboratory test for undrained cyclic behaviour of'soil and insitu test. (1988) Multidirectional irregular loading tests on sand. and Suzuki Y. (1976). John Wiley and Sons. Symp. 649. Symp.. (1986). Soils and Founclations. ASCE. Pastor M. 2944. Simple models for soil behaviour and applications to problems of soil liquefaction. and (4) average shear stress along the vertical line for internal safety.L. Balkhema. Nakamura N. J. 701722. (1988) Liquefactioninduced compaction and settlement of sand during earthquake. A study on building foundation with deepsoil cement mixing method for thermalpower plat. van Impe). Simple method for the determination of the KO value in sandy soil. Undrained deformation and liquefaction of sand under cyclic stresses. 229236 (in Japanese). GT5. Zienkiewicz O. and Dikmen S.C. and Nagase H. and Yamazaki F.302 PREDICTION APPLICATIONS A N D BACK A N A L Y S I S the safety evaluation of the improved soil ground: (1) subgrade reaction of the improved soil ground to the bearing layer.H. Effect of initial shear stress on development of undrained cyclic residual shear strain in saturated sand. (1995). Yasuda S.Kyogikan. Proc. 102 (EM4). Pande and W. 6576. Hatanaka M. Following these conclusions a design procedure based on the seismic intensity method was proposed. qf ASCE.. 111: Earth.Kyogikan report. (1980). and Chan A. 20(1).C. (1975). Murata N. (1998). 3093 14.
B. (1979) Consideration in the earthquake resistant design of earth and rockfill dams. (1995) U . Empirical correlation of soil liquefaction based on SPT Nvalue and tines content. Y. Gpotecl~. F .M.. . Second AsianPasific conference on Computational MechanicslSydney. O N. (GTIO). and Tani S. J.p. Inr. (1984). 7196. Shiomi T .. 12634. Eng. Nurw. Int. and Shiomi T .6001977026.. Sugimoto. (1985). rind A n d . 2 13220. Geotacl~nique. Proceedings of the ThirtyFirst Japan National Conference on Geotechnical Engineering.ric~rl Proccdurt. Shigeno Y. M.. Vaid Y.M ~ t h . 23. DNIU Etzginrering.C.L. Kibayashi M. Zienkiewicz O. Seed H. (1983). 104(GTI). Shigeno. Kimuira T.~ouri pp.. Saito S. (1975) Analysis of slides of the San Fernando dams during the earthquake of February 9. and Hosonii H. Zienkiewicz O. NUIIW. 4.I. Zienkiewicz O.C..C. Lee K. Effect of static shear on liquefaction potential. (1993) Numerical prediction for Model No. California. in partial fulfilment of the requirements for the degree of Ph.F.. and Yoshimi Y. Nuw. (1977) Modeling of leachate and soil interactions in an aquifer. Yoshimi.~.H. Dir.M. 105.~. 2744. 29. 21 5263. K . Nonlinear seismic response and liqueFaction. J. Suzuki K. Con$ Geomec/~at~ics/Mi. 46572.B.C.C.Kon. Method in Geonwcl~. ( r t d AI?N/. Second Int.. pp.. Pyke R. I 1 Nonassociative models for sands.D. (1991) Analysis of Lower San Fernando dam failure under earthquake..H.M. Shiomi T. and Shigeno Y. 477498. Section 3. Simple model for transient soil loading in Anrrl.s. Kim . Shiomi T. ASCE. Seed H. Pyke R.Th. Effect of multidirectional shaking on pore water pressure development on sands... NLIIII. and Hinton E. Gihodo publisher. Dynamic behaviour of saturated porous media. Rollu. (1979). 651688. California11 720 October. 1. t11c.D. Zienkiewicz O. Y.REFERENCES 303 Pastor M.C. Pastor M.. Shiomi T.in Geoniech. ASCE. Issue 4.sure. Shiomi T. 14. 5 4 5 8 (in Japanese) Tokiniatsu K. 951 03. Chang C. h t . and Suzuki.). No. Arulanandan and R. earthquake analysis. (1971). (1991) Influence of liquefaction to pilesoilstructure interaction. Pinder G . and Finn W. thesis presented to the University of California.for Ancr/issisof'Soil Liquclfirction Problems. Van Genuchten M. J. Idriss I. 381404. ( 199 1 ). Scott). (eds. and Saukin W. (1973) Settlement and liquefaction of sands under multidirectional loading... Mrtli. 3'd Annual Municipal Solid Waste Res.P. 3234. GT7. (in Japanese) Zienkiewicz O.P. Berkeley. (1978). 5674. and Zienkiewicz O. It~t. Geolllech. w M(IIIN(I/~ M I I D I A Takenaka Corp.. Muromoto T. Y. Geotech. pp. 2. Geotech. EPA. (1990).C. The generalized Biot formula and its numerical solution. and Martin G.B.J.. Investigation report on the building foundation made of the lattice wall soil improvement for liquefaction prevention. and Zienkiewicz O. J. Geotecl~ Eng. (1993) Consideration of initial shear stress on ground liquefaction...T.. Generalised plasticity and the modelling of soil behaviour. Big. and Leung K. Shigeno Y. 1.. 101. 1511 90. and Makdisi F. 10.3.L. hzfluer~ce f Confnit~g q Pre. Seed H. Verificution c?f'Nutnc.. 10771082. 8. J.C. (1978).M.R. Proceedings q f ' Ve~rijicr~tion Nu~?~erical of' Procwlures for the Antrlysis of Soil Liquefirction ProblemslDavis. (1995). ASCE. Soil unci Fou~~tlutior~. Shiomi T. (1993) Numerical prediction for model No. Div. Liquc~firctiotiqf' Sund (2nd Ed. 9. p. Met11. AIINI. and Xie Y. 307322. Div.. 1. Symp. and Chan A.. Proceedings. J. 12331246. 213219. (1996) Dynamic behaviours of fill dam during earthquake on liquefaction analysis. 2(4).
.
2 Input for earthquake analysis and the radiation boundary.3 Adaptive refinement for improved accuracy and the capture of localized phenomena.Some Special Aspects of Analysis and Formulation: Radiation Boundaries. 48. . The chapter will be divided into three sections corresponding to the topics discussed and each section can be studied independently. We shall introduce these 'finer points' in the present chapter in sufficient detail to allow the reader to follow the current literature and to devise his or her own programme modifications.1 INTRODUCTION In the presentation of the essential theory and the finite element discretization procedures we have deliberately omitted some 'finer points' which on occasion might be essential to obtain more accurate or more generally applicable solutions to realistic engineering problems. 48. These sections are: $8.4 Stabilization of computation for nearly incompressible behaviour with equal interpolation. Adaptive Finite Element Refinement and Incompressible Behaviour 8.
be remarked that. however.1 1 ) and (2. we can proceed as outlined in this book to obtain the solution by time integration of the discretized form of the equations of motion such as those given by equations (2.2 INPUT FOR EARTHQUAKE ANAL YSIS AND RADIA TION BOUNDARY 8. by specifying the input motion at some arbitrary internal boundary shown. it is sometimes convenient to recast the equations of motion in terms of the relative displacement u~ which we define as Known motion specified on all boundaries  Figure 8. The simplest case for the specification of input is illustrated in Figure 8.1 which attempts to model a structure resting on a stratified soil foundation of unlimited extent. It should.13) of Chapter 2.1 Specified earthquake motion: absolute and relative displacements The input for earthquake analysis is based on measured recorded data of actual earthquakes and is generally presented as the values of the displacement u and or of the acceleration ii at the time interval of 0.306 S O M E SPECIAL ASPECTS OF A N A L Y S I S AND FORMULATION 8.* If the time history of the input can be specified.2. In Chapters 6 and 7 we have shown several calculations which correspond to such physical experiments and which model the real phenomena of practice reasonably well. if only uniform motion is specified on the boundaries.02 seconds given for the duration of the earthquake. Such a model corresponds well with such physical models as those of the shaking table or centrifuge where the specified boundary represents a 'box' into which the model is fitted and which moves in a specified manner.1 Specified motion on the boundaries of a 'shaking table box' modelling of an infinite foundation * In the USA such records can be obtained from The Earthquake Research Institute at the University of California at Berkeley and similar sources are available in other countries. With all displacements or tractions at the boundaries specified. . we can use the discretization of Chapter 3 and proceed with the solution of any transient problem.
in the input. We shall discuss this problem in the next section in more detail and suggest how such problems can be dealt with.2 A more realistic model of an 'infinitie' foundation with a specified incoming wave . Figure 8.1 indicating the position of the same limiting boundary but on which the motion will not now be directly specified.2 again shows the problem initially suggested in Figure 8. only the incoming wave motion is specified and that outgoing waves must leave the problem domain unimpeded. In a more realistic treatment of the foundation problem.20) and (2.INPUT FOR EARTHQUAKE ANALYSIS 307 where UE = uE(t) is the prescribed earthquake motion which does not depend on the position. we shall impose somewhat different boundary conditions recognizing the fact that. the numerical computations are identical to those of the absolute displacement if the same initial conditions (e. a few words about the way knowledge of the seismic input wave is obtained. First. however. u = 0) are assumed. The governing equations (2. with the boundary condition on the input boundary being which is replaced by If the relative velocity is used in the finite element discretization of the problem. the input is now the acceleration iiE giving a prescribed body force and this is often more accurately known.g. Incoming plane eathquake wave Figure 8. However.21) of Chapter 2 become now (neglecting the source terms and putting cy = 1).
isotropic wave propagation problem involving the free surface. the transmitted wave is smaller than the reflected one back towards the soil surface. it is very well known that the free surface displacement wave equals the double of the incoming displacement wave. isolated by cut sections AA and BB we note immediately that the problem is onedimensional. Any incoming seismic wave passing from the bedrock to the softer soil layers will amplify depending on the material properties of both bedrock and the soil layers. Clearly if we consider a vertical slice shown in Figure 8.308 SOME SPECIAL ASPECTS OF ANALYSIS AND FORMULATION The seismic signal is usually measured at or near the free surface. and that outside of this homogeneous elastic conditions pertain. Therefore. The significant consequence of the presence of the bedrock lies in the fact that all of the reflected waves are practically trapped inside the soft soil layers. velocity or acceleration) is known at a position corresponding to the model truncation boundary.2. in the following. that the displacements. Here the incoming signal can be easily extracted from the recorded total signal on the undisturbed surface. Even in the case of the elastic nonhomogeneous domain the incoming signal can again be extracted from the total signal recorded on the surface. stresses. In such a case.. horizontal. The need for an arbitrary model truncation emerges in the cases where no distinct base rock exists. The geological conditions at the site will very often be such that the so called 'bedrock' exists as a zone of a significantly more rigid material underneath softer soil layers. homogeneous. The equations governing the problem are still (8. Thus . caused by passing through different material zones and involving a number of internal reflections and refractions at the interfaces between layers of different material. i. If the bedrock is significantly more rigid. 8. a function of the coordinates y and time t . do not vary with the horizontal coordinate x. etc. it will be assumed that the incoming wave (displacement.e. foundation such as we have considered in the previous section but now without a superposed structure. To model such a case correctly it is necessary to reconstruct the incoming seismic wave at the model truncation boundary. or when the extent of the softer soil layers is so great that it would be prohibitive to include the whole zone in a mathematical model. however. as only a small fraction of these can be transmitted back to the bedrock through the interface with the softer soil.2a and b) with the unknown variables remaining as u~ which is now.3. In the simplest case of a onedimensional elastic. Such a situation may also arise when the nonlinear material behaviour can be expected only near the surface and deeper layers (with material properties still far from bedrocklike characteristics) are expected to remain elastic.2 The radiation boundary condition: formulation of a onedimensional problem We return here once again to the problem of a stratified. and it represents the modification of the original seismic wave which is initiated at the earthquake source. the simple fixedbase approach is valid and no transmitting boundary conditions need be imposed on the bedrock level as practically no waves get transmitted into the bedrock.
3 A horizontally stratified foundation subject to vertically propagating compression o r shear waves: (a) the corresponding 1D problem with (b) arbitrary cutoff Thus. Writing the total relative velocity in terms of its components the system (8. (ii) the body forces have been taken into account independently. T o demonstrate the wave nature of the problem we shall assume that in the vicinity of the arbitrary. in the absence of internal flow. 'input' surface CD (and below this surface) the conditions are such that: (i) only isotropic elastic behaviour exists. and finally that (iii) the dynamic phenomena are sufficiently rapid so that the flow in the porous medium can be neglected and k = 0 is assumed.I N P U T FOR EA R T H O U A K E ANALYSIS 309 "An Vertical wave PIopagation tttttt Figure 8. all the derivatives with respect to the x axis are made identically zero.2) reduces. to .
K p= 0 (8. Equation (8. E and u are Young's modulus and Poisson's ratio.. Thus.6b) becomes  d2uY dy2 TU p.310 SOME SPECIAL ASPECTS OF ANALYSIS AND FORMULATION where only total stresses are considered. represents shear waves travelling with velocity . The elastic constitutive relation under isotropic undrained conditions gives and where is the shear modulus and is the restrained axial modulus.9b) Each of the above equations corresponds to the well known scalar wave equation a2Ch which has the solution 1d2d in which c is the wave velocity and 4. 4 = u.6a) becomes on insertion of the above and (8. represent two waves travelling in the positive and negative directions of y respectively (incoming and outgoing waves).and 4.
To obtain the radiation condition we observe the solution sought at the 'cutoff' line C D should represent only an outgoing wave. tends to infinity for fully incompressible solid and fluid situations. Representation of such radiation (or quiet) boundary conditions in the manner presented above was .INPUT FOR EARTHQUAKE ANALYSIS 311 and q5 = u. Using the relationships (8..7) and noting the definitions of (8.13) we will observe that on the boundary C D the tangential traction becomes and the normal traction becomes This is equivalent to the requirement that on the boundary 'dashpots' of suitable strength are imposed in tangential and normal directions.12) and (8. i. We observe immediately that where or that on the boundary to ensure the existence of outgoing waves alone. represents compressive waves travelling with velocity We observe that c.e.
1987a.17(a) and (b) imply dashpots in the tangential and normal direction at any position of the boundary. It is usual to conduct an analysis in terms of the relative displacement UR defined by (8. Many alternative forms of radiation boundary conditions have been developed. The numerical tests of the effectiveness of such a radiation boundary condition are presented by Zienkiewicz et a/.. Here the early work of Smith (1973) which was recently generalised by Zienkiewicz et a1 (1987a) is one possibility.312 S O M E SPECIAL ASPECTS OF A N A L Y S I S AND FORMULATION suggested almost simultaneously by Zienkiewicz and Newton (1969) and Lysmer and Kuhlmeyer (1969). identical results are obtained independently of the arbitrary cutoff position. However. (1987) where it was shown that. (1977).2. on many occasion it has been used effectively on two or threedimensional boundary shapes where the conditions of equations 8.4 where a structure 'perturbs' the simple one dimensional solution of the layered foundation. In the onedimensional case presented here the radiation condition is exact.3 The radiation boundary condition: treatment of twodimensional problems A more general situation of engineering interest is the one illustrated in Figure 8. Alternative methods are discussed by Wolf and Song in their recent text (1996).1) and to apply the radiation condition to this relative displacement only (see Zienkiewicz et al. Kunar and Marti (1981) and also by Zienkiewicz and Taylor (199 1). for a given wave input form. 8. Clough and Penzien.3 perturbed by imposition of a structure (a) and the ID problem (b) .4 Foundation of Figure 8. 1993). Once again the horizontal boundary on which the vertically propagating waves enter the problem domain is treated identically to that of the one dimensional 8 8 ~ e r t i k a ~ prbpakation wave Figure 8. White et a/.
free field. Extension of . solution. identical 'dampers' are placed on this boundary to ensure transmission of the exiting waves (but now these are approximate only as the transmission conditions d o not apply exactly to waves exiting obliquely to the boundary). stresses etc. A homogeneous elastic material is here assumed throughout the space and Figure 8. 1989) as clearly the values of displacement.2. 8.5 Repeatable boundary conditions. Indeed.concluding remarks We have limited our discussion so far to that of the behaviour of the twodimensional foundation layer problems with a vertically propagating wave input. case.4 Earthquake input and the radiation boundary condition .6 we illustrate a test problem where different depths and widths of the analysed domain are used. In the repeatable boundary condition. which is illustrated in Figure 8. More serious difficulties are. Displacement at A=displacement at B. The latter achieves identical results more simply. A possible way of dealing with the boundary conditions on these sections is therefore to impose the radiation damper between the interior region and the onedimensional. acceleration and stress for a typical point at the base of the structure and with different domains of computation. however.INPUT FOR EARTHQUAKE A N A L Y S I S 313 t t t t t t t t t t t t t t Figure 8. Such a treatment is suggested by Zienkiewicz r t 01.7 shows the time histories of displacement. at points which are far away from the superposed structure the solution must be asymptotic to the previously discussed onedimensional one. In Figure 8. Treatment of such repeatable conditions is simple in the finite element context (see Zienkiewicz and Taylor. it is assumed that a sequence of structures is placed on the foundation at regular intervals B.5. (1988) but a simpler alternative is that of repeatable boundary conditions which is also given there. posed on the two vertical boundaries AA and BB and the boundary condition which needs to be imposed on these. are identical on such a section as A or B due to periodicity and the assembly of nodal values at these boundaries is ensured by suitable node numbering. It is surprising to note how little the results are affected by the extent of the domain assumed. Clearly.
of course.1 Introduction to adaptive refinement Accuracy control and adaptive finite element refinement are. the input motion history will be dependent on the position and the determination of this in itself is a major problem. DN a n d SW) the problem to three dimensions for the same wave input is trivial but of course three dampers will now be necessary on the radiation boundaries. Here. We shall not discuss this difficult problem further as it is not frequently encountered in practice. of course. even if the material behaviour is linearly elastic. However. of importance in all analysis problems. . Greater difficulties are presented by problems in which the earthquake (or shock) waves enter the boundary obliquely or indeed horizontally. 8.6 Twodimensional model problem a n d Three mesles (SN. once such motion is established it is possible to apply radiation boundary conditions throughout.3 ADAPTIVE REFINEMENT FOR IMPROVED ACCURACY AND THE CAPTURE OF LOCALIZED PHENOMENA 8.3.314 SOME SPECIAL ASPECTS OF ANALYSIS AND FORMULATION Possible arbitrary model truncation \ \ Shallow and wide mesh (SW) Vertically propagating incoming seismic wave (El Centro) Deep and narrow mesh (DN) Shallow and narrow mesh (SN) Figure 8. However. the need for adaptive refinement is even greater when plastic deformations are pronounced as here often very sharp gradients of displacements can occur.
.
A perforated bar (a) initial mesh (b) final adapted mesh with elongation DOF 1039 (c) initial material configuration (d) final material deformation leading in the limit to localized displacement discontinuities. but also stretched in the direction of this discontinuity which is indicated by the material deformation pattern. wish to find the maximum values of the loads carried by the structure of a particular size at a particular stage of deformation or indeed the maximum loads occurring throughout the deformation history. the exact solution can be approached as closely as possible.8 First adaptive solution of a purely plastic deformation problem. How far the refinement should proceed is a question which is difficult to answer precisely and here we will need to revert to the notion of error tolerance. . However. In adaptive refinement of many engineering problems in which the errors are distributed throughout the domain. In this analysis the mesh was adaptively refined with elements not only being reduced in size near the displacement discontinuity. for instance. Zienkiewicz and Huang 1990). The 'capture' of this discontinuity as it actually develops can be achieved approximately with finite elements using a continuous interpolation. We may. Such an approach is not recommended in studies of plastic deformation or localization as our interest cannot in general be described by a single 'number'. usually as a fixed percentage of the total value of energy norm in the domain.316 SOME SPECIAL ASPECTS OF ANALYSIS AND FORMULATION Figure 8.8 we show a typical plastic deformation pattern occurring in a uniformly stretched tensile specimen with a small perforation (viz. by sufficient refinement. the interest may lie in determining precisely the position of the region where large strains or discontinuities occur. it is convenient to introduce error norms (such as the frequently used energy error norm) and to require that the error of that norm be kept below a certain value. Alternatively. In Figure 8.
(see Zienkiewicz and Zhu. as on many occasions the feature occurring at high gradients is almost onedimensional.ADAPTIVE REFINEMENT 317 For such problems we can separate the process of error determination and of refinement of the mesh.18) allows element elongation to be included in the refinement. Indeed. but alternatively convergence to exact solution can be studied by simply reducing the constant C in refinement. . (1987) and is very effective in the capture of shocks. can be specified more closely.~ could be replaced as I . estimates of error are possible by using various recovery procedures. diverge by an angle 0 then the limit on /. we achieve a solution which captures well all local discontinuities and which is efficient in achieving the progression which gives overall accuracy.. This specification was first formulated by Peraire et ul. At any stage of refinement.. 1992)... the function of interest. the upper limit h..1 and R is the radius of curvature. if the contours separated by the value of h. in a truly onedimensional feature the maximum sizes of an element along its direction would be arbitrary and any reasonable value would be fixed on the maximum element size h. by requiring that such a quantity (or indicator) as where C is constant between all the elements. the elongation of elements can be computed directly in terms of principal curvatures. is their minimum size and s the direction of the maximum gradient of 4. hnli. With curved contours it is often specified where cw is circa 0. as a bound based on the variation of the smallest dimension of the element. These type of procedures are discussed in detail by Zienkiewicz and Wu (1994) in the context of fluid mechanics.. for instance. This quantity can be interpreted as the maximum value of the first term of the Taylor expansion defining the local error of the scalar quantity 4. for instance. However.. Thus the latter can be guided. An alternative refinement indicator has been used for the longer time in fluid mechanics.. if the contours of the function 4 diverge or are curved.. Here.. The use of the indicator defined by (8.. Thus. This is a requirement that specifies the minimum size of elements. By ensuring that the mesh is generated so that the quantity C is constant throughout all elements.
In Chapter 3 we have already mentioned that special conditions have to be satisfied by mixed finite element forms for incompressible.318 SOME SPECIAL ASPECTS OF ANALYSIS AND FORMULA TION It appears that the first indicator (i. . or nearly incompressible..9 Adaptive solution of the problem of foundation collapse with an ideally plasticelastic material (a) ecentrically loaded footing (b) final adapted mesh and deformed configuration showing displacement discontinuity. 1995a and b). Indeed. such behaviour will Rigid and rough footing Figure 8.18)) is most efficient in the capture of narrow discontinuities. but both provide a remeshing which gives a rapid convergence and reduction of both local and global errors.9 shows how an adaptive analysis based on the first indicator can model discontinuity developed during the failure of the foundation under an eccentric load. Here a von Mises type of yield surface is used with ideal plasticity assumptions (Zienkiewicz et a/. Figure 8.e. that of (8. behaviour such as is encountered under undrained conditions.
3. problem illustrated in Figure 8. but the basic difficulties have been overcome many years ago and are described in Zienkiewicz and Taylor (1991). For adequate solution it is always necessary to use here special mixed forms of elements which are outlined in Chapter 3. In the two examples quoted already we used the T6CI3C triangle where six nodes define a quadratic variation of continuous displacement and three nodes interpolate pressures in a continuous manner. In Figure 8.ADAPTIVE REFINEMENT 319 occur in many applications of plasticity using von Mises or Tresca yield surfaces.11. frequently observed in clays. The analysis of plasticity problems with a negative hardening (or softening) modulus. Further. A serious problem with adaptive analysis of nonlinear problems of plasticity in which the results are path dependent is that of data transfer between the various stages of analysis. H. sliding surface type of deformation. the adaptive solution starting from either refinement shows nearly exact values of the collapse load. statistically. etc. In this example we assume plastic. If the extension imposed on the specimen continues beyond that peak then only that one element shows the plastic deformation. It is clearly noted that for the same subdivision the 'bad' mesh gives answers which are always inferior to those of the 'lucky' mesh. thus ensuring a constant degree of accuracy (Zienkiewicz et al. However. the behaviour of a onedimensional. softening. In principle. from the mesh of the previous step to that used in the next increment. necessitates the transfer of history dependent data such as stresses.plastic problem in which a strong localization occurs. the control of the error should be achieved at each load increment separately and this. 8. of course.2 Localization and strain softening: possible nonuniqueness of numerical solutions Strainsoftening behaviour is a phenomenon frequently encountered in soils and invariably it leads to a very localized. the reason for localization only becomes clear if some specific cases are examined. the 'bad' mesh in which these lines are orthogonal to the slip surface. and b) with results shown in Figure 8. (1995a. we perturb the yield stress in an arbitrary element so that only that element yields when the load is applied and thus gives the peak yield stress. However. strains. This is well exhibited in the so called 'slickensides'. 1998). and the other. Depending on the ratio of the . To avoid difficulties we have reanalysed the problems in each of the previous cases from the start of loading for every new mesh developed.10 we show again an analysis of an ideal elasto. is in itself a complex task. Indeed such a procedure has also been used quite effectively in transient analysis of the San Fernando dam by Zienkiewicz and Xie (1991) and Zienkiewicz et ul. Consider. Here two forms of regular mesh are comparedone named 'lucky' mesh in which the triangle subdivision lines follow approximately the slip surface..12. bar type. for example. all others unloading elastically. currently new procedures of transferring data have been developed and it is now possible to change the mesh at each load increment. behaviour and consider analysis of the bar divided into a number of equal elements of length 'lz'. However.
ting T Figure 8. obviously. von Mises. linear continuous pressure) ( a ) geometry data. continuous displacements.320 SOME SPECIAL ASPECTS OF ANALYSIS AND FORMULATION element size h to that of the total length . This can reach negative slope values and allows only elastic unloading which.10 Failure of a rigid footing on a vertical cut. Ideal. (d) Mesh 6 (adaptive solution obtained from Mesh 3). plasticity and triangular T6Cl3C element (quadratic. (e) displacen~ent vectors. is not correct. displacement results for various meshes. a progressively steeper unloading branch of t the load deformation plot will occur. But certainly the steepening of this slope will increase to infinity for a finite value of / I and will imply a displacement discontinuity or full localization. bad) (g) Load. (c) Mesh 3 (coarse. 'bad'). (f) Effective strain contour for Mesh 1 (fine. (b) Mesh 2 (fine 'lucky'). .
(1988) and Belytschko and Tabarrok (1993) describe some of the possible procedures which range from the consideration of material as a Cosserat medium. in other geometrically more complex problems the stress concentration. However.12). Belytschko er al. through so called gradient plasticity. Bazant and Lin (1988). (1987). De Borst er al. (b) adaptive In the example quoted. (1993). the example discussed shows up another feature of the problem. always causing a localization with softening material behaviour. Even discounting the results obtained by the use of the coarse. to a simple failure energy consideration introduced in the last of . bad. mesh as being very unreliable. i.10 for which now a softening modulus has been assumed and different mesh subdivisions used in the solution. Ortiz et al.e. will act in precisely the same manner. While the reason for this has been hinted at it in the simple example of Figure (8..13(a) we illustrate the fairly large discrepancies which occur in the estimate of the maximum load for the problem illustrated in Figure 8. However. the manner in which the problem can be overcome has supplied many researchers with material for exercising their ingenuity.ADAPTIVE REFINEMENT 321 Figure 8. etc. we note a difference of about 20% in the estimate of the maximum load capacity when the simulation is achieved by meshes which for ideal plasticity give almost identical answers. In Figure 8. that of numerical nonuniqueness as the slope of the unloading portion of the displacement load curve depends largely on the size of the element used. the localization was caused by a small weakness due to the statistical nature of the material strength behaviour.11 Earthquake analysis of lower San Fernando Dam (a) initial mesh refinement at t75 seconds (c) adaptive refinement at t = 30 seconds. This nonuniqueness of the problem becomes most serious in multidimensional behaviour in many structural problems.
(a) constant plastic modulus (b) mesh dependent plastic modulus .13 Strain softening (H = 5000): comparison of reaction vs prescribed displacement for various meshes using T6C13C element.322 S O M E SPECIAL ASPECTS OF ANALYSIS AND FORMULATION Figure 8. Figure 8.12 Nonuniquenessmesh size dependence in extension of a homogeneous bar with a strain softening material (peak value of yield stress a. perturbed in a single element). (a) stress u versus strain e for material (b) stress ii versus average strain c = u / L assuming yielding in a single element of length h.
It will be observed that the above discussion leads to two conclusions: .24) as the size of the elements modelling the localization decreases. The procedure considers. It appears therefore necessary to reduce the softening modulus in the manner of (8.ADAPTIVE REFINEMENT Work dissipated in failure per unit volume 323 / Figure 8. in the opinion of the authors. In Figure 8. in the manner common to that of early theories of fracture.14 Work dissipation in failure of the material these references.. The area under the full triangle is the work required to cause this failure and. deals adequately and in a simple manner with the difficulties encountered.e. the work requirement to fail a unit width of the element which must be kept constant i.b) which gives an almost identical failure load obtained by two very different meshes. This indeed was done in the case of a problem illustrated in Figure 8.10(a) and the results are shown in Figure 8. becomes where H i s the softening modulus.13(a. is 1 a.h = constant 2 H  This would be invariant only if where C is a constant. in a unit volume of material. If an element of size h in the direction of maximum straining is to model failure correctly.14 we show a typical stressstrain relation with strain softening in which failure is reached. namely Griffiths (1921). the constancy of work required for failing the material and requires the energy to be independent of the discretization used and therefore to be a pure material property. We shall only refer here to that last procedure which.
while in the second a quadratic approximation is used for the displacement. and (ii) that the softening modulus cannot remain a material constant but must tend to zero (i.324 S O M E SPECIAL ASPECTS OF A N A L Y S I S AND FORMULATION (i) that with strain softening. (1993) and Oliver (1995). (1986). a bilinear interpolation is used for both the u and p variables. tend to give identical results to the adaptive refinement if equation (8.1 The problem of incompressible behaviour under undrained conditions In Chapters 2 and 3 we have already mentioned the difficulties which can be encountered when the standard finite element approximation is used to model incompressible. in recent years much research effort has been devoted to the introduction of stabilizing procedures which would allow arbitrary interpolation (say equal interpolation) of both variables to be used effectively. in the limit.e. Such behaviour will be obtained when the permeability is very small and when the compressibility (1/Q) decreases. Such stabilization can. In the first element of quadrilateral form. Hughes et al. While the use of such correct interpolations is desirable and we have based our code on this assumption. It is clear that such a model will. The adaptive refinements of the type here discussed have been introduced by Zienkiewicz et ul. Brezzi and Pitkaranta (1984). . Here the work of Schneider et a1 (1978). This and other papers in the field. in the up approximation we have mentioned already that satisfactory behaviour of the solution can be obtained under all circumstances when the BabuskaBrezzi condition or the equivalent of the patch test is satisfied. (1997) shows the various approaches suggested in the literature. giving no softening) as the size of the element tends also to zero to present a consistent work estimation. Some permissible interpolations are shown in both Chapters 2 and 3 and in the two examples illustrating this chapter we shall show how the unstable behaviour of the 'illegal' Q4lP4 element can be eliminated by the use of the acceptable Q9/P4 interpolation. or nearly incompressible. 8. In other words. indicate that adaptive refinement is a feature which can improve the results of analysis significantly although with experience reasonable engineering estimates can be obtained without this feature. lead to more efficient and simple formulations and a recent paper by Pastor et ul. This idea can be incorporated in a material model with concentrated localization singularity and has been introduced by Simo et ul. without doubt. this happens when the elastic bulk moduls of the pore water is very high. behaviour.4 STABZLZZA TZON OF COMPUTA TZON FOR NEARL Y INCOMPRESSIBLE BEHA VZOUR WITH MIXED ZNTERPOLA TZON 8. However. (1995a and b) from which the examples and previous figures have been quoted.24) is used. localization will always occur in the failure zone and this will show a continuously decrease in size with the element size h.4.
S T A B I L I Z A T I O N OF COMPUTATION 325 Hafez and Soliman (1991) and Sampaio (1991) suggests many alternatives.21) of Chapter 2 (or (8.27~) noting that a =a  m p and da' = Dds . Thus we have the definition and in each time step it is simple to establish once values of v"+l and p"+l have been computed. operator split procedure which is effective in dealing with the incompressibility problems arising in geomechanics and which follows the methodology originally suggested by Chorin (1967 and 1968) and extended by Zienkiewicz and Codina (1995). It is convenient to introduce the velocity. (1999).4.20) and (2. We can write these governing equations as pv = sT/  Vp + pb and (8. stabilization process In this section we shall outline the semiexplicit timestepping. again neglecting the source term and introducing the new variable v.2a) and (8. In this chapter we shall discuss only this last process of stabilization as it appears to be the most simple and efficient. v.2b) of this chapter) rewritten in terms of effective stresses with CY = 1. as the basic variable and to compute the displacement increment by subsequent integration. 8. Some of these were shown by Zienkiewicz and Wu (1991) to derive very simply from the same roots of timestepping analysis.2 The velocity covvection. The starting points for the development of the algorithm are equations (2. The motivation for most of this work lies in problems of fluid mechanics and their numerical solution and it was shown recently by Zienkiewicz and Codina (1995) that an algorithm using the operator split procedure suggested by Chorin in 1965 automatically provides the desired stabilization. The use of such stabilization in the context of geomechanics was first made by Zienkiewicz and Wu (1994) and extended by Pastor et al.
233.. This can be done in a variety of ways by well known procedures discussed in finite element texts (see.27a) in two steps. and All these matrices are defined in Chapter 3 in (3. Zienkiewicz and Taylor.21) needs to be used for strain calculation. Equations (8. Following the standard procedures of Chapter 3 with we obtain the following after application of the Galerkin process. The operator split algorithm solves (8. In the second part the velocity v* is corrected implicitly in terms of known pressures using The above can only be evaluated after Ap is established if H2 # 0. The determination of the pressure increment A p and hence of p"+l requires the solution of (8. for example. We now write the implicit time approximation as Here various values of 01 can be used but .30) and (8.31) must be discretized in space before proceeding with numerical calculations. In the first the quantity v* is calculated explicitly from where the RHS is computed at t = t. however. mention that the evaluation of both V* and Vn+' is fully explicit if the mass matrix M is diagonalized.27b).326 S O M E SPECIAL ASPECTS OF ANALYSIS AND FORMULATION is the constitutive relation and that (3. 1989). In what follows we shall use O2 = 112 for good accuracy but any values of it in the range can be chosen. We must.. provided that A t satisfies certain stability limits.26) and need not be repeated here.
3 where the limits of applicability of various formulations are tested and for which exact solutions are readily available.16 we show solutions obtained by the use of 2D elements.15 we show the details of the problem and in Figure 8.293. The first of these is the soil layer subject to a periodic surface load. In Figure 8. H and Q are defined in Chapter 3 by (3. Using the computed values of u* and (8. The first uses 20 Q4P4 elements and shows oscillations which are very pronounced. We must remark that with Qr 5 $ no stable solution is possible.32).18 a fully twodimensional problem of a foundation load is solved again showing similar results. The only new matrix occurring now is H* which is the approximation to the Laplacian operator.2.31) we can rewrite (8.S T A B I L I Z A T I O N OF COMPUTATION 327 is particularly convenient and accurate.17 and 8. By the usual procedures we find We shall delay the explanation of the reasons why the split operator procedure permits the use of arbitrary interpolations for u and p (N and Np respectively) and shall first illustrate its effectiveness in examples. The second one shows the very close approximation and suppression of oscillation obtained using the Q9P4 element as well as the new stabilizing algorithm.4. Here we shall only use the up formulation and shall demonstrate how the very oscillatory results obtained by an equal interpolation can be improved by the use of the stabilization just described. . This again proceeds in the manner previously described and we now have In the above the matrices S.3 Examples illustvating the effectiveness of the opevatov split pvocedure Two examples are here quoted. Indeed this problem is identical to the one used in Section 2. In Figures 8.36) as from which A p can be established after discretization. 8.
379 radls Figure 8. = 0 k.333 X 7.e.5 giving a linear form As we mentioned there.31) and (3.2.= 0 11.2 3 2. i.30) tend to zero.~ 10'mn/s 0. PW apw  a.492 10 (pa) 0. when and H . This zero limit leads to a zero diagonal which occurs also in steady state equations of Section 3.0 x 10'(N/m3) The height of the column has been taken as L = 30m and the excitation frequency chosen is W = 3..e.4.328 S O M E SPECIAL ASPECTS OF ANALYSIS A N D FORMULATION H Im 1 q= 100 exp (lwt) t.4 The reason for the success of the stabilizing algorithm In Chapter 3 we have indicated the main reasons for the difficulties encountered in solving the problem where incompressibility is approached.15 Example I a saturated soil layer under a periodic load.0 x I0 (N/m3) 1. We first made a comment on these difficulties when discussing the Jacobian matrix used in the solution of an iterative step by the NewtonRaphson procedure where the matrices of (3.I. . 8. ( ~ ~ 1 ' ) ~ k ~ ~ ~ ' d f l + 0 which occurs when the compressibility and the permeability of both tend to zero. satisfactory solutions can still be obtained but these require that the number of parameters describing the variable u must be greater than these describing the variable p i. n E v p.
" 1. vertical pressure amplitude distribution Note: Exact solution is very close to the stabilized solution .S T A B I L I Z A T I O N OF COMPUTATION 329 Medium compressibility Small compressibility Q * = 10 MPa 4 Q*= I0 MPa 9 ".16 Example 1.5 ~ 1 4 (a) Solution with standard column with 20 Q4P4 elements 0.0 (b) Solution with standard column with 20 Q8P4 elements (c) Stabilized procedure with 20 Q4P4 elements Figure 8.
k = lo' mis. However. As we have recast the problem in terms of velocities we shall linearize using these variables and write where K includes a time integration operator . we shall find that even in the limiting case (i. (c) Q4iP4 elements with stabilized procedure This is a necessary condition for avoiding singularities and can be readily achieved with certain interpolations. (b) transient load applied data Figure 8. (a) direct use of implicit algorithm with Q4P4 elements.17 Example 2: a saturated soil foundation under transient load.2. with zero compressibility and permeability) a nonzero diagonal will be obtained a n d stability can always be achieved.18 Example 2: two dimensional foundation pressure contours computed for small permeability and compressibility Q* = lo9 MPa. if the problem is recast in the manner given in Section 8.330 SOME SPECIAL ASPECTS OF ANALYSIS AND FORMULATION E 30 MPa (a) (b) Figure 8.4. (a) the problem domain. (b) direct use of implicit algorithm with Q8P4 elements.e.
(1984) On the stabilization of finite element approximations of the Stokes problem in Efi'cient solutions uf elliptic prohlenn. and Pitkaranta J. Eng. Brezzi F. 42454265. being the obvious choice.. (1988) Nonlocal yield limit degradation. Appl. and Englemann B. Fish J.. 79.39) by using (8. especially those in which nearly explicit solution is going to be used.48) can be written as and a nonzero diagonal is found to exist in its finite time steps. Int. ( 1 993) Hadaptive finite element methods for dynamic problems with emphasis on localization. although the timestep length is now given by the speed of the shear wave and hence is not too restrictive. 36. J.35) as Eliminating v* from (8. Eng.47) and (8. B. Num.. 5989. The procedure outlined unfortunately results only in conditional stability. Notes on Nzrmerical Fluid Mechanics. . and Lin F. Metl~. E. (1988) A finite element with embedded localization zones.35) we have The two equations (8. Belytschko T. and Tabarrok M. This seems to achieve complete stabilization and any interpolation of the V / u and p variables can be used with equal interpolation.34) and (8. Vieweg. Eng. of course. Comjl. Metll. Metl~. Mec11. We find that the integration of the new stabilization procedure into the computer code is reasonably economic and can well be made use of in many programmes. 18051 823. Int. J. P. Belytschko T.. 26. Wiesbaden. Nunz.REFERENCES 331 In the steady state and we can write the sum of (8. REFERENCES Bazant Z.
. A. Comp. K. Valliappan S. 1226. Vilotte J. Brittle fracture. V. D. 61. Engrg. Appl. Pastor M. and Needleman A. Simo J. Metlz. Oliver J. 1996) Ortiz M . 189214. Pastor M.. (1986) A new finite element formulation for fluid dynamics. Oliver J. C. Meth. A. Balkema. M. (1991) Numerical solution of the incompressible NavierStokes equations in primitive variables on unstaggered grids. Lysmer J. Peraire J. J. 1323. J. Franca L.. Li T. 911561CP.. and Quigley J. Pastor M. AIAA Conf. De Borst R. R.. And. 78 1804. COMPLAS IV. Phy. 90. 647 1. (1991) A PetrovGalerkin Formulation for the Incompressible NavierStokes Equations using Equal Order Interpolation for Velocity and Pressure. 859877. 368379. (1968) Numerical solution of incompressible flow problems. Studies it1 Nunwr. 72. (1921). (see also Int. 59. Comp. 449466. W. P. 8599. and Zienkiewicz 0 .. 62. Li X. Hughes T. B. (1993) Dyntmiics qf Structures (2nd edn). (1997) A mixed displacement pressure formulation for numerical analysis of plastic failure. ASCE E M . (1991) Adaptive remeshing for shear band localization problem. (1978) Finite element analysis of incompressible flow incorporating equal order pressure and velocity interpolation in Nu~nericulMethodsfor Luminur uncl Turbulent Flow.. Leroy Y...332 S O M E SPECIAL ASPECTS OF A N A L Y S l S AND FORMULATION Chorin A. and Marti J. Kunar R. and Kuhlemeyer R . J. Hiihlhaus H. p. h ~ t J. Cotnputationul Pl~ysics. and Armero F. 455479. and Zienkiewicz 0 . 277296. (1969) Finite dynamic model for infinite media. J. Plymouth.. Roy.... Inc. C. Meth. Struct.. Conlp. ASME Engineering Mechanics Division. C. No. (1993) An analysis of strong discontinuities induced by strain softening in rate independent inelastic solids. Appl.. (1991) Adaptive mesh refinement in strain localization problems. Zienkiewicz 0 . McGrawHill. 183204. Raithby G. Archive of Applied Meclzcinicshzgenieur Arclziv. 3039. (1987) A finite element method for localized failure analysis. EM5... Num. 99121. Eng. Vahadati M.. J. 2.. Hafez M.. Mech. Rubio C. (1993) Fundamental issues in finite element analysis of localization of deformation. Eng. and Penzien J. Pentech Press. 12. and Yovanovich M. (1999) Stabilized finite elements with equal order interpolation for soil dynamics problems. 46. Circumventing the BabuskaBrezzi condition: A stable PetrovGalerkin formulation of the Stokes problem accomodating equal order interpolation. 6.. 95. (A). 163. (1967) A numerical method for solving incompressible viscous problems. NO. and Soliman M. L. and Huang M. Metll. A. C. 949965. (1973) A nonreflecting boundary for wave propagation problems. and Zienkiewicz 0. J. Chorin A. and Zienkiewicz 0. and Lee I. Comp. CINME. (1987) Adaptive remeshing for compressible flow computations. 221. 61. and Balestra M. Eng. Morgan K.. Barrelonu. Ortiz M. N. Meth.  . Comput.. A r c l z i ~qj' ~ ~ ~ Computer Methods in Engineering. Appl. Peraire J. Peraire J. 15. 2.. Mech. ASCE E M . Soc. Clough R . Eng. Griffiths A. and Pamin J. Nutw . Smith W. Eng. J. (1977) Unified boundary for finite dynamic models. Sluys L.. Schneider G . B. Mech. White W. E. 492503. 31 No.. Rotterdam. Quecedo M. 11351 149. 10. Proc. (1995) Continuum modelling of strong discontinuities in solid mechanics in Proc. Cornput. EM4. A. De Sampaio P. 103. New York. (1992) Numerical analysis of localization in Nutnericcrl Models in Geotnechunics. Proc. (1981) A nonreflecting boundary for explicit calculation in computational models for infinite domain mediastructure interaction.. Mira P. P. C. R. P h j x . Pastor M. J. Cornput. D. J. Mech. C. Comp..
Comm. Dum Engineering. Part I: The split characteristic based scheme. (1994) Automatic directional refinement in adaptive analysis of compressible flows. Dordrecht. John Wiley & Sons. and Seed H.Volunze I : Basic Formulation and Linear Problems (4th edn).. Conf Nurn. London. and Xie Y. (1998) Error Estimate and adaptivity using recovery procedure. C. (1995) A general algorithm for compressible and incompressible flow. (1987b) Earthquake analysis procedures for dams. Zienkiewicz 0. Zienkiewicz 0. Z. Q. Nurn.. C. 17. 610 July 1987. C. and Taylor R.Volume 2: Solid and Fluid Mcd7cmics. Comm. 37. Mech. Chichester. Int.C.. 307322. Dynamics und Nonlinearity (4th edn). May. and Song C. Pande G. I . P. Z. in Fluids. 127148. Bicanic N. Q. (1995a) Localisation problems in Plasticity using finite elements with adaptive remeshing.. (1987a) Single step averaging generalized Smith boundary .. McGrawHill Book Company. 11841203. Symp.. Con~p.). C. Eng. Shock and Vibration. and Wu S. Mechanics. Appl. (1990) A note on localization phenomena and adaptive finite element analysis in forming processes. Zienkiewicz 0. 869885..C. Nurn. Proc. Eng. Zienkiewicz 0. C. Swansea. C. C. C. and Huang G . Metk. and Shen F. C. 314. Int. McGrawHill Book Company. Eng.C. (1995b) Softening. Meth. and Middleton J. Anal. M. Nunz. (1991) The Finite Element Method . Bicanic N.REFERENCES 333 Wolf J. 98109. Zienkiewicz 0. C. on Finite Element Techniques. 32. and Zhu J. and Zhu J. 13. (1992) The superconvergent patch recovery (SPR) and adaptive finite element refinement. Comp. Wu J. (1991) Incompressibility without tears! How to avoid restrictions of mixed formulations. Mech. (1992) A new algorithm for coupled soilpore fluid problem. St. Zienkiewicz 0. J. Zienkiewicz 0.. Meth. Meth. Appl. Capture of discontinuous solutions. Zienkiewicz 0. London. (1996) Finite Element Modelling o f Unbounded Media. and Wu J. Martinus Nijhoff Publishers. J. 1091 38. (1989) The Finite Element Method .. (1993) A new algorithm for the coupled soilpore fluid problem. 2. (1969) Coupled Vibrations of a Structure Submerged in a Compressible Fluid. Eng. 207224. localisation and adaptive remeshing. and Newton R. Stuttgart. Num.C.C. and Huang M . Shock und Vibration.. . Num. Zienkiewicz 0. N.. Zienkiewicz 0. 1. 215233. Int. and Wu J. Huang M. Meth. 20. 11.) Vol.: Theor. Aj~pl. Huang M. Geomech. Mrth. C. Clough R.. 1.y crnd Applications ( N U M E T A 87) (eds. Pastor M . and Taylor R.. Int. Zienkiewicz 0. Huang M. C. and Wu S. and Pastor M. E. L. L. C. SpringerVerlag. Doltsinis (ed. paper T4911. Zienkiewicz 0.. (1991) Analysis of the lower San Fernando dam failure under earthquake. Boroomand B.. No. Meth.Nurn.. 7176. 19. J. and Shen F. (1988) Earthquake input definition and the transmitting boundary condition in Advtrnces in Conl~~urutioncrl Nonlinear. Zienkiewicz 0. J. Wu J. Int. CIGB ICOLD Bulletin 52. W. 101. Zienkiewicz 0.(a) transmitting boundary for computational dynamics in Proc. pl15. and Codina R. Zienkiewicz 0. B. Zienkiewicz 0. 6. Int. 2189. Zienkiewicz 0.
.
SWANDYNE I1 could be described.2 OUTLINE DESCRIPTION OF DIANAS WAND YNE I1 DIANASWANDYNE I1 is the acronym of Dynamic Interaction And Nonlinear Analysis SWANsea DYNamic version 11. preprocessing program: DYNMESH 2.bham. due to space limitation. 3. The program is an improved version of . only a brief introduction to the finite element program DIANASWANDYNE 11. have been made available on the World Wide Web at URL: http://www. 9. Also available is a limited version of the executable of GLADYS2E which is an explicit implementation of the uw formulation as described in Chapter 3.Computer Procedures for Static and Dynamic Saturated Porous Media Finite Element Analysis 9. postprocessing program: DYNPLT soil model tester: SM2D postprocessing program for soil model tester: SM2DGRH 4.uk/ CivEngIswandyne as well as the associated example data and result files. However.ac. or simply. A limited (not more than 50 elements) executable version of the program together with 1.1 INTRODUCTION This volume on computational geomechanics would not be complete without a description of the finite element implementation.
draining and undrained). The solution process used is the profile solver. It is a twodimensional program which incorporates plane strain and axisymmetric analysis. The programming generally followed the one recommended by Irons. the incremental strain can be further subdivided to increase the accuracy of the incremental stress integration. the bumpy road of Cambridge Geotechnical Centrifuge. most of the relevant properties can be made variable in the program. consolidation and dynamic (drained. the void ratio and the Biot alpha. DIANASWANDYNE I1 is capable of performing analysis for static (drained and undrained). boundary traction or influx and pressure loading on solid phase. it is not often necessary to use a more rigorous formulation.6.336 COMPUTER PROCEDURES FOR STATIC AND DYNAMIC the DIANASWANDYNE I (Chan 1988). except for codes which were obtained from external source. It is intended for static. There are data available for earthquake trace of the following: the El Centro NS earthquake. Finite deformation is accounted for using the Updated Lagrangian Formulation and finite rotation using Jaumann stress rate is included also. the undrained condition can be imposed at element level or simply by using a global nodrainage condition with a drained or draining analysis. the pressure is not known for all locations and it will evolve with time. The variation is useful for initial stress analysis or centrifuge swingup operation. Katona and Ziekiewicz 1985). All the abovementioned can be a function of time. with features such as five characters for variables and six characters for subroutines and functions used. The earthquake is prescribed as a boundary acceleration so that no further assumption is needed for its application. The time integration is done with the Generalised Newmark method (Whitman 1953. Katona 1985. the San Fernando NS earthquake. Both tangential stiffness and BFGS method are available for nonlinear iteration. finite deformation and finite rotation Biot equation with the up simplification with the fluid acceleration neglected as described in Chapter 3. the fluid phase is either neglected or its pressure fixed at constant values. Newrnark 1959. As for draining analysis. External Loadings can be given in the form of boundary displacement or pressure. Thermal phase has also been included. Body force can be applied to each element separately and can be a function of time. For drained analysis. consolidation and dynamic analysis for problems in geomechanics. Lastly for undrained analysis. As the soil models used are primarily small strain models. the fluid density. Both horizontal andlor vertical earthquakes can be applied. As fully nonlinear behaviour is expected in geomechanics applications. Within a given time step. The program used the finite element method with triangular and quadrilateral isoparametric elements in spatial domain. The system dependent functions such as time and date functions are restricted to a limited number of subroutines and they are described in Section 9. earthquake trace from various centrifuge sites via the VELACS project and Bristol shaking table type earthquake. . This includes the solid density. The governing equation that is being solved is the fully coupled. The permeability can be a function of pore pressure and void ratio. The program was written in standard Fortran 77 with a number of minor Fortran 95 enhancements such as the use of IMPLICIT NONE and END DO.
A number of the significant service subroutines are described in Section 9. Non. AlTabbaa 1987). Effective stress state. absolute and relative displacement. Acceleration. Deformed shape. Sander and Rubin 1979) CamClay family: Original CamClay and modified CamClay Kinematic hardening family: AlTabbaa and Wood model (AlTabbaa and Wood 1989. The subroutines are listed in the order that they are first called. Tresca. incremental strain and pore pressure can be given at the gauss points. Hamilton et a1 1998).associative elastoperfectly plastic Mohr Coulomb and Drucker Prager model (Owen and Hinton 1980). This includes displacement. Anisotropic Elastic model (Graham and Houlsby 1983).3. Due to the size of the program.DESCRIPTION OF MAJOR ROUTINES 337 Plotting can be done with all the nodal values of the mesh. total and excess pore pressure can also be plotted from anywhere in the spatial domain. stress distribution and plastic development can also be traced. Basic mesh plotting facilities are available and the program is accompanied with a postprocessing program to handle result processing and plotting. pore pressure contour. Classical Elastoplastic family: von Mises. Elastic model with Moduli varying with mean effective confining pressure and with a Coulomb friction envelope. There are quite a number of constitutive models currently available in the program including: Elastic family: Linear Elastic model. Generalised plasticity model: PastorZienkiewicz MarkI11 model (Pastor and Zienkiewicz 1986) DESCRIPTION OF MAJOR ROUTINES USED IN DIANAS WAND YNE ZI In this section. CAP model (Chen and Mizuno 1990. Two surface kinematic hardening plane strain model for sand (Hamilton 1997.4.1 The top level routines The top two level of the program has been given in Figure 9. .5.1. 9. Simple description of significant constitutive models are given in Section 9. the major routines used in DIANASWANDYNE I1 are described. pressure. It gives the version number and copyright message before calling the main subroutine DYNMAIN. velocity. acceleration and rate of change of pressure. DYNE11 is the dummy program unit. only major subroutines are described.
The names of files for input and output purposes are also determined.3. INMESH inputs the finite element mesh information from DATA and MESH file. Also keywords such as STATIC.4 Figure 9. It finds out the machine constants for the software platform. As shown in Figure 9.2 Subroutines for control and material data input The major subroutines called by subroutine I N D A T A for control a n d material data input are given in Figure 9.2. The major subroutines called by TMSTEP are given in Figure 9.2. CONSTI see Section 9.3. INTIAL checks the hardware and software platform the system is running on.Figure 9. MODNMl compares the input name of the constitutive model with the list of internal names to find out the material model number.3 and are described in Section 9.3. CONSOLIDATION or DYNAMIC are used to specify the analysis intended. The major subroutines called by INDATA are given in Figure 9. MODNAM reads in the name of the constitutive model to be used and finds out its corresponding internal model number.4.2 .3 . The major subroutines called by INMESH given in Figure 9.1.3. TMSTEP is the main analysis routine which controls the whole time stepping and iteration process.2 and are described in Section 9. It then calls subroutine CONSTI with option 1 to read in the constitutive model data.5 .3. it calls four major subroutines to execute the various functions of the program in the following order. INDATA inputs the main control data and material information for the program from the DATA file such as the type of earthquake input and whether or not the global solution matrix is symmetric.4 and described in Section 9.Figure 9.Figure 9. It also checks if the system stores arrays in columnwise format as assumed in the program.338 COMPUTER PROCEDURES FOR STATIC AND DYNAMIC DYNE11 .DYNMAIN INDATA INMESH TMSTEP . It reads from the DATA file the title and user information about the run up to the line starting with EXEC.1 Top two levels of the program DIANASWANDYNE I 1 DYNMAIN is the main calling subroutine in the program. 9. They are described briefly one by one in this section: GETMAT reads in material data and element type information for each material region.
Figure 9.DAMPMD . GETRIG reads in nodes which are connected to the rigid block (if used). GETTIE reads in nodes which are tied in pairs or tied to the rigid block. k = constant 2 . They are described briefly one by one in this section: GETELM reads in element connectivity information and returns the maximum element number used.3. It then calls subroutine DAMPMD with option 1 to read in the damping model data.3. Four functions are expected from the damping models: 1. Reads in the material data for the damping model 2. PERMFN is to determine the type of permeability variation to be used. only two formulae have currently been included: 1. The mass component for the Rayleigh damping is dealt with separately within GETMAT. Forms the symmetric damping matrix 4. It is called and calls to various damping models with the same set of argument. DAMPEL is the damping model for the stiffness component for Rayleigh damping. . At the time of writing. Given the current velocity.2 DAMPNM reads in the name of the damping model to be used and finds out its corresponding internal model number.3 Subroutines for mesh data input T h e major subroutines called by subroutine I N M E S H for mesh d a t a input are given in Figure 9. 9.GETMAT MODNAM DAMPNM PERMFN MODNMI ICONSTI 339 . k cc e3/l + e as suggested in Taylor (1948) The KozenyCarmen equation where k x y . n 3 / p k o ~ ' s ~ .n ) 2 is being considered for (1 future development. calculates the damping force 3.3.8 IDAMPEL DAMPDF Figure 9.2 Subroutines for Section 9.DESCRIPTION O F MAJOR ROUTINES INDATA . Forms the actual damping matrix which could be nonsymmetric DAMPDF is the default damping model with no stiffness component for Rayleigh damping. . DAMPMD is the standard interface for damping models.
GNSIDE generates midside node.3 Subroutines for Section 9. 3 for temperature with 1 dof.3 CHECK1 deduces the phase of each node using the element connectivity and the number of nodes in each phase for each type of elements. CHKTIE checks if the tied nodes are of the same phase. GETNOD gets nodal coordinates from the MESH file.340 COMPUTER PROCEDURES FOR S T A T I C AND D Y N A M I C GETELM GETRIG GETTIE CHECK 1 CHKTIE GNSIDE GETNOD GETRGP CHKNOD CHKRIG CHECK2 GETBOU CHECK3 CHECH8 . There are five phases used in the program: 1 for solid with 2 dofs. 4 for rigid block with 3 dofs and 5 for nodes not referenced therefore with 0 dof.GNSIDl . CHKNOD generates nodal coordinate of midside nodes and other phases of the same element. CHECK7 CHECK4 GETPRE CHECK6 GETPLT WRITEP FDIMEN CHECK5 . 8noded quadrilaterals from 4noded ones. It can also generate higher order elements from a lower order mesh e.CHKPLT Figure 9.NFIND1 7: GENNOD GENFND CHECK9 EQNORD CKCONN MINEQN CKCON2 SLOAN l CKCONl .3. GETRGP gets material properties of the rigid block and generates the area of the rigid block if necessary. GNSIDl keeps track on the midside node number generated for each element side. the coordinate of the midside nodes of the solid phase and of . if only the coordinates of the solid corner nodes of a 84 element are specified.g. For example. 2 for fluid with 1 dof.
Sloan and Ng 1989). Reversed CuthillMcKee method 3. zero if the node is not tied. ' .DESCRIPTION OF MAJOR ROUTINES 341 all the nodes of the fluid phase will be generated. NSW 2308. GENNOD generates coordinates for midside nodes which are not specified. The Fortran 77 coding has been obtained from the author. Department of Civil Engineering. The order of precedence is solid. GETBOU gets boundary condition code from data file.' CHECK7 forms the profile index array for each active equations. Please contact Dr. GETPRE gets timeindependent and timedependent prescribed values from the DATA file. The subroutines are free for academic usage although a nominal charge will be made for commercial usage. University of Newcastle. CuthillMcKee method 2. CHECK3 determines equation number for each dof in the global equation system. MINEQN minimises the profile length using one of the following methods: 1. EQNORD rearranges the equation numbers so that the solid nodes are eliminated before the fluid nodes of the same element. Scott Sloan. NFINDl returns the other tied node of a pair given one of them. This is done to avoid excessive numerical error. the value is used as prescribed and if the dof is free.edu.newcastle. If the dof involved is fixed. CHECKS performs profile minimisation CHECK9 finds the minimum profile length using either the various CuthillMcKee methods or the Sloan method (Sloan 1989.McKee methods in MINEQN. Australia for more information. CKCONN prepares the connectivity list for various Cuthill. then fluid and lastly temperature. CHECK4 forms the element index array from the global profile. Modified reversed CuthillMcKee method CKCON2 prepares the connectivity list for the Sloan's method in subroutine SLOANl. GENFND generates coordinates for nodes of other phases if they are not specified.au. CHECK2 checks which of the phases are present in the analysis. SLOANl (originally called GRAPH) minimises profile length by minimising the diameter of the linked list as given in Sloan (1989) and Sloan and Ng (1989). CKCONl helps subroutine CKCONN to create the connectivity list. His email at the time of writing is cesws @clod. the value is used as (generalised) force. CHKRIG checks if the nodes connected to the rigid block are of solid phase.
FDIMEN generates the location of major arrays on the global integer array and double precision array. . stress. FMCART forms Cartesian derivatives of the shape function for each gauss point. WRTMES writes displacement. CHKPLT checks the plotting informations. timeindependent and time.342 COMPUTER PROCEDURES FOR STATIC AND DYNAMIC CHECK6 checks the boundary code.5. CHECKS checks the boundary code.5.4 Subroutines called by the main control routine for analysis READEQ reads in earthquake acceleration records or coefficients for numerically generated motion such as sinusodial function from ERQK or DATA input file respectively. GETPLT gets information for the plotting output file DISP which records the time history at various userspecified locations.3. incremental strain and internal parameters for the constitutive model at the current time station to output plotting file DISP.47) FMELOl see Section 9. WRTMEl finds the residual force for a dof. CLINT1 performs linear regression for CALINT.dependent prescribed values for the all nodes to see if they are consistent. FORMGL forms the global tangential matrix as given in eqn (3. WRTPLT writes the current quantities requested in GETPLT to plotting output file PLOT. This is done for every time step. pressure. temperature. CONSTI see Section 9.3. 9. calculates and stores the relevant information such as the local coordinate of time history points that do not coincide with a node. WRTEP writes the mesh and other relevant information into the plotting output file PLOT.5 CMPROP computes the value of quantities such as average density.3. GETINT reads from INIT file the initial condition of the analysis. TMCOEF forms the multiplying coefficient for each matrix in accordance to the appropriate order of the Generalised Newmark time stepping scheme. PRCONV prints the convergence criteria used in the analysis.dependent prescribed values for the tied nodes to see if they are consistent. timeindependent and time.2 ASSEMB assembles element matrices into the global profile storage. This file could also be an output file from a previous analysis which is to be restarted.1 FORMEL see Section 9. CALINT calculates simple initial stress state such as linear variation of stress or constant & It also uses subroutine CLINT1 to analyse the initial stress state given.
TRNFRC . .TMVALU FMELOl FORMEL ASSEMB DATRIA . PREACC obtains the prescribed solid acceleration for dynamic analysis.DESCRIPTION OF MAJOR ROUTINES 343 PRCONV FMCART GETINT .5 .CHKREl . For some multitasking operating system. INFLIG opens.WRTMES .CMPROP .CMPINC .ADVALC .QNBFGS .UPDISP .Figure 9.Figure 9.4 DATRIA performs triangular decomposition for a matrix stored in a profile form.TMCOEF CALINT 7: CONSTI . reads then closes an inflight command data file.Figure 9.4 Subroutines for Section 9.FDERIV .WRTPLT .RFNORM .ADVALI .RESID2 SAVFIN .7 .FORMGL .CONVER .UPCORD .OUTCRK . This is adopted from Zienkiewicz and Taylor (1989). TFBOUN calculates the multplicative coefficient for time.CHKRE2 NADVAL ADVALP ADVARE ADVLN2 Figure 9.INCRDP .INFLIG . TMVALU calculates piecewise linear multiplicate coefficient for subroutine TFBOUN.TFBOUN .CHKNOD .DASOLN . EARTHQ calculates the current acceleration for each direction UPDISP updates the current displacement.PREACC .6 .RESIDU .RESIDl . the user would be able to supply a command in the data file leading such as to a soft termination for the analyses.CALINTl .3.8 .EARTHQ .Figure 9.dependent prescribed values.UPPROP .
3 ADVALl interface routine for subroutine ADVALC ADVALC calculates the minimum and maximum alpha values (a measure of mesh distortion suggested by Cheung et al. CHKNOD see Section 9. OUTCRZ outputs cracked Gauss point information on the screen. nonrestraint.3. CHKREl finds the maximum residual error at each dof for each phase CHKRE2 performs insert sort for subroutine CHKREl. ADVLN2 returns the square of the distance between two points.3 TRNFRC transforms residual force vector from containing all dofs to just the active.3.344 COMPUTER PROCEDURES FOR STATIC AND DYNAMIC CMPINC calculates currentlincremental quantities such as gradient of current displacement and incremental temperature.. NADVAL finds the nodes of the required phase within the connectivity of one element ADVALP returns the value of alpha measure of quality of triangle ADVARE returns the area of the triangle. only the master dofs are used. CONVER checks if the convergence criteria is met within the time step. DASOLN finds solution to a system of simultaneous equations stored in profile form and the coefficient matrix already decomposed into triangular form using DATRIA. dofs. This is adopted from Zienkiewicz and Taylor (1989). Roundoff errors are removed if the residual force vector is close to zero. QNBFGS calculates the forward and backward transformation using the quasiNewton BFGS iterative method. It can be used to obtain the initial state for dynamic analysis from a static analysis or restarting an analysis. The code has been adapted from Matthies and Strang (1979). The format is compatible with the input file INIT. incremental norm and residual force norm for each phase.5. RESIDl outputs error norm. UPCORD updates the coordinate for Updated Lagrangian analysis. RFNORM calculates and outputs the error norm for each phase. SAVFIN saves the current state in FINL output file. 1996) for the triangular elements in a mesh. RESIDU see Section 9. INCRDP calculates the incremental displacement (for STATIC). OUTCRK outputs information of cracked Gauss points. UPPROP calculates the updated value of quantities such as average density due to change in void ratio during the finite deformation analysis. For tied nodes. . FDERIV calculates Cartesian derivatives for the shape functions from the derivatives in local coordinates. velocity (for CONSOLIDATION) or acceleration (for DYNAMIC) for the current time step depending on the type of analysis performed. the maximum angle of deviation from centre line is also calculated. For sixnoded elements.
PERMFN see Section 9.3.3.5 FMSTIF forms element stiffness matrix K DAMPMD see Section 9.5.5 Subroutines for the formation of element matrices and residual calculation 9.3.2 FMPMOl forms element permeability matrix H for 3noded element FMCMOl forms mass part of the element damping matrix C for 3noded element FMQMOl forms element coupling matrix Q for 3noded element FMGMOl forms element mass coupling matrix G for 3noded element 9.8 .2 FMASOl forms element mass matrix M for 3noded element PERMFN see Section 9. 9.2 Subroutines for the formation of element matrix (Figure 9.5.5) FMELOl forms element matrices for 3noded element CONSTI see Section 9.3.2 FMPERM forms element permeability matrix H FMELO1 Figure 9.1 Subroutines for the analytical formation of element matrix for 3noded elements (Figure 9.5 Subroutines for Section 9.2 FMMASS forms element mass matrix M CRACKK modifies material property after a Gauss point is cracked.3.6) FORMEL forms element matrices CONSTI see Section 9.3.DESCRIPTION OF MAJOR ROUTINES 345 RESID2 outputs a summary of the current analysis such as the average number of iterations per time step for easy reference.5 DAMPMD see Section 9.3.5.3.1 I CONSTI FMSTO1 DAMPMD FMASO 1 PERMFN FMPMO 1 FMCMO1 FMQMO 1 FMGMO 1  Figure 9.
CONSTI .3 Subroutines for the formation of residual forces (Figure 9.2 ASSEMl assembles into the element force vector the force due to stiffness contribution ASSEM2 assembles into the element force vector the force due to mass and mass damping contribution ASSEM4 assembles into the element force vector the force due to fluid compressibility contribution PERMFN see Section 9.Figure 9. INTSTR subdivides the strain increment for strain integration CONSTI see Section 9.2 FMCOMP forms element compressibility matrix S GXCOND forms element thermal conductivity matrix GXCOMP forms element thermal storage mass matrix GXJMAT forms element fluid and thermal coupling matrix FMQMAT forms element solid and fluid coupling matrix Q FMGMAT forms element solid and fluid mass coupling matrix G GXLMAT forms element solid and thermal coupling matrix 9.GXLMAT Figure 9.FMPERM .DAMPMD .3.7) RESIDU calculates the residual force vector.FMSTIF .GXCOND .3.3.5.8 .5.6 Subroutines for Section 9.3.CRACKK .2 ASSEM3 assembles into the element force vector the force due to permeability contribution ASSEMS assembles the element force vector into the global residual force vector .FMCOMP .FMGMAT .PERMFN .3.GXCOMP .2 CRACKK see Section 9.GXJMAT .346 COMPUTER PROCEDURES FOR STATIC AND DYNAMIC FORMEL .FMQMAT .5 DAMPMD see Section 9.FMMASS .5.
DAMPMD . CFFINP opens a temporary file for free format read for information stored in a character string. CHKDFl checks the difference of two double precision numbers and prints if it exceeds a specified norm.ASSEM3 . CPRINT prints out a matrix stored in profile form.PROPFZ .ASSEMl . BPLOT interfaces the system dependent plotting routine XPLOT with the higher level plotting subroutines and makes them system independent. CLOFIL closes a unit number and outputs a message to that effect. CHRINT writes the value of an integer number into a character string. DOEQO checks if a double precision number is close to zero within the machine precision.ASSEM2 . CTITLZ checks if a specific keyword is present at the beginning of a character string with or without echo of the input. CHKRWl checks if the computer platform uses a columnwise storage scheme as specified in the Fortran 77 standard.3.4 MAJOR SER VICE SUBROUTINES ADDVEC adds two arrays and puts the results into the first one.ASSEMS Figure 9.3 9.CRACKK .0 routine for double precision vector operation of y =y + ax.INTSTR . CTITLl checks if a specific keyword is present at the beginning of a character string with echo of the input. AGAUSS returns the 2D Gauss point location and weight for quadrilaterals and triangles. .ASSEM4 .CONSTI .5. DOEQ checks if two double precision numbers are the same within the machine precision. CFORM8 outputs a double precision array with maximum number of significant figures within the given space using the subroutine CFORM7.PERMFN .MAJOR SERVICE SUBROUTINES RESIDU 347 . CFORM7 formats the output of a double precision number within a given number of character space while maximising the number of significant figures produced. CHREAL writes the value of a double precision number into a character string. DAXPY is the BLAS 3.Figure 9.7 Subroutines for Section 9.8 .
NEOLIN finds the end of text within a character string. LOADVC copies one double precision vector into another. MBTMUL calls BLAS 3. MVECTP prints a double precision vector. IVECTC prints only one digit. MROWPR prints a row of a matrix stored in profile form. i. in compact form. LSAMEl checks if two character strings are the same under case insensitive condition.0 routine DGEMM for double precision matrix multiplication. IOFMTl ensures the same format statement is used in read and write operation for plotting output file PLOT. DMACH obtains the machine constants such as the smallest difference between a double precision number and unity.348 COMPUTER PROCEDURES FOR STATIC AND DYNAMIC DCOPY is the BLAS 3.e. area and the second moment of inertia .0 routine for double precision vector operation of y = x. IVECTP prints an integer array. MSTORE stores message summary for one unified output at the end of the execution. HZSHAP returns twodimensional finite element shape functions for triangular and quadrilateral elements.0 routine for double precision vector operation of d = x. for each element of an integer array. H2CORD returns the local coordinates of the nodes for a twodimensional finite element. It is now an interface for DDOT. LOWERC converts a character string to lower case. INVARl calculates stress invariants for plane strain condition. FRICTM calculates the friction angle required for the current stress state and the given cohesion.y. IALLOC allocates a section of the global integer or double precision array to be used. GIVALU gives a double precision array a particular value. MPRINT prints a double precision matrix. HlSHAP returns onedimensional finite element shape function. DDOT is the BLAS 3. This is now only an interface to DCOPY. HANDLE handles an error condition in the input data and sets a flag so that the execution would stops after the data has been read in. NFACTL returns the factoral function in integer form. PRINT prints an integer matrix. OPENFL is the subroutine to unify the open statements on different computer systems POLYOZ calculates the various properties of a nsided polygon including the centroid. DOTPRD is the former function name for vector dot product. GIVINT gives an integer array a particular value.
RMISPC reduces intermediate spaces within a character string to single space. VECTAS performs addition of two double precision vectors by calling DAXPY. PSTRE3 calculates second order tensor rotation for.. although connected. RMLSPC removes leading spaces from a character string. .2 and 9.5 CONSTITUTIVE MODEL SUBROUTZNES Quite a large number of constitutive models have been made available to the computer program DIANASWANDYNE 11.3.2. SOLVC3 finds the local coordinates for a given point within a given element. SOLVC6 finds the local coordinates for a given point within a given element by looping over all elements within the finite element mesh. UPPERC converts a character string to uppercase. RMCOMM discards the input line if it begins with a letter 'C'. STACKP generates a stack dump for debugging purpose when an error occurs. USAGE1 checks how much of the global integer array has been used. VALMAX returns the absolute element value within a double precision array. RGAUSS gives onedimensional GaussLegendre integration points.5. USAGEC checks how much of the global double precision array has been used. TIMEMG outputs the given message if a specified period of time has elapsed since the last time message has been output from this subroutine to the screen. the number of page faults for the underlying virtual memory system.5. Due to limited space.1. TERMIN termins the analysis after issuing an error message. e. Constitutive models available for general dissemination are described in Section 9.5. Other models available are described in Section 9. Their dissemination is restricted both by licensing conditions and difficulties of implementation. TRIAOl returns the shape function and element matrix components for 3noded triangular element.g.5. STOPCU stops the execution if the maximum allowable CPU time is exceeded. have not been thoroughly tested for general application.5. the subroutines called by the main constitutive model subroutines listed in Sections 9. 9. Some of them. They are all linked to the computer program via a standard constitutive model interface CONSTI which is described in Section 9. WARNED writes out a warning to the screen and stores the message for message summary at the end of the execution. the stress tensor. if supported by the system.CONSTITUTIVE MODEL SUBROUTINES 349 PSTREZ calculates the principal stress and direction.3 will not be shown. RATEPF calculates. VECTSB performs subraction of two double precision vectors by calling DAXPY.
CRISP (Britto and Gunn 1987) via CRSM2D (see AlTabbaa 1996) and LUSAS's MMI. The material interface. The specification has been given in the user manual of the program SM2D (Chan 1995) and it is reproduced in Appendix 9A.g. From CRISP v4. 8. there is still vigorous ongoing research on constitutive models. This is available also for threedimensional applications. The program must provide room for expansion especially in this aspect. The standard interface is called and calls to various constitutive models with the same set of arguments.350 COMPUTER PROCEDURES FOR STATIC AND DYNAMIC 9. Form the symmetric D matrix 4. is also available to ABAQUS via their material model interface (MMI) DMAT. 9.0 onwards. facilities must be made available so that further material models can be installed. .1 Standard Constitutive model interface subroutine CONSTI For a fully coupled soil and porefluid computer program like DIANASWANDYNE 11. 9. Most of the material models available are listed in Figure 9. A few modifications have been made to cater for threedimensional analysis. Initialise the internal parameter array Four other functions are prepared but at the time of writing. Read in the material data for the model 2. besides linking directly to SM2D. internal parameters and incremental strain. not.g. DIANASWANDYNE I1 and GLADYS2E. CAPMODInterface routine with subroutine CAP as given in Sandler and Rubin (1979). Given the current stress state. CONSTI has been adopted as the standard material model interface for the computer program. linear elastic Further suggestions for further standard functions are always welcome. calculate the incremental stress and changes in internal parameters 3. Despite the wide acceptance of the Biot dynamic formulation and numerical implementation as described in this volume. Five standard functions are expected from the constitutive models: 1. fully implemented: 6.8. Form the actual D matrix which could be nonsymmetric 5 . 7.5. consistent tangent operator Constant Dmatrix e.5.2 Constitutive models available for general dissemination (Figure 9. the provision of Gauss point location and models for partially saturation application. The single material interface is performed through the subroutine CONSTI.8) CAP3D1Interface to subroutine CAPMDL as listed on pages 412423 of Chen and Mizuno (1990). Return values for optimisation Output internal values to output unit ICOUT Special Dmatrix e.
Tresca. The variation can be linear.5 CJHMODTwosurface plane strain kinematic hardening model for sand. elastic model. MCOULSClassical Elastoplastic model adopted and modified from Owen and Hinton (1980): von Mises.8 Subroutines for Section 9. The bulk modulus and the shear modulus can vary with mean effective confining stress. square root or generally nonlinear 2. ELAS3DLinear ELASGMgeneral Elastic model which can be used in threedimensional analysis. (Hamilton 1997. Cohesion and the Mohr Coulumb friction envelope is also available ELASTAAnisotropic elastic model for overconsolidated clay (Graham and Houlsby 1983). MODCAMModified CamClay model . A dummy subroutine to provide an easy connection for testing newly implemented constitutive model.CONSTITUTIVE MODEL SUBROUTINES CONSTI 351 ADJCN4 ADJJIM ADJMH4 ALTER0 BRICK1 CAPMOD CJHMOD CSMOOl DEPOIN ELAS3D ELASGM ELASTA EXPERI HASHIS MCOULS MODCAM NCRIS2 SARAH SLIP03 STATE2 TABBA2 TSMODl UNSATS VONBAC VONMIS   Figure 9. nonassociative Mohr Coulomb with associative or nonassociative deviatoric response and DruckerPrager model. in which: 1. Hamilton et a1 1998) CSM001Original CamClay model adapted from Britto and Gunn (1987) DEP08NPastorZienkiewicz mark I11 model (1986) which is a generalized plasticity model for sand as described in Chapter 4 of this volume. EXPERIExperimental new model.
3 Other constitutive models implemented (Figure 9. 1987. VONBACvon parameter based onedimensional model (Muir Wood et a1 1994) surface kinematic hardening model for clay (AlTabbaa and Wood VONMISvon Mises model Mises model with backward Euler integration 9. Alonso et a1 1990) VONOOl to VON008vectorised von Mises models VONADl to VONADSAdaptive strategy subroutines to use VONOOl to VON008 9. ADJCNkInterface to a concrete model (GIBB 1994) ADJJIMInterface to a concrete interface model (GIBB 1995a) ADJMHkInterface to nonassociative Mohr Coulomb model with varying friction and dilatancy angle. kinematic yield surface model (Stallebrass 1990) Twosurface modeladapted from author supplied program UNSATSInterface to MODBUSY and JOSAO1. Roatesi and Chan 1994.5. GETFIL gets the filenames to be used for the current analysis. DOSCOM performs a MSDOS line command.8) The second author would like to take this opportunity to thank all the colleagues who have given kind permission for us to use their constitutive model and supply us with the source code. Roatesi 1995. (GIBB 1995b and 1995c) f i T E R L K i n e m a t i c hardening model proposed by Molenkamp (1982. Backward Euler integration scheme is used in this model together with comer handling strategies. OPNFLL opens a file and connects it to a unit number. . 1990 and 1992) BRICK1Simpson HASHIHashiguchi brick mode1 (1992a and l992b) (1989) model NCRIS1Cristescu saturated sand model (Cristescu 1989 and 1991.6 SYSTEMDEPENDENT SUBROUTINES ANSI01 performs screen control using escape sequence in ANSLSYS for MSDOS. Roatesi and Chan 1998) SARAHThreesurface TSMOD1Pietruszczak TSMAIN. the Barcelona Unsaturated Soil model (Josa 1988. Chan and Roatesi 1998.352 COMPUTER PROCEDURES FOR STATIC AND DYNAMIC SLIPOSSimple plane slip model STATE2State TABBA2Two 1989).
Gibb (1994) Nonlinear Concrete ModelTheory and Validation Report. .D. 4. NUMOG 111. C. Reading. NO. Cristescu N. (1991) Nonassociated elastic/viscoplastic constitutive equations for sand. UK. 3. UK. University of Birmingham. Cambridge. (1995) User Manual for SM2D . 2. T. 9199. Chan A. (1987) Critical State Soil Mechanics via Finite Elements. University College of Swansea. (1998) Finite Element Approach in Viscoplasticity for Cristescu Saturated Sand Model. Chen W.Soil Model Tester for 2Dimensional Application. Gibb (1995b) Nonlinear Soil ModelTheory Report. GETKEY gets one key from the keyboard. Wales. (1990) Nonlinear Analysis in Soil Mechanics . and Josa A. CPU time elapsed. Elsevier. UK. Lo S. 701707. and Roatesi S. E. Kluwer. Chan A. University of Birmingham. MA 02142.REFERENCES 353 SYSFUN returns values for various system functions such as system time. December. (1995) Excess Pore Pressure During Consolidation and Swelling with Radial Drainage. 7 . Gibb Ltd. Reading. 9. H. ZPSYMB draws a symbol on the screen. Roum. Mec. SETSUF sets the name of file extension. Chan A. (1989) An experimentally based 'bubble' model for clay. 165180. H. Niagara Falls.Theory and Implementation. Techn. Gibb Ltd. 4164.. 40. page faults and date. Y. Gens A. Cristescu N.D. M.D. 12. Giotechnique. Chichester. Gibb (1995a) Concrete Construction Joint Interface Model and Validation Report. Dissertation. Ph. (1988) A Unified Finite Element Solution to Static and Dynamic Geomechanics Problems. (1996) Finite Element Implementation. Gibb (1995~) Nonlinear Soil ModelValidation Report. Gibb Ltd. Thesis. DSORTX sorts a double precision array. AlTabbaa A. Plasticity. pen down and opening a graphic device. H. No. K. Int. (1987) Permeability and Stress Strain Response of Speswhite Kaolin. Ph. Blackwell Science. (1989) Rock Rheology. and Gunn M. Reading. 45. Cheung Y. Graham J. Thesis. Ellis Horwood Ltd. Hamilton C. TIMEXX returns time and date for the current time. (1997) A Plane Strain Constitutive Model for Sands under Nonmonotonic Loading. and Muir Wood D. GPotechnique. and Mizuno E. Cambridge University Engineering Department. Dordrecht. Amsterdam. T.. XPLOT maintains a set of basic plotting primitives such as pen up. and Houlsby G. Appl. J. No. F. H. UK. Britto A. J. GCotechnique. and Leung A. C. Reading. C. Birmingham. Gibb Ltd. Ph. 33. Sci. School of Civil Engineering. AlTabbaa A. COMLIN returns the text given on the command line. (1990) A Constitutive Model for Partially Saturated Soils. 405430. J. AlTabbaa A. No..7 REFERENCES Alonso E. UK. (1983) Anisotropic Elasticity of a Natural Clay. Rev.
Rev. (1982) Kinematic Model for Alternating Loading ALTERNAT.. 28. and Zienkiewicz 0. Romania. Num... The Wroth Memorial SymposiumPredictive Soil Mechanics'. Catherine's College. 21. 33. Matthies H. and Ng W. (1980) Finite Elements in PlasticityTheory and Practice. 44. (1994) Numerical integration of a viscoplastic constitutive equation for geomaterials. Belkheir K. UK. (1988) A Mathematical Modification of Two Surface Model Formulation in Plasticity. 26512679. (1979) An Algorithm and a Modular Subroutine for the Cap Model. 8.354 COMPUTER PROCEDURES FOR STATIC AND DYNAMIC Hamilton C.. Romania. (1989) A FORTRAN Program for Profile and Wavefront Reduction. Int. (1994) Technical Note . No. Sandler I. Meth. H. No. Pineridge Press. Num. (1989) A Direct Comparison of Three Algorithms for Reducing Profile and Wavefront.Strain Softening and state parameter for sand modelling. NUMOG 11. H. Roatesi S. G. Roatesi S. Engrg. Balkema. J. Mec.D. C. (ed. 14. A. Wales. W. Sloan S.. Brasov. Molenkamp F. Rotterdam. 24. S. Booth E. J. W. Comparison with the analytical solution for step creep in The XVIII National Conference of Solid Mechanics. Int. (1998) Comparison of Finite Element Analysis and Analytical Solution for Underground Openings Problems in Viscoplastic Rock Mass. Int. 131150. C. (1959) A Method of Computation for Structural Dynamics. 541576. Engrg. Ottawa. Simpson B. Anal. 4. F. G. 887B. Roatesi S. . Ghent. M. Anal. C. Bucharest. 213225. J. ETSICCP. Chan A. and Hinton E. Technical Report. 179186. Katona M. 2729 July. Solids & Structures.. Meth. (1985) A Unified Set of Singlestep Algorithms Part 3: The Betam Method. 13451359. C. Molenkamp F. (1987) Elasto Plastic Model for Analysis of Liquefaction Under Alternating Loading in Workshop on Constitutive Laws for the Analysis of Fill Retention Structures. Ph. 131150. R. and Liu D. J. No. and Strang G. (1988) An Elastoplastic Model for Partially Saturated Soils. 42. Josa A. Comp. Gotechnique. 10. S. Znt. St. (1979) The Solution of Nonlinear Finite Element Equations. and Zienkiewicz 0 . Owen D. Oxford. Num. 3. 2. (1990) Reformulation of ALTERNAT to Minimise Numerical Drift Due to Cyclic Loading. Engrg. Proc. Molenkamp F. and Rubin D.. 3138. and Chan A. (1992a) Development and Application of a New Soil Model for Prediction of Ground Movements. l . Struct. Simpson B. Num. (1992b) Retaining Structures: Displacement and Design. and Muir Wood D. Muir Wood D. Techn. LGM Report Co218598. 6794. Molenkamp F. 16131626. Newrnark N. No. A. (1998) Dynamic Finite Element Analyses of Sand: Structure Interaction Using a New Kinematic Hardening Model in Seismic Design Practice into the Next Century. Manchester. Pastor M. Sloan S. University of Manchester Internal Report. Barcelona (in Spanish). GPotechnique. 9871 001..411419. Meth. J. (1995) Finite Element Approach in Viscoplasticity. a Generalisation of the Newmark Scheme. Appl. 12. Roum. Technology and Research Ministry. Swansea. Katona M. Sci. C. J. 16. 173186. April. J.. NUMETA 85. 335339. Thesis. ASCE. and Chan A. Int. Znt. (1986) A Generalised Plasticity Hierarchical model for Sand Under Monotonic and Cyclic Loading. H. Num. No. (1985) A General Family of Singlestep Methods for Numerical Time Integration of Structural Dynamic Equations. Geomech. (1992) Application for Nonlinear Elastic Model. Hashiguchi K. 628643. NL. J. Delft.). Geomech.
John Wiley and Sons. Ph. so that when reading a particular name.KTEST.DSTRE l. V. two more stress components are required: T. (1948) Fundamentals of Soil Mechanics. the number of stress components is four. please refer to SM2D manual chapter (Chan 1995). The routine can be written in FORTRAN 77 or any other compatible language. d . E. L. W. MIT. the implementation of an elastic and an elastoplastic model von Mises are used as examples. New York. no separate description is required.DSTAN.INTERFACING WITH THE MAIN PROGRAM 355 Stallebrass S. Dissertation. The introduction of parameter NSTRE is to ease the changeover from twodimensional to threedimensional stress state. dyy. Using such would not require the change in the subroutines CONSTI and MODNMl until the model is ready for production..D. For the twodimensional analysis.PARAM. IELEM.. 9A.NTEST) C.1. The interface call is as follow: SUBROUTINEELASTI (PROPD. To implement a new model. City University. Subroutine CONSTI will use the model number to call the specific new model. APPENDIX 9A IMPLEMENTING NEW MODELS INTO SM2D In this chapter.. V O I D R . and T.. Engineers and Education. London. Zienkiewicz 0 ..DMATX. Inc. A model name and a model number has to be assigned to the new model. C. Since it is totally compatible with SM2D.LPRPD.. England.1 INTERFACING WITH THE MAIN PROGRAM The interface of the program is done with a single subroutine call. . London.. 200202. Taylor D.ESTRE.ICOUT. They are ordered as: d. (1989) The Finite Element MethodVolume I: Basic Formulation and Linear Problems (4th edn). subroutine MODNMl will assign the model number to the material array. 2. the way to implement new models onto DIANA SWANDYNEI1 is described.ISWDP. a slight modification to subroutine MODNMl and CONSTI is required. (1953) After Marcuson (1995): An Example of Professional Modesty. To test the soil model routine using program SM2D. In the case of threedimensions. McGrawHill Book Company. Whitman R... and Taylor R. and T. (1990) Modelling the Effect of Recent Stress History on the Deformation of Overconsolidated Soils.LPARA.. The Earth... C L. I C P R T ELASTICCONSTITUTIVEMODELFORVERIFICATION INTEGER NSTRE PARAMETER ( N S T R E = 4 ) The subroutine name EXPERD should be used for material model development. I G A U S ..1 Variable NSTRE NSTRE is the number of stress components in the analysis.. ICDAT. 9A. In this appendix. A subroutine name which is sufficiently individual should be given.
here is a brief overview of its capabilities 1.1 Variable IELEM IELEM is the current element number. check input properties 2. Unit ICOUT should not be used for this purpose. 9A. This is just for information. it could be useful.4 Variable ZSWDP ISWDP is the controlling switch and its function will be fully described later.2 Variable ICPR T ICPRT is the unit number for debugging or information output that can be massive. Form nonsymmetric Dmatrix (if applicable) . .2.2.2 INPUT ONLY VARIABLES The following variables are for input only and their value should not be changed by the subroutine. All of integer variables should be positive. it is the local Gauss point number. INPUT VARIABLES INTEGER IELEM INTEGER I C P R T INTEGER IGAUS INTEGER ISWDP INTEGER LPRPD INTEGER ICDAT INTEGER ICOUT INTEGER NTEST I N T E G E R LPARA d o u b l e p r e c i s i o n D S T A N (NSTRE) DOUBLE P R E C I S I O N V O I D R 9A. The Gauss point is counted for each element i.. . This is just for information. 9A. Form symmetric Dmatrix 4.3 Variable IGA US IGAUS is the current Gauss point number. C . However. If it is desirable to trace the progress of a particular Gauss point. Form incremental stress from the given incremental strain 3.356 COMPUTER PROCEDURES FOR STATIC AND DYNAMIC 9A.e. If it is desirable to trace the progress of a particular Gauss point.2. Read in material data. check length of array. this could be useful.2. 9A.
2. Minor output (less than two lines) per entry is acceptable. The incremental strain is tensile positive. subroutine TABBA2. 9A.INPUT O N L Y VARIABLES 5. The shear strain is the engineering shear strain which is two times the component of the incremental strain tensor.5 Variable LPRPD The dimensions of the material property array.g.2.7 Variable ICOUT The unit number for echoing the material data input. Output internal values to unit icout 8.g.2. tangent operator 9. 9A.g.9 Variable LPARA The dimension of the soil model internal parameter array. Return values for optimisation 7.2. 6.2.8 Variable NTEST The dimension of the soil model condition array KTEST(NTEST). 9A.6 Variable ICDA T The unit number for input data. Special Dmatrix e. 357 Initialise internal variables. 9A. .11 Variable VOIDR The current void ratio based on mechanical deformations. All of the elements from PROPD(1) to PROPD(LPRPD) are available in each material.2. Monitoring of a Gauss point can be done with PRINT statement. see e. it can be implemented with a flag in the propd array.10 Array DSTAN(NSTRE) The array for incremental strain input. This is active only for ISWDP = 2. Constant Dmatrix e.2. All elements from PARAM(1) to PARAM(LPARA) are available at each Gauss point. linear elastic 9A. check current stress state. For initial stress state calculated according to the Gauss point coordinate GPCOD. 9A. 9A.
KTEST(l0) Soil model error counter E: No specific definition yet. 9. 8. If there is a warning state in the current model. 7. C. STOP statement can be used. KTEST(8) Soil model error counter C: Usually used to denote CONSl = nTD. KTEST(9) Soil model error counter D: Usually used to denote the plastic modulus CONSl is less than zero and remedial action is needed. KTEST(6) Soil model error counter A: Usually used to denote tension state is encountered and remedial action is needed. KTEST(11) 0: Elastic Dmatrix. 6. If such a case occurs.1 Away KTEST (NTEST) This is the soil model information array.. If such a case occurs. KTEST(5) Not used. Please use PRINT statement to indicate your definitions. ng is less than zero and remedial action is needed. . 5.358 COMPUTER PROCEDURES FOR STATIC AND DYNAMIC 9A. KTEST(2) The type of subdivision that is active (variable NSUBV) 3. 0 indicates this is a trial incremental. They are defined as follow: 1. increment KTEST(8) by one.. KTEST(1) Valid only for ISWDP = 2. KTEST(3) The number of times MDIVT is exceeded 4. + 10. If such a case occurs.3. increment KTEST(6) by one. 11. . 1 indicates this increment may be taken as the accepted increment 2. INPUT/OUTPUTVARIABLES: INTEGERKTEST (NTEST) d o u b l e p r e c i s i o n PROPD ( L P R P D ) d o u b l e p r e c i s i o n E S T R E (NSTRE) d o u b l e p r e c i s i o n P A R A M (LPARA) 9A. KTEST(7) Soil model error counter B: Usually used to denote allowable stress ratio has been exceeded and remedial action is needed. KTEST(4) Number of warnings from soil models. this can be incremented by one. maybe using it to determine the number of subincrements. 1: Elastoplastic Dmatrix The programmer of new soil model is free to redefine any or all of these soil model error counters. For a serious error. increment KTEST(7) by one. If such a case is encountered. increment KTEST(9) by one.3 INPUT AND OUTPUT VARIABLES Variables and arrays described in this section will carry information into the subroutine or out of the subroutine. .
A check should be used within the routine so that iswdp = 5 will be called automatically if the values in PARAM have not be initialised. ESTRE(final)ESTRE(initia1) when exits. For a given Gauss point. For iswdp = 2.NSTRE) When ISWDP = 2.. Different PARAM(LPARA) will be available for each Gauss point. the value in PARAM should either be the output value of iswdp = 5 or inputloutput value of iswdp = 2.OUTPUT ONLY VARIABLES 359 9A. . This array should be written to only for ISWDP = 1 and for all the other ISWDP.4 OUTPUT ONLY VARIABLES This section describes the arrays for output only. it should contain the incremented stress state which exits. This array should be defined before any other option for iswdp other than one is used. A check could be used to keep track of whether iswdp = I has been called for the material group the Gauss point belongs to. The subroutine should initialise the values within this array for iswdp = 5 and check if the array is long enough for its purpose.. For iswdp = 2. The value of PARAM should not be altered for iswdp = 3 and 4. a temporary file could be used so that a proper readin with iswdp = 1 can be done. This is for inputting information only. For options ISWDP = 2.4. this contains the current effective stress state (tension positive) which enters. the values within the propd(*) array should remain constant throughout the analysis. 5. C.2 Away PROPD(LPRPD) The array for material data input. the ESTRE should either be the output value of iswdp = 5 or inputloutput value of iswdp = 2. 3 and 4. For option ISWDP = 2.3. OUTPUTVARIABLES: d o u b l e p r e c i s i o n D S T R E (NSTRE) d o u b l e p r e c i s i o n D M A T X (NSTRE. 3.. 3 and 4. The array should be read by via iswdp = 1 and internal processing of the input information is allowed. The value of ESTRE should not be altered for iswdp = 3 and 4. 9A. Different material type will have different PROPD even if they share the same model.e. DSTRE(NSTRE) should contain the incremental stress i. So if the material information is read in elsewhere. The array for internal parameters. The subroutine could check if there is sufficient number of elements in the array for the use of the model.
3 Local Variables Any number of local variables can be defined in the subroutine.360 COMPUTER PROCEDURES FOR STATIC AND DYNAMIC 9A. LOCAL VARIABLES LOCALDEBUG doubleprecisionYOUNG doubleprecisionPOISS doubleprecisionRLMDA double precision SHEAR INTEGER ICOUN double precision REAL1.NSTRE DMATX (ISTRE.ISTRE)) & END DO END DO When ISWDP = 4. NSTRE) should contain the nonsymmetric version of the Dmatrix on exit. It is not advisable. 9A. If the soil model is associative.5 ISWDP BRANCHING The simpliest way of branching for ISWDP is to use a computed G O TO. or the associative version of the matrix.4.JSTRE) +DMATX (JSTRE. This can be done either by using the elastic matrix.5DO*(DMATX (ISTRE. C. DMATX(NSTRE. . DMATX(NSTRE.JSTRE) =0.associative then a symmetrical version should be given.. REALO.. to jump from one option to another within the code unless it is a very . All real constants should be given in double precision also to preserve accuracy. NSTRE) When ISWDP = 3.OdO) double precision REAP5 PARAMETER (REAP5=0. it must not retain values that are specific to a Gauss point. then ISWDP = 4 can lead to the same code as ISWDP = 3. value of all the local variables is not necessarily retained for the next entry of the soil model nor retained for different Gauss points in the same material region. however. COMMON should not be used. If the soil model is non. 9A. Note. Except for very good reasons. or averaging so that DOISTRE=l.5DO) CHECK IF (DEBUG)WRITE (ICOUT.VOIDR C. DVOLV PARAMETER (REAL1=1.2 Matrix DMA TX(NSTRE.") 'VOIDRATIO:'.NSTRE DOJSTRE =l... SVOLV.4.OdO.REALO=O. although large arrays are not recommended.NSTRE) should contain the symmetric version of the Dmatrix on exit..OdO) doubleprecisionREAL2 PARAMETER (REAL2=2. If it were used.
*) 'ISWDP:'. ./.3. E11..IS W D P BRANCHING 361 straight forward jump..3. the routine is required to output the incremental stress DSTRE and the new stress state ESTRE. . ISWDP CALLTERMIN ('ELASTI1:UNKNOWNOPTION:') GOT0 2999 Label 2999 is the common exit point 9A. E11. . An error should be given if ISWDP is outside the range and CALL TERMIN will terminate the program after issuing a warning C.5. C. The computed GO TO is used in the same way as the CASE statement in Fortran 95.POISS RLMDA=YOUNG*POISS/( G REAL^REAL~*POISS)*(REAL~+POISS)) SHEAR=YOUNG*REAP5/(REALl+POISS) WRITE (ICOUT. .ERR=3100.. E11. ERROR CONDITION WRITE (ICOUT.. FORMTHE INCREMENTALSTRESS 1200 CONTINUE RLMDA = PROPD ( 1) SHEAR=PROPD ( 2 ) DVOLV=DSTAN(l) fDSTAN(2) +DSTAN(3) SVOLV = RLMDA*DVOLV DSTRE (1) = SVOLV+REAL2*SHEAR*DSTAN (1) DSTRE (2)=SVOLV+REAL2*SHEAR*DSTAN (2) DSTRE (3)=SVOLV+REAL2*SHEAR*DSTAN (3) ESTRE (1)=ESTRE (1)+DSTRE (1) ESTRE (2)=ESTRE (2)+DSTRE (2) ESTRE (3)=ESTRE (3) +DSTRE ( 3 ) .1 ISWDP=l For Material Data Input C. Calculated values are stored in the array 9A.POISS.3 .RLMDA.2 IS WDP = 2 Forming Incremental Stress With the input incremental strain DSTAN.END=3200) YOUNG.3) 1 PROPD(1) =RLMDA PROPD (2) = SHEAR GOT0 2999 2001 Note that the input data is echoed. 'POISS:'. 'RLMDA:'.5.. 1100 READ INMATERIALDATA CONTINUE READ(ICDAT. SHEAR FORMAT ( ' YOUNG:'.*. E11.2001) YOUNG. 'SHEAR:'.
..matrix C.l)=RLMDA DMATX (2..4)=REAL0 DMATX (2.l)=RLMDA DMATX (3. COMPUTER PROCEDURES FOR STATIC AND DYNAMIC SINCEENGINEERINGSHEARSTRAINISUSED DO ICOUN = 4.2)=RLMDA DMATX (1.3) =REAL0 DO ICOUN=4.LPARA .5 IS WDP = 5 Znitialise Internal Variables C.362 C. 9A.5.3 ZSWDP = 3 Symmetric version of Dmatrix As the Dmatrix is symmetric in this case.2)=RLMDA+REAL2*SHEAR DMATX (2...4 ZSWDP = 4 Not Necessary Symmetric D.NSTRE DMATX ( ICOUN. STORE SOMETHINGINTOPARAM 1500 CONTINUE DO ICOUN=l.4)=REALO DMATX (3...3)=RLMDA DMATX (1.l)=RLMDA+REAL2*SHEAR DMATX (1.4)=REAL0 DMATX ( 4...5. 1300 = 4. SYMMETRICVERSIONOFDMATX CONTINUE 9A..2)=REAL0 DMATX (4..5..l) = REAL0 DMATX (4. ICOUN) = SHEAR END DO GOT0 2999 The doloop is used to facilitate change over to three.dimensional analysis 9A. it is merged with option ISWDP C.2)=RLMDA DMATX (3.3)=RLMDA DMATX (2. NOTNECESSARYSYMMETRICVERSIONOFDMATX 1400 CONTINUE RLMDA=PROPD (1) SHEAR=PROPD (2) DMATX (1.3)=RLMDA+REAL2*SHEAR DMATX (3. NSTRE DSTRE (ICOUN)=SHEAR*DSTAN (ICOUN) ESTRE (ICOUN) =ESTRE(ICOUN)+DSTRE(ICOUN) END DO GOT0 2999 The doloop is used to facilitate changeover to threedimensional analysis.
i c d a t .COMMON EXIT PARAM(1COUN) = R E A L 0 END DO GOT0 2 9 9 9 363 9A.1 Interface for partially saturated model A new interface for partially saturated analysis is in the process of development SUBROUTINEMODBUS ( p r o p d . i g a u s . C. dstan. ICDAT CALLHANDLE ( ' E L A S T I . e s t r e . v o i d r . . . * ) ' I C D A T : ' .ntest. gpcod Gauss point coordinate  .6 COMMON EXIT It is advisable to use a common exit for all options. dpore . k t e s t . porep. p a r a m . d m a t x .8. it is used to eliminate error in input data. i c p r t . incremental pore pressure applied with dstan 3. This is especially for debugging purpose.change in pore pressure. 2999 COMMONEXIT CONTINUE RETURN 9A. In this subroutine. i e l e m . porep  pore pressure corresponding to estre 2. i c o u t .. * ) ' I C D A T : ' . C . I C D A T C A L L H A N D L E ('ELASTI2:ERRORINCHANNEL1) RETURN CONTINUE WRITE ( I C O U T . i s w d p . lpara. 3100 ERRORCONDITION CONTINUE WRITE ( I C O U T .3 : E N D O F F I L E I N C H A N N E L : ' ) RETURN END 3200 9A. dstre. gpcod) The new arguments are: 1.7 ERROR EXIT Subroutine HANDLE will output the character message without terminating the program. .8 SOME CURRENT DEVELOPMENTS 9A.. dpore. l p r p d ..
ktest(2)=nsubv type of subdivision d. NTEST) VONMISESELASTOPLASTICMODELWITHLINEARHARDENING C C. . i i t e r . i i t e r .ISWDP.ISWDP.fmt) should be avoided except for debugging purpose. linear elastic response is expected 9A.9 ANOTHER EXAMPLEIMPLEMENTATION An example is von Mises SUBROUTINEEXPERI (PROPD. INTEGER ICHPC COMMON /PCOMOl/ ICHPC The value ICHPC should not be changed.. If NPRIN> 1 then the elastoplastic code will be output during matrix formation. . ktest(1) indicates the state for strain subdivision 0: total increment from final state 1: total increment from initial state 2: subincrement c.364 COMPUTER PROCEDURES FOR STATIC AND DYNAMIC 9A. C C.DMATX. ICOUT.PARAM.DSTRE SUBROUTINEVONMIS (PROPD.LPRPD.LPRPD.8. k e l a s common / t m s t p l / i t i m e .ICPRT 2. ICDAT. .VOIDR.4 Time stepping informationfor the material model A common block was introduced to relay time stepping. 9A.8. . Usually 0 implies elastic response.3 Further error code from the material model for a. KTEST. ktest(3) number of cases where mdivt is exceeded 9A. ICOUT.DSTAN. ktest(l1) can be used as mcode to output a single digit code to indicate the state of the Gauss point. iteration and linear elastic step information to the soil model routine: integeritime. Future development could include outputing the code for iswdp = 2 to indicate the current state for the material model b. k e l a s When kelas is 1. IGAUS.ESTRE. IELEM.DSTRE l. .2 Output to screen A common variable ICHPC is used for the output to the screenlfile determined by the calling program..8. Direct output to the screen using PRINT statement or WRITE(6. ICDAT.LPARA.
CONS2 DOUBLE PRECISIONPMEAN DOUBLE PRECISIONSMALL PARAMETER ( SMALL = 1. OD 10) DOUBLE PRECISIONvsmal PARAMETER (vsmal= 1. 365 updatedon27/3/1991for amistakeinepstn . ..ANOTHER EXAMPLE IMPLEMENTATION C C.. ODO) DOUBLE PRECISIONREAL2 PARAMETER (REAL2= 2..VOIDR INPUT/OUTPUTVARIABLES: DOUBLEPRECISIONPROPD(LPRPD). C.FALSE. . CCOEF DOUBLEPRECISIONEPSTN (NSTRE) DOUBLE PRECISIONREALl. Od30) INTEGER ICOUN DOUBLEPRECISIONACOEF. C.. SVOLV. REAL0 = 0.. KTEST(NTEST) DOUBLEPRECISIONDSTAN(NSTRE). * C.BCOEF.. . OD0 ) DOUBLE PRECISIONREAP5 PARAMETER (REAP5= 0..VOIDR C.... DVOLV PARAMETER (REAL1= 1.... C.PARAM(LPARA) OUTPUTVARIABLES: DOUBLEPRECISIONDSTRE(NSTRE). MSTEP DOUBLE PRECISIONRATIO DOUBLE PRECISIONHMODU DOUBLEPRECISIONCONSl.* ) 'VOIDRATIO:'....ODO... IMPLICIT NONE INTEGER NSTRE PARAMETER (NSTRE = 4) INPUT VARIABLES INTEGER IELEM. IPARA INTEGER ISTEP...DMATX(NSTRE. NTEST INTEGERLPARA.NSTRE) LOCALVARIABLES LOGICAL DEBUG PARAMETER (DEBUG= . .. ICPRT INTEGER IGAUS INTEGER ISWDP INTEGER LPRPD INTEGER ICDAT INTEGER ICOUT.. .ESTRE(NSTRE). C. C. C.REALO.. updatedon15/10/1992toconformwith~alford~ortrancompilercheckingstandard C.) ! YOUNG'S MODULUS DOUBLE PRECISIONYOUNG ! POISSON'S RATIO DOUBLEPRECISIONPOISS ! LAME ' S CONSTANT DOUBLE PRECISIONRLMDA ! SHEAR MODULUS DOUBLE PRECISION SHEAR DOUBLEPRECISIONFSTRE (NSTRE) DOUBLEPRECISIONDEVIA (NSTRE) DOUBLEPRECISIONYIELD DOUBLEPRECISIONCURJ2 DOUBLE PRECISIONREQJ2 INTEGER ISTRE... JSTRE . 5DO) CHECK IF (DEBUG)WRITE (ICOUT.
. 3 2 . . 3 . ' R L M D A : ' . .1400. . ' S H E A R : ' . SHEAR 2 0 0 1 FORMAT ( ' Y O U N G : ' . W D P IS C . . . * .f o r ' i f ( p o i s s . VOLUMETRIC P A R T O F THE I N C R E M E N T . ' E l a s t i c D . O d O else propd ( 1 1 ) = O .POISS. 3 ) \ P R O P D ( 1 ) = RLMDA PROPD ( 2 ) = SHEAR P R O P D ( 3 ) =HMODU PROPD ( 4 ) = Y I E L D PROPD ( 5 ) = CURJ2 GOT0 2 9 9 9 C .HMODU. . 3 . FORM T H E I N C R E M E N T A L S T R E S S 1200 CONTINUE C P R I N T * .1500). E 1 1 . 'W" : P L A S T I C D E V I A T O R I C WORK DONE ' P R I N T * . ' H : HARDENING C O N S T A N T ' p r i n t * .RLMDA./. I S W D P C A L L T E R M I N ('ELASTI1:UNKNOWNOPTION:') GOT0 2 9 9 9 C . . E 1 1 . . . HMODU. l t . ' e x a m p l e e x p e r i . I E L E M .366 COMPUTER PROCEDURES FOR STATIC AND DYNAMIC G O T 0 (1100. E 1 1 . P O I S S .. O d O end i f RLMDA = YOUNG * P O I S S / ( ( R E A L 1 REAL^ * P O I S S ) * REAL^ + POISS)  ) S H E A R = YOUNG * R E A P 5 / ( R E A L 1 + P O I S S ) C U R J 2 = Y I E L D * Y I E L D / 3. T H I S ISTHEDEVIATORICCOMPONENTOFTHE I N C R E M E N T A L S T R E S S DSTRE ( 1 ) = R E A L 2 * SHEAR*DSTAN ( 1 ) D S T R E ( 2 ) = R E A L 2 * SHEAR * DSTAN ( 2 ) DSTRE ( 3 ) = REAL2 * SHEAR * DSTAN ( 3 ) C .YIELD. . . I N VON M I S E S ..ODO WRITE ( I C O U T .m a t r i x is used f o r i t e r a t i o n ' propd ( 1 1 ) = l . E 1 1 . DSTAN C . 2 0 0 1 ) YOUNG. READ I N M A T E R I A L DATA 1100 C O N T I N U E READ ( I C D A T . S I N C E ENGINEERING SHEAR S T R A I N I S USED DSTRE ( 4 ) =SHEAR*DSTAN ( 4 ) C .3 1 . . 'HMODU: ' . O . . 'J2" : SECONDDEVIATORIC INVARIANT' P R I N T *. ./. 'POISS:'. . * ) ' E l a s t i c D . E L A S T I C INCREMENT RLMDA=PROPD ( 1 ) SHEAR=PROPD ( 2 ) HMODU=PROPD ( 3 ) C U R J 2 = PARAM ( 1) C . E 1 1 . . O d 0 ) then poiss = poiss p r i n t *. . ERROR C O N D I T I O N WRITE ( I C O U T .1200. * ) ' I S W D P : ' . . CHECK CURRENT Y I E L D S T R E S S PMEAN=(ESTRE ( 1 ) +ESTRE ( 2 ) +ESTRE ( 3 ) ) /3. .E11. Y I E L D P R I N T *. E R R = 3100. 'VON M I S E S MODEL: J 2 ' ' = Y + H * W' ' ' P R I N T * . 3 .m a t r i x i s u s e d f o r i t e r a t i o n ' w r i t e ( 6 . T H I S I S ALWAYS C ELASTIC DVOLV=DSTAN ( 1 ) +DSTAN ( 2 ) +DSTAN ( 3 ) S V O L V = R L M D A * DVOLV C.ODO DEVIA ( 1 ) = E S T R E ( 1 ) +PMEAN . ' Y I E L D : ' .1300. . E N D = 3 2 0 0 ) YOUNG. I G A U S . .
0DO*SHEAR/ ((HMODU+~. . v s m a l ) THEN RATIO = 1.ODO*CURJZ) I F ( D E B U G ) PRINT * . . 1.O . n s t r e do epstn ( i s t r e ) =epstn ( i s t r e ) / d b l e (max ( 1 .ODO * BCOEF C C O E F = D E V I A ( l ) * D E V I A ( l ) +DEVIA ( 2 ) * D E V I A ( 2 ) 1 + 2 . .MSTEP.ODO*ACOEF*CCOEF)) 1 / (2. l e .MSTEP DSTRE ( 1 ) = 2 .ODORATIO)*DSTAN ( 3 ) E P S T N ( 4 ) = (1. O D O * D E V I A ( 4 ) * D E V I A ( 4 ) +DEVIA ( 3 ) * D E V I A ( 3 ) CCOEF = C C O E F .ODO ELSE RATIO= (BCOEF+SQRT (BCOEF*BCOEF~. . UPDATE THE S T R E S S S T A T E DEVIA ( 1 ) = D E V I A ( 1 ) + R A T I O k D S T R E ( 1 ) DEVIA ( 2 ) = D E V I A ( 2 ) + R A T I O k D S T R E ( 2 ) DEVIA ( 3 ) = D E V I A ( 3 ) + R A T I O k D S T R E ( 3 ) DEVIA ( 4 ) = D E V I A ( 4 ) + R A T I O k D S T R E ( 4 ) EPSTN ( 1 ) = (1.2 . .ODO"ACOEF) RATIO=MIN (RATIO. . 2 . 9 0 0 0 0 0 0 0 0 0 0 1 ~ O ) ) . c o r r e c t i o n on 2 7 / 3 / 1 9 9 1 1 2 4 1 i s t r e = 1. m s t e p ) ) 1241 continue c . L O O P OVER THE NUMBER O F S T E P S . . 0 ) c .ANOTHER EXAMPLE IMPLEMENTATION 367 D E V I A ( 3 ) = E S T R E ( 3 ) +PMEAN D E V I A ( 4 ) =ESTRE ( 4 ) PMEAN=PMEANSVOLV C .ODORATIO)*DSTAN ( 4 ) C . HOW FAR OUT I S THE S T R E S S P O I N T ? R E Q J 2 = 0 . R E Q J 2 . FORM THE C O E F F I C I E N T S ACOEF=DSTRE ( 1 ) *DSTRE ( 1 ) +DSTRE ( 2 ) *DSTRE ( 2 ) 1+2. O D O * S H E A R * E P S T N ( 3 ) SHEAR * E P S T N ( 4 ) DSTRE ( 4 ) = C O N S 2 = D E V I A ( 1 ) *DSTRE ( 1 ) + D E V I A ( 2 ) "DSTRE ( 2 ) 2 .ODO*DSTRE ( 4 ) "DSTRE ( 4 ) + D S T R E ( 3 ) *DSTRE ( 3 ) B C O E F = D E V I A ( 1 ) * DSTRE ( 1 ) + D E V I A ( 2 ) * D S T R E ( 2 ) 1+2. . O D O * SHEAR* EPSTN ( 1 ) DSTRE ( 2 ) = 2 . 0 D O k (DEVIA ( 4 ) + D S T R E ( 4 ) ) * * 2 ) C . e n d of c o r r e c t i o n C P R I N T *. 'MSTEP: ' . . .. . 5 D O * ( ( D E V I A (1)+ D S T R E ( 1 ) ) * * 2 + (DEVIA ( 2 ) +DSTRE ( 2 ) ) * * 2 1 + (DEVIA ( 3 ) +DSTRE ( 3 ) ) * * 2 2 3 + 2 . .. ODO * SHEAR * HMODU "(DEVIA ( 1 ) *EPSTN ( 1 ) +DEVIA ( 2 ) *EPSTN ( 2 ) . O D O * S H E A R * E P S T N ( 2 ) DSTRE ( 3 ) = 2 . DO 1 + 1 1250ISTEP=l.ODO*DEVIA ( 4 ) "DSTRE ( 4 ) +DEVIA ( 3 ) *DSTRE ( 3 ) BCOEF = 2.ODO) END I F C . THE NUMBER O F I N T E G R A T I O N S T E P S RATIO=SQRT (REQJ2 /CURJ2) MSTEP=MAX ( I N T (lO.ODO* ( R A T I O . .ODO*SHEAR) *2. O D 0 * C U R J 2 I F ( a b s ( A C O E F ) . . .ODORATIO)*DSTAN ( 1 ) EPSTN ( 2 ) = (1.ODORAT1OIXDSTAN ( 2 ) EPSTN ( 3 ) = (1. . . 0 D O k D E V I A ( 4 ) " D S T R E ( 4 ) + D E V I A ( 3 ) "DSTRE ( 3 ) CONSl=CONS2*2. . CURJ2 C .
368
2 3
COMPUTER PROCEDURES FOR S T A T I C AND D Y N A M I C
+DEVIA ( 3 ) *EPSTN ( 3 ) + D E V I A ( 4 ) *EPSTN ( 4 ) ) * CURJ2
/ (HMODU+2.ODO * SHEAR) CONS2CONS1*2.ODO
CURJ~=CURJ~+~.ODO*SHEAR*HMODU
"(DEVIA ( 1 ) "EPSTN ( 1 ) +DEVIA ( 2 ) *EPSTN ( 2 ) +DEVIA ( 3 ) * E P S T N ( 3 ) + D E V I A ( 4 ) " E P S T N ( 4 ) ) / (HMODU+2.ODO * SHEAR) DEVIA ( 1 ) =DEVIA ( 1 ) * (1.ODO CONSl) +DSTRE ( 1 ) DEVIA ( 2 ) =DEVIA ( 2 ) * (1.ODOCONS1) +DSTRE ( 2 ) DEVIA ( 3 ) = D E V I A ( 3 ) * ( 1 . O D O C O N S l ) +DSTRE ( 3 ) DEVIA ( 4 ) =DEVIA ( 4 ) * ( 1 . O D O C O N S l ) +DSTRE ( 4 ) REQJ2=0.5DO* (DEVIA ( 1 ) **2+DEVIA ( 2 ) ** 2 1 +2.ODO*DEVIA ( 4 ) * * 2 + D E V I A ( 3 ) * * 2 ) RATIO=SQRT (CURJ2/REQJ2) I F (DEBUG) P R I N T * , RATIO, C U R J 2 , R E Q J 2 DEVIA ( 1 ) =DEVIA ( 1 ) *RATIO DEVIA ( 2 ) =DEVIA ( 2 ) *RATIO DEVIA ( 3 ) =DEVIA ( 3 ) "RATIO DEVIA ( 4 ) =DEVIA ( 4 ) *RATIO 1 2 5 0 CONTINUE PARAM 1) = C U R J 2 F S T R E 1) = D E V I A 1 ) PMEAN F S T R E 2 ) = D E V I A 2 ) PMEAN F S T R E 3 ) = D E V I A 3 ) PMEAN FSTRE 4 ) = DEVIA 4 ) DSTRE 1 ) = FSTRE 1 ) ESTRE ( 1 ) DSTRE 2 ) = FSTRE 2 ) E S T R E ( 2 ) DSTRE 3 ) = FSTRE 3 ) ESTRE ( 3 ) DSTRE ( 4 ) = F S T R E ( 4 ) E S T R E ( 4 ) ESTRE ( 1 ) = F S T R E ( 1 ) ESTRE ( 2 ) = F S T R E ( 2 ) ESTRE ( 3 ) = F S T R E ( 3 ) ESTRE ( 4 ) = F S T R E ( 4 ) GOT0 2 9 9 9 C . . . . SYMMETRIC V E R S I O N O F DMATX 1 3 0 0 CONTINUE C . . . . NOTNECESSARYSYMMETRICVERSIONOFDMATX 1 4 0 0 CONTINUE RLMDA=PROPD ( 1 ) SHEAR=PROPD ( 2 ) HMODU = P R O P D ( 3 ) C U R J 2 =PARAM ( 1 ) DMATX ( 1 , l ) = RLMDA + R E A L 2 * S H E A R DMATX ( 1 , 2 ) = R L M D A DMATX ( 1 , 3 ) = R L M D A DMATX ( 1 , 4 ) = R E A L 0 DMATX ( 2 , l ) = R L M D A DMATX ( 2 , 2 ) = RLMDA + R E A L 2 * S H E A R DMATX ( 2 , 3 ) = R L M D A DMATX ( 2 , 4 ) = R E A L 0 DMATX ( 3 , l ) = R L M D A DMATX ( 3 , 2 ) = R L M D A DMATX ( 3 , 3 ) = R L M D A + R E A L 2 * S H E A R DMATX ( 3 , 4 ) = R E A L 0 DMATX ( 4 , l ) = R E A L 0 DMATX ( 4 , 2 ) = R E A L 0
1 2 3
ANOTHER E X A M P L E IMPLEMENTATION
369
C..
.
DMATX ( 4 , 3 ) = R E A L 0 DMATX ( 4 , 4 ) = S H E A R CHECK C U R R E N T Y I E L D S T R E S S PMEAN=(ESTRE ( 1 ) + E S T R E ( 2 ) + E S T R E ( 3 ) ) /3.ODO DEVIA ( 1 ) = E S T R E ( 1 ) +PMEAN DEVIA ( 2 ) = E S T R E ( 2 ) +PMEAN DEVIA ( 3 ) = E S T R E ( 3 ) +PMEAN DEVIA ( 4 ) =ESTRE ( 4 ) REQJ2=0.5DO* (DEVIA ( 1 ) * * 2 + D E V I A ( 2 ) * * 2 1 +2.ODO*DEVIA(4) **2+DEVIA ( 3 ) ""2) I F (REQJ2/CURJ2.GT.1.ODO+SMALL) T H E N P R I N T * , 'REQJ2 > CURJ2 I N V O N M I S : ' , R E Q J 2 , C U R J 2 STOP 'STOPPEDINVONMIS' END I F I F (REQJ2/CURJ2.GT.O.99DO.and.abs ( p r o p d ( l l ) ) . l e . v s r n a l ) T H E N
CONS1=2.ODO*SHEAR*SHEAR/
(
(HMODU+~.ODO*SHEAR)*REQJ~)
DO1420JSTRE=l,NSTRE 1 4 1 0 I S T R E = 1, N S T R E DO DMATX ( I S T R E , J S T R E ) =DMATX ( I S T R E , J S T R E ) 1 (CONSl*DEVIA (JSTRE)) *DEVIA (ISTRE) CONTINUE CONTINUE END I F GOT0 2 9 9 9 C . . . . STORE SOMETHINGINTOPARAM 1500 CONTINUE D O 1 5 1 0 I P A R A = 1, L P A R A PARAM ( I P A R A ) = O . O D O CONTINUE PARAM ( 1 ) = P R O P D ( 5 ) PARAM ( L P A R A ) =  I 9 GOT0 2 9 9 9 C . . . . COMMON E X I T 2 9 9 9 CONTINUE RETURN C . . . . ERROR C O N D I T I O N 3100 C O N T I N U E WRITE ( I C O U T , * ) ' I C D A T : ' , I C D A T CALLHANDLE ( ' E L A S T I  2 : E R R O R I N C H A N N E L ' ) RETURN CONTINUE WRITE ( I C O U T , * ) ' I C D A T : ' , ICDAT CALLHANDLE ( ' E L A S T I  3 : E N D O F F I L E I N C H A N N E L : ' ) RETURN END
Author Index
Adkin J. E. 18, 50 Advani S. H. 202, 203. 204, 206, 207, 215 Altabbaa A. 337, 350, 352,353 Alonso E. E. 14, 15, 198, 216, 352,353 Anandarajah A. 158, 171 Armero F. 324, 332 Arulanandan K. 5, 15, 218, 230, 231, 247 Arulmoli K. 237, 247 Atkinson J. H. 113, 135, 171 Aubry D. 131 , 159,171 Babuska I. 64, 70, 81 Bacchus D. R . 140, 175 Baggio P. 188, 189, 193, 215 Bahda F. 151, 157, 171 Bailey 184 Baker R. 158, 161, 171 Balakrishna J. 14, 15 Balasubramanian A. S. 134, 135, 138, 139, 171 Balestra M. 324, 332 Banerjee P. K. 158, 172 Bardet J. P. 131, 172 Bathe K. J. 202, 215 Batlle F. 198, 216 Bazant Z. P. 129, 172, 302, 321, 331 Bear J. 14, 15 Belkheir K. 352, 354 Bell 184 Bellman R. 81 Belytschko T. 321, 331 Bergan P. G. 66, 81 Bettess P. 27, 51 Bianchini G. 144, 147, 148, 150, 151, 175 Bicanic N. 18, 36, 52, 84, 188, 209, 212, 216, 312, 313,333
Biot M. A. 8, 12, 15, 18, 49, 50 Bishop A. W. 8, 14, 15, 40, 50 Booker J. R. 202, 207,215 Boroomand B. 3 19,333 Boussinesq J. 6, 15 Bowen R. M. 18, 39,50 Bransby P. L. 113, 171 Brezzi F. 64, 70, 81, 324, 331 Britto A. M. 350, 351, 353 Brooks R. N. 189, 194, 195, 198, 215 Burland J. B. I l l , 123, 172, 175 Cambou B. 172 Carter J. P. 202, 215 Castro G. 144, 145, 153, 154, 172 Chambon R. 133, 172 Chan A. H. C. 18, 25, 26, 50,51, 55, 57, 60, 64, 70, 78, 81, 84, 90, 132, 137, 139, 148, 149, 151, 152, 155, 159, 163, 172, 174, 221, 226, 230, 233, 235, 237, 247, 248, 250, 276, 302, 303, 336, 350, 351,353,354 Chang C. T. 27,51, 62,84, 129, 130, 165, 176, 282, 303 Chaudhry A. R. 134, 135, 138, 139, 171 Chen W.F. 21 1, 337, 350,353 Cheung Y. K. 344, 353 Chorin A. J. 325, 332 Clough R. W. 71, 81,82, 3 12, 332,333 Codina R. 325,333 Cooke B. 221. 248 Corapcioglu M. Y. 14, 15 Corey A. T. 189, 194, 195, 198,215 Coulomb C.A. 85, 100, 175 Coussy 0. 39,50 Cowin S. C. 160, 172 Craig R. F. 25, 50
372 Creager W.P. 284, 295. 302 Crisfield M. A. 63, 82 Cristescu N. 352, 353 Crouse C. B. 236, 249 Cuellar V. 129, 172 Culliganhensley P. J. 221, 248 Dafalias Y. F. 90, 131, 132, 133, 158, 171, 172, 173, 175 Darve F. 90, 133, 159, 172 De Boer R. 8, 15, 18, 39,50 De Borst R. 321. 332 de SaintVenant 85, 175 Derski W. 18, 50 Desai C. S. 36, 50, 158, 161, 171 Desrues J. 133, 172 Dewoolkar M.M. 62,221, 227, 245246,249 Di Prisco C. 158, 172 Dikmen S. U. 255, 261,302 Dobry R. 237, 245,249, 250 Drucker D. C. 86, 111, 120, 144, 172, 188, 215 Ehlers W. 39, 50 Englemann B. E. 32 1, 331 Esrig M. I. 131, 175 Famiyesin 0.0. R. 25,26,50, 55,78,81,226, 233,235,237,248 Felippa C. A. 64. 83 Fillunger P. 6, 7, 10, 15 Finn W. D. L. 129, 267, 303 Fish J. 321, 331 Flavigny E. 133, 172 Franca L. P. 324,332 Frossard E. 135, 172 Fruth L. S. 236, 247 Fuji 267, 302 Fujikawa S. 256, 302 Fukutake K. 256,302 Gantmacher F. R. 82 Gawin D. 40, 51, 188, 189, 193,215 Gens A. 14, 15, 198, 216, 352,353 van Genuchten M. T. 14, 16, 277,303 Ghaboussi J. 25, 50, 72, 82, 131, 158, 172, 255, 261,302 Gibson R. E. 8, 15, 111, 120, 172 Graham J. 337, 351, 353 Gray W.G. 39, 40,42, 50 Green A. E. 18, 39, 50 Griffiths A.A. 323, 332 Gudehus G. 188, 215 Gunn M. J. 350, 351, 353
A UTHOR INDEX
Habib P. 125. 142. 173 Hafez M. 325, 332 Hamilton C. J. 351, 353354 Hashiguchi K. 131, 158, 173, 352, 354 Hassanizadeh M. 39, 40, 42,50 Hatanaka M. 262, 302 Hellawell E. E. 22 1 , 249 Henkel D. J. 11 1, 120, 131, 140,172,173,175 Herrmann L. R. 131, 132,172 Heyliger P. R. 202. 215 Hight D. W. 14, 15 Hill R. 158, 160, 173 Hinds J. 284, 295,302 Hine N. W. 60. 66.84 Hinton E. 62, 64, 83, 84, 129, 130, 165, 176, 282. 303. 337. 351. 354 Hirai H. 131, 158, 173 Holubec I. 128, 174 Hosomi H. 268, 303 Houlsby C . T. 337, 351, 353 Huang G. C. 333 Huang M. 231, 250251, 3 16, 318, 319, 324, 332, 333 Huber M.T. 86, 175 Hughes T. J. R. 324, 332 Hulhaus H. B. 321, 332 Hujeux J. C. 131, 159, 171 Humpheson C. 128, 175, 179, 216 Hushmand B. 236, 237, 249 Hyodo M. 267, 302 Iai S. 23 1 , 250 Idriss I. M. 1. 16, 274, 303 Imamura T. 131, 173 IshiharaK. 125, 142, 151, 152, 153, 163, 164, 173, 175, 237, 249, 255, 256, 302 Ito K. 231. 248 Jauman G. 20, 50 Josa A. 352, 353, 354 Justin J.D. 284, 295, 302 Kaliakin V. N. 131, 173 Kansaikyogikai 257, 302 Karadi G. M. 195, 197,216 Katona M. G. 60.82, 234, 249, 336,354 Kibayashi M. 268, 303
AUTHOR INDEX
373 Makdisi F. I. 274, 303 Marti J. N. 312, 332 Martin G. R. 129, 255, 303 Masuda A. 293,302 Matsuoka H. 159, 173 Matthies H. 63, 82, 344, 354 Mchenry D. 8, 15 Meijer K. L. 202, 216 Meimon Y. 131, 159, 171 Meiri D. 195, 197. 216 Melan E. 86, 173 Meroi E. A. 193, 194, 195, 201, 201, 202,216 Mira P. 332 Mitchell R. J. 221, 248 von Mises R. 86, 175 Miura S. 163, 173 Mizuno E. 21 1, 337, 350, 353 Molenkamp F. 352, 354 Mollener 66, 81 Momen H. 131, 158. 172 Monte J. L. 202, 216 Morgan K. 317, 332 Morland L. W. 18, 39, 51 Mroz Z. 90, 92, 131, 158, 173, 174, 175 Muir Wood D. 25, 26, 50, 55, 78, 81, 128, 174, 226, 233, 235, 237, 248, 337, 351. 352. 353,354 Muller L. 1, 2, 15 Muraleetharan M. M. H. 236, 247 Murata H. 267, 302 Muromoto T. 28 1 , 303 Nagase H. 255,302 Nakamura N. 293,302 Nakazawa S. 64. 70. 84 Narasimhan T. N. 14, 15, 188,216 Narikawa M. 293,302 Needleman A. 321,332 Neuman S. P. 14, 15 Newmark N. M. 60, 82, 336, 354 Newton R. E. 312,333 Ng W. S. 341,354 Norris V. A. I3 1, 158, 173, 174 NovaR. 127, 128,143,158, 160, 161,172, 174 Ohtsuki A. 256, 302 Okada S. 151, 152, 153, 173 Oliver J. 324. 332 Ortiz M. 321, 332 Owen D. R. J. 202, 216, 337, 351,354
Kim C. S. 202. 203. 204. 206, 207, 215 Kimuira T. 268, 303 Kishino Y. 293, 302 KO HonY. 237,250 Kolymbas D. 90, 133, 173, 175 Kondner R. L. 256, 302 Kowalski S. J. 18, 39. 50 Krieg R . D. 131, 173 Krizek R . J. 129, 172, 202, 216,302 Krucinski S. 159, 164, 174 Kuhlemeyer R. L. 312, 332 Kunar R. R . 312, 332 Labanieh S. 90, 133, 159, 172 Lambe P. C. 230,250 Lambe T. W. 25, 50 Lanier J. 158, 172 Lassoudiere F. 131, 159, 171 Law H. K. 231, 250 Ledesma A. 18,36,52,84, 188,209,212,216 Lee F. H. 219, 249 Lee I. K. 312,332 Lee J. H. W. 202,203, 204, 206, 207,215 Lee K. L. 1, 16, 274. 303 Lee T. S. 202, 203, 204, 206, 207, 215 Leliavsky S. 8, 15, 19, 51 Leroy Y. 32 1,332 Leung A. Y. T. 280. 344. 353 Leung K. H. 57, 62, 64,82,83,84, 90, 128, 129, 132, 135, 147. 163, 174. 175. 176 Leung K.H. 303 Levy M. M. 6, 15 Levy M. 86, 175 Lewis R. W. 40, 41, 51, 67, 82, 128, 175, 179, 202,215,216 Li G. C. 36, 50 Li T. 325, 332 Li X. K. 51. 82, 325, 332 Liakopoulos A. C. 14, 15, 36, 51, 188, 189, 215 Liang R. L. 158, 173 Lin F. B. 321, 331 Liu D.F. 352, 354 Lloret A. 14, 15, 198, 216 Lo S. H. 344,353 Luong M. P. 125, 142, 173 Lyell C. 5, 15 Lysmer J. 31 2,332 Madabhushi S. P. G. 240, 249 Makdisi F. I. 1, 16
372 Creager W.P. 284, 295, 302 Crisfield M. A. 63, 82 Cristescu N. 352, 353 Crouse C. B. 236, 249 Cuellar V. 129, 172 Culliganhensley P. J. 221, 248 Dafalias Y. F. 90, 131, 132, 133, 158, 171, 172, 173, 175 Darve F. 90, 133, 159, 172 De Boer R. 8, 15, 18, 39, 50 De Borst R. 321,332 de SaintVenant 85, 175 Derski W. 18, 50 Desai C. S. 36, 50, 158, 161, 171 Desrues J. 133, 172 Dewoolkar M.M. 62, 221,227, 245246, 249 Di Prisco C. 158, 172 Dikmen S. U. 255, 261,302 Dobry R. 237, 245, 249, 250 Drucker D. C. 86, 11 1, 120, 144, 172, 188, 215 Ehlers W. 39, 50 Englemann B. E. 321, 331 Esrig M. I. 131, 175 Famiyesin 0.0.R. 25,26,50,55, 78,81,226, 233,235,237,248 Felippa C. A. 64, 83 Fillunger P. 6, 7, 10, 15 Finn W. D. L. 129, 267,303 Fish J. 321, 331 Flavigny E. 133, 172 Franca L. P. 324, 332 Frossard E. 135, 172 Fruth L. S. 236, 247 Fuji 267, 302 Fujikawa S. 256, 302 Fukutake K. 256,302 Gantmacher F. R. 82 Gawin D. 40, 51, 188, 189, 193,215 Gens A. 14, 15, 198, 216, 352,353 van Genuchten M. T. 14, 16, 277,303 Ghaboussi J. 25, 50, 72, 82, 131, 158, 172, 255, 261,302 Gibson R. E. 8, 15, 111, 120, 172 Graham J. 337, 351,353 Gray W.G. 39,40,42,50 Green A. E. 18, 39, 50 Griffiths A.A. 323, 332 Gudehus G. 188, 215 Gunn M. J. 350. 351, 353
AUTHOR INDEX
Habib P. 125, 142. 173 Hafez M. 325, 332 Hamilton C. J. 351, 353354 Hashiguchi K. 131, 158, 173, 352,354 Hassanizadeh M. 39, 40, 42, 50 Hatanaka M. 262, 302 Hellawell E. E. 221, 249 Henkel D. J. 1 1 1 , 120, 131, l40,I72,l73,175 Herrmann L. R. 131, 132, 172 Heyliger P. R. 202, 215 Hight D. W. 14, 15 Hill R . 158, 160, 173 Hinds J. 284, 295, 302 Hine N . W. 60, 66,84 Hinton E. 62, 64, 83, 84, 129, 130, 165, 176, 282,303, 337, 351,354 Hirai H. 131, 158, 173 Holubec I . 128, 174 Hosomi H. 268, 303 Houlsby G. T. 337, 351, 353 Huang G . C. 333 Huang M. 231, 250251, 316, 318, 319, 324, 332, 333 Huber M.T. 86, 175 Hughes T. J. R . 324, 332 Hulhaus H. B. 321, 332 Hujeux J. C. 131, 159, 171 Humpheson C. 128, 175, 179, 216 Hushmand B. 236, 237, 249 Hyodo M. 267,302 Iai S. 231, 250 Idriss I. M. 1, 16, 274. 303 Imamura T. 131, 173 Ishihara K. 125, 142, 151, 152, 153, 163, 164, 173, 175, 237, 249, 255, 256, 302 Ito K. 231, 248 Jauman G. 20,50 Josa A. 352,353, 354 Justin J.D. 284. 295. 302 Kaliakin V. N. 131, 173 Kansaikyogikai 257, 302 Karadi G. M. 195, 197, 216 Katona M. G . 60,82, 234, 249, 336, 354 Kibayashi M. 268, 303
92. 51. 173 Tokimatsu K. Z. 302. 173 Vahadati M. V. 281. 332 Whitman R .216 Zhu J. 333 Wu W. 21. 131. 25. 336. 139. 179. 51.302 Ueno M. 140. 312. 148. 173. 212. M. 58.83. 303 Tresca H. 1 302 Taylor D. 344. 15. 8. 83. P. 186. 255. 135. 202. 312. 172 Yovanovich M. 256. 128. 175 Uchida A. 82. 128.267. 132.24. P. 199. 354. 162.25. 195. 317. 196. 142. 267. 158. 163. 58. 331 Taboada V. 133. 351. 302 Yoshimi Y. 303 Tanimoto K.302. C. 51 White W. 175 Tani S. 131. 153. 70. 209.52. 235. see Muir Wood D. 355 Terzaghi K. 250. 274. 336. 143. 189. 1 1 I. L. 324. 163. 175. 104. 219. 11 1. Y. 202. 36. 249. 249. 51. 78. 55. 256.A UTHOR INDEX 375 Sugimoto M. 18. M. 319. 324. 175 Whitaker S. B. 190. 84. 230. 280. 175 Yamazaki F. G. 125. 332 Wang Z. 333. 63. Wood W. 163. 147. 173. 316. S. 221. 343. 130. 216 Thompson P. 120. 15. 332. 188. 186. 152. 77. Y . 302 Zeng X. 181.326.84 Wroth C. 250 Vilotte J. 20.250 Tanaka Y.51. 325. 62. 230. 66. 139. 339.303 Suzuki Y. 131. 312.313. 53. M. 256.84. 64. 263. 25. 313. 158. 201. 27. 216. 50 Wilson E.319. 325. 39. S. P. 16. 303. W. 70. 51 Zelasko J. 257. 333 Wu S. 83 Wu J. S. 50.343. 165. 90. 124. 303 Tabarrok M. 133. L. 125. V. 85. 184. W. 257. 147. 82. 151. 142. 175 Willis P. L. 25. 237.250 Z h a n X . 337. 149.302 Yasufuku N. 50. 317. 175 Witherspoon P. 172. 332 Venter K. 175 Tatsuoka F.332 Vaid Y. 83. 237. 63. 8. 78. 250 Thurairajah A. 127. 188. 233. 175 Taylor R. 317. 188. 344. L. 216 Wolf J. 198. 83. 215 Wineman A. 60. 99. 175 Toki S. 319. 143. 174. 66. 264. 155. 121. 144. 264. 312. 129. 281. 332 Zaremba S. 303 Youssif N. 355 . 21.302 Yasuda S.247. 175. 18. 262. 66. 321. Von 16. 137.312. 175 Yamada Y. 276. 26. 282. 60. 146. 64. 318.333. 159. 164. 293. 317. 60.51 Terzaghi K. 176. 355 Taylor P. 21 8. P.333 Zienkiewicz 0. 333 Wood D. 355 Wilde P. 143. 72. 251. 326. A. 268. 234. 67. 175 Wu J. S. 53. 173. 72. 303 Valliappan S. 14.
.
Subject Index ABAQUS 350 Accoustic emission 143 Accumulated shear strain 255 Accuracy control 3 14 Adaptive refinement 305.see effective stress Biot Theory 18. 243 numerical modelling of uniform loose sand bed test 237245 Scaling laws 21 922 1 Silicon oil as substitute fluid 221 use of substitute fluid 220221 validation of numerical solutions 21 8 VELACS study 218. 230245 Characteristic state line 125. 154 Bangkok clay 1381 39 Biot alpha . 148. 158. 151 Cocontinuous 64 Cohesion due to suction 276 Consistency condition 10&105 Consolidation equation 28. 146. 324 Backward Euler integration scheme 352 Balance equation of fluid and air mass for partially saturated soil 38 of fluid mass for fully saturated soil 2223 of fluid mass for partially saturated soil 3435 see ulso general field equation Balance equation for mixture theory macroscopic balance equations 4 2 4 3 microscopic balance equations 42 of energy 4243 of entropy 4 2 4 3 of mass 4 2 4 3 of momentum 4 2 4 3 Banding sand 145.see stress invariants structure tensor 159 transverse isotropic material 1601 6 1 Autogenous volumetric strain 129. 3 14324 Analysing procedure 233234 anisotropic material tensor 158 Anisotropy 151. 159 loadistress induced 151. 66. 67 fully and partially saturated soil column 203206 . 166 BabuskaBrezzi condition 64. 142. 159 modified stress invariants . 70. 242. 144. 1571 63 fabric tensor 159 initial or fabric 158. 256 macroscopic thermodynamical approach 39 Bishop parameter see effective stress Boundary condition 2324 Bulk modulus of fluid 78 of soil matrix 78  capillary pressure 13 Capilliary pressure 40 Centrifuge 6 dynamic compatibility 2 1922 1 numerical modelling of loose and dense sand bed test 237245 numerical modelling of retaining dyke test 221230 numerical modelling of retaining wall test 245247 numerical modelling of submerged quay wall test 240.
351. 78 i ~ h .see SM2D linear elastic 269. 11 91 20 rounded 104. 269 used in densification model for cyclic mobility I65 Pietruszczak twosurface model 352 state parameter based onedimensional model 352 threesurface kinematic yield surface model 352 Tresca 99100.113. 337. 351 parameters for Lower San Fernando dam analysis 276 parameters for Ndam analysis 291. 164 SUBJECT INDEX unloading and cyclic loading 1511 57 for the anisotropic bahaviour of sand 163165 Hashiguchi subloading surface model 131. 352 alternate model 352 anisotropic Elastic 337. 28&285. 351 Barcelona unsaturated soil model 352 benchmark tests 1 10 brick model 352 calibration 1 11 CamClay 120123. 44 Constitutive models (see ah0 plasticity framework) AlTabbaa and Wood 337. 35 1 endochronic theory 129 generalized plasticity model for normally consolidated clays 1341 39 for overconsolidated clays 139141 for sands 141157 Pastor Zienkiewicz mark111 model 132. 165. 337. 149150 inversion 9293 isotropic linear elastic 25. 184. 337. 287. 351 used with or without initial shear stress 267 von MisesHuber 9799. 7576 ~ a ~ l e damping 71. 337 two surface kinematic hardening plane strain model for sand 337. 143. 296 modified for cyclic mobility 1651 71. 352 Constitutive tensor 1920. 120122.378 Consolidation equation (contd. 351 CAP model 337. 91 control volume 43 Cosserat medium 321 Coulomb 85 coupled analysis 4 coupling matrix 59 CRISP 350 critical state line 11 1. 128. 1321 33. 254 determination of soil parameters 253 Drucker Prager 102104. 186. 352 hyperbolic stress streain relation for shear 256 implementing new model subroutine into SWANDYNE . 337. 166 Damping algorithmic (numerical) damping 62 damping matrices 71.337 elastic model with varying moduli 337.187188. 351 modified CamClay 123124. 144 for sands 124.) fully and partially saturated twodimensional soil layer 206209 small strain and finite deformation 202203 Terzaghi theory 204 Constitutive equation for gas (Dalton's law) 4 4 4 5 for soil 19. 153. 184. parameters for VELACS study 236 parametric idenfication 235236 true triaxial path 1491 51. 351 Mohr Coulomb model 100102 applied for LSSGW 269 applied in limit examples 1791 88 implementation in SWANDYNE 337. 282. 143 used in densification model 130. 186. 352 lack of plastic strain before yield surface is reached 1 1 3. 350 computational aspects 1051 10 concrete interface model 352 concrete model 352 constitutive model subroutines 349352 Cristescu viscoelastoplastic model for saturated sand 352 critical state models 120124 extending to sands 12&129 for normally consolidation clays 111115 densification models 130 liquefaction parameters 270. 136. 148 critical state model Cyclic mobility 129. 337.
155. 310 Poisson ratio 27 Young's modulus 27 Equilibrium equation of fluid 22 of mixture 1921 See ulso general field equations Equilibrium line 131 Error indicator 3 17 Failure mechanism investigation using numerical method 254. 298 USA record 306 Earthquake damage Counter measures deep soil mixing 253. 263. 254262 soil layer liquefaction during Niigata 1964 earthquake 253254. 163164 Effective stress path approach 256 negative dilatancy 256 Discontinuity in displacement 3 16 Discrete memory factor 153 Drained Analysis 6769. 328 Lower San Fernando dam failure 253254. 70 Drained behaviour 4 Drucker stability criteria 144 Earth pressure coefficient at rest (KO) 262263. 267. 293302 latticeshaped stiffened ground walls (LSSGW) 267273 Earthquake loading examples earth dam failure in Hokkaido 253254. 39 Bishop parameter 40 from mixture theory under the assumption of incompressible grains 3 9 4 0 in fully saturated porous media 6 1 2 . 274281. 37 Del Monte sand 188 Density fluid 21 solid 21 volume averaged 21 deviatoric stress tensor 95 DIANASWANDYNE 11seeSWANDYNE Dilatancy 122 dilatancy rule for clay 134135 dilatancy rule for sand 128. 297298 Taft (1952) 271. 267273 rock fill dam failure .287293 Earthquake motion input 305. 256. 293302 elastoplastic large strain behaviour 21 1215 379 fully saturated soil column 20921 1. 33 under the assumption of local thermodynamic equilibrium 39 Elastic modulus for generalized plasticity model 1371 38 one dimensional constrained modulus (or restrained axial modulus) 27.see Lower San Fernando dam failure soil layer liquefaction at Kobe Port Island 253254. 287293 AkitaKencho record 288290 Rinkai (1992) artificially generated for Tokyo Bay area 295. 6 Biot alpha 1 112. 265266 in situ measurement 262 Earthquake ElCentro (I 940) 2 10. 141142. 19. 28 1287 effect of deep soil mixing counter measure 253254. 33.262267 soil structure interaction at Kawagishicho during Niigata 1964 earthquake 253254. 269. 273 Kobe (1995) see HyogokenNanbu (1995) Nihonkai Nanseibu (1993) 283 Niigata (1964) 253254. 258. 28 1287 Failures 1771 78 due to loss of suction 276 dynamic failure I failure behaviour 9697. 257. 27 1. 319 quay wall failure in Kobe City 253254. 19 in partially saturated media 1214. 153 local failure 105 . generalized Dashpots 3 1 1 Degree of saturation of air 37 of water 12. 256. 298 HyogokenNanbu (1995) 255256.SUBJECT INDEX viscous damping 71 Darcy's law. 306315 Earthquake record 66 acceleration records 254 input motion recorded during earthquake 254 Edogawa layer 298 Effective stress 4.
see ZarembaJaumann stress changes Kinematic equations 41 Kobe City hotel with earthquake damage counter measure 267268 Ocean Weather Station 271 Port Island 256257 Laminar box numerical n~odellingof 236 Levy 86 Liquefaction 4. 4 Vajont 1. 326 General field equation Linear momentum balance for fluid phases 4 5 4 6 Linear momentum balance for solid phase 4546 Mass balance equation 4 6 4 7 Generalized plasticity Geometric nonlinearity 209 GLADYS2E 55. 144 anisotropic hardening 158 deviatoric hardening 126128. 285 due to Selfweight 262 Type 1 267 Type I1 267 Iteration 6364  Jacobian matrix 63. 335 Governing equation fully saturated behaviour with a single pore fluid 1927 partially saturated behaviour with air flow considered 3639 partially saturated behaviour with air pressure neglected 3 136 Hardening behaviour 9 6 9 7 . 94 Localized phenomena 3 14324 nonuniqueness of numerical solution 3 19324 Galerkin process 54. 330 Quadrilateral Q81P4 330 Quadrilateral Q9/P4 324 Triangular T6CI3C 3 19320. 165  velocity correction 325328 undrained condition 3 2 4 3 2 5 Infinite foundation see radiation boundary condition Inhomogeneity 125 Initial matrix method 69 Initial shear stress 253. 70 Stabilization 305.139. 297 modelled in densification model 130 modelled in kinematic hardening model 131 modelled in modified densification model 166 modelled under cyclic loading in generalized plasticity model for sand 153 modelled under monotonic loading in generalized plasticity model for sand 144 numerical modelling examples 237245 see also earthquake loading examples reason for the development of advanced models 129 Loading criteria 9092.164 isotropic hardening 158 Strain hardening 1 0 4 105 History dependent behaviour 89 Hollow cylinder 150 Hostun Sand 148 Huber 86 Image point 132 Incompressible behaviour 64.) Lower San Fernando dam 13 see also earthquake loading examples see ulso earthquake loading examples static failures 1. 262267.SUBJECT INDEX Failures (contd. 282. 271 criteria for full liquefaction 267 induced by anisotropy 157 liquefaction strength 255. 265. 270. 66 Jaumann stress rate 336 Jaumann stress rate . 322 Finite element discretization 5355 Finite rotation 20 Fluid compressibility 26 Flushing water 256 Free surface see phreatic surface Fuji river sand 153. 164. 32&33 1 operator splitting algorithm 326328 reason for success 328331 . 284. 2 Finite Element Quadrilateral Q41P4 324. 295.147. 327.
322 Mixed interpolation 32433 1 Mixture Theory 18. 195 KozenyCarmen equation 339 See ulso general field equations Phase transformation line 125. 69. 67. 275 Plastic modulus for anisotropic behaviour of sand 163 for bounding surface model 1311 32 for critical state model extended to sand 127 for generalized plasticiy framework 9192 for kinematic hardening models 131 for normally consolidated clay 136 for overconsolidated clay 139 for sand under monotonic loading 143144 for sand under unloading and cyclic loading 1521 53. 1551 56 Plastic potential surface 122 for loose and dense sand 128 Plasticity framework 8586 (see rrlso constitutive models) bounding surface models 1311 32 classical plasticity 91. 143. 254262 multilinear laws 90 Multistep methods 60 Neutral criteria 9092 Newmark method see time stepping scheme Newton Raphson iterative procedure 63. 324 Permeability 65 anisotropic permeability 78 Creager's approach 284. 322 lucky 319. 931 10 pressure dependent criteria 1001 04 pressure independent criteria 97100 critical state framework 110129 generalized plasticity 8792 gradient plasticity 32 1 hypoplasticity and incrementally nonlinear models 90. 3 9 4 9 MuDIAN 281 Multidirection loading 253.SUBJECT INDEX 381 Lode angle 96. 78 Mass matrix 59. 4 4 4 5 . 326 Mesh bad 319. 188189. 1201 21 relation with Plasticity Index 113  Overconsolidation ratio 1 15 partial saturation Partially saturated examples air storage modelling in an aquifer 195197 consolidation of soil column 203206 consolidation of twodimensional soil layer 206209 elastoplastic large strain behaviour 21 1215 flexible footing resting on a partially saturated soil 198201 onedimensional column 36. 77 diagonal form 64. 195 in partially saturated media 14. 188192 subsidence due to pumping from a phreatic aquifer 193195 Patch test 70. 3738 effect of permeability on validity of various assumptions 293 1 for air flow 3738. 158 multilaminate 159 phenomenological aspects 8687 Poncelet 85 Pore water migration 253. 142 Phreatic surface 35. 137. 189. 295 Darcy's Law 22. 132133 kinematic hardening models 13013 1 multisurface kinematic hardening model 130131.4445. 33. 328 Niigata sand 154 Nonassociative flow rule 128 Normal Consolidation Line 113.see balance equation Mass lumping 72. 163 modified for anisotropy 162 LSSGW .see earthquake damage counter measures LUSAS 350 Mass conservation . 274281 Power plant 294 Predictions Class A 21 7218 Class B 2 172 18 Class C 217 Profile length minimizer CuthillMcKee method 341 Sloan method 341 Profile solver 343. 344 . 75.
181183 intermediate constraint on deformation 181. 62. 7778. reloading and cyclic loading 110 Stress ratio 124. I84 small constraint on deformation 1791 8 1 SUBJECT INDEX strong constraint . 62 conditional stability 33 1 unconditional stability 66 Stabilization of staggered scheme 64 Staggered procedure . fully Saturationcapillary pressure relationship 188189. 57. 62. 335337 constitutive model subroutines 349352 element matrices and residual calculation subroutines 345347 Internet URL xi major service subroutines 347349 subroutines for analysis 342345 subroutines for Control and material data input 338339 subroutines for mesh data input 339342 system dependent subroutines 352353 top level routines 337338 Tangent stiffness method 69 Three dimensional modelling 257258. 205 Secant update 63 Shape of yield surface in xplane 98. 119120 conventional triaxial stress path 110. 164 Stress paths anisotropy 110 consolidated drained test 110. 134.undrained behaviour 1821 86 Steady state 6869 Stiffness matrix 63. 115120 modelled by modified densification model 1681 70 isotropic Compression 11 1 isotropic Consolidation 110. 178188 embankment 1791 81 footing 179. 160162. 137. 287293 Tilted building 253.see time stepping Standard penetration test 257 Static analysis 64. 100.see earthquake damage counter measures St. 1 1 6 1 19 consolidated undrained test 110. 3063 15 onedimensional problem 3083 12 twodimensional problem 3 123 15 Rankine 85 Reid Sand 150 Relative fluid displacement 26 Repeated boundary condition 3 13 Representative elementary volume 43 Residual condition 143. 136 threedimensional effects 110 unloading.1 10 modified stress invariants 158. 281 saturated. 64.382 QuasiNewton method 69 BFGS 344 Quay wall 267273 Radiation boundary condition 305. 336 G N 11 scheme 62 GN22 scheme 61 . Venant 85 Stability criteria 60. 147 Roscoe surface 120 Rotation of principal stress axes 1501 51 Sand boiling 256. 287293 Time step length 6667 critical time step length for explicit scheme 78 Time stepping scheme central difference scheme 62 error control 67 explicit scheme 55. 7&76 Stress invariants 95. 317 Simple shear test apparatus 255 Skempton B soil parameter for pore water pressure 2425 SM2D 350 example new model subroutine 364369 implementing new model subroutines 355364 Softening behaviour 9697 localization 3 19324 of sand 147 Soil improvement . 163 SWANDYNE 55. 150. 186188 Shear wave 310 Shock wave 314. 33 1 Generalized Newmark (GNpj) method 6062. 1 17. 163 computational aspects 105. 7778. 72. 60.
2425. 182 Undrained behaviour 4. 273 1. 94 Unloading plasticity 132. 152 up formulation 2527. 7173. 27C277 Vertical Input motion 259261. 27 van Genuchten's formula 14. 182 . 336 383 uU formulation 2627. 5557 explicit u and implicit p scheme 62 fully Implicit scheme spatial discretization 5859 structure of numerical equations 6970 temporal discretization 6065 tensorial form of the equations 7881 Updated Lagrangian Formulation 46. 6769. 336 SSpj 60 staggered procedure 64 trapezoidal scheme 62 Tresca 85 Twophase flow 188 Uncoupled equation 68 Undrained Analysis 64. 70. 77. 256 block diagonal structure 78 fully Explicit scheme 72 spatial discretization 7374 structure of numerical equations 7 4 7 7 uw formulation uwp formulation 1921.SUBJECT INDEX GN32 scheme 61 implicit scheme 55. 66 Newmark method 60. 265 Volume fraction 18 von Mises 86 Wave equation 3 10 Weald clay 140 Westergaard's formula for dynamic water pressure 299 Yield and failure surfaces 9&96 for loose and dense sand 128129 frequently used criteria 971 04 open and closed yield surfaces 1 14 ZarembaJaumann stress changes 20 .see crlso incompressible behaviour Uniaxial behaviour 8687 Unloading criteria 9092.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue listening from where you left off, or restart the preview.