0 1995 Elsevier Science Limited Printed in Great Britain. A11 rights reserved 0266-352X/95/39.50 BEHAVIOUR OF GEOSYNTHETIC REINFORCED SOIL RETAINING WALLS USING THE FINITE ELEMENT METHOD Rajagopal KARPURAPU Department of Civil Engineering, Indian Institute of Technology, Madras, India 600036 Richard J. BATHURST Department of Civil Engineering, Royal Military College of Canada, Kingston, Ontario, Canada K7K 5LO ABSTRACT The Paper describes finite element models that are used to simulate the behaviour of two carefully constructed and monitored large-scale geosynthetic reinforced soil retaining walls. The walls were constructed using a dense sand fill and layers of extensible polymeric (geosyn- thetic) reinforcement attached to two very different facing treatments. The model walls were taken to collapse using a series of uniform surcharge loads applied at the sand fill surface. The Paper demonstrates that correct modelling of the dilatant behaviour of the sand soil is required to give accurate predictions of wall performance. A modified form of hyperbolic constitutive model that includes a dilation parameter is adopted to model the behaviour of the granular soil. Mechanical properties of the constituent components of the large-scale physical models are established using standard laboratory tests including constant load tests on the polymeric reinforcement from which isochronous load-strain-time data is developed. The results of anal- yses show that the finite element model, constitutive models and implementation reported in this study can accurately predict all important features of wall performance. INTRODUCTION The challenge in numerical simulation of geosynthetic reinforced soil wall performance is to quantitatively predict all features of these composite structures using only the results of stan- dard laboratory testing carried out on component materials. The challenge is compounded by the problem of verification due to a general lack of high quality physical data that allows the accuracy of finite element models to be tested against a wide range of measured response. This Paper provides details of the finite element techniques and constitutive models used to simu- late the measured response of two carefully constructed and monitored full-scale geosynthetic reinforced soil walls constructed at the Royal Military College of Canada (RMC). The com- pleted physical models were nominally identical in final appearance and function but differed significantly in facing treatment and construction sequence. Hence, results of these physical models provide a useful database against which to test the accuracy of numerical simulation techniques for candidate wall structures that may be built using a variety of construction tech- niques. 279 280 The most common methods of analysis for geosynthetic reinforced soil wall structures are based on limit-equilibrium methods (e.g. [l-3]). While these methods offer simplicity they are limited in their ability to predict stresses, forces and boundary reactions at working load levels and offer no information on deformations and strains in the structure components. Further- more, current methods of design and analysis for the types of wails described in this Paper can- not distinguish between walls built with different facing treatments and have been proven to be conservative (i.e. excessively safe) [3-41. Composite finite element methods (e.g. [S-S]) h ave the advantage of computational econo- my but cannot provide the detail that is required to isolate mechanisms acting between and within component materials. To date, most finite element analyses of reinforced soil structures have been based on discrete finite element methods (e.g. [9-U]). In the current investigation, a discrete finite element approach is adopted in order to explore mechanisms such as load transfer between reinforcement layers and soil fill. EXPERIMENTAL PROGRAM Two reinforced soil retaining walls were constructed within a test facility 2.4m wide x 6m long x 4m high. The completed structures were 3m high and were constructed with three col- umns of panels composed of two outer sections 0.7m wide and one central instrumented sec- tion 1 m wide. A dense granular sand was used as the backfill. Friction between the contained soil and the vertical sidewalls of the test facility were minimized by using a composite plywood/ Plexiglass/polyethylene sheeting arrangement. The combination of reduced frictional shear- ing resistance at the test facility boundaries and a decoupled central instrumented wall section resulted in a physical model that was close to the idealized plane strain condition assumed in numerical simulations, A complete discussion of edge effects in the RMC retaining wall test facility can be found in the Papers by Bathurst and Benjamin [16] and Bathurst [4] . One physical test was an incremental panel wall structure comprising four rows of panels each 0.75m high that were individually supported during placement and compaction of soil located directly behind the panel [4, 211. Once backfill operations were completed behind a row of panels the external supports were removed and moved to the next row of panels. Layers of soft foam rubber were placed between the stacked panels to providevertical compressibility to the facing system. In the second physical test the panel units were bolted together to create three columns simulating a full-height panel construction [22]. These panels units were sup- ported externally for the duration of fill placement. The supports were removed once fill place- ment was complete. Both walls were reinforced with four layers of a biaxial geogrid oriented in the weak direction and extending 3m into the backfill. The different construction techniques and wall facing type can be anticipated to result in qualitatively and quantitatively different wall response. The use of incremental panel wall construction means that outward wall deformations are developed as construction proceeds 281 from base to wall crest. Hence, tensile loads in the reinforcement are generated during the construction phase. In addition, vertical deformation of the facing column is permissible due to the compressibility of the joint material between panel rows. In contrast, the full-height pan- el structure that is braced externally for the duration of construction prevents mobilization of the reinforcement tensile capacity until the full height of fill is placed and the props removed. In addition, the monolithic panel construction and pinned toe connection results in constrained wall deformations. Despite obvious differences in wall construction it is interest- ing to note that current limit equilibrium-based analysis methods cannot distinguish between the generic forms of construction just described. Following construction, the walls were subjected to a series of surcharge pressure incre- ments by inflating air bags which were confined between the backfill soil and top of the test facility. Each pressure increment was maintained at a constant magnitude for at least 100 hours to measure time-dependent deformations in the wall structures. A schematic of the wall and the instrumentation that was used as part of the monitoring pro- gram are shown in Figure 1. Approximately 300 electronic instruments were installed in each wall [23]. CONSTITUTIVE MODELS FOR COMPONENT MATERIAU A number of different constitutive models are required to represent the mechanical beha- viour of the backfill soil, polymeric reinforcement, wall facing units, and the interfaces be- tween components [9-111. displacement Dotentiometer surcharae T load 3m 0.5m T E ring - D load cell f-i *-_$(- - reinfircement Layer 2 1 ,,_,___.~p I Din FIGURE 1 Cross-section view of incremental panel wall test 282 The properties of sheet polymeric reinforcement under tensile load are strongly dependent on the rate of loading, duration of load application and temperature [24,25]. Matsui and San [13] have explicitly included time as a parameter in a model that can simulate the reinforce- ment stiffness as a function of both strain and time. If the creep effects of soil are not significant (e.g. dense well-graded granular soils), then isochronous stiffness data can be used directly to model the time dependency of geosynthetic reinforcement in soil retaining wall structures [9-111. The strength and stiffness properties of granular soils have been modelled using simple lin- ear elastic models, no-tension models, and more rigorous elastic-plastic models [5-15,26,27J. The major disadvantagewith advanced constitutive models for soils is that the models are for- mulated in terms of functions containing parameters whose numerical values cannot be easily determined using routine laboratory tests. In the current investigation a modified form of hy- perbolic constitutive model was developed and was shown to capture stiffness, strength, and plasticity features of granular soil behaviour with a minimum number of parameters. Further- more, these parameters can be estimated from routine laboratory testing. The modification to hyperbolic models of the type originally proposed by Duncan et al. [28] is a dilatancy param- eter that is required to accurately simulate laboratory shear test data and the behaviour of large-scale retaining walls constructed with the same granular soil. Pullout tests performed on geogrid reinforcement and soil materials similar to those used in the current study clearly dem- onstrate that dilatancy of soil in contact with the reinforcement occurs during shear transfer. This soil dilatancy is responsible in part for the bond capacity that can develop in the anchorage zone [29]. The interfaces between various components in reinforced soil structures have been simu- lated using stick-slip type models and hyperbolic models where independent parameters in- clude normal pressure acting at the interface [9-151. The interface strength properties in these earlier investigations have been obtained from direct shear and pullout tests performed with candidate reinforcement and backfill soils tested over the same range of normal pressures and at the same soil densities as in the reinforced structures. FINITE ELEMENT SIMULATION The numerical simulation work reported in this Paper was performed using a finite element program and ancillary utilities (GEOFEM) developed by the authors at the Royal Military College of Canada [30]. The package contains a general purpose program that includes some special nonlinear constitutive models developed for analysis of soil-polymeric reinforcement interaction. The finite element mesh used for the numerical simulations is shown in Figure 2. The mesh is made up of quadrilateral and triangular continuum elements, interface elements, uniaxial elements, and nodal link elements to represent components of the reinforced soil wall. Figure 283 temporary external props during construction anel-soil interface I 3m 1 OSm T FIGURE 2 Finite element mesh for full-height panel wall test 3 shows mesh details at the reinforcement-panel connections and the nodes associated with each element at these junctions. This arrangement of elements was adopted after testing sev- eral trial meshes for numerical accuracy and was also found to be efficient for use with an auto- matic mesh generation utility that is part of the GEOPEM suite of programs. The finite ele- ment mesh for each wall consists of approximately 1700 nodal points, 650 elements, and 3300 degrees of freedom. The solution scheme employed in the computer code updates the stiffness matrix at every iteration. Large deformation effects are accounted for in numerical simulations by using the linearised updated Lagrangian method. In this method, the coordinates of nodes are updated by adding the corresponding displacements of nodes at every load step. Without this scheme it is not possible to model important effects such as the additional tensile resistance due to the membrane action of the reinforcement as it deforms to a concave shape immediately behind the full-height panel wall columns. This effect in physical models is the result of relative down- ward movement of the soil immediately behind full-height retaining wall facing units. The stiffness matrix is modified by the nonlinear stress correction terms as shown in Equa- tion 1: r 1 BE D B, dv + I BE oj B, dV I {du)i = { P, ] i - [BIT {U } j (1) V V 284 L zero thickness Element 2 3 4 5 6 7 6 9 10 Type Nodes quadrilateral 9-l -3-l l-6-2-7-10 quadrilateral 11-3-5-13-7-4-8-12 interface 16-14-9-11-15-10 interface 20-18-11-13-19-12 quadrilateral 26-14-16-28-21-15-22-27 interface 17-29-28- 16-23-22 uniaxial bar 17-29-23 interface 18-30-29-17-24-23 quadrilateral 30-18-20-32-24-19-25-31 nodal link 11-17 FIGURE 3 Details of finite element mesh at panel-geosynthetic connections Here, BL and BN are linear and non-linear strain-displacement relations as discussed by Bathe et al. [31], D is the constitutive matrix, and i is the current load step number. oj corresponds to the stresses at the previous load step (i-l) for the first iteration at a load step. For subsequent iterations within a load step, oj corresponds to the stresses at the previous iteration. The load vector in GEOFEM is formulated as the difference between the external loads and the internal forces computed from the element stresses in the previous iteration as shown on the right hand side of Equation 1. This formulation ensures that any out-of-balance force is carried forward during the analysis thus satisfying the equilibrium of the total system at all stages of analysis. The pin connection at the base of the panels was modelled using a six-noded inverted trian- gular element as shown in Figure 4. The stiff foam layers separating the individual facing panel units in the incremental wall were modelled using solid elements with a bulk modulus deter- mined from physical tests. This arrangement allowed for independent movement of panel units in the numerical simulation. The stiffness of the facing panel in bending and compression was determined from physical testing prior to construction. 285 pin +-- stiffneSSpin +f-- a) full-height panel wall b) incremental panel wall FIGURE 4 Finite element models for wall facing panels The incremental construction technique was simulated by placing rows of elements in se- quence and gradually turning on gravity-induced body forces over several load steps (typically ten for each layer and K, = 0.5). The external props used during wall construction were simu- lated using springs with a linear axial stiffness value determined from measured prop forces and wall displacements recorded during construction. The surcharge pressure on the wall was applied in increments of 0.25 kPa per load step. The solution was iterated until the out-of-bal- ance force norm Nf, defined in terms of out-of-balance forces Sfi and the applied forces fi (Equation 2) was less than 0.5%. J 1 Sfi 6f, N, = 5 fi fi N (2) Eight-noded quadrilateral elements were used to model the backfill soil in the wall. The stiffness matrix and other element matrices corresponding to the soil were computed using nine-point numerical integration rule. This numerical analysis satisfies the kinematic constraints required to accurately model the plastic flow as described by Nagtegaal et al. [32]. A modified form of hyperbolic stress-strain model was employed in the current investigation to model the stiffness behaviour of the backfill soil. The constitutive matrix D is formulated in terms of tangent Youngs modulus Et and tangent bulk modulus Kt. Modulus values are re- lated to the confining pressure as in the hyperbolic model originally proposed by Duncan et al. [28]. These values are applicable only under monotonically increasing load conditions. In the current study the friction angle of the soil was assumed to be constant. Poissons ratio v at any stage of analysis was computed using the tangent Et and Kr values and was allowed to vary 286 between 0 and 0.495. When the magnitude of Poissons ratio exceeds these limits, the magni- tude of Kr is adjusted according to the value of I$ and the limiting value of v. The hyperbolic model is simple to use and has the advantage that model parameters can be easily determined fromstandard laboratory test data. This model has been used extensively for the analysis of many soil structures as reported by Duncan et al. [28]. However, this model can only be used for monotonically increasing load conditions as it is not applicable for simula- tions involving unload-reload conditions. Soil dilation is an important mechanism that controls the strength of soil and the efficiency of load transfer from the reinforcement to the soil during shear deformations in reinforced soil structures [29]. Conventional hyperbolic models lack the ability to simulate the dilation beha- viour of granular soils. This deficiency in the original model can be overcome by using it in conjunction with classical plasticity models. Elastic-perfectly plastic models which are based on associated flow rules predict excessive dilation of soil and hence it is common to employ a plastic potential function defined in terms of a dilation angle II, and use a non-associated flow rule to compute the plastic strains of soils. Zienkiewicz et al. [36] have suggested a suitable form for the plastic potential function by using dilation angle I# in place of the friction angle I$ in the Mohr-Coulomb yield function. The same yield and potential functions have been employed in the current investigation. The stress state at any stage is computed by correcting the stresses along the flow direction defined by the dila- tion angle in the potential function. The constitutive matrix D which relates the incremental stresses and strains is formulated using the current values of Et and Kr. The incremental stresses are added to the total stresses from the previous step to update the current stress state. If the updated stress state does not satisfy the yield criterion, the excess stresses are released along the flow direction using the dilation angle W_ This technique is similar to the initial stress methods. However, in our method, the stiffness matrix is continually updated as the stress state changes during the analy- sis rather than remaining constant as in initial stress methods. The advantage with this ap- proach is that the stiffness matrix remains symmetric leading to significant savings in storage space and computational effort. The results from this model compare well with similar results reported by Byrne and El- dridge [33] which were obtained using another form of dilatant hyperbolic model. The current hyperbolic model has been observed to give accurate predictions of many classical elastic-plas- tic problems in geomechanics such as simple shear, direct shear and bearing capacity of foot- ings. Figure 5 shows some typical results from the simulation of a simple shear test using prop- erties for the sand used in the RMC walls that has a peak friction angle $ of 53 and a dilation angle II, of 1_5[17]. The ultimate ratio of shear stress t to normal stress u, in simple shear can can be expressed using the well-known relationship given in Equation 3 [ 181: 287 v 25 shear strain (%,) a) normalized stress-strain behaviour p -M 25 shear strain (%) b) volumetric behaviour modified model original model FIGURE 5 Simple shear behaviour using modified hyperbolic model wh), = sin Cp cos$ 1 - sin+ sinr$ The modified hyperbolic model gives failure stress ratios of 0.972 for 9 = 15 and 0.798 for q = O, which are in agreement with Equation 3. With zero dilation, peak strengths are 18% lower than those predicted by the hyperbolic model with $I = 15. The slope of the volu- metric strain curve dsY/dy predicted by the current model is very close to tan+ In contrast, the conventional hyperbolic model predictsvolumetric compression before yield and zerovol- ume change during plastic deformation. If the dilation angle I) is set to zero in the current hy- perbolic model it degenerates to the original non-dilatant hyperbolic model. The results of standard triaxial compression tests [19] carried out at densities lower than those achieved in the as-built structures were used as a starting point for the estimation of hy- perbolic soil parameters in the current FEM simulations of RMC walls. These parameters were adjusted based on the results of direct shear tests from two different laboratories carried out on sand specimens prepared at representative densities [17,34]. Plate bearing tests were carried out at the surface of the sand backfill after the full-height propped panel wall test was completed and these results were used as a further independent check on the accuracy of the estimated parameters. Comparisons between experimental and FEM-predicted behaviour 288 TABLE 1 Hyperbolic parameters for RMC Soil Parameter Value Parameter Value Parameter Value K, 950 n 0.5 C 0 m 0.65 Rf 0.75 A+ O0 % 250 @0 53O 21 W from direct shear tests and a conventional plate bearing test are shown in Figures 6 and 7. Val- ues of hyperbolic parameters used in the simulations are summarized in Table 1. Reinforcement The reinforcement layers were modelled using three-noded uniaxial elements. These ele- ments represent a linear strain variation along their length. This order of strain variation is compatible with that in the surrounding interface and soil elements. The constitutive behaviour of reinforcement layers is modelled using a nonlinear equation developed from isochronous load-strain-time test data. The isochronous load-strain data is in- terpreted from constant load creep test results according to the method reported by McGown et al. [24]. The creep load tests were performed by subjecting virgin reinforcement samples to a constant tensile load for an extended period of time. The results of constant load (creep) tests are shown in Figure 8a. The isochronous curve in Figure 8b for any elapsed time is constructed from corresponding load and strain values as illustrated in the figures. In this investigation, the lOOhour isochronous curve was used in the simulations since the surcharge pressure increments were applied in roughly 100 hour time steps in the physical ex- periments. The results of in-isolation tests [24,25] have shown that under staged loading the polymeric materials exhibit a cumulative load-deformation response at the end of any incre- ment of load that is equivalent to the deformation recorded as if the final load had been applied in a single step. In the numerical model, the load P in the reinforcement layer and the tangent stiffness Jr are related to the strain E in the geosynthetic as shown in Equations 4 and 5. P = As + BE= (4) J t = a = A + ABE de An excellent curve-fitwas obtained by usingvalues of 60 and -126 for constants A and B respec- tively as shown in Figure 8b. Compressive forces are not allowed to develop within the rein- forcement elements as the geosynthetic reinforcement layers behave as fabric sheets. 289 0 2.5 5.0 7.5 10.0 horizontal displacement (mm) a) normalized stress versus displacement response 2.6 E g 1.5 E g S 1.0 S Ef- 0.5. - test data e --- 3 FEM .g 0.6 2 -0.5 0 2.5 5.0 7.5 10.0 horizontal displacement (mm) b) vertical displacement versus horizontal displacement FIGURE 6 FEM simulation of direct shear test data 0 - test data --- FEM 15 4 0 100 200 300 400 500 600 plate pressure (kPa) FIGURE 7 Comparison of experimental and predicted pressure-settlement behaviour in plate bearing test 290 Interface models The interfaces between the reinforcement layers and soil and those between wall panels and soil were modelled using six-noded joint elements of zero thickness. These elements have linear shear strain variation along their length. The six noded joint elements were developed by extending the four node joint element formulation reported by Ghaboussi et al. [35]. The global stiffness matrix of these elements is formulated in terms of two independent stif- fness values, one in the tangential (shear) direction and the other in the normal direction. When the normal stress on the interface is compressive, perfect bond is assumed in the normal direction and when the normal stress becomes tensile, the normal stiffness is assigned a small value to allow debonding at the interface. The shear strength and stiffness behaviour of interfaces between the wall panels and back- fill soil were determined using data from direct shear tests carried out on physical models of the sand/wall panel interface [ 161. The shear stiffness of interfaces between the soil and rein- forcement was modelled using stick-slip type formulation in which perfect bond was assumed when the shear stress is less than the shear strength defined by the Mohr-Coulomb model. When the shear stress exceeds the shear strength, the shear stiffness was reduced to a small residual value to allow for relative movement between the reinforcement and soil. Based on the experimental observation that the interface friction angle between well-compacted granu- lar soil and most types of geogrids is higher than that of the soil alone and that failure occurs within the soil [29], it was decided to use greater shear strength values for the interface than those for the soil alone. The properties used for all interface elements are reported in Table RESULTS AND DISCUSSION Selected experimental results The physical models were monitored until collapse due to surcharging or until incipient collapse was suspected [4,17,20-221. Both structures revealed a well-developed internal fail- ure plane through the reinforced soil zone at the end of each test. The collapse surcharge pres- sures for the full-height and incremental wall tests were 80kPa and 70kPa respectively. Incipi- ent collapse was manifest as accelerated lateral panel displacements (e.g. Figure 9) and elevated reinforcement strains indicating load transfer from soil to reinforcement. In the full- height panel wall test the uppermost reinforcement layer ruptured at the panel connection. The strains within the reinforcement layers revealed a saddle-shaped distribution with a peak at the panel connections and another peak at about the location of the internal soil failure plane. The incremental panel wall failed in two distinct steps identified as initial shear failure of the soil in the reinforced soil zone followed by load transfer to the reinforcement. The period between initial soil failure and reinforcement rupture was about 250 hours during which large 291 30 20 Q e .L t v) IO 0 / 6.25 kN/m 0 200 400 600 600 l( 100 time ( hrs ) a) constant load (creep) test data 6 r 2i / - test data or -y- ~~~+Ztion IO 0 0.05 0.10 0 strain E (mm/mm) b) curve-fit for 100 hour isochronous test data FIGURE 8 100 hour isochronous load-strain behaviour of geosynthetic reinforcement TABLE 2 Properties of interface elements panel/soil interface soillgeosynthetic interface initial shear stiffness 1000 kN/m3 lo6 kN/m3 initial normal stiffness lo6 kN/m3 lo6 kN/m3 residual shear stiffness 10 kN/m3 10 kN/m3 residual normal stiffness 100 kN/m3 100 kN/m3 friction angle 2o 5.9 creep deformations were measured. The peak strains were observed to occur within the rein- forced soil mass rather than at the connections as recorded in the full-height panel wall. The distribution of reinforcement strains at incipient collapse in both tests are compared with pre- dicted values later in the Paper (see Figure 12). Numerics One set of numerical analyses was performed with a soil dilation angle v=O and the other using a value of v,= 15 based on laboratory direct shear test results described earlier. The nu- merical analyses with q=O predicted much greater panel displacements and larger reinforce- ment strains. In some cases the over-prediction was a factor of two greater than measured val- ues as demonstrated in Figures 10 and 11. 100 end of construction F 80 L E geosynthetic rupture ~4 0 400 800 1200 1600 elapsed time (hours) FIGURE 9 Panel displacements during RMC full-height panel wall test Figure 10 shows measured and predicted lateral displacements at the mid-height of each wall during surcharge steps. The measured displacements in the figure correspond to the val- ues at the end of each lOOhour increment of surcharge. The finite element results show that the incremental wall failed between 60kPa and 75 kPa pressure and the full-height panel wall failed at slightly greater than 80kPa pressure. The predicted collapse pressures are very close to those observed in the physical experiments using*= 15. In addition, it is important to note that the finite element approach adopted in this study resulted in accurate predictions of dis- placements at working load levels (e.g. at 20-40 kPa surcharge pressure). Measured and predicted lateral displacement facing profiles are shown in Figure 11. The measured panel displacements are reported at two stages during the final (maximum) sur- charge load increment corresponding to conditions at soil failure and incipient rupture of geo- synthetic reinforcement. Displacements corresponding to soil failure are accurately predicted by the numerical simulations using W= 15 . Measured and predicted reinforcement tensile strains are illustrated in Figure 12. The peak strain magnitude and trend in strain distribution profiles along the length of the reinforcement layers in the rigid and flexible facing systems are captured by the numerical results. For exam- ple, the elevated strain levels recorded at the connections for the relatively rigid facing test are evident in the numerical results. Similarly, the peak strain levels occurring within the rein- forced soil zone are close to the observed internal failure zone. The predicted reinforcement strains at the panel connections agreed within f 1% strain of the measured values in excess of 200 - measured ) --- FEM I 0 20 40 60 80 100 surcharge (kPa) a) incremental panel wall 7 0 20 40 60 80 100 surcharge (kPa) b) full-height panel wall FIGURE 10 Lateral panel displacements at the end of surcharge increments 1% strain at all surcharge levels. The 1% strain threshold is considered by the authors to be the minimum value for which significant strains in the reinforcement can be identified. Experimentally observed and numerically predicted failure surfaces (planes) in both walls were found to closely match the failure surfaces given by Rankine theory using the peak friction angle of the backfill soil (Figure 13). The failure surfaces in the numerical models are inferred from either the location of peak reinforcement strains or from the shear strain contours as shown in Figures 13a and 13b. This finding gives confidence to the widely used design assump- tion that the reinforced soil zone can be divided into active and resistant zone based on classi- cal earth pressure theory [l-2]. Figure 14 shows measured and predicted forces developed at the base of the wall facing units. The comparison is reasonably good indicating that the interface elements in the vicinity 294 soil failure 3.0 2.5 r prior to geosynthetic rupture 70 kPa surcharge --- FEM - measured 0, I I I I 1 I I I , 0 50 100 150 200 2 panel displacement (mm) a) incremental panel wall I soil failure 3.0- I, ,L ~=15 / --- FEM - measured 0 50 100 150 200 2 panel displacement (mm) b) full-height panel wall io FIGURE 11 Measured and predicted panel displacements of the wall facing performed satisfactorily through the entire loading range. The data illus- trates that significant load shedding is developed at the base of the wall and hence the facing elements in combination with a rigid footing are responsible for a significant portion of the surcharge capacity of the trial walls. The magnitude and distribution of vertical pressures act- ing at the base of the reinforced soil zone can be expected to be influenced by load shedding to the wall facing. Figure 15 shows that this effect is pronounced at large surcharge pressures. The trend observed in physical experiments is captured in the numerical simulations. CONCLUSIONS This Paper presents details of discrete type finite element modelling for geosynthetic rein- forced soil walls together with the material models employed to simulate the behaviour of vari- ous components in these structures. The modified form of hyperbolic model used by the au- thors is shown to account for soil shear strength increase due to dilation. The results presented in the Paper show that it is possible to accurately simulate all significant performance features of geosynthetic reinforced soil walls at both working load and collapse conditions. The Paper 295 Layer 4 incremental wall Layer 3 - full-height panel wall 0 0.5 1.0 1.5 2.0 2.5 3.0 8- full-height panel wall Layer 2 ;i 5 .5 s UY 01 1 0 0.5 1.0 1.5 2.0 2.5 3.0 3- T 2.J full-height panel wall 5 Layer 1 .E \ S - \ incremental wall * I- cfull-height panel wall I. I - measured \ \ ~_---\--,___ __ --- FEM 0, 1 0 0.5 1.0 1.5 2.0 2.5 3.0 FIGURE 12 Strain in geosynthetic reinforcement layers at incipient collapse 296 observed failure surface in test distance from toe (m) a) incremental panel wall v layer 1 ------------ OV 0 0.5 1 .o 1.5 2.0 distance from toe (m) ---- FEM M observed failure surface in test b) full-height panel wall FIGURE 13 Predicted and measured Internal failure surfaces further demonstrates that construction-induced differences in behaviour can be simulated us- ing the approach adopted in this investigation. An important conclusion of the work described here is that the strength and stiffness properties of component materials can be determined from the results of independent routine laboratory tests and then successfully implemented in finite element modelling of the composite structure. ACKNOWLEDGEMENTS Financial support for the work reported here was provided through the Academic Research Program (ARP) program and by the Chief of Construction and Properties, Department of Na- tional Defence, Canada. 1. 2. REFERENCES Bathurst, R.J., Geosynthetics for Reinforcement Applications in Retaining Walls, &&r Canadian Geotechnicul Conference, Calgary (1991) Vol. 2, p. 74.1-74.10. Christopher, B.R., Gill, S.A., Giroud, J-P, Juran, I., Schlosser, F., Mitchell, J.K., and Dunnicliff, J., Reinforced Soil Structures, Volume I. Design and Construction Guidelines, Federal Hiphwqvs Administration, Washington USA, Report No. FHWA-RD-89-043 (1989) pp. 287. 297 OS_ I b 1 I I i 0 10 20 30 40 50 60 70 80 Rh-cm surcharge (kPa) I a) incremental panel wall f% 0 10 20 30 40 50 60 70 80 surcharge (kPa) I R b) propped panel wall FIGURE 14 Predicted and measured reactions at pinned toe restraint 75- ------------- 50- end of construction - measured -- 25- FEM 0 I I I I 0 1 2 3 4 distance from toe (m) i FIGURE 15 Distribution of earth pressures at base of backfill (incremental panel wall) 298 3. Bathurst, R.J., Simac, M.R., and Berg, R.R., Review of the NCMA Segmental Retaining Wall Design Manual for Geosynthetic Reinforced Structures, Transoortation Research Re- d, (1993) (in press). 4. Bathurst, R.J., Investigation of Footing Restraint on Stability of Large-scale Reinforced Soil Wall Tests, 46th Canadian Geotechnical Conference, Regina, September (1993) 10 p. 5. Romstad, K.M., Herrmann, L.R., and Shen, C-K., Integrated Study of Reinforced Earth-I: Theoretical Formulation, J. Geotech. Erw. Div. ASCE, (1976) Vol. 102, No. 5, 457-471. 6. Romstad, K.M., Herrmann, L.R., and Shen, C.K., Integrated Study of Reinforced Earth- II: Behavior and Design, J. Geotech. Erg. Div. ASCE, (1976) Vol. 102, No. 6,577-590. 7. Herrmann, L.R., and Al-Yassin, Z., Numerical Analysis of Reinforced Soil Systems, Proc. ofASCE Svmp. ofEarth Reinforcement, Pittsburgh, (1978) 428-457. 8. Herrmann, L.R., Welch, K.R., and Lim, C.K., Composite FEM Analysis for Layered Sys- tems, J. Geotech. Eng. Div. ASCE, (1984) Vol. 110, No. 9, 1284-1302. 9. Karpurapu, R.G., and Bathurst, R.J., Analysis of geosynthetic reinforced soil wall by the finite element method, Proc. of fourth Znt. Svmp Num. Models in Geomech., Swansea, Ed. by G.N. Pande and S. Pietruszczak, A.A. Balkema, Rotterdam, (1992) 861-870. 10. Bathurst, R.J., Karpurapu, R., and Jarrett, PM., Finite Element Analysis of a Geogrid Reinforced Soil Wall, Groutinp. Soil I mprovement and Geosythetics, ASCE Geotechnical Special Publication No. 30, (1992) Vol. 2, 1213-1224. 11. Karpurapu, R.G., Bathurst, R.J. and Jarrett, PM., Finite Element Analysis of an Incre- mentally Constructed Geogrid Reinforced Soil Wall, Proc. of44th Canadian Geotechnical Conference, Calgary, (1991) Vol. 2, 82.1-82.8. 12. Kutara, K., Gomadou, M. A., and Takeuchi, T., Deformation Analysis of Geotextiles in Soils using the Finite Element Method, Geotextiles and Geo membranes, (1986) Vol. 4, 191-205. 13. Matsui, T., and San, KC., Prediction of Two Test Walls by Elastoplastic Finite Element Analysis, Znt. Svmu. on Geosyrthetic-Reinforced SoilRetaining Walls, Denver, Ed. by J.T.H. wu (1991) 143-151. 14. Karpurapu, R.G., and Bathurst, R.J., Numerical Investigation of Controlled Yielding of Soil Retaining Wall Structures, Geotextiles and Geomembranes, (1992) Vol. 11, 115-131. 15. Rowe, R.K., and Soderman, K.L., Stabilization of Very Soft Soils Using High Strength Geosynthetics: the Role of Finite Element Analyses, Geotextiles and Geomembranes, (1987) Vol. 6,53-80. 16. Bathurst, R.J., and Benjamin, D.J.R.S., Preliminary Assessment of Sidewall Friction on Large-Scale Wall Models in the RMC Test Facility,A_oplication o_fPolvmeric Reinforcement m Soil Retaining Structures, Ed. by PM. Jarrett and A. McGown, Kluwer Academic Pub- lishers, Dordrecht, (1987) 181-191. 17. Bathurst, R.J., Jarrett, PM., and Benjamin, D.J.R.S., A Database of Results from an In- crementally Constructed Geogrid-reinforced Soil Wall Test, Proceedinas ofSoil Reinforce- ment: Full Scale Exueriments of the 80s, ISSMFEENPC, Paris, France, November (1993) 401-430. 18. Davis, E.H., Theories of Plasticity and the Failure of Soil Masses. In: Soil Mechanics: Se- lected Tonics, Ed. by I.K. Lee, Butterworths, London (1968) 341-380. 19. Bathurst, R.J., Wawrychuk, W.F. and Jarrett, PM., Laboratory Investigation of Two Large- Scale Geogrid Reinforced Soils Walls, TheAvulication o_fPolvmetic Reinforcement in Soil Retaining Structures, Ed by PM. Jarrett and A. McGown, NATO Advanced Study Institutes Series, Kluwer Academic Publishers, (1987) 75 - 125. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 299 Bathurst, R.J., Benjamin, D.J., and Jarrett, PM., Laboratory Study of Geogrid Rein- forced Soil Walls, Geosythetics for Soil I mprovement Ed. by R. Holtz, ASCE Special Publication 18, (1988) 178-192. Bathurst, R.J., Benjamin, D.J., and Jarrett, PM., An Instrumented Geogrid Reinforced Soil Wall, Froc XI I nt Conf on Soil Mech. and Found. Engg., Rio de Janerio, Brazil, (1989) 1223-1226. . Bathurst, R.J., and Benjamin, D.J., Failure of a Geogrid Reinforced Soil Wall,mnotia- tion Research Record 1288, Washington, D.C., (1990) 109-116. Bathurst, R.J., Instrumentation of Geogrid Reinforced Soil Walls, J Yransnortation Re- search Record 1277, Washington, DC, (1990) 102-111. McGown, A., Andrawes, K.Z., Yeo, K.C., and Dubois, D.D., The Load-Strain-Time Be- haviour of Tensar Geogrids, Proc. Svmn. Po&mer Grid Reinforcement in Civil Entineeting, London, Paper No. 1.2 (1984) 11-17. Yeo, K.C., J heBeh vio ro_fPolvmenc G ds forSoilReinforcement, Ph.D. Thesis, Univer- sity of Strathclyde, &asiow, Scotland (lG85) 185 p. Pastor, M., Zienkiewicz, O.C., and Chan, A.H.C., Generalized Plasticity and the Model- ling of Soil Behaviour, Znt. J . Num. Anal. Meth. in Geomechanics, (1990) Vol. 14, 151-190. Lade, PV, Elasto-Plastic Stress-Strain Theory for Cohesionless Soil with Curved Yield Surfaces, I nt. J . Solids and Structures, (1977) Vol. 13, 1019-1035. Duncan, J.M., Byrne, P, Wong, K.S., and Mabry, I?, Strength, stress-strain and bulk modu- lus parameters for finite element analysis of stresses and movements in soil masses, Beport No. UCBI GTl80-01, University of California, Berkely, (1980) 70 p. Johnston, R.S., and Romstad, K.M., Dilation and boundary effects in large scale pull-out tests, Proc. XI I 1 I nt. Conf Soil Mech. and Found. En_% Rio de Janeiro, (1989) 1263-1266. Bathurst, R.J., and Karpurapu, R.G., Large-Scale Reinforced-Soil Wall Tests and Nu- merical Modelling of Geosynthetic Reinforced Soil Structures, report submitted to the Chief of Construction and Projects, Department of National Defence, Military Enaineeting Research Group, Royal Military College of Canada, 1990 & 1991. Bathe, K.J., Ramm, E., and Wilson, E.L., Finite element formulation for large deforma- tion dynamic analysis, J nt. .I . Num. Meth. in Erg., (1975) 9, 353-386. Nagtegaal, J.C., Parks, D.M., and Rice, J.R., On Numerically Accurate Finite Element Solutions in the Fully Plastic Range, .I. Camp. Meth. App. Mech. and Enp, (1984) Vol. 4, 153-177. Byrne, PM., and Eldridge, T.L., A three parameter dilatant elastic stress-strain model for sand, Proc. Znt. Svmp. Numerical Models in Geomechanics, Zurich, (1982) 73-80. Jewell, R.A., Analysis and Predicted Behaviour for the Royal Military College Trial Walls, . . . Apniication of he PO ym nc Rernforcement in Sot Re mung Stru es, Ed. by PM. Jarrett and A. McGo&, Kl\wEr Academic PublishersjbdPdrecht, (1%; 193-235. Ghaboussi, J., Wilson, E.L., and Isenberg, J., Finite Element for Rock Joints and Inter- faces. J. of Soil Mech. and Found. Division, ASCE, (1973) Vol. 99 No. 10, 833-848. Zienkiewicz, O.C., Humpheson, C., and Lewis, R.W., Associated and nonassociated visco-plasticity and plasticity in soil mechanics, GeotechniQye, (1975) Vol. 25, No. 4, 671-689. Received 22 June 1993; revised version received 3 March 1994; accepted 30 March 1994