Professional Documents
Culture Documents
Prediction of Field Behavior of Reinforced Soil Wall Using Constitutive Model
Prediction of Field Behavior of Reinforced Soil Wall Using Constitutive Model
Abstract: A geosynthetic-reinforced soil retaining wall using full-height concrete wall facing panel was constructed at Tanque Verde
Road site for grade-separated interchanges in Tucson, Ariz. Numerical simulation of this wall was performed using a finite element code
Downloaded from ascelibrary.org by New York University on 05/12/15. Copyright ASCE. For personal use only; all rights reserved.
called DSC-SST-2D. The program allows for plane strain, plane stress, and axisymmetric idealizations including simulation of construction
sequences. The wall was modeled as a plane strain, two-dimensional problem. Material parameters used in the analysis were obtained
from experimental results from conventional triaxial compression tests for backfill soils and cyclic multidegree-of-freedom shear tests for
interfaces. The soils and interfaces were modeled using the disturbed state concept and hierarchical single surface plasticity models, and
the geogrid reinforcement was simulated by a linear elastic model. The interfaces between the reinforcement layers and soil were modeled
using the thin layer element. The results of the finite element analysis were in good agreement with the measured field behavior of the
wall. Comparison involved vertical and lateral stress transferred to reinforcements and wall face movements. It was found that the use of
the unified constitutive model in a nonlinear finite element method provided satisfactory predictions for the field performance of the
Tensar geogrid reinforced soil wall.
DOI: 10.1061/共ASCE兲1090-0241共2005兲131:6共729兲
CE Database subject headings: Walls; Soil structure; Geosynthetics; Finite element method; Constitutive models; Soils; Interfaces;
Predictions; Comparative studies.
Construction Simulation
The in situ stress was introduced in the foundation soil by adopt-
ing coefficient K0 = 0.4. Then the backfill was constructed into 11
layers 关Fig. 2共b兲兴, as was done in the field 共Fig. 1兲. The compacted
soil was included in each layer, and the reinforcement was placed
on a layer before the next layer was installed. The compacted soil
in a given layer was assigned the material parameters according
to the stress state induced after installing the layer. The comple-
tion of the sequences of construction is referred to as “end of
construction.” Then the surcharge load due to the traffic of 20 kPa
Fig. 2. Coarse and part of fine mesh
was applied uniformly on the top of the mesh, 共Fig. 2兲; this stage
is referred to as “after opening to traffic.” The concrete pavement
was not included in the mesh. However, since it can have an
viscoelastic, plastic, and disturbance 共softening兲 can be chosen for influence on the behavior of the wall, in general, it is desirable to
the analysis. The wall was modeled as a plane-strain, two- include the pavement.
dimensional problem for the finite element analysis. Since the
tensar reinforcement is continuous normal to the cross section,
共Fig. 1兲, the plane strain idealization is considered to be appropri-
Materials Properties and Models
ate. DSC-SST-2D uses the finite element approach and requires
material properties to explicitly model the soil, facing panels,
reinforcement layers, and the interfaces. The program was written The properties of the materials used in the analysis were obtained
to allow incremental fill placement to be simulated 共i.e., rows of from experimental results from conventional triaxial compression
elements added sequentially as fill placement兲. tests for soil backfill, and cyclic multidegree-of-freedom
Two finite element meshes, coarse and fine, were used. Fig. 共CYMDOF兲 shear tests for interfaces 共Desai and Rigby 1997兲.
2共a兲 shows the coarse mesh with 184 nodes and 167 elements The material models play a major role in any solution methods
including 10 wall facing, 18 interface between soil and reinforce- such as the finite element analysis. So, it is very important to
ment, and 9 bar 共for reinfircement兲 elements. In the coarse mesh, adopt suitable material models for soil and interfaces.
only three layers of reinforcement were considered. The fine mesh
contained 1,188 nodes, and 1,370 elements including 480 inter- Soil and Interface Modeling
face, 35 wall facing, and 250 bar elements; it contained 11 layers
as in the field. The properties of reinforcement in the coarse and The soil 共backfill, foundation, and retained fill兲 and the interfaces
fine mesh were assumed to be the same. The fine mesh was con- between reinforcement and soil were modeled using the DSC,
sidered to contain a great number of nodes and elements; hence, which included the HISS plasticity model. Details of the DSC for
proved correlation with the field data, compared to those from the
coarse mesh. Hence, from now on the results from the fine mesh
are presented and analyzed.
The measured and predicted vertical soil stress near the wall
face is generally less than the overburden value 共i.e., v = ␥h,
where h = height to elevation= 1.53 m兲. This can be due to the
relative motions between the backfill and reinforcement. It is seen
that the vertical stress distribution along the reinforcement layer is
nonlinear. The vertical pressure increases in the zone away from
the facing panel until reaching a maximum value at a distance of
about 152 cm from the wall face. Thereafter, it shows a decrease.
Also, shown in Fig. 3 are the trapezoidal vertical stress distri-
Fig. 6. Comparison between field measurements and predictions for Fig. 7. Comparison between field measurements and predictions for
horizontal soil strains at elevation= 2.44 m vertical soil strains at elevation= 1.08 m
Displacements
Fig. 10 shows predicted and measured wall movements. The cor-
relation is satisfactory near the lower heights of wall; however, it
is not satisfactory elsewhere. For example, near the top of the
wall the predicted value of about 42 mm is not in good agreement
with the measured value of about 76 mm. The finite element
analysis using the linear elastic model reported the maximum
displacement of about 30 mm 共Fishman and Desai 1991兲. With
the present nonlinear soil and interface models, the maximum
Downloaded from ascelibrary.org by New York University on 05/12/15. Copyright ASCE. For personal use only; all rights reserved.
Fig. 9. Comparison of field measurements and predictions for Fig. 10. Comparison between predicted and measured wall face
horizontal stress carried by geogrid near wall face movement after opening to traffic
Summary and Conclusions ported, in part, by the National Science Foundation, Washington,
D.C., Grants Nos. CMS 9115316 and 9732811.
The nonlinear finite element procedure with the unified and real-
istic DSC model is used to predict the field performance of a
Tensar reinforced wall. The DSC model parameters for soil and Appendix. Disturbed State Concept Model, Testing,
interfaces are found based on comprehensive soil tests by using and Parameters
triaxial testing and interface tests by using the CYMDOF shear
device. The measured behavior of the wall at the end of construc-
tion and after opening to traffic has been compared with the pre- Disturbed State Concept
dictions from the finite element analysis with respect to lateral In the DSC, it is assumed that a material 共soil or interface兲 ele-
stresses on the wall facing, soil strains, geogrid strains, horizontal ment at any stage of loading is composed of the relative intact
and vertical soil stresses, lateral stresses carried by geogrids, and 共RI兲 and fully adjusted 共FA兲 parts. The basic incremental equation
wall displacements. Overall, the realistic DSC model in nonlinear is given by 共Desai 2001兲
finite element procedures provides very good correlations be-
tween measured and predicted results for all quantities, except the
wall displacements; comments regarding possible errors in mea- da = 共1 − D兲Cidi + DCcdc + dD(c − i) 共2兲
surement of displacements are provided.
where a, i, and c denote observed, RI, and FA responses, respec-
> i and C
tively; C > c denote constitutive matrices for RI and FA parts,
Acknowledgments respectively; > = stress vector; > = strain vector; D denotes distur-
bance; and d denotes increment or rate.
The financial support provided to Dr. K. E. El-Hoseiny by Men- The RI 共C > i兲 behavior can be modeled by using elastic 共linear
oufia University, Egypt, is gratefully acknowledged. Dr. A. F. or nonlinear兲 and plasticity models. Here, it is simulated by using
Youssef provided valuable guidance and support. The research on the HISS plasticity model in which the yield function for soils is
the development of the CYMDOF device used herein was sup- given by
a1
␣= 共4兲
n1
where ␣1 and 1 = growth parameters; and = plastic strain trajec-
Downloaded from ascelibrary.org by New York University on 05/12/15. Copyright ASCE. For personal use only; all rights reserved.
Fi = 2 + ␣inn − ␥2n = 0 共5兲 Fig. 12. Configurations for interface layer at top and bottom
where and n = shear and normal stresses, respectively; ␥
= ultimate parameter; and 关Eq. 共4兲兴 = trajectory of plastic tangen-
tial and normal relative displacements. Interface Tests
The disturbance function is given by The CYMDOF device 共Desai and Rigby 1997兲 was used to test
Z interface behavior between the Tensar reinforcement and the
D = Du共1 − e −AD
兲 共6兲 backfill soil under direct shear testing.
where Du, A, and Z = disturbance 共softening parameters兲 and It should be noted that direct shear tests can simulate the in-
Dtrajectory of deviatoric plastic strains. terface behavior approximately. The interface behavior is influ-
enced by complex interaction and interlocking of grains around
the grid openings 共Jewell et al. 1984兲. In order to identify the
Nonassociative Behavior effect of openings 共apertures兲, two types of tests were performed
The behavior of soils and interfaces were found to be frictional using flat geogrid and geogrid with openings.
and nonassociative. Hence, modification in the hardening function The diameter of the specimen is 共165 mm 6.5 in.兲. The desired
␣ to account for the nonassociative response introduced an addi- initial density 共about 18 kN/ m3兲 of the soil was achieved by com-
tional parameter 共Desai 2001兲. pacting the specimen in five layers using about 25 blows.
The geogrid sample was prepared by cutting the specimen
共165 mm兲 from a large sheet and attaching it to a steel block. It
Laboratory Tests was placed on the soil 共dry or bulk with 8% moisture content兲 in
two configurations, 共Fig. 12兲, which were used to study the effect
Comprehensive series of triaxial, hydrostatic and shear tests were of the location of the interface in the test device.
performed for soils, and CYMDOF shear tests were performed A normal stress was applied and then the horizontal shear
for interfaces between reinforcement and soil. Brief details and displacement or stress was applied with a number of loading,
typical results are presented below. unloading and reloading cycles. Table 1 shows the interface fric-
tion angles and adhesion for different configurations, 共Fig. 12兲 for
geogrids with openings and flat geogrid. Because of the interlock-
Soils ing effect, the values for geogrids with openings are higher than
The soil 共backfill兲 samples were collected from the site. The soil those for the flat geogrids. The results for the bulk soil 共with
was classified as SP poorly graded or gravelly sand. The follow-
ing are the index properties: specific gravity= 2.64; D10 , D30 , D60
= 0.48, 1.00, 1.75 mm; emax = 0.71, emin = 0.37; ␥d max Table 1. Summary of Interface Tests and Configurations
= 18.84 kN/ m3; ␥d min = 15.35 kN/ mm3; ␥d共field兲 ⬇ 18.0 kN/ m3; Interface
and optimum moisture content= 8.0%. friction Adhesion
The soil sample 共71 mm diameter and 142 mm height兲 was angle ␦ ca
prepared by compacting it in a split model into six layers until the Interface tested 共degree兲 共kPa兲
desired initial density 共⬇18 kN/ m3兲 was obtained. The sample
Dry soil/geogrida at top 关Fig. 12共a兲兴 38.5 41
was installed in the triaxial device and the initial confining stress
Dry soil/geogrid at bottom 关Fig. 12共b兲兴 38.2 40
was applied. Then the deviatoric stress was applied with the strain
rate of about 0.03% per millimeter. Bulk soil/geogrid at bottom 关Fig. 12共b兲兴 32.2 58
The triaxial shear tests including loading, unloading, and re- Bulk soil/geogrid at top 关Fig. 12共a兲兴 33.8 66
loading were performed on the samples under different initial Dry soil/flatb geogrid at top 关Fig. 12共a兲兴 26.5 29.5
normal stresses, 3 = 17.5, 35.0, 52.0, 70.0, 140.0, 210.0, 345.0, Bulk soil/flat geogrid at top 关Fig. 12共a兲兴 27.0 37.0
and 420.0 kPa. The maximum confining pressure relates to the Bulk soil/flat geogrid at bottom 25.2 23.0
approximate field pressure of about 480 kPa. Typical test results, 关Fig. 12共b兲兴
hydrostatic and triaxial shear 共3 = 35, 210, and 420 kPa兲, are
a
Geogrid implies geogrid with openings.
b
shown in Figs. 11共a–d兲. Flat implies geogrid without openings.
Fig. 13. Cyclic multidegree-of-freedom shear tests for interface bulk soil/geogrid at top
moisture content of about 8.0%兲 with the geogrid at the top of the Material Parameters and Validations
soil were found to be realistic and are used for modeling and
computer analysis. The DSC/HISS model parameters were obtained by following the
The shear tests were performed under loading, unloading, and procedures published in various publications 共e.g., Desai 2001兲.
reloading at various normal stresses: n = 35.0, 70.0, 140.0, 350.0, The soil exhibited softening behavior, 关Figs. 11共b–d兲兴; hence, the
and 700.0 kPa. Typical results for normal and shear behavior for disturbance 共softening兲 model was used. The interface did not
n = 70.0, 140.0, and 700.0 kPa are shown in Figs. 13共a–d兲. exhibit significant softening behavior; hence the HISS plasticity
Table 2. Material Parameters Used in Finite Element Method Analysis of Tucson Wall
Material constant Symbol Soil Interface
Elastic E or Kn f 1 共3兲a f 2 共n兲b
v or Ks 0.3 f 3 共n兲
Plasticity—ultimate ␥ 0.12 2.3
 0.45 0.0
Phase change parameter n 2.56 2.8
Growth parameters a1 3.0E − 05 0.03
1 0.98 1.0
Nonassociative constant 0.2 0.4
Disturbance parameters Du 0.93 —
A 0.37
Z 1.60
Angle of friction and adhesion / ␦ / ca = 40° ␦ = 34°
ca = 66 kPa
Unit weight 共field兲 ␥ 18.00 kN/ m3 —
Coefficient of earth pressure at rest K0 0.4 —
a
E = 62⫻ 103 0.28
3 .
b
ks 共shear stiffness兲 = 30⫻ 103 0.28
n ; kn 共normal stiffness兲 = 18⫻ 10 n .
3 0.29