Vous êtes sur la page 1sur 10

78

Active pressure on gravity walls supporting purely


frictional soils
D. Loukidis and R. Salgado

Abstract: The active earth pressure used in the design of gravity walls is calculated based on the internal friction angle of
the retained soil or backfill. However, the friction angle of a soil changes during the deformation process. For drained load-
ing, the mobilized friction angle varies between the peak and critical-state friction angles, depending on the level of shear
strain in the retained soil. Consequently, there is not a single value of friction angle for the retained soil mass, and the active
earth pressure coefficient changes as the wall moves away from the backfill and plastic shear strains in the backfill increase.
In this paper, the finite element method is used to study the evolution of the active earth pressure behind a gravity retaining
wall, as well as the shear patterns developing in the backfill and foundation soil. The analyses relied on use of a two-surface
plasticity constitutive model for sands, which is based on critical-state soil mechanics.
Key words: finite elements, plasticity, retaining walls, sands.
Résumé : La pression active des terres utilisée dans la conception des murs gravitaires est calculée à partir de l’angle de
friction interne du sol ou du remblai retenu. Cependant, l’angle de friction d’un sol change durant le processus de déforma-
tion. Dans le cas d’un chargement drainé, l’angle de friction mobilisé varie entre l’angle de friction au pic et celui à l’état
critique, dépendant du niveau de déformation en cisaillement dans le sol retenu. En conséquence, il n’y a pas de valeur
unique d’angle de friction pour une masse de sol retenue, et le coefficient de pression active des terres varie à mesure que
le mur se sépare du remblai et que les déformations plastiques en cisaillement augmentent dans le remblai. Dans cet article,
la méthode par éléments finis est utilisée pour étudier l’évolution de la pression active des terres derrière un mur de soutène-
ment gravitaire, ainsi que les patrons de cisaillement qui se développent dans le remblai et dans le sol de fondation. Les ana-
lyses sont réalisées à l’aide d’un modèle constitutif de plasticité à deux surfaces pour des sables, qui est basé sur la
mécanique de l’état critique des sols.
Mots‐clés : éléments finis, plasticité, murs de soutènement, sables.
[Traduit par la Rédaction]

Introduction Caquot and Kerisel (1948) produced solutions in tabulated


form, assuming slip surfaces with logarithmic spiral shape.
The active earth pressure acting on the back of a retaining
More recently, Paik and Salgado (2003) estimated the active
wall controls its design. The active earth pressure is ex-
earth pressure behind rigid walls by improving the formula-
pressed as the product of the vertical effective stress s v0 in
tion of Handy (1985), which considers soil arching concepts.
the retained soil mass or backfill1 and the active earth pres-
Limit analysis has also been used to study the active earth
sure coefficient KA. The earliest and simplest methods for the
pressure problem. Rigorous upper bound values for KA estab-
calculation of the active earth pressure for purely frictional
lished by Chen (1975) and Soubra and Macuh (2002) using
backfills are those based on the Coulomb and Rankine theo-
limit analysis are in very close agreement with the values of Ca-
ries. For a backfill with horizontal surface, the Rankine solu-
quot and Kerisel (1948). Sokolovskiĭ (1965) solved the problem
tion is mathematically exact for a vertical and smooth wall
of active and passive earth pressure using the method of charac-
backface. Coulomb’s solution assumes a planar slip surface
teristics. More recently, Lancellotta (2002) provided a rigorous
and is equivalent to an upper bound solution. For a horizon-
lower-bound solution for active pressures in closed form:
tal backfill and a vertical wall backface, Coulomb’s solution

yields cosd
½2 KA ¼ ðcosd
1 þ sinf
½1 KA ¼
cos 2 f pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2  sin 2 f  sin 2 dÞ e½darcsin ðsind=sinfÞ tanf
cosdf1 þ ½sin ðf þ dÞ sinf=cosdg

Received 21 March 2011. Accepted 26 September 2011. Published at www.nrcresearchpress.com/cgj on 20 December 2011.
D. Loukidis. Department of Civil and Environmental Engineering, University of Cyprus, Nicosia 1678, Cyprus.
R. Salgado. School of Civil Engineering, Purdue University, W. Lafayette, IN 47907-1284, USA.
Corresponding author: D. Loukidis (e-mail: loukidis@ucy.ac.cy).
1The paper is not restricted to backfilled walls. To call attention to applicability of the discussion to walls supporting natural ground as

well as completely backfilled walls, the terms “retained soil mass” and “backfill” are used interchangeably throughout.

Can. Geotech. J. 49: 78–97 (2012) doi:10.1139/T11-087 Published by NRC Research Press
80 Can. Geotech. J. Vol. 49, 2012

Fig. 1. Evolution of coefficient of lateral earth pressure and total resisting horizontal force, Hr, and moment, Mr, and total destabilizing hor-
izontal force, Hd, and moment, Md, with displacement of wall crest.

the factor of safety (FS) would be to account for design un- responds to the WLS, since the resistances can’t increase at a
certainties and with serviceability limit states, which must be rate that matches the increase in driving actions, and thus any
handled separately. At early stages of wall movement away further increase in the driving actions leads to wall collapse.
from the backfill, the total horizontal resistance Hr and mo- If the system were left to respond on its own (i.e., without
ment resistance Mr increase at high rates, since the strains in the application of artificial external forces), the system would
the foundation soil are still small and the soil stiffness is have reached the stationary state and remained in it. It should
therefore large. At the same time, the driving horizontal force be noted that the wall is marginally stable (on the verge of
Hd and moment Md either decrease, as the earth pressure co- failure) with respect to one driving action but may be stable
efficient decreases from K0 to KA, min, or increase at small with respect to the other at the WLS.
rates after the earth pressure coefficient bottoms and starts to If FS > 1, the wall is stable, meaning that equilibrium
increase from KA, min to KA, cr (Fig. 1). The variations of these (Hd = Hr and Md = Mr) is reached before the stationary state
quantities with wall displacement can be written mathemati- or limit state is reached (Fig. 1). Artificial external forces
cally as dHr > dHd and dMr > dMd. Beyond a certain point would need to be applied to the wall–soil system to bring it
in the process (e.g., a certain amount of wall crest displace- to the limit state, which we define as identical to the station-
ment u), the resistance starts increasing at a lesser rate than ary state first reached by the wall (i.e., if, by the addition of
the driving action. This happens first for one of the two resis- external force, the Hd – Hr reaches its stationary state before
tances (Hr or Mr), so that this stage of the loading process is Md – Mr, then the limit state is defined by the horizontal
mathematically identified as the state at which either dHr < force, not moment). This is analogous to having a foundation
dHd or dMr < dMd for the first time. This stationary point in element (e.g., a footing or pile) supporting a vertical load Qd
the Hd – Hr (or Md – Mr) versus u curve (point F in Fig. 1, less than its limit bearing capacity. To bring the foundation
where dHr = dHd or dMr = dMd) corresponds to a stationary element to its bearing capacity ULS, we must apply an artifi-
state of the wall soil system. If the problem configuration cial external force Qext to the foundation element until it col-
(the combination of wall dimensions, soil weight, and soil lapses, which happens when the foundation resistance Qr
strength) is such that FS = 1, the point F lies on the horizon- attains its maximum value QL. At this stage, both the Qr ver-
tal u axis, i.e., Hd = Hr (or Md = Mr). So, for the case of a sus settlement curve and the Qext = Qr – Qd versus settlement
wall with FS = 1, it is obvious that the stationary point cor- curve reach stationary (maximum) points (since Qd is con-

Published by NRC Research Press


Loukidis and Salgado 81

stant). In the case of a retaining wall, the load to be sup- It is well known that analyses involving materials that
ported is not constant because of the dependence of KA on soften and follow a nonassociative flow rule suffer from the
wall displacement. Therefore, the WLS can only be identified problem of solution nonuniqueness. This means that, as the
by the stationary point of the Hr – Hd (or Mr – Md) versus u mesh gets refined, the FE analysis results change, and con-
curve and not of the Hr (or Mr) versus u curve. If FS < 1, vergence to a unique solution does not happen. To tackle
Hd – Hr (or Md – Mr) is always greater than zero and the this problem, FE analyses should either employ a regulariza-
wall is unstable, meaning that artificial external forces would tion approach (such as Cosserat or gradient plasticity) or use
need to be applied to the wall to establish equilibrium. In meshes with element sizes consistent with the known shear
other words, wall movement never manages to mobilize band thickness. The thickness of the soil elements inside the
enough resistance to balance driving forces. In this case, the shear bands simulating slippage between a rough structure
stationary point corresponds to the minimum value of these and granular soil is an important factor for the accurate pre-
external forces or, equivalently, the maximum value of Hr – diction of the shear resistance acting on the structure (Louki-
Hd (or Mr – Md). The WLS is therefore the state at which dis and Salgado 2008). Hence, the thickness of the backfill
the rate of increase (mobilization) of wall base resistance (in soil elements that are in contact with the wall backface is set
terms of either force or moment) becomes smaller than the equal to 5–20 times the mean particle diameter of the sand
rate of increase of the destabilizing actions (mainly the earth (D50). This is roughly the thickness of the shear bands that
thrust). form in sandy soils, as observed in a number of experimental
The goal of this paper is to study the evolution of KA with studies (e.g., Uesugi et al. 1988; Vardoulakis and Sulem
wall displacement and establish an indication of the KA value 1995; Nemat-Nasser and Okada 2001). Due to restrictions in
(and the appropriate f value for its calculation) at WLS con- memory allocation and analysis runtime, the element sizes in
ditions. For this purpose, we perform FE analyses of the re- other locations where shear bands are expected to develop (i.e.,
taining wall stability problem using a two-surface constitutive inside the sliding wedge and in the foundation soil) were
model for sands, which is based on CS soil mechanics. The larger than 5–20D50. Element size inside the sliding wedge
FE analyses are performed for a rigid gravity wall with a forming behind the wall was of the order of 500D50. As
rough vertical backface supporting soil with level surface shown later in the paper, this choice of the element size
and purely frictional nature. The analyses apply to the typical has only a small impact on the analysis accuracy.
case of backfilled walls but also to walls retaining natural
ground if sandy or gravelly in nature. Both the retained soil Constitutive model
and foundation soils are sands. The analyses do not simulate The constitutive model used in this study is the two-surface
the several complex stages involved in the construction of plasticity model based on CS soil mechanics developed
gravity walls (such as backfill laying and compaction), which originally by Manzari and Dafalias (1997). The model was
would lead to different initial stress conditions, but such is subsequently modified by Li and Dafalias (2000), Papadimi-
not the focus of the analyses, which aim instead at bringing triou and Bouckovalas (2002), Dafalias et al. (2004), and
out the details of the mechanics of wall loading not ad- Loukidis and Salgado (2009). The model parameters were
dressed in the literature and the implications and insights determined by Loukidis and Salgado (2009) for two sands:
that they offer. air-pluviated or dry-deposited Toyoura sand (Iwasaki et al.
1978; Fukushima and Tatsuoka 1984; Lam and Tatsuoka
FE methodology 1988; Yoshimine et al. 1998) and water-pluviated or slurry-
deposited clean Ottawa sand (Carraro et al. 2003; Carraro
FE mesh 2004; Murthy 2006; Murthy et al. 2007). Toyoura sand is
The analyses use unstructured meshes consisting of eight- a fine sand (D50 ≈ 0.2 mm) with angular to subangular par-
noded, plane-strain quadrilateral elements with 12 quadrature ticles, while Ottawa sand is a medium-sized sand (D50 ≈
points. A typical FE mesh is shown in Fig. 2. It includes the 0.4 mm) with rounded to subrounded particles. The model
wall, the backfill soil, and the foundation soil. The wall has considers four distinct surfaces having the form of open
a rectangular cross section, with width B and height H. The cones in stress space: the bounding surface, dilatancy sur-
thickness of the backfill soil layer is equal to H. The retain- face, CS surface, and yield (loading) surface. Bounding and
ing wall is embedded a distance D into the foundation soil. critical surfaces represent peak and CS shear strengths, re-
All analyses start from an ideal state of the retained soil, spectively. The dilatancy surface divides the stress space
reached without the wall having moved or rotated (as if the into two regions: inside the dilatancy surface, the soil plas-
backfill soil had been placed in one lift instantaneously). tic behavior is contractive; outside it, it is dilative. The
The wall is modeled as a block of linear elastic material yield (loading) surface defines a very narrow conical do-
with very large Young’s modulus so that it can be consid- main inside which the soil develops no plastic strain. The
ered rigid. yield surface hardens kinematically upon shearing, leading
No interface elements are placed between the soil and the to the development of plastic strains prior to failure.
wall; i.e., wall and soil share the same nodes along the corre- Through this feature, the model simulates accurately the be-
sponding contact planes. As a consequence, slippage between havior of the soil at small and large strains. The constitutive
the wall and backfill occurs due to the formation inside the model takes into account the inherent anisotropy of sands
soil mass of a shear band parallel to the wall backface. This through the use of a fabric tensor (Dafalias et al. 2004),
roughness condition is realistic given the rough materials and the assumption that the position of the CS line in the
commonly used for gravity walls, such as masonry, concrete, void ratio (e) – mean effective stress (p′) space depends on
and cribs containing stone. the direction of loading relative to the axis of sand deposi-

Published by NRC Research Press


Loukidis and Salgado 83

cause of the external application of a horizontal force Fext,0 = for a practical problem that is considerably complex once ex-
RC,0 + RT,0 and a moment Mext,0 = RC,0/H. The analysis pro- amined using rigorous mechanics.
ceeds by the application of outward horizontal displacement In most field cases, the active state will be mobilized grad-
increments DuC and DuT (i.e., displacements pointing away ually, and the wall base will translate and rotate as the back-
from the backfill) at nodes C and T (Fig. 3), while monitor- fill is constructed before reaching full height. Moreover, the
ing the values of external force Fext = RC + RT, where RC backfills placed behind gravity walls in practice are com-
and RT are external reactions, and external moment Mext = pacted, resulting in initial stress conditions in each layer
RC/H. Applying static equilibrium principles, the excess of larger than the K0 conditions assumed in this paper due to
the driving forces throughout the process of wall movement locked-in stresses (which are difficult to simulate, requiring
must be balanced by artificial external forces, which exist ex- three-dimensional FE analysis). These factors would generate
clusively for the purpose of performing the analysis. It can be different stress paths in the soil mass than those produced in
shown that Fext = Hd – Hr and Mext = Md – Mr, given that the our analyses. Problems involving materials that soften and
unbalanced forces at the end of each increment of the follow a nonassociative flow rule exhibit path dependence, i.e.,
Newton–Raphson solution are minimal (less than 1% of the the results depend on the stress paths followed at the stress
external forces). As the wall moves, Fext and Mext decrease integration points of the mesh. Hence, it is expected that
progressively. The wall is allowed to move vertically, since the results of these analyses would be somewhat different
no restraints are imposed on its nodes in the vertical direc- if the exact backfill construction process were simulated.
tion. The prescribed displacements uC and uT are not constant However, discrepancies due to wall motion during backfill
during the analysis and are not equal to each other. Their construction are believed to be small because most of the
magnitude varies in such a way that Fext and Mext (and conse- wall displacement will occur when the backfill height is
quently RC and RT) change in the same proportion. This is near the wall height, since the earth thrust increases at least
achieved by setting DuT = 0 (a pure rotation step) after any quadratically with the rate of backfill height, taking also
analysis increment, resulting in Mext/Mext,0 > Fext/Fext,0, and into account that the soil friction angle would decrease due
DuT = DuC (a pure translation step) after any analysis incre- to the increase in mean effective stress as the backfill rises.
ment, resulting in Mext/Mext,0 ≤ Fext/Fext,0. This scheme relies Discrepancies due to non K0 initial conditions would exist
on the fact that the wall rotation has a stronger effect on the mostly during the early stages of the predicted response, de-
rate of increase of Mext than of Fext, while wall translation has creasing as the active state were approached.
a stronger effect on the rate of increase of Fext than of Mext.
The increment DuC is always equal to a specified value of Results of FE simulations
the order of 10–6H. Hence, the analyses consist of alternating
phases of pure wall rotation and pure wall translation. Appli- Finite element analyses were performed for values of wall
cation of the loading in this manner, combined with the very width B, ranging from 1.5 to 3 m and wall height ranging
fine incrementation used in the present analyses, results in from 6 to 8 m. The sand unit weight g was set equal to
Mext/Mext,0 ≈ Fext/Fext,0 (both ratios thus denoted by the single 18 kN/m3. The wall unit weight was also set equal to 18 kN/m3,
variable Y) throughout the analysis. As a result, if the wall is which corresponds more closely to the unit weight of ma-
stable or marginally stable, Mext and Fext become equal to sonry, gabion, or a crib wall rather than a concrete wall.
zero simultaneously at which point the wall is completely un- The coefficient K0 was set equal to 0.5 in all analyses. No
supported by external reactions (which means that this be- surcharge is placed on the backfill free surface. The range
comes a point of equilibrium at which the wall comes to a of the wall dimensions was chosen such that the wall FS is
rest). In addition, Mext/Mext,0 and Fext/Fext,0 reach their mini- not excessively high or excessively low. As will be shown
mum value (Ymin) simultaneously, which happens when later, the FS of the wall configurations analyzed is in the
dMd = dMr and dHd = dHr. Therefore, referring to our pre- 0.5–2.0 range. Analyses are performed for Toyoura and Ot-
vious discussion of the WLS, Ymin is reached at the WLS. tawa sands, with relative density DR ranging from 30% to
It should be noted that there is an infinite number of load- 90%. For the sake of simplicity, the foundation soil is as-
ing path formulations that can bring the wall to a limit state, sumed to be of the same type and density as the backfill
and the formulation presented here is just one of them. The soil.
present wall loading formulation makes it possible to perform
displacement-controlled analyses, instead of load-controlled Collapse mechanism patterns
analyses that drive the wall to its limit state by increasing Most of the analyses were performed with the loading
the soil unit weight or a surcharge pressure. Displacement scheme described in the previous section, which subjects the
controlled analyses allow the wall to move beyond the limit wall to both rotation and horizontal translation in such a way
state all the way to CS (a requirement of this study) for all that the stabilizing external reactions RC and RT decrease pro-
possible outcomes (stable, marginally stable, and unstable portionally to each other. For comparison purposes, analyses
walls). In contrast, load-controlled analyses cannot proceed were performed with both the wall rotating about its heel
past the point of limit state. This is because any increase of without translating horizontally (pure rotation case) and
the applied load past this point results in unbalanced forces translating horizontally without rotating (pure translation).
that cannot be mitigated, since they increase with each Figure 4 shows contours of the incremental maximum shear
Newton–Raphson iteration. The formulation used here offers strain gmax (= ɛ1 – ɛ3, where ɛ1 and ɛ3 are the major and mi-
also simplicity, allowing clear understanding of the mechan- nor principal strains, respectively) from analyses with pure
ics involved and straightforward derivation of conclusions rotation, pure translation, and combination of rotation and

Published by NRC Research Press


90 Can. Geotech. J. Vol. 49, 2012

Fig. 11. Evolution of key problem variables, with increasing wall crest displacement for (a) loose and (b) dense Ottawa sand and wall with
H = 7 m, B = 1.5 m, and D = 0.5 m.

only at very large wall displacements (Figs. 10a and 11a). thrust increases (i.e., dHd > 0 and dMd > 0) with wall dis-
The WLS is reached when the slow base resistance develop- placement past the KA, min state. As a consequence, the wall
ment can no longer match the increase in destabilizing ac- displacement required to reach the WLS is smaller than that
tions due to the increase of KA towards KA, cr. The WLS can required to reach base failure.
be described mathematically as the state at which dHr = dHd Although WLS and base failure state occur at distinctively
or dMr = dMd; beyond the WLS, gains in resistance do not different wall crest displacements, the Hb and Mb values at
match gains in driving forces (i.e., dHr < dHd or dMr < WLS are practically identical to the peak Hb and Mb values.
dMd). If the active earth thrust were constant (dHd = 0 and This occurs because, after attainment of the WLS, Hb and Mb
dMd = 0), then the wall would keep moving until the bearing increase at very small rates towards their peak values due to
capacity were reached, and the WLS would coincide with the the development of regions of intense plastic straining in the
state of peak Hb or peak Mb (base failure state). In our retain- foundation soil. Consequently, the peak Hb and Mb values,
ing wall problem, this does not happen because the earth which can be determined in practice with relative ease based

Published by NRC Research Press


Loukidis and Salgado 91

Failure mode
on bearing capacity calculations, can be used as good ap-

Horizontal

Horizontal

Horizontal
Horizontal

Horizontal
Horizontal
Horizontal

Horizontal
Horizontal

Horizontal
proximations of the values of Hb and Mb at the WLS.

Moment

Moment
Moment

Moment
Moment
Moment

Moment

Moment
Table 2 summarizes the values of certain key variables of
interest in design at the state of minimum active pressure and
at the WLS: the normalized coefficient KA/K0, the mobilized
1.04
1.38
1.65
0.47
0.62
1.43
2.01
0.40
0.50
0.82
1.23
1.73
1.42
0.66
1.49
1.98
0.83
1.38
friction angle at the wall–backfill interface dmob, normalized
FS

crest displacement uC/H, and the relative height Deax/H of


the point of application of the active earth thrust from the
Deax/H
0.260
0.266
0.334
0.311
0.290
0.254
0.257
0.319
0.296
0.292
0.292
0.300
0.269
0.286
0.260
0.269
0.277
0.306
heel of the wall. Table 2 also contains the value of the FS
mobilized at the WLS. The reported FS is taken as the small-
est value of the FS against overturning FSM (based on mo-
0.021
0.016
0.007
0.163
0.032
0.020
0.019
0.160
0.054
0.025
0.015
0.013
0.021
0.018
0.023
0.020
0.021
0.015
uC/H

ment equilibrium about the wall toe) and “sliding” FSH


(based on horizontal equilibrium). The FS values are calcu-
lated using the following equations:
fc; TX (°)

Note: Rot+hor, rotation and horizontal; fc; TX , CS friction angle corresponding to triaxial compression conditions; feq , equivalent sand internal friction angle.
X
feq 

16.0

12.9
18.2

14.1
5.9
8.5
5.1
1.2
4.4
9.9

1.6
2.2
7.1

6.8
8.7
4.5
4.9
8.2
stabilizing forces Hb
½3a FSH ¼ X ¼
destabilizing forces E A; x
30.8
30.8
30.8
30.9
30.8
30.9
31.0
29.6
29.6
29.6
29.6
29.8
30.8
30.8
30.8
30.8
30.8
30.8
RC þ RT
dmob
(°)

¼1
EA; x
KA, LS/K0

X
0.383
0.344
0.396
0.461
0.407
0.326
0.247
0.483
0.472
0.391
0.306
0.240
0.370
0.342
0.405
0.399
0.349
0.270
WLS
Table 2. Summary of FEM results with respect to state of minimum lateral earth pressure coefficient and wall limit state.

stabilizing moments
½3b FSM ¼ X
overturning moments
Deax/H
0.337
0.267
0.358
0.319
0.325
0.342
0.336
0.328
0.324
0.327
0.330
0.329
0.335
0.332
0.340
0.337
0.344
0.354

WðB=2Þ þ MbO RC H
¼ ¼1
EA; x Deax  EA; y B EA; x Deax  EA; y B
0.004
0.016
0.001
0.006
0.003
0.004
0.004
0.010
0.006
0.006
0.006
0.006
0.003
0.003
0.003
0.003
0.003
0.005
uC/H

where EA, x and EA, y are the horizontal and vertical compo-
nents of the active earth thrust, W is the wall self-weight,
Deax is the height from the base of the point of application
fc; TX (°)

of the earth thrust, and MbO is the moment of the base resis-
feq 

15.1

13.0

20.2
26.3

12.7
18.0
23.7
14.7
14.4
14.9
15.1
15.0
20.4
8.6

4.2
9.6

3.1
7.8

tance taken about the toe of the wall (not to be confused with
Mb). The peak values of the mobilized FS values do not hap-
dmob (°)

pen at the WLS (Figs. 10 and 11). This is because the ex-
Minimum KA state

35.9
30.8
38.0
31.4
34.1
37.7
38.5
29.7
31.3
31.8
31.7
31.7
37.7
36.1
37.9
36.6
38.0
33.9

trema of the FSH and FSM depend on the denominators in


eqs. [3a] and [3b]. The minimum value Ymin would occur si-
multaneously with the peak mobilized FS and the peak total
KA, min/K0

base horizontal or moment resistance only if these denomina-


0.249
0.344
0.269
0.408
0.321
0.192
0.136
0.456
0.374
0.304
0.239
0.178
0.250
0.257
0.248
0.248
0.247
0.195

tors were constant during the analysis.


Figure 12 compares KA, min, KA, LS, and the inferred KA, cr val-
Toyoura
Toyoura
Toyoura
Toyoura
Toyoura
Toyoura
Toyoura

Toyoura
Toyoura
Toyoura
Toyoura
Toyoura
Toyoura

ues for analyses with rotating and translating walls. We see that
Ottawa
Ottawa
Ottawa
Ottawa
Ottawa
Sand

the KA, LS values lie approximately halfway between the mini-


mum and CS values of KA. The value of KA, LS is 25%–80%
greater than KA, min, with the differences increasing with increas-
D (m)

ing relative density. Figure 12a also plots results from analyses
0.5
0.5
0.5
0.5
0.5
0.5
0.5
0.5
0.5
0.5
0.5
0.5
0.5
0.5
0.5
0.5
0.2
0.5

Nos. 13–17 (Table 2), which have different wall dimensions (H,
B) and embedment D but the same sand relative density (DR =
DR (%)

75+45

60%). Figure 12a indicates that the wall dimensions and embed-
60
60
60
30
45
75
90
30
45
60
75
90
60
60
60
60
60

ment have a more pronounced effect on KA, LS than KA, min.


W (m)

Wall displacement to reach characteristic states


1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
2.0
2.5
1.5
1.5

The ratio uC/uT, resulting from the loading scheme adopted


for the analyses of walls moving in both rotation and transla-
H (m)

tion, is in the 3–8 range for MPS. Beyond the MPS, uC/uT
7
7
7
7
7
7
7
7
7
7
7
7
6
8
7
7
7
7

increases a further 20%–40% by the time the WLS is


reached, remaining practically constant for the remainder of
Horizontal
Rot+hor

Rot+hor
Rot+hor
Rot+hor
Rot+hor
Rot+hor
Rot+hor
Rot+hor
Rot+hor
Rot+hor
Rot+hor
Rot+hor
Rot+hor
Rot+hor
Rot+hor
Rot+hor
Rotation
Loading

the analysis, with the exception of the simulation for DR =


mode

90% for which uC/uT can reach values in the 10–12 range.
In the analyses in which the wall is allowed to rotate and
translate, the crest displacement required for reaching the
No.

10
11
12
13
14
15
16
17
18

MPS is in the 0.003H–0.006H range (Fig. 13), with the ex-


1
2
3
4
5
6
7
8
9

Published by NRC Research Press


92 Can. Geotech. J. Vol. 49, 2012

Fig. 12. Minimum, limit state, and critical active earth pressure Fig. 13. Wall crest displacement required to reach MPS (uCp) and
coefficient from analyses with (a) Toyoura sand and (b) Ottawa sand. WLS (uC, LS) from analyses with Toyoura and Ottawa sands. Analy-
sis numbers follow the numbering shown in Table 2.

horizontal and vertical directions, in contrast with laboratory


tests. In Fig. 9, we see that, as expected, the initial, roughly
linear, response prior to MPS is much stiffer for dense sand
than for loose sand. As K approaches KA, min, the response
becomes nonlinear and is smoother for large DR values than
for small ones. As a consequence, KA, min for dense sands is
reached at roughly the same wall displacement as for loose
and medium dense sands. This is a consequence of the pro-
gressive failure developing prior to the attainment MPS. The
progressive failure is evident by the fact that the mobilized
friction angle on the wall–soil interface is very close to the
CS value by the time MPS is reached, as demonstrated later
in the paper. Progressive failure is more intense in dense than
in loose sands, counterbalancing the effect of sand stiffness
ception of Ottawa sand, with DR = 30%. According to the on the wall displacement required to reach MPS.
Canadian Geotechnical Society (1992), the uC required for The crest displacement uC,LS required to reach WLS is in
developing the active earth pressure corresponding to peak the range of 0.013H–0.026H, except for DR = 30%–45%. In
soil strength (i.e., uCp) is 0.001H for dense and 0.004H for addition, for a given sand and relative density, the ratio uC,LS/
loose sand. Our results are of the same order of magnitude uCp increases with increasing FS. The uC,LS lies in the range
but don’t exactly match the values proposed by the Canadian of displacements for which the wall foundation has not yet
Geotechnical Society (1992). In addition, our analyses sug- collapsed (i.e., the peak base resistance has not yet been
gest that the relative density has no significant effect on uCp reached) but is very compliant, yielding considerably for
unless the tendency for strain softening is very weak to prac- even small changes in foundation load.
tically absent (e.g., DR = 30%). One would expect that uCp In most of the analyses with a rotating and translating wall,
should decrease with increasing relative density as the failure the crest displacement required to reach the peak Hb is in the
strain decreases with increasing relative density in laboratory 0.01H–0.09H range (corresponding to toe displacement of
tests, which in turn is due to the fact that the sands stiffness 0.01B–0.065B). The mobilization of the full horizontal base
increases with relative density. However, in the present prob- capacity requires displacements that exceed those required to
lem, the stress and deformation conditions vary in both the reach the WLS in the retained soil by 10%–150%. The peak

Published by NRC Research Press


Loukidis and Salgado 93

moment base resistance requires even larger wall movement The variation of the earth pressure coefficient K with uC
in all analyses performed in this study. can be described mathematically by the following equation:

f½ðc  1ÞðK0  KA; cr Þ=ðK0  KA; min ÞguCp  cuC


½4 K¼ þ KA; cr
½ðc  1ÞuCp =ðK0  KA; min Þ þ uC c =ðKA; cr  KA; min ÞuCp c1

The input parameters in eq. [4] are the characteristic values wall backface forms well before the main shear band, the mo-
for the earth pressure coefficient, K0, KA, cr, and KA, min, the bilized frictional angle dmob at MPS is clearly larger than dc
crest displacement uCp at which KA, min is attained, and a fit- (Fig. 15) for both rotating and translating walls and purely
ting parameter c. According to eq. [4], K is equal to K0 for translating walls. The ratio dmob/dc ranges from 1.0 to 1.07
uC = 0 and tends to KA, cr asymptotically for large values of for Ottawa sand, while for Toyoura sand, it ranges from 1.02
uC. The parameter c is introduced to control the rate at which to 1.25 (corresponding to dmob from 31.4° to 38.5°). The uCp
K increases towards KA, cr after the attainment of KA, min. By for a purely rotating wall is so large that, by the time MPS is
fitting eq. [4] to the results of this study, shown in Figs. 8 reached, dmob has become equal to dc.
and 9, we obtain c in the 1.7–2.1 range, with an average
value of two. An estimate of the KA, LS to use for wall design Equivalent value of sand friction angle for calculation of
calculations could be obtained using eq. [4], with c = 2.0, KA at limit state
uC = 0.025H for medium dense and dense sands, and 0.06H It is of practical interest to assess what the appropriate
for loose sands. (equivalent) value of the sand internal friction angle feq is for
use in the calculation of KA, min and KA, LS using an analytical
Point of application of lateral earth thrust formula widely used in practice, such as Coulomb’s solution
An important parameter for retaining wall stability calcula- (eq. [1]), to obtain a value of KA, LS that is in agreement with
tions is the location along the wall height at which the lateral the present numerical simulations. Figure 16 shows the differ-
earth thrust EA, x acts. Figure 14 shows the values of the ver- ence between feq and the CS friction angle fc; TX corresponding
tical distance Deax of the point of application of EA, x from the to triaxial compression conditions. The feq values are back-
wall base obtained from the analyses of a wall that both ro- calculated using eq. [1] from the KA values resulting from
tates and translates. We see that, for the MPS, Deax is roughly the FE analyses. We consider fc; TX instead fc; PS because it
equal to the widely used value of (1/3)H. Beyond that state, is easier to estimate it through either empirical relationships
Deax starts decreasing, reaching a minimum value almost or a few triaxial compression tests. Even the frequently per-
coincidentally with the WLS. This decrease is negligible for formed shear box tests would yield fc estimates that are
loose sand, but it can be up to 25% for dense sand. The closer to fc; TX than fc; PS . According to Fig. 16a, the feq
vertical distance Deax subsequently increases but at a very for MPS is 3°–26° larger than fc; TX , depending on the value
small rate. of the sand relative density. However, for calculating KA, LS,
The same trends of Deax with increasing wall displacement feq is only 1°–18° larger than fc; TX (Fig. 16b). Figure 16b
were observed in the experiments of Fang and Ishibashi shows that existing walls are not necessarily poorly designed,
(1986). Specifically, for a purely rotating wall, Deax starts de- even if the design is based on the prevailing practice of as-
creasing from an initial value of 0.333H towards a minimum suming the soil to be perfectly plastic with a peak value of
value of 0.22H and then rises slowly (but never exceeding (triaxial compression) friction angle to calculate KA: practi-
0.28H, even at uC = 0.008H). Similar trends are found in tioners would rarely use friction angles exceeding fc; TX by
Ichihara and Matsuzawa (1973). Fang and Ishibashi (1986) more than 15° for a dense sand or more than 2° for a loose
also present data that supports the fact that Deax for purely sand. So, whether by accident or proper intuition and judg-
rotating walls is less than 0.3H, while it is around H/3 if the ment by engineers working on this problem years ago, stand-
motion has a translational component. ard practice uses friction angles that are roughly consistent
with WLS rather than the state of mobilization of peak
Mobilized resistance along wall–backfill interface strength in the backfill. Figure 17 shows the difference be-
The mobilized friction coefficient m (= tand) on the wall tween feq and the CS friction angle fc; PS corresponding to
face reaches a peak value at very early stages of the analyses, plane-strain conditions. The values plotted in Fig. 17 are
before the attainment of the MPS, suggesting the vertical about 4.5°–5° smaller than those in Fig. 16. In Fig. 17b, we
shear band along the wall backface forms well before the see that, for loose sand, feq for KA, LS is smaller than fc; PS .
main shear band delimiting the sliding wedge. This peak At first sight, this would seem to be a violation of the basic
value is strongly dependent on the relative density of the principle of soil mechanics that the minimum value of the
sand (Figs. 10 and 11). After the peak, m decreases quickly sand friction angle is that for CS, but all analytical methods
towards a residual value that is consistent with the develop- for calculating KA presented in the introduction produce re-
ment of CS (f ¼ fc; PS , j = 0) inside the thin backfill soil sults that are valid for an associated flow rule (f = j) and,
elements that are in contact with the wall. The residual values thus, underestimate the actual KA by roughly 20% (see Table 1).
for the angle d are 30.8° and 29.6° for Toyoura sand and Ot- On the other hand, the FE simulations discussed in this paper
tawa sand, respectively. These are consistent with the theoret- use a model that captures the sand dilatancy realistically. Had
ical values calculated as dc = arctan(sinfc; PS ) (Loukidis and we had a formula that predicted KA for realistic j values, all
Salgado 2008). Although the vertical shear band along the resulting feq  fc; PS values would have been positive.

Published by NRC Research Press


94 Can. Geotech. J. Vol. 49, 2012

Fig. 14. Vertical distance of point of application of active earth thrust Fig. 15. Ratio of d mobilized along wall–backfill interface at mini-
from wall base at (a) minimum earth pressure state and (b) WLS. mum earth pressure state to d corresponding to CS conditions.

4. The limit state KA (KA, LS) lies between KA, min and the
corresponding CS value KA, cr. The KA, LS values are lar-
ger than KA, min by a factor of 1.1–1.8, with the differ-
ences increasing with increasing relative density.
5. The WLS does not necessarily coincide with the mobiliza-
tion of the maximum base resistance, which may require
much larger wall displacements.
6. The height of application of the lateral earth thrust at limit
state conditions is less than one-third, ranging from 0.25
to 0.32, suggesting that the current design practice is
slightly conservative.
7. The equivalent friction angle to be used for the calculation of
KA values consistent with WLS design can be up to 18°
higher than the soil CS friction angle under triaxial
compression conditions.
The results regarding the WLS depend on the base stiff-
ness and strength. Hence, our findings regarding the WLS
are strictly applicable to foundation soils that are like the re-
tained soil (i.e., purely frictional soils). It is expected that, for
walls founded on stiff clay or weak rock, the WLS may coin-
Conclusions cide with or even precede the MPS and the WLS, given the
high stiffness and brittleness of such geomaterials. Although
This paper presents the results of a set of FE analyses of a in all analyses the foundation soil had the same relative den-
gravity wall subjected to the action of a mass of sand that the sity as the backfill, the results are expected to hold even for
wall initially retains at a state of rest. After the wall is al- cases in which the relative densities are different. This is be-
lowed to move due to the action of the retained soil mass, cause the displacement required to reach KA, min is practically
the pressures on the wall evolve towards active pressures. independent of the density of the backfill (Fig. 9). In addi-
Based on the FE results and focusing on walls that are free tion, we see in Fig. 9 that the shape of the curves is the
to rotate and translate, we arrive at the following conclusions: same for all densities. Hence, what matters regarding the dis-
1. The attainment of the minimum value of the active earth placement required to attain WLS is the density of the foun-
pressure coefficient (KA, min) requires wall crest displace- dation soil. Therefore, the displacement needed to attain
ments of the order of 0.001H–0.010H. WLS for a loose backfill will not be much different from
2. The attainment of KA, min corresponds to a mobilized wall– that for a dense backfill as long as the density of the founda-
soil interface friction angle d that is larger than the CS tion soil is the same.
interface friction angle dc by a factor ranging from 1.0 From a practical standpoint, this study suggests that the
(loose sand) to 1.3 (dense sand). minimum active earth pressure state is of limited relevance
3. For dense and medium dense sand, the crest displacement to ULS design, since it happens for wall crest displacements
required to bring the wall to its limit state is in the of the order of only 0.5% the wall height; it is possibly repre-
0.013H–0.026H range. At the WLS, the mobilized inter- sentative of a serviceability limit state (SLS). Given that the
face friction angle has already reached the CS value of dc. active earth pressure coefficient is a function of the wall dis-

Published by NRC Research Press


Loukidis and Salgado 97

Appendix A
Table A1. Values of constitutive model parameters for Toyoura and clean Ottawa sands.

Parameter value
Parameter Clean
Parameter symbol Toyoura sand Ottawa sand Comment
Small-strain (“elastic”) n 0.15 0.15* Poisson’s ratio
parameters Cg 900 611 Parameter controlling the magnitude of the maximum shear
modulus Gmax
ng 0.400 0.437 Exponent controlling the rate of increase of Gmax with effec-
tive confining stress
g1 0.0010 0.000 65 Parameters controlling the decrease of “para-elastic” shear
a1 0.40 0.47 modulus G with shear strain
CS Gc 0.934 0.780 Intercept of CS line in e–p′ space
l 0.019 0.081 Parameter controlling inclination of CS line in e–p′ space
x 0.70 0.196 Parameter controlling curvature of CS line in e–p′ space
Mcc 1.27 1.21 Critical stress ratio in triaxial (TX) compression conditions
Bounding surface kb 1.5 1.9 Parameter controlling the increase of friction angle with
sand density
Dilatancy Do 0.90 1.31 Inclination of the stress–dilatancy curve
kd 2.8 2.2 Parameter controlling the stress ratio at phase transformation
Plastic modulus h1 1.62 2.20 Parameters controlling the magnitude of plastic modulus
h2 0.254 0.240
elim 1.00 0.81 Upper limit for void ratio for which the plastic modulus be-
comes zero
m 2.0 1.2 Parameter controlling stress ratio in undrained instability
state
Stress-induced c1 0.72 0.71 Ratio of the critical stress ratio in TX extension to that in
anisotropy TX compression
c2 0.78 0.78* Parameter controlling the value of the magnitude of inter-
mediate principal stress relative to the two other principal
stresses under plain-strain conditions
ns 0.35 0.35* Parameter controlling the magnitude of the friction angle in
plane-strain conditions relative to the friction angle in TX
compression
Inherent anisotropy a 0.29 0.31 Parameter controlling the intercept of CS line in e–p′ space
under conditions other than TX compression
kh 0.11 0.39 Parameter controlling the variation of plastic stiffness, with
the direction of loading relative to the axis of sand de-
position
m 0.05 0.05 Radius of conical yield (loading) surface
*Assumed.

Published by NRC Research Press

Vous aimerez peut-être aussi