Computers and Geotechnics 65 (2015) 164–174

Contents lists available at ScienceDirect

Computers and Geotechnics

journal homepage: www.elsevier.com/locate/compgeo

Research Paper

Influence of choice of FLAC and PLAXIS interface models on reinforced

soil–structure interactions
Yan Yu a,1, Ivan P. Damians b,2, Richard J. Bathurst a,⇑
a
GeoEngineering Centre at Queen’s-RMC, Department of Civil Engineering, 13 General Crerar, Sawyer Building, Royal Military College of Canada Kingston, Ontario K7K 7B4, Canada
b
Department of Geotechnical Engineering and Geo-Sciences (ETCG) and Institute for Sustainability (IS.UPC), Universitat Politècnica de Catalunya-BarcelonaTech (UPC), Spain

a r t i c l e i n f o a b s t r a c t

Article history: The choice of structure element to simulate soil reinforcement and soil–structure interaction details for
Received 9 September 2014 numerical modelling of mechanically stabilized earth (MSE) walls can have a significant influence on
Received in revised form 12 December 2014 numerical outcomes. Program FLAC (finite difference method) offers three different options (beam, cable
Accepted 13 December 2014
and strip element) to model the reinforcement and program PLAXIS (finite element method) has two
Available online 3 January 2015
(beam and geogrid element). Both programs use different models and properties to simulate the mechan-
ical behaviour of the interface between dissimilar materials. The paper describes the details of the linear
Keywords:
elastic Mohr–Coulomb interface model available in the two software packages to model material inter-
Interface
Soil–structure interaction
action and how to select model parameters to give the same numerical outcomes. The numerical results
Reinforced soil walls quantitatively demonstrate the conditions that give good agreement between the two programs for the
Numerical modelling same steel strip reinforced soil–structure problem and the situations where they do not. For example, the
FLAC paper demonstrates that results can be very different depending on the type of structure element used to
PLAXIS model horizontal reinforcement layers that are discontinuous in the plane-strain direction.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction Two numerical methods are generally used to model MSE walls:
(a) finite element method (e.g., [3–8]), and (b) finite difference
Mechanically stabilized earth (MSE) walls have advantages method (e.g., [9–15]).
with respect to ease of construction and cost over traditional The numerical modelling of MSE walls requires the use of inter-
concrete gravity and cantilever retaining walls, and are now face boundaries to simulate the discontinuity and transfer of nor-
used widely around the world. However, MSE walls are compli- mal and shear stresses from the soil to the reinforcement and
cated mechanical structures with multiple design limit states for facing components. However, the different treatment of the inter-
internal, external and facing stability modes of failure. Further- nal boundaries in commercially available programs using these
more, the interactions between the backfill soil and the facing two different numerical techniques and choice of reinforcement
and reinforcement components strongly affect the performance structure element available in the programs may result in different
of MSE walls. The conventional approach to internal stability numerical predictions for the nominally identical MSE wall.
design of these structures is to use closed-form solutions based The objective of this paper is to examine numerical modelling
on classical notions of active earth pressure theory (e.g., [1,2]). details of the load transfer within a segment comprising a precast
However, this approach is restricted to simple structures with concrete panel with steel strip soil reinforcement using the finite
simple boundary conditions, geometry and materials. For more difference method (FLAC; [16]) and the finite element method
complicated project conditions or for performance-based design, (PLAXIS; [17]). Both programs are widely used by geotechnical
geotechnical engineers often resort to advanced numerical mod- engineers and researchers to solve soil–structure interaction prob-
elling techniques. lems including MSE wall systems. A method to develop equivalent
interface property values for both programs is presented. The
paper also demonstrates the influence of choice of structure ele-
⇑ Corresponding author. Tel.: +1 613 541 6000x6479; fax: +1 613 541 6218.
ment on numerical outcomes using beam, cable and strip options
E-mail addresses: [email protected] (Y. Yu), [email protected] (I.P.
Damians), [email protected] (R.J. Bathurst). in FLAC for the soil reinforcement and the beam and ‘‘geogrid’’
1
Tel.: +1 613 541 6000x6347; fax: +1 613 541 6218. options in PLAXIS. Finally, the paper identifies situations where
2
Tel.: +34 93 401 1695; fax: +34 93 401 7251. the two programs can give very different results.

http://dx.doi.org/10.1016/j.compgeo.2014.12.009
0266-352X/Ó 2014 Elsevier Ltd. All rights reserved.
 Y. Yu et al. / Computers and Geotechnics 65 (2015) 164–174 165

2. Interface modelling where usoil, csoil, wsoil, rt,soil, and Gsoil are the friction angle, cohesion,
dilation angle, tensile strength, and shear modulus of the surround-
For soil and structure zones in direct contact, two options are ing soil, respectively.
available to model soil–structure interaction in advanced numeri- The second option treats the interface as a separate soil zone
cal models [18]: (a) interface elements with zero thickness to (with zero thickness). The interface property values are also calcu-
transfer shear and normal stresses from the soil to the structure; lated using Eqs. (4)–(11) but the soil property values are for the
and (b) continuum elements with finite thickness. The focus of this interface and thus can be different from the properties of the sur-
paper is on interface elements with zero thickness that are avail- rounding soil. This is a more flexible approach with respect to
able in FLAC and PLAXIS. equivalency between parameters used in FLAC and PLAXIS models,
especially when the shear stiffness is available from laboratory
2.1. Interface model and properties in FLAC tests or assumed from FLAC modelling as discussed below. It
should be noted that Poisson’s ratio is fixed with vi = 0.45 in PLAXIS
The interfaces in FLAC [16] can be defined as glued, unglued, or for interfaces which results in the normal stiffness kn = 11ks for all
bonded interfaces depending on the application. For the purpose of interfaces.
comparison with PLAXIS [17], unglued interfaces (where the slip
or/and opening of interfaces is allowed and the plastic shear dis- 2.3. Equivalent interface properties for FLAC and PLAXIS
placement occurs after the shear stress exceeds a maximum shear
strength) are used in this paper. The interface properties are fric- The interface friction angle, cohesion, dilation angle, and tensile
tion angle (ui), cohesion (ci), dilation angle (wi), tensile strength strength in FLAC are the same as those in PLAXIS and the same
(rt,i), normal stiffness (kn), and shear stiffness (ks). The interface parameter values can be set directly in both programs. If the
shear strength is governed by the Mohr–Coulomb failure criterion: normal stiffness and shear stiffness from FLAC are known and
kn = 11ks, the equivalent interface properties in PLAXIS can be
ss;max ¼ ci þ rn tan ui ð1Þ
found using the following equations:
where ss,max is the maximum shear stress at the interface under
normal stress (rn).
ð3kn  4ks Þks ti
Ei ¼ and mi ¼ 0:45 ð12Þ
The normal stress and shear stress (ss) are calculated based on kn  ks
the interface normal displacement (un) and shear displacement (us)
using the following equations: Eoed;i ¼ kn t i ð13Þ

rn ¼ kn un ð2Þ Gi ¼ ks t i ð14Þ

ks us ks us 6 ss;max where ti is the virtual thickness of the interface which is related to
ss ¼ ð3Þ average element size and virtual thickness factor in PLAXIS (the
ss;max ks us > ss;max
exact value used during calculation can be found in the OUTPUT
program – a post-processor in PLAXIS). For cases where kn – 11ks,
2.2. Interface model and properties in PLAXIS no equivalent interface properties can be found for FLAC and
PLAXIS.
Interfaces using the linear elastic model with Mohr–Coulomb If Young’s modulus and fixed Poisson’s ratio vi = 0.45 (or
failure criterion in PLAXIS [17] are considered here for comparison oedometer modulus and shear modulus) at the interface with
with FLAC [16]. These interfaces have properties of friction angle, Ri = 1.0 (using the second option for setting interface property val-
cohesion, dilation angle, tensile strength, Young’s modulus (Ei), ues) are provided from PLAXIS, the following equations can be used
and Poisson’s ratio (vi). Young’s modulus and Poisson’s ratio can to compute the equivalent interface properties in FLAC:
be replaced by using oedometer modulus (Eoed,i) and shear modu-
lus (Gi). The values of interface properties in PLAXIS can be set Ei ð1  mi Þ Eoed;i
kn ¼ ¼ ð15Þ
using two options. The first option uses a reduction factor ð1 þ mi Þð1  2mi Þt i ti
(Ri 6 1.0) applied to the soil material when defining soil property
values (the default value is Ri = 1.0, i.e. a fully-bonded interface). Ei Gi
ks ¼ ¼ ð16Þ
Hence, the interface property values are directly related to the 2ð1 þ mi Þt i ti
mechanical properties of the soil forming the interface as:
It should be noted that Eqs. (4)–(11) are used in this investiga-
ci ¼ Ri csoil ð4Þ tion to calculate interface property values including those in Eqs.
(15) and (16) that are used in turn to compute kn and ks for FLAC
tan ui ¼ Ri tan usoil ð5Þ simulations. If the soil Poisson’s ratio is not 0.45 and reduction fac-
 tor Ri < 1.0 are assumed for the interfaces, Young’s modulus, Pois-
0 Ri < 1:0 son’s ratio, oedometer modulus and shear modulus for Eqs. (15)
wi ¼ ð6Þ
wsoil Ri ¼ 1:0 and (16) are computed using Eqs. (9), (8), (10) and (7), respectively,
in PLAXIS simulations.
Gi ¼ R2i Gsoil ð7Þ
3. Problem definition and parameter values
mi ¼ 0:45 ð8Þ
3.1. Unit cells
Ei ¼ 2Gi ð1 þ mi Þ ð9Þ
Fig. 1 shows the unit cells (dimensions of 1 m  1 m) that were
1  mi modelled in this paper. The unit cell approach with concrete in the
Eoed;i ¼ 2Gi ð10Þ
1  2mi top cell was found to be the simplest method to examine equiva-
lent interface properties for the same geometry and boundary
rt;i ¼ Ri rt;soil ð11Þ conditions using FLAC and PLAXIS programs. Two cells were
166 Y. Yu et al. / Computers and Geotechnics 65 (2015) 164–174

Surcharge load q Surcharge load q were calculated using Eqs. (12)–(14). The number of zones in FLAC
(and the number of elements in PLAXIS) has no effect on the

Displacement
Displacement

Concrete numerical results because of the very high elastic modulus

Concrete

1.0 m
1.0 m
assigned to the concrete and the fixed boundary conditions of
Interface Interface the soil.

3.2. Single precast concrete panel segment

1.0 m
Soil 1.0 m
Soil
y y To better understand the load transfer from the backfill soil to
x x the adjacent structures, a single precast concrete panel segment
was simulated as shown in Fig. 2. The panel has a height of
1.0 m 1.0 m
1.5 m, a thickness of 0.18 m, and an out-of-plane width of
(a) (b) 1.35 m. These dimensions fall within the range of panel dimen-
sions reported in the literature for steel reinforced soil wall sys-
Fig. 1. Schematic showing unit cells with concrete–soil interface: (a) boundary
tems. However, actual dimensions are not critical to the
conditions and finite difference numerical grid with FLAC and (b) boundary
conditions and finite element mesh with PLAXIS (Note: for boundary conditions and qualitative outcomes in this investigation. The modelled backfill
concrete material modulus examined, the number of zones in FLAC and elements in soil zone is 5.0 m long and 1.5 m high and is supported by a
PLAXIS does not affect numerical results). smooth rigid foundation.
Three cases were examined. For all cases, the top of the backfill
soil was free in both the x- and y-direction and the bottom of the
backfill soil was fixed in the y-direction. The right side of the back-
considered for each test. The material in the bottom cell was soil
fill soil was fixed in the x-direction. Other boundary conditions and
and the top cell was concrete. The property values for the concrete
geometric details are given below for each case:
and the interface are given in Table 1. The soil property values are
unrestricted because of the fixed boundary conditions. The sur-
charge load was applied to the top surface of the upper cell. All
boundaries of the lower cell (including the top boundary of the
Surcharge load q
lower cell) were fixed in both the x- and y-direction. All boundaries
of the upper cell were free in both the x- and y-direction. On the
Facing panel

Backfill soil
left side of the upper cell, prescribed displacements were applied

1.5 m
in the x-direction after surcharging. The concrete was modelled (a) Interface
as a linear elastic medium. The Mohr–Coulomb failure criterion y
was applied to the interface. Other details for FLAC and PLAXIS
x
simulations are given below:

 In FLAC (Fig. 1a), each cell was modelled by one zone. The zero-
5.0 m
thickness interface was located between the upper and lower
Surcharge load q
cells.
Connecon part

 In PLAXIS (Fig. 1b), each cell was modelled using two 15-node

Backfill soil
triangle elements. One 10-node interface element with zero Interfaces

1.5 m
thickness was located between the upper and lower cells.
y Steel strips
The interface normal stiffness and shear stiffness were first x
assumed in FLAC and the equivalent interface properties for PLAXIS
4.0 m 0.25 m
5.0 m
Backfill soil

(b)
1.35 m

Table 1
0.675 m

Property values for unit cells (Note: the property values for the soil are unconstrained Steel strips
0.1 m

because all boundaries of the bottom soil cell are fixed in x- and y-direction).
z
Parameter Value Programa x
Concrete
Unit weight, cconc (kN/m3) 0 F and P
Young’s modulus, Econc (GPa) 32.0 F and P
Poisson’s ratio, vconc (–) 0.15 F and P x =0
Interface y = 0.75 m
Interface friction angle, ui (°) 40.0 F and P
Adhesion, ci (kPa) 1.0 F and P
Dilation angle, wi (°) 0 F and P
Tension strength, rt,i (kPa) 0 F and P
x = 0.05 m
Young’s modulus, Ei (MPa) 0.82 P x=0 y = 0.745 m
(c)
Poisson’s ratio, vi (–) 0.45 P y = 0.75 m
Compression modulus, Eoed,i (MPa) 3.11 P
Shear modulus, Gi (MPa) 0.283 P
Virtual interface thickness, ti (m) 0.283 P Fig. 2. Schematic showing single precast concrete panel wall segment: (a) without
Normal stiffness, kn (MPa/m) 11.0 F steel strips, (b) with steel strips defined horizontally, and (c) with steel strips
Shear stiffness, ks (MPa/m) 1.0 F defined specifically at the connection (Note: the extension of interfaces between the
steel strips and backfill is applied only when using PLAXIS; x is the horizontal
a
F and P denote FLAC and PLAXIS computer programs, respectively. direction; y is the vertical direction; z is the out-of-plane direction).
 Y. Yu et al. / Computers and Geotechnics 65 (2015) 164–174 167

 Case 1 (Fig. 2a): the panel was fixed in the x-direction and the interfaces between the geogrid elements and backfill soil
bottom of the panel was fixed in the y-direction. The purpose (x = 0.05–4.0 m) were extended to x = 0.025 and 4.25 m to avoid
of Case 1 is to model the transfer of normal and shear stresses stress concentration near x = 0.05 and 4.0 m, respectively.
from the backfill soil to the facing panel without the
reinforcement. As shown in Table 2, all interfaces have the same interface prop-
 Case 2 (Fig. 2b): two 4 m long by 0.1 m wide steel reinforcement erty values equivalent to those for Ri = 0.3 applied to the backfill
layers (the thickness of each strip is 0.0023 m) were attached to soil in PLAXIS. It should be noted that Ri = 0.3 is generally lower
the back of each facing panel at y = 0.75 m. These dimensions than that commonly used for retaining walls with concrete facing.
correspond to a steel strip reinforced soil wall reported by Chida However this lower reduction factor can be justified to account for
and Nakagaki [19]. Today most of these steel strip reinforce- the effect of light compaction equipment that is recommended
ment products are narrower (e.g., 50 mm; [20]). However, qual- immediately behind the facing in current reinforced soil wall con-
itative comparisons are unaffected by the choice of steel strip struction practice. However, the general conclusions made in this
reinforcing elements in this range. The steel reinforcement is paper remain valid when the reduction factor for the interfaces
located horizontally in the backfill soil with x = 0–4.0 m and between the facing and backfill soil and between the steel rein-
y = 0.75 m. No restriction was applied to panel movement in forcement and backfill soil is set to other values (e.g., the com-
the x-direction other than the bottom of the facing was fixed monly used reduction factors are in the range Ri = 0.6–0.9).
in both x- and y-direction. Interface property values other than those listed in Table 2 are
 Case 3 (Fig. 2c): the only difference between Case 2 and Case 3 examined later and numerical outcomes are investigated in the
is that the steel reinforcement in Case 3 is located in the backfill corresponding sections. The equivalent interface properties for
soil by defining three points (point one at x = 0 and y = 0.75 m, FLAC were evaluated using Eqs. (15) and (16). The small strain
point two at x = 0.05 m and y = 0.745 m, and point three at mode was used in both programs. Uniformly distributed surcharge
x = 4.0 m and y = 0.745 m). load was applied to the top surface of the backfill soil at three dif-
ferent magnitudes (q = 10, 50 and 100 kPa). The backfill soil was
In both FLAC and PLAXIS, the facing panel is modelled using initially brought to equilibrium using K0 = 1  sin(usoil) = 0.305
beam elements and the Mohr–Coulomb model is applied to the for both programs. Parameter values used in both programs are
backfill soil. Element (or zone) size smaller than that used in this shown in Table 2. It should be noted that, when cable and strip ele-
study for both programs was shown to have only minor effect on ments in FLAC are used to model the steel reinforcement, their
the numerical results reported later. The details using FLAC are
provided below:
Table 2
 For Cases 1, 2 and 3, a total of 20 beam elements for the facing Property values for model with single precast concrete panel segment and single layer
panel and 2000 zones for the backfill soil were employed. The of soil reinforcement.
interface was applied between the facing beam elements and
Parameter Value Programa
backfill soil.
Concrete panel
 For Cases 2 and 3, three different types of structure elements
Unit weight, cconc (kN/m3) 24.0 F and P
were used (beam, cable and strip type) with a total of 80 ele- Young’s modulus, Econc (GPa) 32.0 F and P
ments defined using x- and y-coordinates to simulate the steel Poisson’s ratio, vconc (–) 0.15 P
strips. It should be noted that when using beam elements in Cross-sectional areab, Ap (m2) 0.18 F
FLAC, the extension of the interface is not necessary because Moment of inertiab, Ip (m4) 4.86  104 F
Axial stiffnessb, EconcAp (GN/m) 5.76 P
beam elements are defined using coordinates in this paper.
Bending stiffnessb, EconcIp (MN/m2/m) 15.6 P
 For Case 3 with beam elements for the reinforcement, no inter-
Backfill soil
face was applied between x = 0 and 0.05 m (the interfaces are
Unit weight, csoil (kN/m3) 18.0 F and P
applied on both sides of beam elements between 0.05 and Friction angle, usoil (°) 44.0 F and P
4.0 m). Cohesion, csoil (kPa) 1.0 F and P
Dilation angle, wsoil (°) 14.0 F and P
Tension strength, rt,soil (kPa) 0 F and P
In PLAXIS simulations, using a beam element with near zero
Young’s modulus, Esoil (MPa) 5.0 F and P
bending stiffness for the steel reinforcement is equivalent to the Poisson’s ratio, vsoil (–) 0.3 F and P
geogrid element which is used hereafter. The following are details
Steel reinforcement
using PLAXIS: Young’s modulus, Esteel (GPa) 200 F
Scaled cross-sectional areac, As (m2/m) 3.41  104 F
 For Cases 1, 2 and 3, the panel was modelled using 5-node beam Moment of inertia, Ip (m4) 0 F
elements (total of 8 elements) and the backfill soil was mod- Scaled axial stiffnessc, EsteelAs (MN/m) 68.2 P

elled using 15-node triangle elements (total of 202 elements). Interface

The 10-node interface elements with zero thickness were Friction angle, ui (°) 16.2 F and P
Cohesion, ci (kPa) 0.3 F and P
applied between facing beam elements and backfill soil
Dilation angle, wi (°) 0 F and P
elements. Tension strength, rt,i (kPa) 0 F and P
 For Case 2, the steel reinforcement is modelled using 5-node Young’s modulus, Ei (MPa) 0.502 P
geogrid elements with a total of 21 elements and the interfaces Poisson’s ratio, vi (–) 0.45 P
Compression modulus, Eoed,i (MPa) 1.90 P
between the geogrid elements and backfill soil were extended
Shear modulus, Gi (MPa) 0.173 P
to x = 4.25 m (the end of the reinforcement is at x = 4.0 m). Virtual interface thickness, ti (m) 0.0383 P
 For Case 3, one ‘‘anchor’’ was applied for the short connection Normal stiffness, kn (MPa/m) 49.7 F
portion of the steel reinforcement between x = 0 and 0.05 m Shear stiffness, ks (MPa/m) 4.51 F
(Fig. 2c; note that the anchor in PLAXIS only transfers load a
F and P denote FLAC and PLAXIS computer programs, respectively.
between two points). The remainder of the steel reinforcement b
Based on out-of-plane width of 1 m and the unit of the variable based on cor-
was modelled using 21 5-node geogrid elements (both anchor responding computer program manual.
c
and geogrid elements have the same axial stiffness). The The value is scaled to 1.0 m out-of-plane width for plane strain calculation.
168 Y. Yu et al. / Computers and Geotechnics 65 (2015) 164–174

Table 3 4. Results
Cable and strip property values in FLAC.

Parametera Value 4.1. Modelling of interfaces using unit cells

Cable element
Exposed perimeter, perimeter (m) 0.2 (1.3) Table 4 shows interface normal displacements and shear stres-
Cross-sectional area, area (m2) 2.3  104 ses under different applied surcharge loads and displacements. The
Grout stiffnessb, kbond (MN/m/m) 0.903 (6.09)
results from both FLAC and PLAXIS are compared with the analyt-
Grout cohesionc, sbond (kN/m) 0.06 (0.405)
Grout frictional resistance, sfriction (°) 16.2 ical solutions (Eqs. (2) and (3)). For the applied surcharge load
Spacing, spacing (m) 0.675 q = 10 kPa, the exact normal displacement at the interface (with
Out-of-plane stress component, szz (–) Off kn = 1.1  107 Pa/m) using Eq. (2) is un = rn/kn = 10,000/1.1/107
Strip element = 9.09  102 m = 0.909 mm. The maximum shear stress using Eq.
Calculation width, calwidth (m) 0.675 (1) is calculated to be ss,max = ci + rn tan ui = 1000 + 10,000 
Number of strips per calculation width, nstrips (–) 1 tan(40°) = 9.39  103 Pa = 9.39 kPa. Thus the exact shear stress at
Initial apparent friction coefficient, fstar0 (–) 0.3
the interface (with ks = 1.0  106 Pa/m) using Eq. (3) under the
Minimum apparent friction coefficient, fstar1 (–) 0.3
Strip/interface shear stiffnessd, strkbond (MN/m/m) 0.903 (6.09) applied shear displacement us = 5 mm is ss = ksus = 1.0  106 
Strip/interface cohesione, strsbond (kN/m) 0.06 (0.405) 5  103 = 5.0  103 Pa = 5.0 kPa (<9.39 kPa; the shear stress is at
Strip width, strwidth (m) 0.1 (0.675) the elastic state). For the applied shear displacement us = 10 mm,
Strip thicknessf, strthickness (m) 2.3  103 (3.41  10–4)
the exact shear stress is ss = ss,max = 9.39 kPa (due to ksus = 1.0 
a
Italicized parameter names are used in the FLAC manual. 106  10  103 = 10.0  103 Pa = 10.0 kPa > 9.39 kPa; the shear
b
kbond = ks  perimeter. stress is at the plastic state). The same procedure is used to evalu-
c
sbond = ci  perimeter. ate the normal displacements and shear stresses under other
d
strkbond = ks  2  strwidth.
e
strsbond = ci  2  strwidth.
applied surcharge loads and displacements. The numerical results
f
To keep the same cross-sectional area of 2.3  104 m2 (strwidth  strthickness) in Table 4 show that the calculated normal displacements and
and therefore the same axial stiffness. shear stresses from FLAC and PLAXIS analyses agree very well with
the analytical solutions.

property values are calculated from Table 2 based on the definition 4.2. Interface normal and shear stresses between facing and backfill
of these properties in FLAC (Table 3). soil for Case 1
In the simulations to follow the out-of-plane width of the rein-
forcement is 0.1 m and the total out-of-plane width modelled for Fig. 3a shows the normal stresses acting at the interface
the facing panel and backfill soil is 0.675 m (Fig. 2). For the rein- between the facing and backfill soil for Case 1 (without steel strips)
forcement using FLAC beam elements and PLAXIS geogrid ele- and three different surcharge pressures. For q = 10 kPa, the normal
ments, the modelled interface between the steel reinforcement stress from PLAXIS was 0.64 kPa at the top of the interface (com-
and backfill soil has an out-of-plane width of 0.675 m (this is the pared to 0.83 kPa using FLAC) and increased to 13.2 kPa at the
only choice for these two element types). When using FLAC cable bottom of the interface (compared to 13.0 kPa – FLAC). Increasing
elements and strip elements, the out-of-plane width of the inter- the surcharge to q = 50 kPa, the normal stresses at the top and
face can be less than 0.675 m. For example, the true width of the bottom of the interface using PLAXIS increased to 6.31 kPa
steel strip in this paper is 0.1 m corresponding to 15% area cover- (9.01 kPa – FLAC) and 32.3 kPa (32.1 kPa – FLAC), respectively.
age ratio. However, depending on the steel reinforcement product When the surcharge was q = 100 kPa, the normal stresses from
this coverage ratio could be as high as 50% for some steel bar mat PLAXIS were 13.5 kPa at top (19.7 kPa – FLAC) and 56.1 kPa at
and welded wire products ([20,21]). For the case of geosynthetic bottom of the interface (55.9 kPa – FLAC). The normal stresses from
sheet reinforcement products the coverage ratio is 100%. It should PLAXIS and FLAC are judged to be in generally good agreement. The
be noted that for cable and strip elements in FLAC, the interface- small visual differences in normal stresses near and at top of the
related properties are part of the cable and strip element proper- interface are due to the large plastic deformations in this region
ties. In the simulations to follow the above conditions apply unless that resulted in small differences in predicted normal displace-
noted otherwise. ments between programs.

Table 4
Numerical results for the soil–concrete interface response between unit cells.

Applied surcharge load (kPa) Applied displacement (mm) Soil–concrete interface (kn = 1.1  107 Pa/m, ks = 1.0  106 Pa/m)
Normal displacement (mm) Shear stress (kPa) Shear stress state
FLAC PLAXIS Analytical FLAC PLAXIS Analytical
q = 10 5 0.909 0.910 0.909 5.00 5.00 5.00 Elastic
10 9.39 9.40 9.39 Plastic
15 9.39 9.40 9.39 Plastic
20 9.39 9.40 9.39 Plastic
q = 50 20 4.55 4.55 4.55 20.0 20.0 20.0 Elastic
40 40.0 40.0 40.0 Elastic
60 43.0 43.0 43.0 Plastic
80 43.0 43.0 43.0 Plastic
q = 100 80 9.09 9.10 9.09 80.0 80.0 80.0 Elastic
100 84.9 85.0 84.9 Plastic
120 84.9 85.0 84.9 Plastic
140 84.9 85.0 84.9 Plastic
 Y. Yu et al. / Computers and Geotechnics 65 (2015) 164–174 169

1.5 (14.0 kPa from FLAC at y = 0.53 m). The slight difference for shear
PLAXIS stresses near and at y = 1.5 m between PLAXIS and FLAC was
FLAC
because the predicted normal stresses from both programs were
Case 1
slightly different in this area (Fig. 3a) which resulted in different
q = surcharge load computed maximum shear stress (Eq. (1)). The difference in shear
1.0 Esoil= 5 MPa
stresses using the two programs is greatest near the location of
Elevation (m)

kn = 49.7 MPa/m
ks = 4.51 MPa/m maximum shear stress, especially for q = 100 kPa. This is due to dif-
ferences in predicted shear displacements when slippage (interface
shear failure) occurred.
q = 10 kPa
0.5 q = 50 kPa q = 100 kPa

4.3. Reinforcement and facing panel axial loads for Case 2

The reinforcement axial loads for Case 2 with PLAXIS geogrid

0.0 elements and FLAC beam elements are shown in Fig. 4. Using PLAX-
0 10 20 30 40 50 60 70 80
IS, the reinforcement connection load was about 6.33 kN/m for the
Normal stress (kPa) applied surcharge load q = 10 kPa. It increased to 21.1 and 39.8 kN/
(a) m when increasing the surcharge load to q = 50 and 100 kPa,
respectively. For FLAC, the reinforcement connection loads were
1.5 6.55, 22.7, and 42.3 kN/m for surcharge loads q = 10, 50, and
PLAXIS 100 kPa, respectively. The results for the reinforcement axial load
FLAC (Case 2) from PLAXIS generally agreed well with those from FLAC.
Case 1 Fig. 4 also showed that the predicted reinforcement axial load
q = surcharge load decreased to near zero at the tail of the reinforcement in both
Esoil= 5 MPa
1.0
kn = 49.7 MPa/m PLAXIS and FLAC simulations. This must be the case at this bound-
Elevation (m)

q = 10 kPa ks = 4.51 MPa/m ary and thus serves as a check on the validity of numerical
outcomes.
Fig. 5 shows the facing panel axial loads for Case 2 with PLAXIS
q = 100 kPa
0.5
q = 50 kPa geogrid elements and FLAC beam elements. Recall that in Case 2
the reinforcement was located at y = 0.75 m (Fig. 2b) and the mesh
and reinforcement position in PLAXIS and FLAC were not updated
during calculations using the small strain option. Thus the rein-
forcement generates only horizontal tensile load. The down-drag
0.0 force (i.e., vertical load) on the facing panel from the reinforcement
0 5 10 15 20
is zero. However, a sharp increase in facing axial load at the rein-
Shear stress (kPa) forcement elevation was observed for all three surcharge loads
(b) (Fig. 5) using both programs. For example, using PLAXIS the facing
axial load at y = 0.75 m jumped from 4.08 to 4.76 kN/m when
Fig. 3. Load transfer from backfill soil to facing panel (Case 1): (a) normal stress and q = 10 kPa, from 5.74 to 8.72 kN/m when q = 50 kPa, and from
(b) shear stress.
7.80 to 13.6 kN/m when q = 100 kPa. These jumps in facing axial
load were not from reinforcement down-drag forces, but are the
The shear stresses on the interface between the facing and result of unbalanced vertical force between the upper and lower
backfill soil for Case 1 are shown in Fig. 3b. When the surcharge sides of the reinforcement as shown in Fig. 6. This is the result of
load was q = 10 kPa, the shear stress using PLAXIS was 0.48 kPa the two interface nodes (one above the reinforcement and the
at top of the interface (0.50 kPa – FLAC), increasing to a maximum other one below the reinforcement) sharing the same physical
value of 2.60 kPa at y = 0.84 m (2.57 kPa from FLAC at y = 0.83 m), position with the beam node at x = 0 and y = 0.75 m.
and thereafter decreasing to zero at bottom of the interface. For For Case 2 shown in Fig. 5, the calculated facing axial loads from
q = 50 kPa, the shear stress at y = 1.5 m was 2.13 kPa from PLAXIS PLAXIS generally agreed well with those from FLAC. The slight dif-
(2.80 kPa from FLAC) and the maximum shear stress was ference in axial loads for the facing panel below the reinforcement
8.47 kPa from PLAXIS at y = 0.52 m (7.94 kPa from FLAC at layer (y 6 0.75 m) was due to slightly different normal stress on
y = 0.53 m). Increasing the surcharge load to q = 100 kPa increased the upper side of the reinforcement at x = 0 and y = 0.75 m calcu-
the shear stress at y = 1.5 m to 4.21 kPa from PLAXIS (5.75 kPa from lated by the two programs. For example, when q = 100 kPa the nor-
FLAC) and the maximum shear stress to 15.1 kPa at y = 0.47 m mal stress on the upper side of the reinforcement at x = 0 and

50
Reinf. axial load (kN/m)

q = 100 kPa Case 2 PLAXIS (Geogrid)

40 FLAC (Beam)
q = surcharge load
30 Esoil = 5 MPa
q = 50 kPa kn = 49.7 MPa/m
20 ks = 4.51 MPa/m

10 q = 10 kPa

0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Distance from back of facing (m)

Fig. 4. Reinforcement axial loads (Case 2) using PLAXIS geogrid elements and FLAC beam elements.
170 Y. Yu et al. / Computers and Geotechnics 65 (2015) 164–174

1.5 brevity the results with cable elements are not reported in this
PLAXIS (Geogrid)
FLAC (Beam)
paper.
Fig. 7 shows the reinforcement axial loads for Case 2 using
Case 2
PLAXIS geogrid elements and FLAC strip elements. The reinforce-
q = surcharge load ment connection loads at different surcharge loads from FLAC with
1.0 Esoil= 5 MPa
strip elements were similar to those from FLAC with beam ele-
Elevation (m)

kn = 49.7 MPa/m
ks = 4.51 MPa/m ments (Fig. 4). Good agreement between calculated reinforcement
tensile loads using PLAXIS with geogrid elements and FLAC with
strip elements can be seen in Fig. 7.
0.5 q = 100 kPa The calculated facing panel axial loads for FLAC with strip ele-
ments are shown in Fig. 8 and compared with those from PLAXIS
with beam elements. The FLAC results show the facing panel axial
q = 50 kPa
load increasing gradually from top to bottom of the facing with no
q = 10 kPa jump. This confirms that the sudden change in the facing axial load
0.0
0 5 10 15 20 25 30 at y = 0.75 m (Figs. 5 and 8) was due to the unbalanced vertical
force at x = 0 and y = 0.75 m between the upper and lower inter-
Facing panel axial load (kN/m)
faces of the reinforcement when normal stiffness was applied
Fig. 5. Facing panel axial loads (Case 2) using PLAXIS geogrid elements and FLAC (Fig. 6).
beam elements.

4.5. Reinforcement and facing panel axial loads for Case 3

y = 0.75 m was about 380 kPa from PLAXIS and about 284 kPa from
FLAC as shown in Fig. 6. The modelling also showed that the calcu- The previous sections have shown that the calculated facing
lated interface normal stresses on the upper and lower sides of the axial loads can be different when using different structure ele-
reinforcement using PLAXIS were in generally good agreement ments in FLAC, and they are also different when using strip ele-
with those from FLAC (Fig. 6). The difference near the tail-end of ments in FLAC and geogrid elements in PLAXIS with the same
the reinforcement was because of the extended interface adopted out-of-plane width. Hereafter, the short connection segment
in PLAXIS to avoid stress concentration near the tail. between the reinforcement and facing panel shown Fig. 2c was
included in numerical simulations using both programs.
4.4. Reinforcement modelled by structure elements without normal Fig. 9 shows the reinforcement axial loads for Case 3 with PLAX-
stiffness for Case 2 IS geogrid elements and FLAC beam elements. For the surcharge
load condition q = 10 kPa, the reinforcement connection load was
The previous section used beam elements in FLAC to model the 6.52 kN/m and the reinforcement axial load gradually decreased
steel reinforcement. However, more often cable and strip structure to near zero at x = 4.0 m. The steel strip connection load was 22.3
elements are used in FLAC for this type of application [14]. In this and 42.2 kN/m when q = 50 and 100 kPa, respectively. The results
section, both cable and strip elements are assumed to have an show that the reinforcement axial loads predicted from both pro-
interface on each side of the reinforcement with out-of-plane grams are in very good agreement using geogrid and beam
width of 0.675 m as in the previous section (the influence of true elements.
out-of-plane width of 0.1 m for typical steel strip reinforcement For Case 3, the facing axial loads due to the interface shear
is examined later). It should be noted that the cable and strip ele- stresses from backfill soil and down-drag loads from the reinforce-
ments in FLAC only have shear stiffness (no normal stiffness is ment are shown in Fig. 10. Recall that the steel strips were mod-
specified) and the backfill soil can move through the plane of rein- elled using geogrid elements in PLAXIS and beam elements in
forcement without restriction when using cable and strip ele- FLAC. For the surcharge load q = 10 kPa, the axial load increased
ments. The results using cable and strip elements were the same from zero at y = 1.5 m to 4.09 kN/m at y = 0.75 m (just above the
for all cases and conditions examined in this paper and thus for reinforcement). The down-drag load (0.66 kN/m) from the steel

400
380 kPa (PLAXIS)
Case 2 PLAXIS (Geogrid)
FLAC (Beam)
300
284 kPa (FLAC) q = 100 kPa
Esoil= 5 MPa
Normal stress on the upper side
Interface normal stress (kPa)

200 kn = 49.7 MPa/m

Reinforcement ks = 4.51 MPa/m

100

0
Reinforcement
0

-100

-200 Reinforcement

Normal stress on the lower side

-300

-400
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Distance from back of facing (m)

Fig. 6. Interface normal stresses on the upper and lower sides of the reinforcement using PLAXIS geogrid elements and FLAC beam elements.
 Reinf. axial load (kN/m) Y. Yu et al. / Computers and Geotechnics 65 (2015) 164–174 171

50 1.5
q = 100 kPa Case 2 PLAXIS (Geogrid) PLAXIS (Geogrid)
40 FLAC (Strip) FLAC (Beam)
q = surcharge load
30 Esoil = 5 MPa
q = 50 kPa kn = 49.7 MPa/m
ks = 4.51 MPa/m Case 3
20 q = surcharge load
q = 10 kPa
1.0 Esoil= 5 MPa
10

Elevation (m)
kn = 49.7 MPa/m
0 ks = 4.51 MPa/m
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Distance from back of facing (m)

0.5 q = 100 kPa

Fig. 7. Reinforcement axial loads (Case 2) using PLAXIS geogrid elements and FLAC
strip elements.
q = 50 kPa
q = 10 kPa

0.0
1.5 0 5 10 15 20 25
PLAXIS (Geogrid) Facing panel axial load (kN/m)
FLAC (Strip)
Fig. 10. Facing panel axial loads (Case 3) using PLAXIS geogrid elements and FLAC
Case 2 beam elements.
q = surcharge load
1.0 Esoil= 5 MPa
Elevation (m)

kn = 49.7 MPa/m
ks = 4.51 MPa/m 4.6. Reinforcement modelled by structure elements without normal
stiffness for Case 3

0.5 q = 100 kPa The use of different structure elements in FLAC resulted in dif-
ferent facing axial loads for Case 2. In this section the effect of dif-
ferent structure elements on the reinforcement tensile loads and
facing axial loads for Case 3 are examined. The reinforcement axial
q = 10 kPa
q = 50 kPa
loads are shown in Fig. 11 for the reinforcement using strip ele-
0.0 ments in FLAC. Again very good agreement was observed between
0 5 10 15 20 25
the strip elements in FLAC (Fig. 11) and beam elements in FLAC
Facing panel axial load (kN/m) (Fig. 9) for reinforcement axial loads. When comparing the pre-
dicted reinforcement axial loads from FLAC (with strip elements)
Fig. 8. Facing panel axial loads (Case 2) using PLAXIS geogrid elements and FLAC with those from PLAXIS (with geogrid elements), the maximum
strip elements.
difference was within 2% for q = 100 kPa.
Fig. 12 shows the facing axial loads using strip elements in FLAC.
strips resulted in a jump in facing axial load from 4.09 to 4.75 kN/m Differences in facing axial loads are minor when using beam
at y = 0.75 m. The reinforcement connection segment in Fig. 2c has (Fig. 10) and strip (Fig. 12) elements in FLAC. When comparing the
a slope of 0.005 m/0.05 m = 1/10. Thus, based on the reinforcement predicted facing axial loads from FLAC (with strip elements) with
connection load of 6.52 kN/m at q = 10 kPa the down-drag force those from PLAXIS (with geogrid elements), the maximum differ-
from the reinforcement is 6.52  sin(tan1(1/10)) = 0.65 kN/m. ence was within 10% near y = 0.75 m for the surcharge load
This confirms that the sharp change in facing axial load is from q = 100 kPa. The modelling results show that for Case 3, the rein-
the down-drag load in the reinforcement due to the connection forcement tensile loads and facing axial loads between different
geometry in Fig. 2c. The facing panel axial load continued increas- structure elements and between different programs agree very well.
ing to 9.15 kN/m at bottom of the facing (y = 0). For q = 50 kPa, the
down-drag load from the reinforcement was about 2.27 kN/m and 4.7. Effect of backfill soil modulus for Case 3
the maximum facing axial load was about 15.9 kN/m at y = 0.
When the surcharge load was increased to q = 100 kPa, the Fig. 13 shows the effect of the backfill soil modulus value on the
down-drag load from the reinforcement increased to 4.28 kN/m reinforcement axial loads using FLAC and PLAXIS. Compared to the
and the maximum facing load increased to 24.3 kN/m. In conclu- results with Esoil = 5 MPa in Fig. 9, increasing the backfill soil mod-
sion, the results from both programs agreed very well when geo- ulus to Esoil = 50 MPa (Fig. 13) decreased the tensile loads in the
grid elements in PLAXIS and beam elements in FLAC were used reinforcement when other conditions remained the same (i.e.,
to model the reinforcement for Case 3. interface property values kept the same using the second option

50
Reinf. axial load (kN/m)

q = 100 kPa Case 3 PLAXIS (Geogrid)

40 FLAC (Beam)
q = surcharge load
30 Esoil= 5 MPa
q = 50 kPa kn = 49.7 MPa/m
20 ks = 4.51 MPa/m

10 q = 10 kPa

0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

Distance from back of facing (m)

Fig. 9. Reinforcement axial loads (Case 3) using PLAXIS geogrid elements and FLAC beam elements.
172 Y. Yu et al. / Computers and Geotechnics 65 (2015) 164–174

50

Reinf. axial load (kN/m)

q = 100 kPa Case 3 PLAXIS (Geogrid)
40 FLAC (Strip)
q = surcharge load
30 Esoil= 5 MPa
q = 50 kPa kn = 49.7 MPa/m
20 ks = 4.51 MPa/m

10 q = 10 kPa

0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

Distance from back of facing (m)

Fig. 11. Reinforcement axial loads (Case 3) using PLAXIS geogrid elements and FLAC strip elements.

1.5 1.5
PLAXIS (Geogrid) PLAXIS (Geogrid)
FLAC (Strip) FLAC (Beam)

Case 3 Case 3
q = surcharge load q = surcharge load
1.0 Esoil= 5 MPa 1.0 Esoil= 50 MPa

Elevation (m)
Elevation (m)

kn = 49.7 MPa/m kn = 49.7 MPa/m

ks = 4.51 MPa/m ks = 4.51 MPa/m

0.5 q = 100 kPa 0.5 q = 100 kPa

q = 10 kPa q = 50 kPa
q = 10 kPa q = 50 kPa

0.0 0.0
0 5 10 15 20 25 0 5 10 15 20 25

Facing panel axial load (kN/m) Facing panel axial load (kN/m)

Fig. 12. Facing panel axial loads (Case 3) using PLAXIS geogrid elements and FLAC Fig. 14. Facing panel axial loads (Case 3) using PLAXIS geogrid elements and FLAC
strip elements. beam elements with higher Young’s modulus of the backfill soil.

described in Section 2.3). The slight differences in tensile loads 4.8. Effect of interface stiffness for Case 3
occur for the beam elements in FLAC and geogrid elements in
PLAXIS when q = 50 and 100 kPa and are due to the small differ- The magnitude of the normal and shear loads transferred from
ences in predicted displacements. Thus the results show that the backfill soil to the reinforcement and facing panel is controlled
greater backfill modulus and larger surcharge load have more by the interface stiffness and shear strength. Fig. 15 shows the
effect on the differences in predicted reinforcement tensile loads effect of the interface stiffness on the reinforcement tensile loads.
between FLAC and PLAXIS than cases with lower values. When compared to the tensile loads of the reinforcement with
Fig. 14 shows the facing panel axial loads for the case of backfill kn = 49.7 MPa/m and ks = 4.51 MPa/m in Fig. 9, the increase in
soil modulus Esoil = 50 MPa. The higher modulus of the backfill soil interface stiffness (kn = 497 MPa/m and ks = 45.1 MPa/m) increased
resulted in lower facing axial loads when compared to those with the reinforcement tensile loads (Fig. 15). The data in Fig. 15 show
Esoil = 5 MPa in Fig. 10 and is attributed to less soil deformation due that the reinforcement tensile loads using FLAC agree very well
to the greater backfill soil modulus. The facing axial loads using with those using PLAXIS even for cases with higher interface
FLAC generally agreed well with those from PLAXIS. The slight stiffness.
differences in facing axial loads were because of small differences The effect of the interface stiffness on the facing axial loads is
in down-drag loads predicted using the two programs as shown in shown in Fig. 16. The higher interface stiffness (kn = 497 MPa/m
Fig. 13. and ks = 45.1 MPa/m) in Fig. 16 resulted in larger facing axial loads

50
Reinf. axial load (kN/m)

Case 3 PLAXIS (Geogrid)

40 q = 100 kPa FLAC (Beam)

q = surcharge load
30 Esoil= 50 MPa
q = 50 kPa kn = 49.7 MPa/m
20 ks = 4.51 MPa/m

10 q = 10 kPa

0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Distance from back of facing (m)

Fig. 13. Reinforcement axial loads (Case 3) using PLAXIS geogrid elements and FLAC beam elements with higher Young’s modulus of the backfill soil.
 Y. Yu et al. / Computers and Geotechnics 65 (2015) 164–174 173

50

Reinf. axial load (kN/m)

q = 100 kPa PLAXIS (Geogrid)
Case 3
40 FLAC (Beam)

30 q = surcharge load
q = 50 kPa Esoil= 5 MPa
20
kn = 497 MPa/m
10 q = 10 kPa ks = 45.1 MPa/m

0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

Distance from back of facing (m)

Fig. 15. Reinforcement axial loads (Case 3) using PLAXIS geogrid elements and FLAC beam elements with higher interface stiffness.

1.5 1.5
PLAXIS (Geogrid) PLAXIS (Geogrid)
FLAC (Beam) FLAC (Strip)

Case 3 Case 3
q = surcharge load q = surcharge load
1.0 Esoil= 5 MPa 1.0 Esoil= 5 MPa

Elevation (m)
kn = 49.7 MPa/m
Elevation (m)

kn = 497 MPa/m
ks = 45.1 MPa/m ks = 4.51 MPa/m

0.5 q = 100 kPa 0.5 q = 100 kPa

q = 50 kPa q = 50 kPa
q = 10 kPa q = 10 kPa
0.0 0.0
0 5 10 15 20 25 30 35 0 5 10 15 20 25

Facing panel axial load (kN/m) Facing panel axial load (kN/m)

Fig. 16. Facing panel axial loads (Case 3) using PLAXIS geogrid elements and FLAC Fig. 18. Facing panel axial loads (Case 3) using PLAXIS geogrid (out-of-plane
beam elements with higher interface stiffness. width = 0.675 m) and FLAC strip (out-of-plane width = 0.1 m).

when compared to results in Fig. 10 with kn = 49.7 MPa/m and discontinuous structures in the out-of-plane direction. In this
ks = 4.51 MPa/m. The predicted facing axial loads from FLAC agree section, the out-of-plane width of 0.1 m for the steel strip was
well with those from PLAXIS as shown in Fig. 16 for the higher modelled using strip elements.
interface stiffness. The reinforcement axial loads from FLAC with strip elements
are shown in Fig. 17. The FLAC results clearly show that, when
4.9. Effect of contact area between soil and reinforcement for Case 3 the 0.1-m wide steel strip was modelled, the reinforcement axial
loads were lower than those assumed using a 0.675-m wide
The results reported in previous sections for Case 3 are very steel strip in PLAXIS (for the same reinforcement axial stiffness
encouraging because the predictions for facing axial loads, rein- computed, i.e., Esteel  As). The reduced contact area between the
forcement axial loads are generally in very good agreement using 0.1-m wide steel strip and backfill soil was the main reason for
both programs. However, the out-of-plane width of the steel rein- the lower reinforcement axial loads when compared to the
forcement is 0.1 m (less than 0.675 m as noted earlier in the paper assumed 0.675-m wide steel strip in PLAXIS.
and the interface with out-of-plane width of 0.675 m was assumed Fig. 18 shows the facing panel axial loads using strip elements
in the previous sections for Cases 2 and 3). The geogrid elements in with the out-of-plane width of 0.1 m in FLAC. Predicted facing axial
PLAXIS and beam elements in FLAC with interfaces between the loads using PLAXIS with geogrid elements (out-of-plane width of
structure elements and backfill soil assume that the reinforcement 0.675 m), were similar to facing axial loads over the range
is continuous in the out-of-plane direction. However, the cable and y = 0.75–1.5 m but facing axial loads were visibly lower for y =
strip elements in FLAC can be used to model both continuous and 0–0.75 m using FLAC with strip elements (out-of-plane width of

50
Reinf. axial load (kN/m)

q = 100 kPa Case 3 PLAXIS (Geogrid)

40 FLAC (Strip)
q = surcharge load
30 Esoil= 5 MPa
q = 50 kPa kn = 49.7 MPa/m
20 ks = 4.51 MPa/m

10 q = 10 kPa

0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Distance from back of facing (m)

Fig. 17. Reinforcement axial loads (Case 3) using PLAXIS geogrid elements (out-of-plane width = 0.675 m) and FLAC strip elements (out-of-plane width = 0.1 m).
174 Y. Yu et al. / Computers and Geotechnics 65 (2015) 164–174

0.1 m). These differences increased with increasing surcharge load.  For the case of soil reinforcement materials which are discontin-
The differences in facing axial loads were due to lower reinforce- uous in the out-of-plane direction, program PLAXIS with geogrid
ment loads (Fig. 17) resulting in less down-drag forces using FLAC elements and program FLAC using beam elements use larger
with 0.1-m wide steel strip when compared to PLAXIS with interface area and therefore predict greater reinforcement axial
assumed 0.675-m wide steel strip. loads than program FLAC using cable and strip elements.

Despite the potential for different quantitative predictions

5. Conclusions depending on which program is used and which options and con-
stitutive models available in each program are adopted, both pro-
A number of commercial software programs are available to grams have been used to reproduce the measured performance of
geotechnical design engineers and researchers to predict the instrumented full-scale walls to acceptable accuracy by adjusting
behaviour of reinforced soil walls (MSE walls). Most programs soil parameter values within reasonable limits to improve agree-
are based on the finite element method. An example program is ment (e.g., [10–12,8]).
PLAXIS [17]. Another widely used software program is FLAC [16]
which is based on the finite difference method. The treatment of Acknowledgements
soil–facing interfaces and the inclusions used to model the rein-
forcing layers in MSE walls also vary between and within the pro- The work reported in this paper was supported by grants from
grams. Potential quantitative differences in numerical predictions the Natural Sciences and Engineering Research Council of Canada
for nominally identical wall cases using these two programs are (NSERC) awarded to the third author. The second author wishes
of interest to both designers and researchers. This is the motivation to acknowledge the support of the Universitat Politècnica de
for the work described in this paper. Catalunya (UPC-BarcelonaTech) and the funding received through
Numerical predictions using both programs were focused on the research project BIA2010-20789-C04-01 from the Ministry of
reinforced soil–facing panel interaction with equivalent interface Education and Innovation of Spain.
properties based on the Mohr–Coulomb failure criterion. The
numerical analyses using unit cells showed that for the same sur- References
charge loads and applied lateral displacement for the upper cell the
[1] AASHTO. LRFD bridge design specifications, American Association of State
predicted normal and shear stresses and normal displacements Highway and Transportation Officials (AASHTO), 6th ed., Washington, DC; 2012.
from both computer programs agreed very well. A MSE wall model [2] BS8006 2010. Code of practice for strengthened/reinforced soil and other fills.
segment with and without steel reinforcement was used to dem- UK: Milton Keynes, British Standards Institution (BSI).
[3] Cai Z, Bathurst RJ. Seismic response analysis of geosynthetic reinforced soil
onstrate how the reinforcement can be modelled in both programs segmental retaining walls by finite element method. Comput Geotech
using interfaces with zero thickness to capture soil–structure 1995;17(4):523–46.
interactions. Based on the cases and conditions examined, the fol- [4] Karpurapu RG, Bathurst RJ. Behaviour of geosynthetic reinforced soil retaining
walls using the finite element method. Comput Geotech 1995;17(3):279–99.
lowing conclusions can be made:
[5] Rowe RK, Ho SK. Continuous panel reinforced soil walls on rigid foundations. J
Geotech Geoenviron Eng 1997;123(10):912–20.
 The predicted normal and shear stresses between the facing panel [6] Yoo C, Jang YS, Park IJ. Internal stability of geosynthetic-reinforced soil walls in
tiered configuration. Geosynth Int 2011;18(2):74–83.
and backfill soil using FLAC generally agreed well with those from
[7] Damians IP, Bathurst RJ, Josa A, Lloret A, Albuquerque PJR. Vertical facing loads
PLAXIS (Case 1 shown in Fig. 2a). The slight differences in normal in steel reinforced soil walls. J Geotech Geoenviron Eng (ASCE)
and shear stresses between the two programs were due to very 2013;139(9):1419–32.
small differences in predicted plastic displacements. [8] Damians IP, Bathurst RJ, Josa A, Lloret A. Numerical analysis of an instrumented
steel reinforced soil wall. Int J Geomech (ASCE) 2014.
 Considering the steel reinforcement with out-of-plane width of [9] Hatami K, Bathurst RJ, Di Pietro P. Static response of reinforced soil retaining
0.675 m (Case 2 shown in Fig. 2b), the tensile loads of the rein- walls with non-uniform reinforcement. Int J Geomech 2001;1(4):477–506.
forcement agreed well between FLAC (beam, cable, strip ele- [10] Hatami K, Bathurst RJ. Development and verification of a numerical model for
the analysis of geosynthetic reinforced soil segmental walls under working
ments) and PLAXIS (geogrid elements). However, the facing stress conditions. Canad Geotech J 2005;42(4):1066–85.
panel axial loads using FLAC with cable and strip elements were [11] Hatami K, Bathurst RJ. Numerical model for reinforced soil segmental walls
different from those using FLAC with beam elements and PLAX- under surcharge loading. J Geotech Geoenviron Eng (ASCE) 2006;132(6):673–84.
[12] Huang B, Bathurst RJ, Hatami K. Numerical study of reinforced soil segmental
IS with geogrid elements. walls using three different constitutive soil models. J Geotech Geoenviron Eng
 When the steel reinforcement was assumed to be 0.675 m wide (ASCE) 2009;135(10):1486–98.
(Case 3 shown in Fig. 2c), both programs predicted similar [13] Huang B, Bathurst RJ, Hatami K, Allen TM. Influence of toe restraint on
reinforced soil segmental walls. Canad Geotech J 2010;47(8):885–904.
results for facing and reinforcement axial loads even though
[14] Abdelouhab A, Dias D, Freitag N. Numerical analysis of the behaviour of
the reinforcement was modelled using different structure ele- mechanically stabilized earth walls reinforced with different types of strips.
ments (geogrid in PLAXIS, and beam, cable, and strip in FLAC). Geotext Geomembr 2011;29:116–29.
[15] Damians IP, Bathurst RJ, Josa A, Lloret A. Numerical study of the influence of
 Increasing the backfill soil modulus reduced both the reinforce-
foundation compressibility and reinforcement stiffness on the behavior of
ment tensile loads and facing axial loads. Small differences in reinforced soil walls. Int J Geotech Eng 2014;8(3):247–59.
the reinforcement tensile loads and facing axial loads were [16] Itasca. FLAC: fast lagrangian analysis of continua. Version 7.0 [computer
observed between FLAC and PLAXIS results for larger backfill program]. Minneapolis, Minn: Itasca Consulting Group Inc; 2011.
[17] PLAXIS. Reference manual, 2D – version 9.0, PLAXIS. Delft, Netherlands: Delft
soil modulus values and larger surcharge loads (other condi- University of Technology; 2008.
tions being equal). [18] Ng PCF, Pyrah IC, Anderson WF. Assessment of three interface elements and
 Increasing the interface stiffness increased both the reinforce- modification of the interface element in CRISP90. Comput Geotech
1997;21(4):315–39.
ment tensile loads and facing axial loads. The numerical results [19] Chida S, Nakagaki M. Test and experiment on a full-scale model of reinforced
from both FLAC and PLAXIS agreed very well. earth wall. In: Proceedings of international conference on soil reinforcement,
 Modelling the true out-of-plane width of 0.1 m for the steel Paris; 1979, vol. II, pp. 533–8.
[20] Allen TM, Bathurst RJ, Holtz RD, Lee WF, Walters DL. A new working stress
strip with cable and strip elements in FLAC resulted in lower method for prediction of loads in steel reinforced soil walls. J Geotech
reinforcement axial load and less down-drag forces on the fac- Geoenviron Eng (ASCE) 2004;130(11):1109–20.
ing panel when compared to results from both FLAC and PLAXIS [21] Yu Y, Bathurst RJ. Analysis of soil-steel bar mat pullout models using a
statistical approach. J Geotech Geoenviron Eng (ASCE) 2015. http://dx.doi.org/
using the assumed out-of-plane width of 0.675 m (using the
10.1061/(ASCE)GT.1943-5606.0001281.
same reinforcement axial stiffness).