INTERNATIONALJOURNALFOR NUMERICAL AND ANALYTICAL METHODS IN GEOMECHANICS. VOL.

15, 301-3 15 ( 1991)

INTERFACE ELEMENTS FOR SINGULAR PLASTICITY POINTS

H. VAN LANGEN AND P.A. VERMEER

Department of Civil Engineering, Dewt Universiiy of Technology, PO Box 5048, Dew[, The Netherlandr

SUMMARY Zero-thickness interface elements are applied to solve plasticity problems involving singular points where stresses and displacements are non-unique. Singular points occur in soil-structure interaction, e.g. at the ridges of footings, buried structures and foundation piles. Here the introduction of potential displacement discontinuitiesappears to give quite accurate finite element solutions. This is demonstrated for indentation problems both for plane strain and axisymmetric situations. In all problems the soil behaviour is modelled by the elastic-plastic Mohr-Coulomb model.

INTRODUCTION In the analysis of soil-structure interaction the development of displacement singularities and zones of high distortion requires particular attention. Special finite elements can be used to model these zones. The first use of such interface elements or joint e1ernentsl-l was made in the analysis of rock masses with predefined joints. Later, interface elements were also applied to the problem of soil-structure interaction.'-'' Independently from this, the analysis of concrete fracture led to the development of lumped interface elements in which the contacting faces are coupled by nodeto-node spring boxes. Roughly speaking, two approaches can be found in the literature. One approach is to treat the interface as a direct problem of compatibility.'-'' Compatibility requirements are enforced exactly, e.g. by means of Lagrangian multipliers, or approximately by means of a penalty formulation. The second approach'-4* "-'' uses the physical concept of an interface element. Here a constitutive equation relates contact stresses to slip and compression. From a mathematical point of view this approach and the penalty approach coincide. A detailed description of interface formulations is given by Hohberg." This paper concentrates on a very special use of interface elements, namely the modelling of singular points in plasticity problems as discussed in the next section. After that the formulation of a zero-thickness interface element is given. A punch problem is solved numerically to show the effect of singular plasticity points. It is shown that the introduction of a potential slip line into the finite element mesh improves the computational results considerably. A section on the trapdoor problem touches on possible mesh dependence. Finally, a section is devoted to the analysis of a pile penetration problem. For such an axisymmetric problem the use of potential slip lines is shown to be essential.

'

SINGULAR PLASTICITY POINTS Singular plasticity points are best illustrated by considering problems involving perfectly plastic material. Consider for example the indentation of a rigid die into a purely cohesive material that

0363-9061 1 91/050301-15W7.50 0 1991 by John Wiley & Sons, Ltd.

Received 1 June 1990 Revised 8 October 1990

302

H.VAN LANGEN AND P.A. VERMEER

obeys the Tresca yield criterion. Here the limit state of stress and deformation can be solved exactly by using the method of slip line fields. This problem was first investigated by Prandtl'' and later by Hill". Figure I(a) shows Hill's field of slip lines for a smooth die. The velocity pattern is fully symmetric. The material between OPR slides downwards along OP and the triangle RQS slides upwards along QS. The material in the centre sectors PQR does not move as a rigid block. There is a singularity at R, the edge of the die. The singular velocity at R is plotted in Figure l(b); it is a fan radiating in different directions. Such a singularity cannot be modelled by a conventional finite element computation which involves regular displacements for all material points. Despite this shortcoming of the conventional finite element method, the Prandtl-Hill solution can be captured reasonably well by using a fine mesh near the edges of the punch. Indeed, for this particular problem numerical collapse loads are within a few per cent of the exact limit load of (2 + n)c, where c is the cohesive strength. The effects of singular plasticity points become more severe when considering buried structures in a soil mass or piercing problems in metal plasticity. A well-known problem of deep penetration is shown in Figure 2. Here the plane strain problem of piercing by a flat punch into cohesive material is analytically solved by Hi1119. The difference from the previous problem of surface penetration is not only the embedment but also the roughness of the flat punch. Since a rough

(a)

(b)

Figure 1. Theoretical slip line solution for a smooth die: (a) slip line field (b) singular velocity at comer point

I a1

I bl

Figure 2. Theoreticalslip line solution for a rough rigid punch penetrating into cohesive soil (a) slip line field; (b) singular velocity at corner point

INTERFACE ELEMENTS FOR SINGULAR PLASTICITY POINTS

303

punch is considered, the material below the punch does not slip to the sides but only moves down. In contrast, the material along the shaft moves upwards. Hence an extremely strong displacement discontinuity occurs at the edges, as indicated in Figure qb). In a later section it will be shown that strong discontinuities like this tend to reduce the accuracy of finite element solutions. Here the numerical collapse load is much too high. Even a fine mesh is constrained too much, resulting in high stress concentrations at the edges. To overcome this effect of singular plasticity points, it would seem that we need to incorporate some potential slip lines into the finite element mesh. Provision for such slip lines is easily incorporated using interface elements in between two common finite elements. It should be emphasized that such potential slip lines are not zones of weakness since they have the same strength properties as the rest of the material. Internal interface elements were used before by Simmons in a paper on the finite element analysis of an excavation problem. Simmons used internal interfaces to obtain information on the elastic stress distribution near the comer at the bottom of a trench. FORMULATION OF AN INTERFACE ELEMENT In this section a brief description is given of the interface formulation. This description involves the virtual work equation, the constitutive relation for the interface and the finite element discretization. An interface may be seen as a displacement discontinuity surface in a continuum. Koiter gives the virtual work equation for a continuum with such a discontinuity:

where u is the Cauchy stress tensor, 6u and 68 are virtual displacements and strains, t, are tractions transmitted by the interface, y and t, are body forces and surface tractions and 6 {Au} is the virtual displacement discontinuity. The volume Vand surface S, extend over both parts of the continuum which are separated by the discontinuity. Note that we are considering a real discontinuity, i.e. a zero-thickness interface. In the earliest concept of an interface, slipping will occur when the interface traction t, reaches some threshold value. Below this value no relative movement occurs and, as a result, the interface term in equation (1) will disappear. This behaviour can be directly implemented in a finite element formulation using the Lagrangian multiplier method. In fact, the Lagrangian multipliers are the unknown nodal contact forces. Another approach to the interface problem is the formulation of an interface element. Instead of prescribing zero relative displacements up to slipping or debonding, any relationship between the interface stresses and the relative displacements can be introduced. Such interface elements and constitutive laws can be implemented straightforwardly in the stiffness formulation of a finite element code. In the literature on interfaces several authors have already presented interface elements, both of the thin-continuum type and of the zero-thickness type. Figure 3(a) shows a 10noded zero-thickness element. This particular element was chosen for its compatibility with 15noded triangular elements. In this study we assume an elastoplastic relationship which combines linear elasticity and perfect plasticity. In this model the rates of interface traction, i,, and displacement discontinuity, A i , are related by

i, = DAU

=D(Ai

- Aip)

(2)

304

H.VAN LANGEN AND P.A. VERMEER

soil

'1

\
Figure 3. (a) Geometry and (b) yield locus for a 10-noded interface element

where superscripts are used to distinguish between elastic and plastic parts. For the elasticity matrix D in equation (2) the following simple expression is chosen:

Note that the elastoplastic stiffness matrix, which is derived from further development of equation (2), will generally have off-diagonal terms. For computations in the fully plastic range the definition of the D-matrix is basically not very important. Here the elastic slip becomes negligibly small and, as a result, i, vanishes. Interface stiffnesses k, and kn should be chosen such that the initial slopes of load4isplacement curves closely resemble those which are obtained without the use of interfaces. In this way the influence of interfaces is limited to the case of true plastic slip, where it enhances the flexibility of the model. Upper bounds to the values of k, and k, are determined by possible ill-conditioning of the stiffness matrix. Hence suitable values are difficult to choose in advance and, as a result, the elastic solution should be checked with the above considerations in mind. Presently we are relating interface stiffnesses to the mean element length I, the shear modulus of the soil, G, and Poisson's ration v in the following way:

For the present calculations we used p = 50. In the present study we consider only a simple Mohr-Coulomb yield function with a nonassociated plastic potential as depicted in Figure 3(b):
f(t) = t,

+ t,tan4,

- c,,

g(t) = t,

+ t , tan JI,

(5)

The angle of friction, 4,, and the adhesion c, govern the strength. The dilation angle JI, controls the plastic dilation. The use of a non-associated flow rule with JI, < &c prevents overprediction of the plastic dilation. In confined problems this would yield an overestimation of contact pressures and thus of the shear strength. The rate of plastic slip is derived from the plastic potential function g:
AUP

= EX (dg/at)

(6)

INTERFACE ELEMENTS FOR SINGULAR PLASTICITY POINTS

305

where the switch-on/switch-off coefficient a is defined as

a =0

if f c 0 or [df/dt]'DAU

c0

(74

a = 1 if f = 0 with [dfldt]* DAU 2 0 (7b) The multiplierX can now be solved from the consistency conditionf = 0. Equations (6) and (7) can also be used for finite increments of stress and relative displacements. Here we will not discuss the methods for solving the non-linear virtual work equation, since a detailed description can be found in the l i t e r a t ~ r e . All ~ ~ the -~~ solutions presented in this paper were obtained by using an initial stress method for the equilibrium iterations.zs A Newton-Cotes schemeZ3is used for the numerical integration of the interface stiffness matrix and related vectors. The position of the five integration points coincides with that of the node ' ~ HohbergZ6 show that interface stresses can be improved if such a scheme pairs. Gens et ~ 1 . and is employed. Essentially this integration yields a lumped interface stiffness matrix, which suppresses possible stress oscillations. For further information the reader is referred to References 15 and 26. THE PENETRATION O F A FLAT RIGID PUNCH IN COHESIVE MATERIAL In this section the quality of interface stresses will be analysed for the problem shown in Figure 4. In the context of soil mechanics this problem might represent a model test for a punch penetrating into clay. The constitutive behaviour of the material is elastic-perfectly plastic according to the well-known Tresca model. The material parameters are as follows: shear modulus Poisson's ratio shear strength interface stiffness adhesion
G = lo00 kPa v = 0.49 c = 10 kPa k, = 50 k, = 2500MNn1-~ c, = 10 kPa.

Note that the base of the punch is fully rough whereas the side is fully smooth.
vertical boundaries

Tresca Material

0-2d

Figure 4. Rough rigid punch jmetrating into cohesive material

306

H.V A N LANGEN AND P. A. VERMEER

The theoretical solution given by Hill for the limit load is based on the slip line field on the lefthand side of Figure 5. The theoretical contact pressure at the base of the punch depends on the ratio d / D , where d is the width of the punch and D is the size of the container. We will consider the special ratio d / D = 0-5 since it gives precisely a constant limiting contact pressure of p = (n + 2)c. Figure q a ) shows the finite element schematization of the punch problem. The 15-noded triangular elements involve a fourth-order approximation of displacements. This type of element was first advocated by Sloan and Randolph for its capability of yielding accurate results in cases of nearly incompressible axisymmetric deformation. In the case of plane strain conditions this element tends to yield fairly smooth stress distributions.28 Along the base of the punch two 10-noded interface elements were added to enable representation of contact pressures. The loaddisplacement curve as obtained by the non-linear finite element calculation is shown in Figure 7. A true limit load is obtained. This load is 14% in excess of the exact solution. The computed stress at the base of the punch are shown in Figure 8(a). The quality of the computed stresses is quite poor; according to the theoretical solution, there should be no stress concentrations near the edges. The finite element calculation was repeated for a mesh with potential discontinuities as indicated in Figure qb). The load-displacement curve (Figure 7), shows that this schematization indeed computes a better collapse load. Now the error is only 2% as compared to the exact solution. The quality of interface stresses (Figure 8(b)) is also improved substantially; stress concentrations have virtually disappeared. Closer inspection of the results shows that at collapse the shear stress in the internal interface element is equal to the shear strength; it thus represents a true slip line. Figure 5 shows the velocity field at collapse for the improved schematization. At the edge of the punch the computed velocity actually becomes discontinuous as in the theoretical solution.

/
/

/ / I I
I

t
. . . .

I
/ /
/

I
I
/

I
I

I
/ I
I I

--.--.4cc/.

D=2d
Figure 5. Velocity field at failure for a rough rigid punch

INTERFACE ELEMENTS FOR SINGULAR PLASTICITY POINTS

307

Ia )

Ib )

Figure 6. Finite element mesh for a punch indentation problem: (a) element subdivision; (b) potential slip plane

60

Exact Solution IT1.2 I

Simple Mesh

597

50-

t
. n

LO-

30w

20-

10-

0
0

25

SO

15
Punch ~ndentrtion Imnl

10 0

12 5

no

Figure 7. Load-displaccmentcurves for a rough rigid punch penetrating into cohesive soil

Figure 8. Computed contact pressure under a rigid punch (a) simple mesh; (b) improved mesh with singular velocity point

308

H.VAN LANGEN AND P.A. VERMEER

A SOLUTION TO THE TRAPDOOR PROBLEM

The plane strain problem of a passive trapdoor (Figure 9) has been analysed by several researchers.28.29In the context of this paper the treatment of the boundary discontinuity at the edge of the trapdoor is of interest. In previous analyses this discontinuity was 'smoothed out' by taking a linear variation of prescribed displacements in a small transition zone (Figure lqa)). In the present analysis the mesh contains an interface element to anticipate the singular velocity (Figure lqb)). The influence of this singularity is now-as in reality-concentrated in a very thin zone. The strength properties of the interface element coincide with those of the soil. The plasticity model for the soil is based on the Mohr-Coulomb yield criterion. The yield function consists of six intersecting planes in principal stress space. The plastic potential function, from which plastic strains are derived, has a similar shape and, as a consequence, integration of the stress-strain relation involves the treatment of vertices for the direction of plastic flow.A detailed formulation and discussion of this fully angular model is given by Smith and GriffithsZ2 Here only an outline of the plasticity model is given by omitting the treatment of vertices. For convenience we consider only a single regular yield function and a single plastic potential function, namely f= ) l o , - cr31 t(o,+ o,)sin$ - ccos$ < 0

g=

410,

- 031+ +(ol+ a3)sin$

so

I \ I \ tt11rt

J, = 0"

,$,
D=2m

Figure 9. Finite element mesh and velocity field at failure for a trapdoor problem

fa1

lb)

Figure 10. Solution to the trapdoor discontinuity: (a) conventional approach; (b) singular point approach

INTERFACE ELEMENTS FOR SINGULAR PLASTICITY POINTS

309

Here 4 denotes the friction angle of the material and c the cohesive strength. The plastic potential involves the dilation angle JI. In this example the soil and interface properties are assumed to be as follows:

k , = 2 5 k , = 2500MNm-3

4, = fp = 30"
c, = c = 0 kPa

*, *
=

= 0".

Figure 11 shows the load4isplacement curve for the uplift of the trapdoor. Now it might be argued that an interface in the interior of the mesh will erroneously attract the failure mechanism. To weaken this argument, a second analysis was made for an associated material with a dilatancy angle of JIc = JI = 30". The velocity field at collapse of this computation is shown in Figure 12. In this case the shear band has an inclination with respect to the vertical axis owing to the dilation of the material. It is clear that in this case the interface element only creates a velocity discontinuity at the boundary of the mesh. The shear band that originates at this point does not follow the direction of the interface. ANALYSIS OF A PILE PENETRATION PROBLEM IN COHESIVE SOIL The empirical formulae for the bearing capacity of piles make a distinction between a tip contribution and a shaft contribution to the limit load. In this section we will concentrate on the end bearing for a pile with a flat tip placed in a prebored hole (Figure 13). The pile is assumed to have been installed in a homogeneous clay layer. The properties are chosen in such a way that

co

4
c

J, i30"
I

"

I0/

10

20

30
Uplift Trapdoor tar1

40

50

Figure 1 1 . Loaddisplacement curves for a trapdoor problem

3 10

H.VAN LANGEN AND P.A. VERMEER

Figure 12. Velocity field at failure for a trapdoor problem in frictional material with high dilation

they correspond to undrained stiff clay behaviour: ratio of shear modulus over cohesion Poisson's ratio
G/c = 100
v = 0-49.

Since friction is absent, the Coulomb yield criterion degenerates to the Tresca criterion. The contact between pile and soil is assumed to be completely rough, which means that the adhesion is equal to the shear strength of the soil. The finite element schematization of this problem is shown in Figure 13. Again a mesh of 15noded elements is used, supplemented by 10-noded interface elements along the tip and shaft of the pile. The latter interfaces are merely used to represent contact pressures. In a first schematization of the problem no special measures are taken to anticipate the discontinuous displacement at the edge of the tip. In a second schematization, however, the technique of adding internal interfaces, as developed in the previous sections, is employed. In this way two potential slip planes are created at the edge of the tip (Figure 14). Actually the singular point C in this figure is now split into four points. Starting at the prescribed vertical displacement of C , , point C2 slides horizontally away from C, . Relative to this, point C3 slides vertically. The net displacement of point C , is upwards, allowing for slip along the shaft. Bearing factors for the simple and the improved mesh are compared in Figure 15. The bearing factor N , is the ratio of the average contact pressure over the shear strength. Empirical factors reported in the literature3' range between 10 and 20. Both analyses fall within this range. However, the simple mesh yields a 25% higher limit load as compared to the improved mesh. This difference is even more dramatic than in the analysis of the punch problem, where a 12% difference was obtained. Note that open hole computations3' yield values of the order of N , = 9. However, in such analyses the soil at the shaft is free to move inwards. In the simple schematization of the Hill punch problem the overshoot of the limit load was caused by a stress concentration near the edge of the punch. The same proves to be the case in the present analysis, as can be seen from Figure lqa). Again the simple mesh shows a stress concentration near the edge of the tip. For axisymmetric configurations this contributes even

INTERFACE ELEMENTS FOR SINGULAR PLASTICITY POINTS

311

Figure 13. Finite element mesh for a pile penetration problem

more to the limit load than in the case of plane strain, since the area increases with radius. In addition to this, the pressure on the lower part of the shaft obtained with this simple mesh shows a non-physical raggedness. The improved mesh, on the other hand, is found to yield much better results, as indicated in Figure lqb). The tip pressure distribution is very smooth and does not culminate in a peak value near the edge. The same smoothness characterizes the shaft pressure. Just as in the Hill problem, the simple schematization for the pile problem suffers from a lack of flexibility near the edge of the tip. This can also be demonstrated by looking at the contour plots of the velocity intensity at failure (Figure 17). This intensity is defined as
0=J

( u ~ + v;)/Vpilc

(10)

where 0, and u, denote the Cartesian components of the velocity at a material point. For both meshes the soil movement appears to be concentrated near the tip of the pile. The only, albeit significant, difference between the two sets of contour lines is the soil velocity field at the lower

312

H. V A N LANGEN AND P.A. VERMEER

la1

Ibl

Figure 14. Introduction of potential slip planes in the finite element mesh for a pile

20

Simple Mesh

187

15
0 1 ,

Improved Mesh 150

=
0 c u LL

10

0
0
100 200

300
Normalized Displacement 6 =

600

DC

Figure 15. Tip reaction curve for a penetrating pile with a Rat tip

part of the shaft. In the result for the simple schematization (Figure 17(a)) a concentration of contours is obtained, indicating a zone with a high velocity gradient. In this region the soil elements distort without any slip at the shaft, even though interface elements were used. Apparently the boundary condition at the corner point prevents direct slipping. As explained before, splitting the corner into four separate points allows for slip along the shaft in the improved schematization. As a result the contour plot for this case (Figure 17(b)) does not show a high velocity gradient in the soil elements near the shaft. Now the mesh offers enough flexibility to cope with the corner singularity. Some researchers28s32 performed similar small-strain computations for the bearing factor. A striking difference between the present analysis and those in the literature is the normalized displacement u at which collapse is assumed. This displacement is defined as

INTERFACE ELEMENTS FOR SINGULAR PLASTICITY POINTS

313

0.5 D

0.5 D

i
0
v\

10 C"

Figure 16. Contact pressure along tip and shaft for a penetration problem at failure: (a) simple mesh; (b) improved mesh with singular point treatment

where 6 is the tip displacement, D is the pile diameter, G is the shear modulus and c is the cohesion. A typical value in the literature is of the order of u = 25, whereas collapse occurs here at a value of approximately u = 300. Apparently collapse is assumed when the tangent to the tip reaction curve becomes very small as compared to the initial one. Indeed, at a value of u = 30 the present analysis already yields a slope reduction of 98%. However, from the tip reaction curves it may be seen that at this point the tip factor is still far from its ultimate value. In fact, when collapse is assumed at this point, a bearing factor of about nine is obtained.

CONCLUSIONS
A zero-thickness interface element has been developed which can be used to solve problems of soil-structure interaction. Results are presented for a problem of plane plastic flow, analytically solved by Hill. The predicted limit load and stress distribution improve substantially if interface elements are used to create a boundary discontinuity at failure.

314

H. VAN LANGEN A N D P.A. VERMEER

v =1.0
a5

->

O5

(a)

(b)

Figure 17. Velocity intensity contours for a pile at failure: (a) simple mesh; (b) improved mesh with singular point treatment

The technique of introducing an interface in the interior of the mesh is used in the solution of the trapdoor problem. An accurate result for the limit load is now obtained with a coarse finite element mesh. The interface in the interior of the mesh does not cause mesh dependence with respect to the failure mechanism. The end bearing capacity of a pile can be predicted with a relatively coarse mesh if interfaces are added along the shaft and the tip. The quality of stresses on the tip and on the lower part of the shaft is improved by introducing additional interfaces to allow for singular plasticity points. In a conventional approach, without additional interfaces, high concentrations of stress are predicted near such points. For axisymmetric situations this would yield a severe overprediction of the limit load.
ACKNOWLEDGEMENT

The authors wish to express their gratitude to Mr. R.B. J. Brinkgreve for his help in implementing the interface element in the PLAXIS finite element program.

INTERFACE ELEMENTS FOR SINGULAR PLASTICITY POINTS

315

REFERENCES
1. R. E. Goodman, R. L. Taylor and T. L. Brekke, 'A model for the mechanics of jointed rock, J . Soil Mech. Found. Diu. A S C E , 94, (SM3). 6 3 7 4 5 9 (1968). 2. R. E. Goodman and J. Dubois, 'Duplication of dilatancy in analysis ofjointed rocks', J. Soil Mech. Found. Diu. ASCE,

98, (SM4). 3 9 9 4 2 2 (1972). 3. 0. C. Zienkiewicz, B. Best, C. Dullage and K. C. Stag& 'Analysis of non-linear problems in rock mechanics with particular reference to jointed rock systems', P roc. 2nd Congr. ofthe International Society /or Rock Mechanics Vol. 3, Beograd. 1970, pp. 501-509. 4. J. Ghaboussi, E. L. Wilson and J. Isenberg, 'Finite element lor rock joints and interfaces', J. Soil Mech. Found. Dio. A S C E , 99, (SM10). 833-848 (1973). 5. P. Nath. 'The analysis of composite structures with prescribed frictional conditions at the interfaces', Int. j . numer. anal. methods geomech., I. 387-396 (1977). 6. N. Okamoto and M. Nakazawa, 'Finite element incremental contact analysis with various frictional conditions', Inr. j . numer. methods eng.. 14, 337-357 (1979). 7. M. Boulon and P. Foray, 'Physical and numerical simulation of lateral shaft friction along offshore piles in sand', Proc. 3rd Int. Con& on Numerical Methods in OflThore Piling, Nantes, 1986, pp. 127-147. 8. N. Chandrasekaran, W. E. Haisler and R. E. Goforth, 'A finite element solution method for contact problems with friction', Int. j . numer. methods eng., 24, 477-495 (1987). 9. P. R. Heyliger and J. N. Reddy, 'A mixed computational algorithm for plane elastic contact problems- 1. Formulation', Comput. Struct. 26, 6 2 1 4 3 4 (1987). 10. M. G. Katona, 'A simple contact-friction interface element with application to buried culverts', Int. j . numer. anal. methods geomech., 7, 371-384 (1983). 1 1 . G. N. Pande and K. G. Sharma, 'On jointiinterface elements and associated problems of numerical ill-conditioning', Inr. j . numer. anal. methods geomech., 3, 293- 300 (1979). 12. F. E. Heuzk and T. G. Barbour, 'New models for rock joints and interfaces'. J. Geotech. Eng. Diu. ASCE, 108, (GTS), 757-776 (1982). 13. C. S. Desai, M. M.Zaman, J. G. Lighter and H. J. Siriwardana. 'Thin-layer element for interfaces and joints', Inr. j . numer. anal. methods geomech.. 8. 19-43 (1984). 14. C. S. Desai and S. Sargand, 'Hybrid FE procedure for soil-structure interaction', J. Geotech. Eng. Diu. ASCE. 110, 473-486 (1984). 15. A. Gens, 1. Carol and E. E. Alonso, 'An interface element formulation for the analysis of soil reinforcement interaction', Comput. Georech., 7, 133-151 (1988). 16. D. Ngo and A. C. Scordelis, 'Finite element analysis of reinforced concrete beams', J. Am. Concrere Inst., 64, (14). 152-163 (1967). 17. J. Hohberg. 'Trennflachenformulierungen fur die statische und dynamische Berechnung von Bogenstaumauern', Bericht Nr. 163, Institut fur Baustatik und Konstruktion, ETH Zurich, 1988. 18. L. Prandtl, *Uber die HHrte plastischer KBrper', N a r h r . Ges. Wiss., M a t h . Phys. Kl., GGttingen, 74. 19. R. Hill, The Mathematical Theory o f Plasriciry, Clarendon, Oxford, 1950. 20. J. V. Simmons, 'Assessment of the numerical performance ol quadratic isoparametric joint elements'. Proc. 5th Int. C o n j on Numerical Methods in Geomechanics, Nagoya, 1985. pp. 199-206. 21. W. T. Koiter, 'General theorems for elastic-plastic solids', in 1. N. Sneddon and R. Hill (eds), Progress in Solid Mechanics, Vol. 1. North-Holland, Amsterdam, 1960, pp. 165-221. 22. 1. M. Smith and D. V. Griffiths, Programming the Finite Element Merhod, 2nd edn, Wiley, Chichester and New York, 1988. 23. K. J. Bathe. Finite Element Procedures in Engineering Analysis, Prentice-Hall, Englewood Cliffs, NJ, 1982. 24. H. van Langen and P. A. Vermeer, 'Automatic step size correction for non-associated plasticity problems', Inr. j. numer. methods eng., 29, 579-598 (1990). 25. P. A. Vermeer and H.van Langen, 'Soil collapse computations with finite elements', Ingenieur-Archiu, 59, 221-236 (1989). 26. J. Hohberg. 'A note on spurious oscillations in FEM joint clements', Earthq. Eng. Struct. Dyn., 19, 773-779 (1990). 27. S. W. Sloan and W. F. Randolph, 'Numerical prediction of collapse loads using finite element methods', Inr. j . numer. anal. methods geomech., 6.47-76 (1982). 28. R. de Borst and P. A. Vermeer. 'Possibilities and limitations of finite elements for limit analysis', Geotechnique. 3 4 ,(2). 199-210 (1984). 29. N. C. Koutsdbeloulis and D. V.Griffiths, 'Numerical modelling of the trap door problem', Geotechnique. 39, (I), 77-89 ( 1989). 30. J. de Ruiter. 'The static cone penetration test: state-of-the-art-report', Proc. E S O P T I / , Vol. 2, Amsterdam, 1982, pp. 38945. 31. D. V. Griffiths, 'Elast~plastic analyses of deep foundations in cohesive soil', Int. j . numer. anal. methods geomech., 6, 211-219 (1982). 32. S. M. Willson, 'Finite element analysis of piles and penetrometers', Dissertation. University of Manchester, 1985.