Vous êtes sur la page 1sur 15

Powrie, W. (1996). GeÂotechnique 46, No.

4, 709±723

Limit equilibrium analysis of embedded retaining walls

W. P OW R I E 

Embedded retaining walls are often designed on Les murs de souteÁnement encastreÂs sont habi-
the basis of limit equilibrium stress distribu- tuellement concËus sur la base de la distribution
tions. Although this type of analysis has been des contraintes aÁ l'eÂquilibre limite. Ce type
used widely for many decades, there is con- d'analyse qui a eÂte utilise durant depuis
siderable scope for confusion and disagreement plusieurs deÂcennies, induit de nombreuses er-
concerning the values of soil strength and soil± reurs et confusions concernant le choix des
wall friction which should be adopted, and the valeurs de reÂsistance du sol et de frottement
way in which a factor of safety should be mur-sol, ainsi que dans le choix d'un facteur de
applied. In this paper these issues are investi- seÂcurite correct. Dans cet article, ces probleÁmes
gated with reference to previously-published seront examineÂs graÃce aÁ des essais en labora-
laboratory and centrifuge model test results, toire de modeÁles centrifugeÂs et d'observations
and the observed behaviour of real walls at in-situ du comportement de murs sur trois sites
three sites. It is shown that, for embedded diffeÂrents. Il a eÂte observe que pour les murs de
cantilever retaining walls which are either souteÁnement cantilever encastreÂs, qu'ils soient
unpropped or propped near the crest, the onset pourvu ou non d'une chaise en teÃte, l'amorce de
of large deformations is reasonably well pre- grandes deÂformations est assez bien preÂvue par
dicted by classical limit equilibrium stress la distribution des contraintes aÁ l'eÂquilibre
distributions, using earth pressure coef®cients limite, en utilisant les coef®cients de pression
based on critical state angles of soil friction with des terres, avec un angle de frottement du sol
wall friction angle ä ˆ ö9 ˆ ö9crit on both sides aÁ l'eÂtat critique et en prenant un angle de
of the wall. The case data also tend to con®rm frottement sur le mur ä ˆ ö9 ˆ ö9crit sur les
that the procedures given in BS 8002 should deux coteÂs du mur. L'eÂtude de ces cas tend a
lead to depths of wall embedment which are con®rmer que les proceÂdures donneÂes dans le
suf®cient to guard against excessive movements BS 8002 doivent donner des profondeurs d'en-
under working conditions, provided that the foncement des murs qui sont suf®santes pour se
most onerous geometrical, loading and ground- proteÂger des mouvements excessifs lors de la
water conditions are correctly identi®ed, and phase de travaux, pour vu que la geÂomeÂtrie la
that the soil strength is selected with care. plus couÃteuse, le chargement et les conditions
hydrogeÂologiques soient correctement identi®eÂs
KEYWORDS: case history; diaphragm and in situ et que la reÂsistance du sol soit choisie avec soin.
walls; failure; limit state design/analysis; retaining
walls; sheet piles and cofferdams.

INTRODUCTION level, butÐif the retained height is modestÐwalls


Embedded retaining walls are frequently used to are often unpropped and act as free cantilevers.
retain the sides of road and railway cuttings, and to Unpropped embedded retaining walls rely en-
form temporary cofferdams to enable the construc- tirely on the passive resistance of the soil in front
tion of underground structures. Temporary coffer- of the wall below formation level to counter the
dams are usually formed of interlocking driven overturning effect of the lateral stresses in the
steel sheet piles, while more permanent walls may retained ground. Assuming that a structural failure
be constructed in situ using reinforced concrete. of the wall does not occur, the collapse mechanism
Walls may be supported by props at or near crest for an unpropped wall would be expected to
correspond to rigid body rotation about a hori-
zontal axis within the plane of the wall at some
depth zp below formation level.
Manuscript received 15 September 1994; revised manu-
script accepted 25 August 1995. A lower bound stress distribution for an
Discussion on this paper closes 3 March 1997; for further unpropped embedded cantilever wall on the verge
details see p. ii. of collapse may be determined by introducing
 University of Southampton. frictionless stress discontinuities running vertically

709
710 POWRIE

on both sides of the wall and horizontally through able wall in a uniform soil (following Symons,
the pivot and the toe. At failure, the lateral 1983) is shown in Fig. 1(b).
effective stresses in zones where the wall is Two unknowns must be calculated from the
moving away from the soil are at the active limit stress distributions shown in Figs 1(a) and 1(b),
using the conditions of horizontal and moment
ó9h ˆ K a ó9v ; K a ˆ (1 ÿ sin ö9)=(1 ‡ sin ö9) equilibrium. These are the depth of embedment d
(1) required for limiting equilibrium, and the depth to
the pivot zp . If steady-state pore water pressures
where ó9h and ó9v are the horizontal and vertical are represented as shown in Fig. 1(b), the
effective stresses, Ka is the active earth pressure equilibrium equations for an unpropped wall are
coef®cient and ö9 is the soil angle of shearing simultaneous and quartic in the two unknowns. For
resistance, and in zones where the wall is moving this reason, an approximate effective stress dis-
into the soil they are at the passive limit tribution is often used in which the lateral stresses
below the level of the pivot are replaced by a point
ó9h ˆ K p ó9v ; K p ˆ (1 ‡ sin ö9)=(1 ÿ sin ö9) force Q (Fig. 1(c)). The two unknowns in this case
(2) are the force Q and the depth zp , and the
equilibrium equations can be uncoupled by taking
where Kp is the passive earth pressure coef®cient. moments about the pivot. The calculated value of
The limiting lateral effective stress distribution for zp is conventionally increased by 20% to estimate
an unpropped wall is shown in Fig. 1(a). Pore water the required depth of embedment. It should be
pressures, which might correspond to steady-state checked that this additional depth of embedment is
seepage around the wall from a high groundwater actually suf®cient to generate the calculated value
level on the retained side into drains below the of Q, under the action of passive pressures behind
formation, must be considered separately. A pore the wall and active pressures in front (Pad®eld &
water pressure distribution that approximates to Mair, 1984).
steady-state seepage conditions around an imperme- For most real walls, the stress analysis indicated

Wall rotation h=h

Datum for h
h h=0
measurement
Active σ'h = Kaσ'v of excess
Passive head h
σ'h = Kpσ'v d
d zp
Pivot

Active Passive d
Frictionless
(
h = h h + 2d ) u = 2γ d (hh++2dd)
w

stress (b)
discontinuities
for true lower
bound
(a)

h
h

Active

Passive d*
zp
d

Q
(c) (d)

Fig. 1. Idealized stress distributions for unpropped embedded cantilever retaining wall: (a) limiting lateral
effective stresses; (b) pore water pressures, according to the linear seepage approximation (Symons, 1983);
(c) approximate limiting effective stress distribution for simpli®ed analysis; (d) net total lateral stresses
EMBEDDED RETAINING WALLS 711
in Fig. 1(a) (with the values of Ka and Kp given Prop force F
by equations (1) and (2)) will lead to over-
conservative results because the effects of soil± h Wall rotation
wall friction are neglected. Where there is only
Active σ'h = Kaσ'v
one zone of soil at failure (either active or
passive), modi®ed earth pressure coef®cients may Passive
be derived that take account of the rotation in the d σ'h = Kpσ'v
direction of principal stresses between the soil±wall
interface and the free soil surface (e.g. Caquot &
Kerisel, 1948; Kerisel & Absi, 1990). The use of Frictionless stress discontinuities
for true lower bound
these modi®ed earth pressure coef®cients in the
effective stress distribution shown in Fig. 1(a) is
not rigorous, unless the complexities of the stress Fig. 2. Idealized effective stress distribution at failure
distributions around the pivot and at the toe of the for an embedded wall propped at the crest
wall are taken into account. None the less, the
in¯uence of soil±wall friction on the stability of unpropped wall, because only the depth of em-
embedded cantilever walls is often taken into bedment features in the equation of moment
account in this way. The analysis is no longer a equilibrium about the prop.
true lower bound, but might be termed a limit The designer of an embedded cantilever retain-
equilibrium calculation. It should be noted that the ing wall using the effective stress distributions
modi®ed earth pressure coef®cients are normally shown in Fig. 1(a) (for an unpropped wall) or
de®ned as ó9h /(ãz ÿ u) rather than ó9h /ó9v (where Fig. 2 (for a wall propped at the crest), with earth
ã is unit weight and u is pore water pressure): this pressure coef®cients modi®ed to take account of
is because the vertical effective stress adjacent soil±wall friction, must address three fundamental
to the wall is not equal to ãz ÿ u in the presence questions
of shear stresses at the soil±wall interface. This
(a) what value of ö9 should be used (i.e. ö9peak or
approach to design is approximate, in that it does
ö9crit ) as a predictor of collapse?
not consider in detail the complex and inscrutable
(b) what value of soil±wall friction angle ä is
behaviour of the soil around the wall as it moves
appropriate?
towards a state of active or passive failure (cf.
(c) how should the collapse calculation be factored
Rowe & Peaker, 1965).
in order to ensure that a structure will be
The inef®ciency of an unpropped embedded
suf®ciently remote from collapse not to deform
cantilever as an earth retaining system is demon-
excessively in service?
strated by Fig. 1(d), which shows the net total
stress distribution resulting from the effective In this paper, these three issues are discussed
stresses and pore water pressures indicated in with reference to previous experimental and ®eld
Figs 1(a) and 1(b). It may be seen that the net work and new case data.
overturning force actually increases with depth
below formation level until an embedment of d is
reached. It is only below this level that the net APPLICATION OF FACTOR OF SAFETY
lateral stress begins to act so as to maintain the The effective stress distributions shown in Figs
stability of the wall. For retained heights in excess 1(a) and 2, with values of Ka and Kp for a smooth
of about 5 m, the depths of embedment required to wall given by equations (1) and (2), represent
ensure that an unpropped cantilever wall remains rigorous lower bound plasticity solutions (provided
remote from collapse can begin to become uneco- that the possibility of a bearing failure at the toe
nomic, and it is usual to install one or more rows of the wall is investigated: see e.g. Powrie & Li,
of props in order to help to support the wall. For 1991). The stress state of the soil everywhere is in
retained heights of less than about 5 m, the cost of equilibrium, without violating the failure condition
installing props, together with the restrictions they ô/ó9 ˆ tan ö9.
impose on programming and access, may outweigh In structural engineering, plasticity theory is
their bene®t in reducing the depth of embedment used to estimate the relation between the collapse
of the wall. loads Wc and the fully plastic moment (i.e.
For an embedded wall propped at the crest, the bending strength) Mp of beams and frames. The
idealized effective stress distribution at failure is design load for a structure is the collapse load
shown in Fig. 2. This stress distribution corre- divided by a load factor ëc . It follows from the
sponds to rotation about the position of the prop. unique relationship between Wc and Mp that the
In this case, the two unknowns are the prop load maximum permissible bending moment under
and the depth of embedment of the wall. Solution working conditions is Mp /ëc . In the case of an
of the equilibrium equations is easier than for an embedded cantilever retaining wall, the loads that
712 POWRIE

cause collapse are not as readily identi®able as and the depth of embedment is chosen such
they are in the analysis of structural frames. This that the moment about the prop of the net
had led to the development of a number of differ- pressures in front of the wall is equal to a
ent methods of applying a factor of safety to the factor Fnp times the moment of the net
collapse calculation, as follows. pressure behind the wall. The pressure dis-
tribution used to obtain the design depth of
(a) The calculated depth of embedment at collapse embedment is not in equilibrium. Bending
is multiplied by an empirical factor Fd. This moments may be calculated for the upper part
method is unscienti®c, and will yield a stress of the wall only, using the depth of embedment
distribution that is in equilibrium only if the at which the wall is just stable. A further
total stresses in front of and behind the wall drawback is that an apparently satisfactory
below the depth required just to maintain numerical value of Fnp corresponds to a much
stability are equal and opposite. smaller factor on soil strength Fs (Burland,
(b) The equilibrium calculation is carried out with Potts & Walsh, 1981).
the soil strength tan ö9 reduced by a factor Fs. ( f ) The `revised' method (Burland et al., 1981):
This method is consistent with the way in the fully active moment directly attributable to
which factors of safety are applied in structural the retained soil is equated to the moment of
analysis. It is essentially the approach adopted the net available passive resistance below
in BS 8002 (British Standards Institution, formation level (calculated from the net
1994), except that the factor of safety Fs is pressure coef®cient (Kp ÿ Ka )), divided by a
termed a strength mobilization factor M, in factor Fr (Fig. 6 of Burland et al., 1981).
recognition of the fact that its main purpose is Burland et al. show that this method gives
to limit wall movement under working condi- numerical values of Fr for propped walls that
tions. correspond to a reasonably consistent value of
(c) The equilibrium calculation is carried out with Fs over the range ö9 ˆ 58 to 358, but do not
active pressures in the soil behind the wall, and investigate its applicability to unpropped walls
the passive pressures reduced by a factor Fp. in the same detail. While the approach might
This procedure is traditionally used for sheet be interpreted as analogous to the structural
pile walls in sands, for which it is generally load factor ëc , the same effect is achieved
accepted that the stresses behind the wall will rather more transparently by the application of
fall to their active values after only a very a factor Fs to the soil strength.
small movement, while the movement required
for the stresses in front of the wall to rise to For embedded walls propped at the crest, the
their passive limit would be unacceptably various methods of applying a factor of safety are
large. The application of the factor of safety discussed by Burland et al. (1981). They show that
to an earth pressure coef®cient (rather than the most sure, rational and consistent approach is
to the soil strength directly) can, however, to apply the factor of safety to the soil strength
be misleading: a value of Fp ˆ 1´5 implies directly. This is not to say that the same value of
Fs ˆ 1´6 at ö9 ˆ 208 but Fs ˆ 1´2 at ö9 ˆ 358 Fs will be applicable in all soils, or indeed must
(assuming full wall friction ä ˆ ö9 and pas- be applied to the soil on both sides of the wall. In
sive pressure coef®cients given by Caquot & any particular case, the allowable mobilized soil
Kerisel (1948)). strength (and hence the numerical value of Fs or
(d) For unpropped walls, the British Steel Piling mobilization factor M) would ideally be chosen
handbook (British Steel General Steels, 1988) on the basis of the allowable soil strain according
appears to suggest a factor of safety of unity to an appropriate mobilized strength±shear strain
(pp. E2±E3). Although in the example given relationship (Bolton, Powrie & Symons, 1989,
on pp. E4±E5 wall friction ä is neglected, the 1990a, 1990b). Loose soils, with a less stiff
method relies entirely on the designer to select response (i.e. a less rapid mobilization of soil
conservative soil parameters and worst-case strength with shear strain) would be expected to
geometrical and groundwater conditions. In the require a greater factor of safety or mobilization
case of a cofferdam retaining water (for which factor on soil strength than dense soils, in order to
there is no wall friction), the depth of embed- limit soil strains under working conditions to an
ment calculated using this method is almost acceptable level.
certain to be inadequate. The mobilized strength±strain response of the
(e) For walls propped at the crest, the `net soil will also depend on the stress path followed,
pressure method' is described in the Piling and may well be different for the soil on each side
handbook (British Steel General Steels, 1988, of the wall. This might impinge particularly on the
pp. E8±E9). A net pressure diagram using fully calculation of bending moments from the equili-
active and fully passive pressures is plotted, brium stress distributions. For example, in low K0
EMBEDDED RETAINING WALLS 713
soils, Fs ˆ 1 could be used in the retained soil to sand, they recommend ämob ˆ ämax on the passive
correspond to the traditional assumption (as in side, where ämax is given by equation (3). For
method (c) above) that fully active stresses will sheet pile walls bearing on rock, however, where
be developed under working conditions. In very the likely magnitude of downward wall movement
high K0 soils in which the wall is installed with is small, the recommended values of ämob are
minimal lateral stress relief, Fs ˆ 1 might be much reduced.
appropriate for the soil in front of the wall The results of Rowe & Peaker (1965) indicate
(Symons, 1991). In each case, the allowable earth that, in the limit equilibrium analysis of an
pressure coef®cient on the other side of the wall embedded retaining wall, the assumption of a
would be chosen on the basis of the factored soil uniform soil±wall friction angle on each side is
strength, rather than by the less certain procedure probably merely a convenient ®ction, without
of applying a factor to the passive (or active) earth which the analysis would be almost impossible.
pressure coef®cient. One possible concern with this approach is that the
implied bearing capacity below the base of the
wall may be unrealistically high. If there is any
SOIL±WALL FRICTION doubt about this (for example, if the wall is thin or
According to Rowe (1963), the maximum soil± the sub-stratum is weak), the base bearing stress
wall friction angle ämax has two components associated with the development of the assumed
ämax ˆ öw ‡ r (3) wall friction should be calculated (by considering
the vertical equilibrium of the wall), to ensure that
where öw is the true friction angle between the soil it does not exceed the ultimate bearing capacity.
grains and the material of the wall, and r is the wall
roughness angle. Rowe notes that öw is generally
close to öì , the true friction angle between soil SOIL STRENGTH
grains, and that the angle of soil±wall friction ä is Rowe (1951) reported the results of a series of
unaffected by the component of soil strength due large scale laboratory tests at normal gravity (lg)
to dilatancy, except insofar as the peak angle of on unpropped embedded cantilever retaining walls
shearing resistance ö9peak represents an extreme in dry sand. Conditions at collapse were reasonably
upper limit. For a smooth (a term not synonymous well represented by the idealized distribution of
with frictionless) wall, r could be of the order of effective stresses shown in Fig. 1(a), with Ka and
0´58: real walls made from steel sheet piles, Kp based on the peak angle of shearing resistance
concrete or timber would be expected to be ö9peak and wall friction ä ˆ 0´67ö9peak . (In a later
signi®cantly rougher. In general, the roughness of paper, however, Rowe (1963) cautioned against the
the wall will compensate at least approximately for expression of ä as a proportion of ö9peak .) Remote
the difference between the critical state angle of from collapse, a triangular pressure distribution in
shearing resistance ö9crit and the true friction angle front of the wall with the full passive pressure
between the soil grains öì (ö9crit ÿ öì  58 for coef®cient Kp reduced by a factor Fp was found
quartz sand), so that for practical purposes to overestimate the lateral stresses near the toe of
ämax  ö9crit . the wall, and hence bending moments. This was
An alternative line of argument, followed by probably because with a wall rotating about a
Bolton & Powrie (1987), leads to the same point close to the toe, there is only limited
conclusion. A surface on which there is insignif- movement of the wall into the soil at this level,
icant opportunity for dilation, but whose roughness restricting the development of lateral stresses in
is comparable with the particle size of the soil, front of the wall.
would be expected to mobilize ämax ˆ ö9crit . If the For embedded cantilever sheet pile walls in dry
wall is very rough, so that a rupture surface is sand, restrained by anchors near the crest, Rowe
forced to develop within the body of the soil rather (1952) found that the effective stress distribution
than along the interface, some or all of the dilatant shown in Fig. 2, with active and passive pressures
strength of the soil might be mobilized, giving an based on ö9peak and ä ˆ 0´67ö9peak , was represen-
upper limit of ämax ˆ ö9peak . In general, however, tative of conditions at collapse. These walls were
it is more appropriate (and more conservative) to not propped rigidly, and a horizontal movement of
take ämax ˆ ö9crit . the crest of the wall into the excavation of (h + d)/
Rowe & Peaker (1965) show that the wall 1000, where (h + d) is the overall height of the
friction actually developed depends on the direc- wall, was suf®cient to destroy arching and allow
tion and degree of wall movement. They suggest the lateral stresses in the retained soil to fall to the
that in carrying out an analysis of conditions at active limit. For stiff walls remote from collapse,
collapse, it is necessary to decide on the value of the measured maximum bending moment was
mobilized wall friction angle ämob at a movement consistent with the stress distribution shown in
of 5%. For sheet pile walls freely embedded in Fig. 2, with fully active pressures behind the wall
714 POWRIE

and smaller-than-passive pressures in front. This centrifuge: a range is therefore given, correspond-
corresponds to the application of a factor of safety ing to the last stable excavation depth, and the
Fp to the passive pressure coef®cient (method (c) removal of a further 0´5 m of soil (at ®eld scale)
above). For more ¯exible walls, where the move- from in front of the wall, which resulted in
ment into the excavation at formation level collapse. It may be seen that, in the tests on dense
exceeded that at the toe, the generation of high sand, the calculations using the peak angles of
lateral stresses in the soil immediately below shearing resistance quoted by King & McLoughlin
formation level led to a reduction in the maximum (1993) consistently overestimate the retained height
bending moment. It was to take account of this at failure. In the loose sand tests, the calculated
effect that Rowe (1952, 1955) proposed his retained heights at collapse are close to or just
moment reduction curves. outside the upper limits observed in the centrifuge
King & McLoughlin (1993) summarized the tests. If ö9peak for the loose sand were only slightly
results of a series of centrifuge model tests on in excess of ö9crit , then the use of earth pressure
unpropped embedded cantilever walls, retaining coef®cients based on ö9crit in the stress distribution
dry sand. The model walls represented a Froding- shown in Fig. 1(a) would lead to a generally
ham no. 5 section sheet pile wall of bending correct or only slightly conservative prediction of
stiffness 10´2  104 kN m2 /m and total length collapse. Although the retained heights at collapse
11 m, at a scale of 1:92. Tests were carried out in the dense sand tests were greater than in the
on walls with three different surface ®nishes, corresponding loose sand tests, the differences
giving maximum angles of soil±wall friction were not as great as suggested by the use in the
ämax ˆ 0, ämax ˆ ö9peak and intermediate values limit equilibrium calculation shown in Fig. 1(a) of
of ämax ˆ 0´36ö9peak to ämax ˆ 0´40ö9peak , retaining earth pressure coef®cients corresponding to the
either dense sand (mass density r ˆ 1631 kg/m3 , quoted peak strengths.
ö9peak ˆ 49´58) or loose sand (mass density r ˆ King & McLoughlin (1993) also observed that
1448 kg/m3 , ö9peak ˆ 40´08). Bending moments the deformations of all walls at Fp ˆ 1´5 (based on
were measured using strain gauges glued to the peak strengths) were much larger than would be
wall. The soil in front of the wall was excavated in tolerated in reality. This demonstrates one of the
stages by stopping and restarting the centrifuge. major disadvantages of using a factor of passive
This procedure is considered by King & McLough- pressure coef®cient Fp in design, which is that the
lin (1993) not to have affected the behaviour of the additional embedment required to increase the
model in any signi®cant way. Given the stabilizing numerical value of Fp from 1´0 to 1´5 or even
effect of the likely increase in ö9peak at centrifugal 2´0 may be very small, especially when the angle
accelerations below the test value (due to the of shearing resistance of the soil ö9 is high
increased potential for dilation), this conclusion is (Simpson, 1992). The problem is exacerbated by
probably not unreasonable. the use of ö9peak , rather than ö9crit , as a design
Table 1 compares the retained heights at parameter.
collapse observed by King & McLoughlin (1993) Bolton & Powrie (1987, 1988) described a series
with those obtained using the stress distribution of centrifuge model tests carried out to investigate
shown in Fig. 1(a), using Caquot & Kerisel's the behaviour of diaphragm walls in overconsoli-
(1948) earth pressure coef®cients for the values of dated kaolin clay (r  1780 kg/m3 , ö9crit  228,
ö9peak and ä stated. It was not possible to identify ämax  ö9crit measured in a shearbox). Walls
precisely the retained height at failure in the modelling a retained height of 10 m at a scale of
Table 1. Comparison of actual and theoretical retained heights at collapse for centrifuge model tests of King &
McLoughlin (1993)
ä Retained height at collapse: m at ®eld scale
King & McLoughlin (1993) Fig. 1(a) stress analysis with
centrifuge test Caquot & Kerisel (1948) earth
pressure coef®cients using ö9peak
and ä in columns 1 and 2
0 7´0±7´5 8´00
Dense sand (ö9peak ˆ 49´58) 17´78 8´0±8´5 8´54
49´58 8´5±9´0 9´51
0 6´0±6´5 6´89
Loose sand (ö9peak ˆ 408) 15´88 7´0±7´5 7´46
408 8´0±8´5 8´36
EMBEDDED RETAINING WALLS 715
1:125 were made from either 9´5 mm or 4´7 mm further point is that in small-scale laboratory tests
aluminium plate, corresponding to full-scale ¯ex- at lg, the range of stress is not great and the use of
ural rigidities of approximately 107 kN m2 /m and a uniform value of ö9peak in back-analysis is
1´2  106 kN m2 /m respectively. An equilibrium probably justi®able. (Alternatively, it is possible
calculation based on the effective stress distribu- that the earth pressure coef®cients based on ö9peak
tion shown in Fig. 1(a) and measured pore water and ä ˆ 2ö9peak /3 might be numerically equal to
pressures was found to provide a close indication those based on ä ˆ ö9 ˆ ö9crit for the dense
of the embedment of an unpropped wall at col- materials. This is less likely for the loose materials
lapse, using Caquot & Kerisel's (1948) earth pres- where ö9peak might have been little greater than
sure coef®cients based on the critical state soil ö9crit . However, the process of collapse, and the
strength and wall friction angle ä ˆ ö9crit . For an accompanying geometry changes, might have been
embedded cantilever wall propped at the crest, the suf®ciently gradual to admit either possibility for
effective stress distribution shown in Fig. 2 (again the back-analysis of conditions at `failure'.)
with Caquot & Kerisel's earth pressure coef®cients At ®eld scale, or in a centrifuge model, the
and ä ˆ ö9 ˆ ö9crit ) gave a slightly conservative range of stress is much larger than in a lg test, and
indication of collapse. This was considered to be the applicability of a uniform value of ö9peak is
due to the kinematic restraint imposed by the prop, unlikely. In these circumstances, the use of ö9crit
which prevents horizontal movement at the crest of rather than ö9peak (with Caquot & Kerisel's (1948)
the wall. This is not taken into account in the earth pressure coef®cients and soil±wall friction
derivation of the earth pressure coef®cients used in ä ˆ ö9crit ) would appear to give a more reliable,
the limit equilibrium calculation. In practice, a albeit possibly conservative, indicator of collapse.
less-than-rigid prop might not impose this kine- (Having said this, the peak strength ö9peak , meas-
matic restraint, and the additional margin of safety ured at the highest applicable stress, does serve as
would be lost. an indicator of the rate of mobilization of soil
The results of the stress analyses shown in Figs strength with shear strain.)
1(a) and 2 may be presented in terms of the These issues are now examined further, with
embedment ratio d/h at failure as a function of reference to three case studies.
the angle of shearing resistance ö9 of the soil. The
unit weight of the soil does not affect the
embedment ratio at collapse, so that a model wall
that does not fail at lg should in theory be stable CASE STUDY 1
at any gravity level to which it is subjected in a The ®rst case study concerns a retaining wall,
centrifuge, provided that the maximum reliable approximately 30 m long with an 8 m return at
angle of shearing resistance of the soil does not each end, constructed using Frodingham 3N Grade
change. In reality, the peak angle of shearing 50A steel sheet piles. The wall was installed to
resistance of a soil of a given void ratio (density) retain one side of an excavation during the
will decrease as the applied effective stress construction of the basement of a new building,
increases, due to the suppression of dilation as adjacent to an existing structure. Ground condi-
the void ratio at the appropriate critical state tions comprised approximately 7´4 m of granular
decreases. Thus a small-scale model may be stable material (described as sandy to very sandy sub-
at lg, but fail in an enhanced acceleration ®eld angular and sub-rounded ®ne, medium and coarse
because the peak strength is reduced. If this gravel, generally medium dense with occasional
happens, the implication is that a large retaining loose zones), underlain by 4´7 m of laminated
wall of a given embedment ratio d/h will be less clayey sandy silt, and microdiorite rock. The
stable than a smaller wall of the same embedment natural water table was 1´6 m below original
ratio in the same material, because the peak angle ground level. A cross-section, showing the ground
of shearing resistance which maintains the stability conditions and the geometry of the excavation, is
of the smaller wall cannot be mobilized at the shown in Fig. 3. Analysis is complicated slightly
higher stresses that exist in the soil around the by the 168 upward slope of the retained ground,
larger wall. and the presence behind the wall of a number of
This suggests that considerable caution must be 1´2 m  1´2 m footings at 13 m centres, supporting
exercised in the selection of ö9 values in the the columns of the existing building. It is
extrapolation of lg model test results to ®eld scale estimated that these foundations each exerted a
structures. Rowe (1951, 1952) appears to have vertical surcharge of approximately 150 kPa.
based his back-analyses on ö9peak . In the applica- Following installation of the wall, excavation
tion of his analysis to a larger structure, ö9peak proceeded to the level of the underside of the
should be measured at the highest applicable stress basement ¯oor slab (4´4 m below original ground
level, in which case it might be closer to the level). The excavation was dewatered by means of
critical state angle of shearing resistance ö9crit . A a sump pump. Some concern had been expressed
716 POWRIE

1.2 m 2.3 m

Original
ground level
Column footings
1.2 m x 1.2 m x 0.6 m
deep at 13 m centres 150 kPa

1.6 m
16°

Original ground water level (GWL)


0.325 m

2.8 m

GWL in front of wall


maintained by sump
Sandy GRAVEL pumping

3m

Clayey
sandy 3m
SILT

Fig. 3. Cross-section showing retaining wall geometry and ground conditions, case study 1

that the groundwater level behind the sheet pile (range 6´6 to 22´5) for the granular material
wall should not be allowed to fall, in order to indicated ö9peak ˆ 338, according to Stroud
minimize the possibility of settlement damage to (1989), assuming that the soil is normally con-
the existing building. On completion of the solidated. A critical state friction angle ö9crit of
excavation, a series of small holes was drilled into 308±328 for the granular material would not seem
the sheet pile wall, which established that the unreasonable (Stroud, 1989). The critical state
groundwater level in the retained soil had fallen to friction angle of the silt would be expected to be
less than 0´9 m above formation level. In order to in the range 258±278 (Bowles, 1988).
raise the groundwater level behind the wall, the The cross-section in Fig. 3 was back-analysed
discharge from the sump pump was diverted on to using an effective stress distribution similar in
the retained soil surface. Some time later, the principle to that shown in Fig. 1(a) with different
retaining wall moved forward slightly, causing a earth pressure coef®cients in each stratum of soil
noticeable settlement of the existing building. to re¯ect their different strengths. Four different
The site investigation data indicated that the cases were examined to investigate the effect
granular material had a unit weight of 22 kN/m3 , (separately and in combination) of both the
and the laminated silt 20 kN/m3 . No experimental groundwater level in the retained soil and the
data concerning frictional strengths were given, but surcharge from the existing building. These were
an average standard test (SPT) blowcount (cor-
rected for the effects of overburden using the (a) low groundwater level (4´4 m below original
expression N1 ˆ CN  N, where CN ˆ 0´77 log10 ground level (OGL)) in the retained soil, no
[2000/ó9v (kPa)]: Skempton, 1986) of N1 ˆ 16 surcharge from existing building
EMBEDDED RETAINING WALLS 717
(b) low groundwater level (4´4 m below OGL) in B A
the retained soil, with surcharge from existing
building
q
(c) high groundwater level (1´6 m below OGL) in
the retained soil, no surcharge from existing
building A (1 − Ka)
(d) high groundwater level (1´6 m below OGL) in 2√Ka
the retained soil, with surcharge from existing A/√Ka
building.
Total area
The pore water pressures in the granular = qB√Ka
material were taken as hydrostatic below the water
table on each side of the wall. In the case of the B/√Ka
high groundwater level on the retained side, all the
head loss was assumed to occur in the silt, and to
be distributed linearly around the wall (Fig. 4). The
effect of the surcharge was modelled using the
procedure suggested by Pappin, Simpson, Felton & ≤ qKa
Raison (1986), which is illustrated in Fig. 5.
The equilibrium equations associated with the
idealized effective stress and pore water pressure
distributions were assembled on to a spreadsheet, Fig. 5. Additional lateral effective stress acting on the
and solved iteratively. The aim of the calculation back of the wall due to a surcharge (after Pappin
was to achieve equilibrium with a depth of et al., 1986)

u=0

2.8 m

Hydrostatic Datum for


h = 2.8 m u=0 excess head
h

GRAVEL

Hydrostatic
3m h=0 GRAVEL

u = 56.9 kPa

u = 29.4 kPa

SILT (Downward seepage) (Upward seepage)

3m SILT

h = 1.4 m, u = 72.6 kPa

Fig. 4. Pore water pressure distribution used in back-analysis of case study 1 (high groundwater level behind the
wall)
718 POWRIE

embedment of 6 m, a certain value of zp , and earth measured values, rather than the linear seepage
pressure coef®cients corresponding to mobilized approximation (Symons, 1983): Fig. 6(b)
soil strengths ö9mob consistent with approximately (b) the carriageway was assumed to act as a
the same factor of safety Fs (ˆ tan ö9crit /tan ö9mob ) surcharge of 14 kPa on the excavated soil
in all zones. The earth pressure coef®cients given surface
by Caquot & Kerisel (1948) were used, with an (c) the analysis was carried out with ä ˆ ö9mob on
angle of soil±wall friction ä ˆ ö9mob . Behind the both sides of the wall, rather than ä ˆ 2ö9mob /
wall, the slope ù (ˆ 168) of the retained soil sur- 3 (active) and ä ˆ ö9mob /2 (passive) as used by
face was taken into account by using the tabulated Carder & Symons (1989).
earth pressure coef®cients appropriate to the value
of ù/ö9mob in each soil layer. According to the revised calculation, the factor
The results of the calculations are summarized of safety on soil strength (i.e. tan ö9crit /tan ö9mob ) is
in Table 2, and are discussed below. 1´16. This is probably still on the low side, as any
possible propping action of the carriageway slab
(which might occur in reality, even though it was
CASE STUDY 2 never intended in design) has been neglected.
The second case study is the embedded canti-
lever retaining wall on the A329(M) near Reading,
Berkshire (Carder & Symons, 1989). This wall has CASE STUDY 3
performed perfectly satisfactorily since its con- The third case study is an embedded sheet pile
struction in 1972. It is included in the present wall propped near the crest, forming a cofferdam
study because the calculations presented by Carder of plan dimensions 14 m  10 m. The retained
& Symons (1989) (which follow the design height was 7´5 m and the embedded depth was
recommendations given by Pad®eld & Mair, 3´5 m. Ground conditions comprised storm beach
1984) suggest that the wall has a factor of safety gravels overlying marine sands, as indicated on the
Fs of about unity. (It should be noted, however, cross-section shown in Fig. 7. Failure occurred
that Pad®eld & Mair (1984) recommend the use of over almost the entire length of one of the short
wall friction angles ä ˆ 2ö9mob /3 behind the wall sides of the cofferdam by rotation about the prop,
ä ˆ ö9mob /2 in front, rather than ä ˆ ö9mob as used shortly after excavation to formation level.
in this paper). Fig. 6(a) (after Carder & Symons, The storm beach gravels were described in the
1989) shows an idealized cross-section of the site investigation report as a predominantly uni-
retaining wall, indicating the ground conditions form sandy ®ne to coarse gravel with occasional
and soil parameters based on remoulded strengths. cobbles, and the marine sands as a silty ®ne to
The calculations carried out by Carder & coarse sand. No data concerning unit weights or
Symons (1989) are essentially those that would friction angles were given. At the borehole nearest
have been used in design. In order to estimate the to the side of the cofferdam that failed, the average
actual factor of safety of the wall, a back-analysis corrected SPT blowcount N1 was 20´6 for the
has been carried out, which differs from the storm beach gravels (range 8´5±48´0) and 41´6 for
calculation by Carder & Symons (1989) in the the marine sands (range 32´2±55´5). These imply
following ways peak friction angles ö9peak of 34´58 for the storm
beach gravels and 418 for the marine sands,
(a) the pore water pressures were based on the assuming that the materials are effectively nor-

Table 2. Results of equilibrium analyses, case study 1


Analysis

a b c d
Groundwater level Low Low High High
Surcharge Absent Present Absent Present
ö9mob (sandy gravel) 22´38 26´98 258 298
ö9mob (silt) 18´38 22´28 20´58 248
Fs (based on ö9crit ˆ 308 for the sandy 1´42 1´14 1´24 1´04
gravel and 258 for the silt)
Fs (based on ö9crit ˆ 328 for the sandy 1´54 1´24 1´35 1´13
gravel and 278 for the silt)
EMBEDDED RETAINING WALLS 719
Depth: m
Surcharge 12 kPa

0 Topsoil/made ground
γ = 17.3 kN/m3, φ' = 30°
GWL
1.52
2.0 γ = 18.9 kN/m3, φ ' = 42°
2.59 Gravel and clayey sand

6.9 m
Firm grey clay
γ = 20.1 kN/m3, φ' = 21°

Surcharge 14 kPa
6.0

GWL
Firm grey clay
γ = 20.1 kN/m3, φ' = 22°

(8.9 m below OGL)

11.5 m

14.5

Very stiff grey clay


γ = 20.1 kN/m3, φ' = 32°

18.4

(a)

Depth: m
0
GWL

GWL

10

15

120 100 80 60 40 20 20 40 60 80 100 120

0
Pore water pressure: kPa
Spade cell piezometers
Standpipe piezometers
Used in analysis
(b)

Fig. 6(a). Idealized cross-section through A329(M) reataining wall near Reading (after Carder & Symons, 1989);
(b) measured and idealized pore water pressures, (measured values from Carder & Symons, 1989)
720 POWRIE

0.4 m PROP
PROP
3.3 m 3.05 m
Mean
GWL Maximum
GWL
7.5 m h = 4.45 m
7.3 m Tidal
Storm variation Storm beach
beach ±0.25 m Gravels
GRAVELS GWL in front of
wall maintained 4.25 m
by pumping from
wells
h=0
GWL Datum
for h
Marine SANDS
3.5 m

Marine sands
3.7 m
Downward Upward
Fig. 7. Cross-section showing retaining wall geometry seepage seepage
and ground conditions, case study 3
h = 1.36 m u = 47.7 kPa

mally consolidated (Stroud, 1989). Critical state Fig. 8. Pore water pressure distribution used in back-
analysis of case study 3
angles of friction might reasonably be taken as 328
for the predominantly single-size, sub-angular to
sub-rounded storm beach gravels, and up to 368 for wall, and a uniform unit weight in both strata of
the well-graded marine sands (Stroud, 1989). The 19 kN/m3 , the calculated factor of safety Fs was
SPT blowcounts from 14 boreholes across the site 1´08.
displayed considerable variation, with those closest A second cofferdam, 8´5 m wide, with a
to the location of the failure being generally at the retained height of 7´3 m and a depth of embedment
low end of the range, at least above the level of of 3´7 m, in which the props were installed at a
the toe of the wall. depth of 2´8 m below the crest of the wall, showed
In situ tests indicated permeabilities generally in no signs of distress. The factor of safety Fs in this
the range 1 to 7  10ÿ5 m/s for the beach gravels case was 1´24. This may be an underestimate,
and 0´5 to 2´5  10ÿ5 m/s for the marine sands. because the cofferdam was quite narrow. In these
The measured permeability of the beach gravels is conditions, the proximity of the two sidewalls
somewhat smaller than would be inferred from the represents an additional kinematic restraint which
typical particle size distribution curve using, for is not taken into account in the limit equilibrium
example, Hazen's rule (Hazen, 1892). This is calculation.
probably due to a loss of ®nes on sampling. As
the measured permeabilities of the two materials
do not differ signi®cantly, it is reasonable to treat DISCUSSION
them as a single stratum for the purpose of One of the striking features of the two cases in
estimating the steady-state pore water pressures which the sheet pile retaining walls failed is the
due to seepage. The natural groundwater level was paucity of data from the site investigation con-
approximately 3´3 m below original ground level, cerning essential parameters such as ö9crit and unit
with a tidal variation of 60´25 m. The groundwater weight. SPT data can provide useful empirical
level inside the cofferdam was maintained at just con®rmation of the applicability in situ of
below formation level by a system of pumped laboratory test results. It is surely unsatisfactory,
wells. however, when the SPT blowcount is the only way
An equilibrium calculation was carried out, of estimating a parameter so fundamental to
based on the effective stress distribution shown in retaining wall stability as the strength of the soil.
Fig. 2 with different earth pressure coef®cients in Although typical critical state strengths of sands
the storm beach gravels and the marine sands to and gravels are well documented (e.g. Bolton,
re¯ect their different strengths. The pore water 1986; Stroud, 1989), a reliance on generic data
pressure distribution was calculated according to carries with it a need to err on the side of caution.
the linear seepage model for the upper extreme of Unfortunately overconservatismÐparticularly in the
the tidal range, as indicated in Fig. 8. Using the design of temporary worksÐis not conducive to
earth pressure coef®cients given by Caquot & winning contracts, which are awarded primarily on
Kerisel (1948) with ä ˆ ö9mob on both sides of the the basis of the lowest price.
EMBEDDED RETAINING WALLS 721
Apart from the quanti®cation of basic soil shown in Figs 1(a) and 2, using Caquot & Kerisel's
parameters such as unit weight and strength, the (1948) earth pressure coef®cients with ä ˆ ö9mob ).
case histories demonstrate that, even in back- A less conservative interpretation indicates that the
analysis, there is considerable scope for uncertainty minimum factor of safety required may be as low
concerning the actual in-service conditions. The as 1´15 (case 1, analysis b, lower values of ö9crit ;
issues that have arisen here include the presence of case 2, although the true factor of safety here is
discontinuous surcharges on the retained side of probably signi®cantly higher because the propping
the wall; the support offered to potentially weaker action of the carriageway slab has been neglected),
or more heavily loaded cross-sections by stronger but is unlikely to be very much less than this value
or less heavily loaded parts of the wall; the (case 1, analysis d, higher values of ö9crit ; case 3,
possible stabilizing effect of basement or carriage- high-level prop).
way slabs acting as partial props; and variations in For permanent works in stiff clays, Pad®eld
groundwater levels (which may be tidal, seasonal & Mair (1984) recommend Fs ˆ 1´2 applied to
or man-made). In design, the new UK Code of ö9crit , with ä ˆ ö9mob /2 in front of the wall
Practice BS 8002 (British Standards Institution, and ä ˆ 2ö9mob /3 behind. (Incidentally, this is the
1994) requires some allowance for events such as a opposite way round to Terzaghi (1954), who
minimum surcharge on the retained side, and an suggests ä ˆ ö9mob /2 behind the wall and
additional unplanned excavation below formation ä ˆ 2ö9mob /3 in front.) Inspection of Caquot &
level. Kerisel's (1948) tables shows that the calculation
A further uncertainty concerns the identi®cation of earth pressure coef®cients using ä ˆ ö9mob /2
of failure. The movement in case 1, although large (rather than ä ˆ ö9mob ) is equivalent to an addi-
enough to cause distress to the building adjacent to tional factor of safety of approximately 1´11 over
the excavation, was a serviceability failure rather the range 188 , ö9mob , 368. Using ä ˆ 2ö9mob /3,
than an outright collapse. Case 3 was closer to an the additional factor of safety is reduced to
obvious collapse, but the retaining wall remained approximately 1´05. Thus the use of Fs ˆ 1´2 with
standing after the movement had ceased. In reduced wall friction as recommended by Pad®eld
general, the amount by which a wall moves during & Mair (1984) will in general correspond approxi-
excavation would be expected to be related to its mately to Fs ˆ 1´2  1´08  1´3 with full wall
factor of safety. If the factor of safety is high friction ä ˆ ö9mob .
enough, the wall will not move very much. As the Also for stiff clays, Simpson (1992) presents an
factor of safety is reduced, wall movements will analysis which suggests that wall displacements
increase. Serviceability failures would be expected increase only gradually with decreasing factor of
to be associated with factors of safety slightly safety Fs , and are unlikely to become critical at
greater than 1: this is borne out by the behaviour Fs > 1´2 (assuming wall friction ä ˆ ö9mob ).
of the retaining walls in cases 1 and 3. BS 8002 (British Standards Institution, 1994)
In calculating factors of safety for the retaining requires that an embedded retaining wall be
wall in case 1, the possibility of longitudinal load designed to be in equilibrium mobilizing the lesser
transfer along the wall was discounted. The cross- of (tan ö9peak ) 4 1´2 and tan ö9crit . Wall friction is
sections next to the pad foundations were assumed taken as tan ä/tan ö9mob ˆ 0´75, giving an addi-
not to receive support from the less severely tional factor of safety compared with taking
loaded parts of the wall. Unless the wall and its ä ˆ ö9mob of approximately 1´03 on the active
walings are very stiff, this assumption is not un- side. On the passive side the additional factor of
reasonable. However, the complete neglect of safety is very small (cf. Rowe & Peaker, 1965);
longitudinal load transfer might mean that the true decreasing from 1´02 to less than 1´01 as ö9mob
factors of safety are slightly higher than calculated increases from 158 to 358.
for cases 1b and 1d, and slightly lower for cases 1a It is also necessary to allow for an additional
and 1c (Table 2). unplanned excavation of at least 10% of the re-
Bearing in mind that the factors of safety tained height h. This changes the design embed-
calculated by back-analysis depend on the esti- ment ratio from d/h to approximately [0´9  (d/h)
mated soil strengths and in-service conditions, a ÿ 0´1]. Using the nomograms presented by Bolton
conservative interpretation of the case records et al. (1990b), together with some supplementary
described in this paper suggests that, for embedded calculations using the stress distributions shown in
cantilever walls that are either unpropped or Figs 1(a), 1(b) and 2, it can be shown that in a
propped near the crest, a factor of safety on soil uniform soil with unit weight 㠈 2ãw and ground-
strength of 1´25 is suf®cient to guard against a water levels at the retained and excavated soil
serviceability failure (case 1, analysis b, higher surfaces, this change in the embedment ratio is
values of ö9crit ; case 3, the wall with the lower equivalent to an additional factor of safety on soil
level prop. Limit equilibrium calculations are strength Fs of at least 1´12 over the range
based on the lateral effective stress distributions 108 < ö9mob < 358, for both unpropped walls and
722 POWRIE

walls propped at the crest. In addition, a 10 kPa walls with Fs > 1´25 (or possibly Fs > 1´15)
surcharge must be assumed to act on the retained would be expected to perform satisfactorily in
soil surface. practice.
Even if the surcharge requirement is ignored, a (c) The real factor of safety for a particular wall
wall designed in accordance with BS 8002 will may be higher or lower than that calculated in
have an equivalent factor of safety Fs based on either back-analysis or design, depending on
ö9crit and ä ˆ ö9crit of between about 1´02  1´12 how conservatively the soil parameters and
ˆ 1´14 for dense soils and (1´2  1´02  1´12) ˆ various other uncertainties have been quanti-
1´37 for loose soils in which ö9peak ˆ ö9crit . On the ®ed.
basis of the data presented in this paper, these (d) The application of Fs ˆ 1´25 with wall friction
values would seem to be adequate. In the case of a ä ˆ ö9 ˆ ö9mob is in general slightly less
dense soil, however, there may be comparatively conservative than the procedure recommended
little margin for error in the assessment of factors by Pad®eld & Mair (1984) for stiff clays.
such as groundwater conditions and soil strength. However, the procedure recommended in BS
8002 for dense soils may be only slightly more
conservative than the application of Fs ˆ 1´15
CONCLUSION
with wall friction ä ˆ ö9 ˆ ö9mob . This means
The aim of this paper is to set out a rational that particular care must be taken in the
basis for the selection of soil parameters and identi®cation of the most onerous geometrical,
factors of safety for use in the conventional limit loading and groundwater conditions, and in the
equilibrium stress analysis of embedded retaining quanti®cation of soil strength.
walls. The success of the method depends on a
number of factors in addition to the selection of
appropriate soil parameters and an appropriate NOTATION
factor of safety. These include the identi®cation CN correction factor applied to SPT blowcount
of the most onerous loading and geometrical d depth of retaining wall embedment
conditions, and the fullest consideration of ground- d depth below formation level at which net total
water effects. The difference between a wall that lateral stress on unpropped cantilever wall is zero
performs satisfactorily and a wall that suffers a Fd factor of safety on wall embedment
serviceability failure may be very slight: the Fnp factor of safety using net pressure method
Fp factor of safety applied to passive earth pressure
calculations summarized in Table 2 suggest that a
coef®cient
10% reduction in the real factor of safety Fs Ð Fr factor of safety: `revised' method (Burland et al.,
which could result from the overestimation of ö9crit 1981)
by only 28Ðmight make all the difference. Bearing Fs factor of safety applied to soil strength
these points in mind, the following conclusions g acceleration due to earth's gravity (9´81 m/s2 )
may be drawn. h retained height
hÅ excess head
(a) The collapse of embedded cantilever retaining Ka active earth pressure coef®cient
walls that are either unpropped or propped Kp passive earth pressure coef®cient
near the crest is in general well or slightly K0 in situ earth pressure coef®cient
conservatively predicted by classical limit M mobilization factor on soil strength (British
equilibrium stress distributions, using the earth Standards Institution, 1994)
Mp fully plastic moment of steel beam
pressure coef®cients given by Caquot &
N SPT blowcount
Kerisel (1948) with ä ˆ ö9 ˆ ö9crit on both N1 corrected SPT blowcount
sides of the wall. The ability of the soil below Q equivalent toe force in simpli®ed stress analysis
the base of the wall to maintain vertical for unpropped wall
equilibrium without suffering a bearing failure r wall roughness angle
should be checked, particularly if the wall is u pore water pressure
thin or the sub-soil is weak. Wc set of collapse loads for structure
(b) It is logical to base the calculation of a design zp depth of pivot (unpropped wall) below formation
depth of embedment on the limit equilibrium level
ã unit weight of soil
stress distribution which predicts collapse, with
ãw unit weight of water
the soil strength reduced by a factor of safety ä soil±wall interface friction angle
Fs or, in the parlance of BS 8002 (British ämax maximum soil±wall interface friction angle
Standards Institution, 1994), a mobilization ämob mobilized soil±wall interface friction angle
factor M. This involves the use of earth r mass density
pressure coef®cients based on ä ˆ ö9 ˆ ó9h horizontal effective stress
ö9mob , where tan ö9mob ˆ (tan ö9crit )/Fs. The ó9v vertical effective stress
case data examined in this paper suggest that ö9 soil angle of shearing resistance
EMBEDDED RETAINING WALLS 723

ö9crit critical state angle of shearing resistance Kerisel, J. & Absi, E. (1990). Active and passive pressure
ö9mob mobilized angle of shearing resistance tables. Rotterdam: Balkema.
ö9peak peak angle of shearing resistance King, G. J. W. & McLoughlin, J. P. (1993) Centrifuge
model studies of a cantilever retaining wall in sand.
öw true friction angle between soil grains and wall In Retaining structures (edited by C. R. I. Clayton).
material London: Institution of Civil Engineers.
ö9ì true friction angle between soil grains Pad®eld, C. J. & Mair, R. J. (1984). Design of retaining
walls embedded in stiff clays, report 104. London:
Construction Industry Research and Information
REFERENCES Association.
Bolton, M. D. (1986). The strength and dilatancy of Pappin, J. W., Simpson, B., Felton, P. J. & Raison, C.
sands. GeÂotechnique 36, No. 1, 65±78. (1986). Numerical analysis of ¯exible retaining walls.
Bolton, M. D. & Powrie, W. (1987). Collapse of Proceedings of symposium on computer applications
diaphragm walls retaining clay. GeÂotechnique 37, in geotechnical engineering. Birmingham: Midland
No. 3, 335±353. Geotechnical Society.
Bolton, M. D. & Powrie, W. (1988). Behaviour of Powrie, W. & Li, E. S. F. (1991). Analysis of in situ
diaphragm walls in clay prior to collapse. GeÂotech- retaining walls propped at formation level. Proc. Inst.
nique 38, No. 2, 167±189. Civ. Engrs 91, Part 2, 853±873.
Bolton, M. D., Powrie, W. & Symons, I. F. (1989). The Rowe, P. W. (1951). Cantilever sheet piling in cohesion-
design of stiff in situ walls retaining overconsolidated less soil. Engineering 51, Sept., 316±319.
clay part 1: short-term behaviour. Ground Engng 22, Rowe, P. W. (1952). Anchored sheet pile walls. Proc.
No. 8, 44±47. Instn Civ. Engrs 1, Part 1, 27±70.
Bolton, M. D., Powrie, W. & Symons, I. F. (1990a). The Rowe, P. W. (1955). A theoretical and experimental
design of stiff in situ walls retaining overconsolidated analysis of sheet pile walls. Proc. Instn Civ. Engrs 4,
clay part 1: short-term behaviour (continued). Ground Part 1, 32±69.
Engng 22, No. 9, 34±40. Rowe, P. W. (1963). Stress±dilatancy, earth pressures and
Bolton, M. D., Powrie, W. & Symons, I. F. (1990b). The slopes. J. Soil Mech. Fdn Div. Am. Soc. Civ. Engrs
design of stiff in situ walls retaining overconsolidated 89, SM3, 37±61.
clay part 2: long-term behaviour. Ground Engng 23, Rowe, P. W. & Peaker, K. (1965). Passive earth pressure
No. 2, 22±28. measurements. GeÂotechnique 15, No. 1, 57±78.
Bowles, J. E. (1988). Foundation analysis and design. Simpson, B. (1992) Retaining structures: displacement
New York: McGraw-Hill. and design. 32nd Rankine lecture. GeÂotechnique 42,
British Steel General Steels (1988). Piling handbook, 6th No. 4, 541±576.
edn. Skempton, A. W. (1986). Standard penetration test
British Standards Institution (1994). Code of Practice for procedures and the effects in sands of overburden
earth retaining structures, BS 8002. London: BSI. pressure, relative density, particle size, ageing and
Burland, J. B., Potts, D. M. & Walsh, N. M. (1981). The overconsolidation. GeÂotechnique 36, No. 2, 425±447.
overall stability of free and propped embedded Stroud, M. A. (1989). The standard penetration testÐits
cantilever walls. Ground Engng 14, No. 5, 28±38. application and interpretation. In Penetration testing
Caquot, A. & Kerisel, J. (1948). Tables for the in the UK, pp. 29±49. London: Thomas Telford.
calculation of passive pressure, active pressure and Symons, I. F. (1983). Assessing the stability of a propped
bearing capacity of foundations. Paris: Gauthier in situ retaining wall in overconsolidated clay. Proc.
Villars. Instn Civ. Engrs 75, Part 2, pp. 617±633.
Carder, D. R. & Symons, I. F. (1989). Long-term Symons, I. F. (1991). Discussion of Limit equilibrium
performance of an embedded cantilever retaining wall methods for free embedded cantilever walls in
in stiff clay. GeÂotechnique 39, No. 1, 55±75. granular materials by A. V. D. Bica and C. R. I.
Hazen, A. (1892). Physical properties of sands and Clayton. Proc Instn Civ. Engrs 90, Part 1, 213±216.
gravels with reference to their use in ®ltration. Report Terzaghi, K. (1954). Anchored bulkheads. Trans. Am.
to the Massachusetts State Board of Health. Soc. Civ. Engrs 119, 1243±1280.

Vous aimerez peut-être aussi