Potts Et Al 2015 Delayed Collapse of Cut Slopes in Stiff Clay

Potts, D. M., Kovacevic, N. & Vaughen, P. R. (1997). GeÂotechnique 47, No.
5, 953±982
Delayed collapse of cut slopes in stiff clay
D. M . P OT T S , N. KOVAC E V I C { a n d P. R . VAU G H A N
Delayed failure of railway cutting slopes in stiff La rupture diffeÂreÂe de talus de deÂblai ferroviaires
clays excavated in the nineteenth and early tailleÂs dans de l'argile raide au 198 sieÁcle et au
twentieth centuries has been studied for some deÂbut du 208 sieÁcle est eÂtudieÂe depuis une
forty years. Known failures have generally been quarantaine d'anneÂes. En geÂneÂral, les ruptures
deep-seated. More recently failures have been a dont on a connaissance se sont produites en pro-
problem in the slopes of motorway cuttings and fondeur. Plus reÂcemment, on a assisteÂ aÁ la
embankments, although generally these have rupture de talus de deÂblai et de remblai auto-
been shallow. The average operational strength routiers, mais, geÂneÂralement, aÁ faible profondeur.
at failure in these slides has been signi®cantly Au moment de la rupture, la reÂsistance opeÂr-
less than the peak strength, and progressive ationnelle moyenne eÂtait bien infeÂrieure aÁ la reÂ-
failure has been postulated as the probable sistance de pointe, et l'on suppose qu'une rupture
cause of this. Progressive failure can now be progressive en a eÂteÂ la cause. Il est maintenant
analysed using advanced numerical techniques. possible d'analyser la rupture progressive aÁ l'aide
A series of coupled ®nite element analyses have de techniques numeÂriques avanceÂes. L'article
been conducted assuming strain-softening soil preÂsente les reÂsultats d'une seÂrie d'analyses
with properties based on the Brown London jumeleÂes des eÂleÂments ®nis portant sur un
Clay, and the results are reported. They show sol qui s'amollit sous l'effet des contraintes et
that progressive failure is considerable, and fully dont les proprieÂteÂs s'apparentent aÁ celles de
explain the observed ®eld behaviour. The delays l'argile brune de Londres. Ces reÂsultats montrent
experienced in the ®eld are also recovered by que la rupture progressive est consideÂrable et
the analyses. Progressive failure is generated expliquent entieÁrement les pheÂnomeÁnes observeÂs
primarily by the high lateral stresses in the soil sur le terrain. Les analyses ont eÂgalement repro-
prior to excavation. The rupture surface gener- duit le retardement de la rupture sur le terrain.
ated spreads horizontally from the toe as the La rupture progressive est causeÂe principalement
soil swells, and differs signi®cantly from the par de fortes contraintes lateÂrales preÂsentes dans
critical surface predicted by limit equilibrium le sol avant les travaux d'excavation. A mesure
analysis. The average strength on it at collapse que le sol gon¯e, la surface de rupture se propage
is signi®cantly lower than that obtained by back sur le plan horizontal aÁ partir du pied du talus et
analysis by limit equilibrium methods. It be- est fort diffeÂrente de la surface critique preÂdite
comes lower as lateral stress increases, but the par l'analyse d'eÂquilibre limite. Au moment de
effect is compensated by the increasing depth of l'affaissement, la reÂsistance moyenne sur cette
rupture. The roll of the hydraulic surface surface est bien infeÂrieure aÁ la valeur obtenue
boundary condition, which is controlled by par reÂtro-analyse aÁ l'aide des meÂthodes d'eÂqui-
climate, is important in controlling collapse. libre limite. Elle baisse aÁ mesure qu'augmentent
The implications of the ®ndings for monitoring les contraintes lateÂrales, mais cet effet est com-
and for remedial works in existing slopes is penseÂ par la profondeur de plus en plus grande
discussed. de la rupture. L'eÂtat limite de la surface hydraul-
ique, qui est reÂgi par le climat, joue un roÃle
important dans le preÂvention de l'affaissement.
Les auteurs examinent l'application de ces reÂsult-
KEYWORDS: clays; consolidation; numerical model- ats aÁ la surveillance des talus et aux travaux de
ling and analysis; slopes. reÂ¯ection.
INTRODUCTION
Manuscript received 21 November 1995; revised manu- The delayed collapse of old railway cutting slopes in
script accepted 15 August 1996.
Discussion on this paper closes 2 March 1998; for further stiff clay by the formation of deep seated slides has
details see p. ii. been studied extensively (DeLory, 1957; Henkel,

Imperial College of Science, Technology and Medicine, 1957; Skempton, 1948, 1964, 1970, 1977; James,
London. 1970). These studies have shown that the average
{ Geotechnical Consulting Group, London. drained strength operating at collapse was consider-
953
Downloaded by [ Hongkong Univ Of Science and Technology] on [03/10/24]. Copyright © ICE Publishing, all rights reserved.
954 POTTS, KOVACEVIC AND VAUGHAN
ably less than the peak strength measured in the where ô is the average shear stress at collapse and
laboratory. This has been attributed to several causes, ôP and ôR are the average peak and residual shear
such as loss of strength with time, and progressive strengths at the current normal effective stress
failure. More recently, delayed slips of both cutting along the eventual rupture surface.
and embankment slopes in plastic clays have become For progressive failure to play a role in the
a problem on British motorways, although so far delayed collapse of cut slopes the soil must exhibit
most of these slips have been shallow (Perry, 1989). brittle properties and nonlinear strains must be
A programme of ®nite element analysis to examine mobilized along any potential rupture surface. Sim-
the potential causes of these slips has been con- ple methods of analysis, such as limit equilibrium,
ducted recently, supported by the Ministry of Trans- are incapable of dealing which such complexities
port. The ®rst stage of this work, in which the and therefore little progress has been made theoreti-
development of deep-seated slides was examined to cally in quantifying the possible effects of progres-
test the effectiveness of the method of analysis, is sive failure. However, the use of the ®nite element
reported here. The soil was assumed to have the method to analyse the recent end-of-construction
properties of Brown London Clay, as most is known slip in the Carsington embankment (Potts et al.,
about slopes and soil properties for this material. 1990; Chen et al., 1992) and other analyses of
Parametric studies into the effects of the initial in embankments (Dounias et al., 1988, 1989; Vaughan
situ stress, the hydraulic boundary conditions and the et al., 1989) have shown that progressive failure can
slope geometry were investigated, and the results be accurately reproduced and quanti®ed. However,
compared with ®eld data. these previous studies involved large embankments
Collapse is delayed primarily by slow swelling which were subjected to loading and consolidation.
and pore pressure equilibration (Vaughan & Wal- Here the same techniques are applied to cuttings
bancke, 1973). An excavation in saturated stiff which involve unloading and swelling. In addition
plastic clay of low permeability is undrained. It the analyses include coupled consolidation which
unloads the soil and pore water pressures become enable the history of swelling and, in particular, the
depressed both beneath and adjacent to the excava- time to collapse to be predicted.
tion. With time the soil swells as the pore water
pressures equilibrate to the long term steady see-
page regime imposed by the hydraulic boundary SOIL PROPERTIES ADOPTED
conditions on the excavation slopes. With swelling The constitutive model employed to model soil
or equilibration, mean effective stresses reduce and behaviour is the same as that used previously by
the stress state approaches failure. A detailed dis- Potts et al. (1990). This is an elasto-plastic model
cussion of this process is given by Bishop & in which softening behaviour is accounted for by
Bjerrum (1960). allowing the angle of shearing resistance, ö9, and
Progressive failure refers to the non-uniform the cohension intercept, c9, to vary with the devia-
mobilization of shear strength along a potential toric plastic strain invariant, åDp , as shown in Fig. 1.
rupture surface. If brittle soil is loaded non-uni- The deviatoric plastic strain invariant, åDp , is de-
formly, some elements of soil will reach peak ®ned as
strength before others and a rupture surface will
begin to develop. With further loading the post-peak (åpD)2
strains along the rupture surface increases, and the 2[(åp1 ÿ åp2 )2 (åp2 ÿ åp3 )2 (åp3 ÿ åp1 )2 ]=3 (2)
strength reduces from peak towards residual. Final
collapse of the soil mass can occur before the The model requires the speci®cation of peak
rupture surface has fully developed. At collapse, (ö9P , c9P ) and residual (ö9R , c9R ) strength, the angle
part of the rupture surface has formed and lost of dilation (ø), pre-peak stiffness (E), stiffness in
strength post-peak, and a part has not yet formed. unloading (Eu ) and the rate at which strength is lost
Thus, the average strength of the soil mass at with strain (åpDP , åpDR ). As the analysis involves
collapse (the operational strength) must be less than swelling it is also necessary to specify a permeabil-
the peak strength and greater than the residual ity (k). In the previous analyses examining progres-
strength. The mechanism of progressive failure was sive failure (Dounias, 1987; Potts et al., 1990) it
recognized a long time ago (Terzaghi & Peck, 1948; was found that stability was not affected strongly by
Taylor, 1948). The signi®cance in stability problems the pre-failure stiffness used. A simple elasto-plastic
of the different peak and residual strengths of soils soil model is used here, with stiffness prior to
was emphasized by Skempton (1964), who intro- failure given by an elastic model in which Poisson's
duced a residual factor R as a measure of the ratio, ì, is constant, and Young's modulus, E, varies
difference between the average ®eld strength at col- with mean effective stress, p9, but not with shear
lapse and the strengths measured in the laboratory: stress level. No distinction is made between unload-
ing and loading. Consolidation and swelling follow
R (ôP ÿ ô)=(ôP ÿ ôR ) (1) approximately the same modulus in stiff clays and
DELAYED COLLAPSE OF CUT SLOPES IN STIFF CLAY 955
φ9, c9
φ9P, c9P
φ9R, c9R
0 εPDP εPDR εPD
Fig. 1. Variation of the angle of shearing resistance ö9, and the

cohesion intercept c9, with the deviatoric plastic strain invariant åPD
(Potts et al. 1990)
therefore the Young's modulus used in the present shows the data adopted for peak strength. These are
analysis is based on the appropriate swelling mod- taken directly from triaxial tests by Sandroni (1977)
ulus. on 150 mm and 265 mm diameter samples from the
The parameters used are listed in Table 1. Fig. 2 Broad Oak dam site in Kent. The clay had a ®ssure
Table 1. Material properties assumed in the analyses

Property Assumed value
Bulk unit weight, ã 18´8 kN=m3
Peak strength (intact) c9P 20 kPa, ö9P 208
Discontinuity strength c9S 0 kPa, ö9S 208
Peak strength (bulk) c9P 7 kPa, ö9P 208
Residual strength c9R 2 kPa, ö9R 138
Plastic strain at peak åpDP 5%
Plastic strain at residual åpDR 20%
Poisson's ratio, ì 0´20
Young's modulus, E 25( p9 100), (min: 4000 kPa)
Angle of dilatation, ø 0 (or 208)
Coeff. of permeability, k varies, see Fig. 5
Coeff. earth pres. at rest, K 0 1´0, 1´25, 1´5, 1´75, 2´0 or 2´15
p9 is the mean effective stress in kPa
Vertical 
 265 mm dia.
Inclined 
Vertical 
 150 mm dia.
Inclined  Possible intact strength
Vertical 71 mm dia. c9 5 20 kPa, φ9 5 20°
100
Inclined 38 mm dia.
(σ1 2 σ3)/2: kPa
Average strength
265 mm dia. samples
c9 5 7 kPa, φ9 5 20°
Possible discontinuity strength

c9 5 0, φ9 5 2 0°
0
0 100 200 300
(σ91 1 σ93)/2: kPa
Fig. 2. Peak strength in terms of effective stress from undrained triaxial compression tests on large
samples of ®ssured Brown London Clay (Sandroni, 1977)
spacing of the order of 20 mm or less, and the tests strains and strain-softening are strongly in¯uenced
on larger samples gave a strength envelope accord- by the strain energy released by swelling. Very
ing to Mohr±Coulomb of c9P 7 kPa, ö9P 208, little experimental data are available for swelling
which was thought representative of the ®ssured along different stress paths. Some swelling data
structure. Well de®ned shear surfaces were formed, from odeometer tests (Apted, 1977) are shown on
at about 408 to the vertical. Fig. 4. Also shown is the swelling curve obtained
Measurements of residual strength are sum- by plotting depth=water content pro®les (Skempton
marized on Fig. 3. Both ring shear and ®eld data & Henkel, 1957) as an intrinsic swelling line
are shown. The strength back-calculated from ®eld (Burland, 1990). The swelling lines predicted by
slips is slightly higher, probably due to the less the simple elasto-plastic constitutive model adopted
planar shear surfaces usually formed in the ®eld. are also shown.
The ®eld strength of c9R 2 kPa, ö9R 138 was The coef®cient of permeability was assumed to
adopted. vary with depth as shown on Fig. 5, where ®eld
The clay in a slope swells with time, and the data for both the Upper Lias Clay and the London
50
13°
, φ9R 5 ses
Back-analysed slides 2 kPa ment analy
c9R 5
Shear strength τf: kPa
(Chandler, 1984a; Bromhead, 1978) it e e le

fi n
d in
as sume
Stress range
ear test
in analyses Ring sh t al., 1981)
e
(Lupini
0
0 50 100 150
Normal effective stress σ9n: kPa
Fig. 3. Drained residual strength of Brown London Clay
10 Field swelling from

water content/depth profiles
(Skempton & Henkel, 1957)
Laboratory oedometer
tests (Apted, 1977)
Finite element predictions

Volume change ∆v/v : %
K0 5 1.0
K0 5 1.5
K 5 2.0
0
5
0
0 100 200
Vertical effective stress σ9v: kPa
Fig. 4. Observed and predicted swelling of Brown London Clay
Coefficient of permeability, k : m/s
10212 10211 10210 1029
0
Brecciated Upper Lias clay, Empingham.

Laboratory and in situ tests.
(Chandler, 1974)
London Clay, Wraysbury.

Laboratory test on large sample.
(Garga, 1970)
10
London Clay, Wraysbury.

In situ tests (Garga, 1970)
London Clay, University of Kent.
Change in permeability with depth
(assuming k 5 1029 m/s at GL)
(Vaughan et al., 1983)
20
Assumption made in analyses
Depth: m
Fig. 5. Variation in permeability with depth assumed in the analyses compared with ®eld and
laboratory data
Clay are also shown. This permeability, together ment is 0´5 m thick. A displacement scale equiva-
with the swelling modulus from Fig. 4 for lent to this is shown in the ®gure.
K 0 1:5, gives a coef®cient of swelling, cs Figure 7 shows a comparison between the
2:7 m2 =year at 5 m depth. Walbancke (1976) stress±strain behaviour observed in two unconsoli-
quotes ®eld rates of swelling equivalent to cs dated undrained triaxial tests on 265 mm diameter
3:2 m2 =year for the Brown London Clay (see also samples of Brown London Clay (Sandroni, 1977)
Chandler (1984a)). and the predictions made from the model adopted.
The rate at which such a clay strain-softens The comparison is reasonable, although the genera-
post-peak is dif®cult to establish. A reasonable tion of pore water pressure during shear is under-
assumption was made based on precedent from estimated.
other analyses. Parametric studies, performed as
part of the present investigation, have shown that
this property, when varied over a realistic range, FINITE ELEMENT ANALYSES
does not have a major in¯uence on the results Coupled consolidation analyses were performed
obtained (Kovacevic, 1994). One or more shear using the Imperial College Finite Element Program
surfaces are likely to form as peak strength is (ICFEP). Plane strain eight-noded isoparametric
mobilized in real plastic clays, and loss of strength elements with reduced integration were used.
post-peak occurs as a consequence of sliding on While all eight nodes of an element had displace-
these surfaces. Such thin discontinuities are not ment degrees of freedom only the four corner
reproduced in the ®nite element analysis, in which nodes had pore water pressure degrees of freedom.
the minimum thickness of a rupture surface is An accelerated modi®ed Newton±Raphson scheme
approximately half the thickness of an element. with a sub-stepping stress point algorithm was
This must be taken into account when specifying employed to solve the nonlinear ®nite element
the rate at which strength is lost post-peak. The equations (Potts & Ganendra, 1994). Table 2 sum-
strength±displacement plot for a shear zone of marizes details of all the analyses performed. A
thickness T in simple shear is shown in Fig. 6. typical ®nite element mesh is shown in Fig. 8. No
Typically, in the present investigation, half an ele- horizonal displacement was allowed on the vertical
σ9 5 137 kPa
100
∆ τ
75 T γ 5 ∆/T
Shear stress, τ (kPa)
50
25
0
0 5 10 15 20 25 30
Shear strain, γ (%)
0 0.05 0.1 0.15

Displacement across a 0.5 m thick layer, ∆ (m)
[1/2 the thickness of a typical element]
Fig. 6. Assumed stress±strain and stress±displacement relationships

in simple shear
boundaries while the bottom boundary was com- adequate and in addition the residual stresses of all
pletely ®xed in both the vertical and horizontal integration points were monitored and kept small.
direction. The initial pore pressure was assumed to For the early stages of an analysis residual stresses
be hydrostatic with a surface boundary value of were kept below 2 kPa, however, when collapse
ÿ10 kPa, the average surface value indicated by was approached more stringent conditions were
®eld measurements (Vaughan, 1994). introduced and they were kept below 0´1 kPa. To
Excavation was simulated by removing layers of achieve this level of accuracy small time steps
elements at a uniform rate. For a 10 m high (typically 0´0025 year) and a large number of
cutting, excavation was complete in an overall time iterations (typically 200) were required towards the
of 3 months. There was no signi®cant swelling end of the analyses. The time at collapse was
during excavation. deduced by plotting horizontal mid-slope displace-
After excavation swelling was allowed with a ment against time. Once the collapse time was
surface boundary suction of ub ÿ10 kPa at the exceeded the solution became unstable, since the
excavated slope, and on the original ground sur- slope could not then be in equilibrium. This was
face. Seasonal ¯uctuations in surface boundary always con®rmed by running the ®nal increment
conditions are not modelled in the analyses re- for a small time step of 0´0025 year and a large
ported here. The vertical and bottom boundaries of number of iterations (typically 400). Instability was
the mesh were assumed to be impermeable. Swel- indicated when deformations increased according
ling was simulated in the ®nite element analyses to the number of iterations without an improve-
by applying increments of time, Ät, with no ment in convergence (Kovacevic, 1994).
change to the other boundary conditions. For the
®rst time step after excavation Ät 0:25 year, for
the second Ät 0:5 year. For subsequent steps NON-SOFTENING ANALYSES
Ät 1 year until collapse was approached, when Six analyses of a 10 m high 2:1 slope were
the time step had to be reduced substantially. performed with non-softening soil, with residual
Monitoring of convergence is dif®cult for ana- strength equal to the peak strengths (NS1 to NS6
lyses involving strain-softening. A procedure simi- in Table 2). Analyses of 3:1 and 2´5:1 slopes, 10 m
lar to that employed by Potts et al. (1990) was high, were also performed with a non-softening
used. The norms for iterative nodal displacements, soil but collapse was not predicted.
loads and pore water pressures were all kept less In analyses NS1 to NS3 the angle of dilation
than 1% of the norms of the associated incremental was assumed to be ø 0, and the initial in-situ
values. However, this procedure alone was not stress was assumed to be given by K 0 1, 1´5 and
75
Finite element
Shear stress, (σ1 2 σ3)/2 (kPa) predictions
50
σ 39 5 137 kPa
Measured behaviour σ 39 5 112 kPa
25 (Sandroni, 1977)
0
0 5 10 15
Axial strain, εa (%)
0 5 10 15
325
350 Finite element

predictions
Pore pressure, u (kPa)
σ 39 5 137 kPa
375
σ 39 5 112 kPa
400
Measured behaviour
(Sandroni, 1977)
425
Fig. 7. Measured and predicted behaviour in undrained triaxial tests on Brown

London Clay
2. In each case collapse was predicted during the by performing a separate long term steady state
swelling process, with a well de®ned rupture zone, seepage analysis. The associated change with time
of the same shape in each analysis. Figure 9(a) of the mid-slope horizontal displacement, äh , is
shows the change with time of the predicted aver- given in Fig. 9(b). It is evident from these plots
age pore water pressure on this rupture surface, that collapse occurs suddenly, after a delay which
expressed as pore pressure ratio, ru u=ó v , depends on the value of K 0 assumed.
where u pore water pressure, and ó v is the Analyses NS4 to NS6 were similar to NS1 to
predicted vertical stress on the rupture surface. NS3 except that the angle of dilation was ø ö9.
This de®nition differs slightly from the conven- The results from these analyses are shown in Fig.
tional one. ru is the average value along the 10. Well de®ned rupture surfaces were again pre-
rupture surface. All three analyses gave ru dicted, of the same shape in each analysis, but of
0:235 at collapse. Also shown on this ®gure is the different shape to that predicted with ø 0. This
equilibrium value of ru 0:3. This is the value is why there is a slight difference between the
that would have been obtained after a very long steady state equilibrium values of ru on Figs 9(a)
time if collapse had not occurred. It was calculated and 10(a). The ®nal values of the average pore
960
Table 2. Results of analyses performed
Analysis Height Slope K0 ø (deg) ôf (kPa) ó 9n (kPa) ru ru ru R (A) tf (years) Notes
(m) (cot â) excavation collapse steady state (B)
NS1 10 2 1´00 0 21´2 42´1 ÿ0´068 0´236 0´300 ± 12´0 (1)
NS2 10 2 1´50 0 21´4 42´2 ÿ0´193 0´235 0´300 ± 20´0 (1)
NS3 10 2 2´00 0 21´5 42´2 ÿ0´347 0´234 0´300 ± 36´0 (1)
NS4 10 2 1´00 20 21´4 38´7 ÿ0´098 0´250 0´290 ± 51´0 (1)
NS5 10 2 1´50 20 21´4 38´9 ÿ0´252 0´251 0´290 ± 111´0 (1)
NS6 10 2 2´00 20 21´4 38´6 ÿ0´462 0´253 0´290 ± 201´0 (1)
POTTS, KOVACEVIC AND VAUGHAN

S1 10 3 1´00 0 16´3 40´0 ÿ0´169 0´240 0´315 0´436 23´8
S2 10 3 1´25 0 18´3 52´7 ÿ0´057 0´181 0´340 0´583 13´5
S3 10 3 1´50 0 20´2 61´5 0´044 0´238 0´375 0´623 14´5
S4 10 3 1´75 0 21´6 67´5 0´039 0´265 0´385 0´643 23´0
S5 10 3 2´00 0 22´0 62´7 ÿ0´047 0´264 0´380 0´500 31´6
S6 10 3 2´15 0 22´0 62´9 ÿ0´098 0´266 0´380 0´508 44´7
S7 10 3 2´00 0 21´2 68´4 ÿ0´036 0´243 0´300 0´693 100´0 (2)
S8 10 2 1´50 0 23´4 64´8 ÿ0´105 0´027 0´329 0´434 3´45
S9 10 2 2´00 0 23´1 75´2 ÿ0´119 0´070 0´686 5´67
S10 10 2´5 2´00 0 22´5 70´6 ÿ0´136 0´087 0´636 8´7
S11 10 4 2´00 0 21´0 67´1 0´077 0´396 0´425 0´680 203´0
S12 15 2 1´50 0 35´6 105´2 ÿ0´064 ÿ0´026 0´368 0´412 3´3
S13 15 2 2´00 0 34´6 114´8 ÿ0´141 ÿ0´034 0´622 4´6
S14 15 3 2´00 0 35´8 113´7 0´049 0´263 0´437 0´539 51´25
S15 15 4 2´00 0 30´2 97´5 0´049 0´361 0´445 0´606 145´0
S16 6 2 2´00 0 14´2 31´9 ÿ0´384 0´107 0´393 12´65
S17 6 2´5 2´00 0 13´7 36´1 ÿ0´172 0´236 0´285 0´591 36´0
S18 6 3 2´00 0 (3)
(A) Residual factor; (B) time to failure; (1) no strain-softening; (2) surface suction 20 kPa; (3) no collapse after 150 years.
Free boundary: constant uB
DELAYED COLLAPSE OF CUT SLOPES IN STIFF CLAY

Elements excavated
in layers
impermeable
impermeable
∆h 5 0
∆h 5 0
Constant uB
After excavation
Fixed boundary: impermeable
0 Scale 20 m
Fig. 8. Typical ®nite element mesh
961
0.5
at equilibrium: r u* 5 0.30
0.25 K0 5 1.0 (NS1) K0 5 1.5 (NS2) K0 5 2.0 (NS3)

Average pore pressure ratio, r u*
0
5 10 15 20 25 30 35 40 45 50
Time since excavation, t (years)
20.25
(a)
20.5

0 5 10 15 20 25 30 35 40 45 50
0
Displacement, δh (m)
K0 51.0 (NS1)
K0 51.5 (NS2)
0.5
K0 52.0 (NS3)
(b)
Fig. 9. Non-softening analyses (NS1), (NS2) and (NS3): 2:1 slope, 10 m high, K 0 1´0, 1´5 and 2´0,
angle of dilation `zero'. (a) Development of average pore pressure ratio on the rupture surface, and
(b) mid-slope horizontal displacement with time
water pressure at collapse were all ru 0:25, pressure is controlled by the permeability of the
which is slightly higher than for the ®rst three clay and the thickness of the rupture zone. This
analyses. accounts for the gradual increase in displacement
Figure 10(b) shows the change of mid-slope which is predicted with ø ö9 once a rupture
displacement with time. When rupture develops surface has formed, and for the higher average
and the average pore water pressures become con- pore water pressure at collapse. To check the
stant the displacements do not accelerate as in Fig. validity of the later statement an attempt was made
9(b). Instead there is a gradual increase in dis- to restart analyses NS4 to NS6 but with the angle
placement with time. This is due to two compen- of dilation, ø, reduced to zero. In each case it was
sating effects. During swelling pore water pressures not possible to maintain equilibrium of the solution
are generally increasing with time, whereas in the and very large displacements occurred. The pore
thin rupture zone the soil is dilating and the pore water pressures generated by dilation essentially
water pressures are reducing. At collapse the rate act as a partial brake on post-rupture movements.
of decrease in pore water pressure in the rupture This effect is of great practical signi®cance, since
zone is exactly matched by the increase due to it helps to prevent rapid post-collapse movement of
swelling. The generation of negative pore water slides in strain-softening clays (Vaughan, 1994).
0.5
K0 5 1.0 (NS4) K 5 1.5 (NS5)
0 K0 5 2.0 (NS6)
0.25
0
50 100 150 200
20.25
(a)
20.5

0 50 100 150 200
0
K0 5 1.0 (NS4)
K0 5 1.5 (NS5)
0.5
K0 5 2.0 (NS6)
(b)
Fig. 10. Non-softening analyses (NS4), (NS5) and (NS6): 2:1 slope, 10 m high, K 0 1´0, 1´5 and 2´0,
angle of dilation 208. (a) Development of average pore pressure ratio on the rupture surface, and (b)
mid-slope horizontal displacement with time
In reality, dilation would be restricted in the ®eld the rupture surface predicted, and the stresses act-
to a much thinner rupture zone than that predicted ing on it. The angles which the predicted rupture
by the analyses, where the rupture surface is ap- surfaces (velocity characteristics) make with the
proximately half an element thick (typically 0´5 m). plane on which the major principal stress acts are
Equilibration of pore water pressure would occur (458 ø=2). The shear stresses acting on these
more rapidly in the ®eld. Thus, if it is assumed that rupture surfaces are given by:
dilation occurs post-peak and only on thin shear
surfaces, as occurs in laboratory tests (Sandroni, c9 cos ö9 cos ø ó 9n sin ö9 cos ø
ôf (3)
1977), the analyses in which ø 0 are likely to 1 ÿ sin ö9 sin ø
give a better representation of undrained effects in
clay in which discontinuities develop post-peak, When ø ö9 equation 3 reduces to the Cou-
unless an unrealistically large number of elements lomb failure criterion, namely:
are used to allow a thin rupture zone to develop.
Consequently, for analyses involving strain-soften- ô f c9 ó 9n tan ö9 (4)
ing, an angle of dilation, ø 0, was adopted. If ø 6 ö9 the strength on the rupture surface
The angle of dilation also effects the position of from equation (3) does not reduce to the Coulomb
criterion. For instance, if ø 0, equation (3) re- it is therefore not surprising that the rupture sur-
duces to: faces from the ®nite element analyses are depen-
dent on ø. It may be noted that at the soil surface
ôf c9 cos ö9 ó 9n sin ö9 (5) the major principal effective stress is vertical and
For clays the differences are small, but they the rupture surfaces should be inclined at an angle
need to be taken into account when comparisons of (458 ÿ ø=2) to the vertical. Inspection of Fig.
are made between the results of ®nite element 11 shows this to be so.
analyses and limit equilibrium calculations in From the limit equilibrium analysis the average
which the Coulomb equation is adopted (Potts et pore water pressure ratio on the slip surface was
al., 1990). If the rupture surface is kinematically ru 0:233 at collapse. This is in good agreement
constrained, then the differences are likely to be with the ®nite element results (ru 0:235 for
mainly in the strength on the rupture surface. If ø 0 and ru 0:25 for ø ö9). The average
the rupture surface is not constrained, the differ- values of the shear and normal stress acting on the
ences are likely to be mainly in the location of the rupture surfaces at failure are plotted in Fig. 12
rupture surface itself. along with equations (4) and (5). As would be
As noted previously, the rupture surfaces pre- expected the results from the ®nite element ana-
dicted by the non-softening ®nite element analyses lyses with ø 0 agree with equation (5) whereas
depend on ø. The two surfaces predicted for the analyses with ø ö9 and the limit equilibrium
ø ö9 and ø 0 are shown in Fig. 11. Also analysis agree with equation (4).
shown is the critical circular slip surface obtained In the softening analyses reported here equation
from a limit equilibrium calculation using the (5) is used. Since there is little apparent kinematic
Bishop rigorous method of slices (Bishop, 1955). constraint on the rupture surfaces, and ø 0 is
In the cut slope problem there is little kinematic assumed, a small difference between the predicted
restraint on the position of the rupture surface and rupture surfaces and those in the ®eld is likely.
Ψ 5 φ9 Ψ 5 0
LE
1:2 FEM 10 m
Fig. 11. Comparison of predicted rupture surfaces from non-softening analyses
25 τf 5 c p9 1 σ n9 tan φ9p
τf 5 c p9 cos φ9p 1 σ9n sin φ9p
Average shear stress at collapse, τ: kPa
20
Legend:
15 Limit equilibrium
FEM – with dilation
FEM – no dilation
10
25 30 35 40 45 50
Average normal effective stress, σn9: kPa
Fig. 12. Average shear stress at collapse predicted by non-softening ®nite element
analyses and calculated by limit equilibrium analysis
SOFTENING ANALYSES of the slip surface has developed after 9 years, and
Results from a typical analysis that the complete slip has developed after 14´5
Results are now shown from a typical analysis years.
of a 10 m high, 3:1 slope, with a value of the Contours of deviatoric strain, åpD (see equation
initial coef®cient of earth pressure at rest, K 0 (2) are given in Fig. 14. The 5% and 20% contours
1:5 (analysis S3). This slope has a conventional represent the start and end of strain softening. A
long term factor of safety of 1´4 based on a limit horizontal shear zone propagates from just below
equilibrium calculation with peak soil strengths the toe of the slope, in a manner similar to that
(c9P 7 kPa, ö9P 208). Fig. 13 shows predicted observed in the ®eld by Burland et al. (1977). This
vectors of current (incremental) displacement 9 base shear rapidly drops from peak to residual as
years after excavation, and 14´5 years after excava- the rupture zone develops.
tion, just before collapse is predicted. The vectors The associated velocity characteristic directions
show the current mechanism of movement and are shown in Fig. 15 as crosses indicating the two
indicate the eventual collapse mechanism. The conjugate potential rupture planes. These are at
absolute values of the displacements are of no right angles because ø 0 was assumed. They are
signi®cance. The ®gures show that the outer part generated at integration points of elements where
(a) 9 years after excavation
Rupture surface at peak
Rupture surface between peak and residual
Rupture surface at residual
0 Scale 25 m
(b) 14.5 years after excavation – just before collapse
Fig. 13. Typical analysis (S3): 3:1 slope, 10 m high, K 0 1´5, surface suction 10 kPa. Vectors of current
displacement
Note: Accumulated deviatoric plastic strains

during excavation εpD , 5%.
Note: Strain-softening starts when εDp 5 5%.

It is complete when εDp 5 20%.
εDp 5 5%
εDp 5 20%
Rupture surface
Rupture surface
εDp 5 5%
εDp 5 20%
εDp 5 50%
0 Scale 25 m
Fig. 14. Typical analysis (S3): 3:1 slope, 10 m high, K 0 1´5, surface suction 10 kPa. Contours of
accumulated deviatoric plastic strain, åpD
local failure is occurring currently. Fig. 15(a) Pore water pressure development is shown as
shows a rather general area of failure near the contours in Fig. 17, with some vertical pro®les on
inner end of the extending rupture surface. Fig. which both changes towards equilibrium and equi-
15(b) shows this area has become much narrower librium pressures are shown. In the middle of the
by the time collapse occurs. As the strength has slope, pore water pressure changes due to excava-
reduced with strain softening, failure is concen- tion are almost constant with depth. For this value
trated in a narrow zone, and the clay on either side of K 0 , they change by an amount almost equal to
has been unloaded and is no longer failing. At the weight of soil removed from the slope above.
collapse there is a more general active failure zone At the crest of the slope the reduction is also
at the back of the slip, where the ®nal rupture nearly linear with depth. It varies from 20 kPa to
surface has not started to strain-soften. 30 kPa due to lateral stress relief, although there is
Figure 16 shows the volumetric strains which no removal of weight at this section. Below the toe
develop due to swelling. The strains within the slip of the slope high lateral stresses remain after
are quite large, and are equivalent to a change in excavation. The reduction in mean total stress is
water content at the toe of the slope before col- less, and the reduction in pore water pressure is
lapse of the order of 1±2%. The analysis predicts signi®cantly less than the weight of soil removed.
an abrupt change in volumetric strains across the The interaction between changes in total stress and
rupture surface. pore water pressure is clear. After 9 years, the pore
Rupture surface
0 Scale 25 m
Fig. 15. Typical analysis (S3): 3:1 slope, 10 m high, K 0 1´5, surface suction 10 kPa. Velocity character-
istics
water pressures at depth have reduced due to pore water pressure is equivalent to ru 0:375.
undrained unloading as the slope deforms outwards Thus, on the rupture surface, the pore water pres-
and releases the locked-in lateral stresses, rather sure has changed 60% of the way towards ®nal
than increasing due to swelling. equilibrium.
These results predict that a complete thin rup- Figures 18(a) and 18(b) show the change with
ture zone has formed just before collapse with the time of the predicted average pore water pressure
stresses lined up to allow rupture on a single ratio on the rupture surface, and the horizontal
surface. Residual strength has been reached along displacement at mid-height of the slope. The aver-
all the horizontal part of the zone (Fig. 13). The age pore water pressure has got close to the value
amount of progressive failure is substantial, with at collapse before the rupture surface at the
the residual factor R 0:623, i.e. the average base of the slide starts to form. The mid-height
strength has dropped 62´3% of the way from peak displacement then starts to accelerate. Swelling
to residual (see equation (1)). The average opera- then involves reducing lateral total stress rather
tional strength along the eventual rupture surface at than increasing pore water pressure. The coupled
collapse is given by ôf =ó n 9 0:328, which is effect of the reducing lateral total stress causes a
equivalent to c9 0, ö9 18:28. The average pore small decrease in average pore water pressure,
water pressure ratio on the ®nal rupture surface although the clay is continuing to swell. There is
has increased from ru 0:044 after excavation to only a relatively small increase in pore water
ru 0:238 at collapse. The long term equilibrium pressure in the ®ve years preceding collapse. Thus
22.0 20.5
22.5 E D
23.0 C
G E C
G F E C
I H G F E 21.5 21.0 ∆v /v %
IH G E B
F D
I
H HI I
H
I
H H
H
G F D B 20.1 A
G G G G
F F E E E
E C C B A A
F F G F E E D D D C B A
F E D D C CB B B B
E E D A
D D C B
C B A
C B A A A A A A
C B A A
B
B A A A A
B A A
A
A
A
B B B B B
B B 20.5
B
B
21.0
F B
C
∆v /v %
H
H G
22.5 21.5
B
23 0 . G E
D
B
H G F E B A
G
I I
I H 22.0 D C B AA
I I G
H HI H H H G F E D
C CB A A
G G H G G G FE E D C D B
F G F F
E F
E F
E F
E D D C B C
B A
F E D D D C C B A
E D C A A
E E D C C B A
D D D
C C
C
B B B
A A A
A 20.1
B A A A
C
C B
B B A
B B A
0 Scale 25 m
Note: In a saturated clay of γ 5 18.8 kN/m3, water content 32%, a volumetric

strain of 1.0% is equivalent to a water content change of 0.7%.
Fig. 16. Typical analysis (S3): 3:1 slope, 10 m high, K 0 1´5, surface suction 10 kPa. Contours of
volumetric strain, åv Äv=v
the analysis implies that the measurement of pore the strain at `b' has become suf®cient for the
pressure in the ®eld is a poor way of monitoring strength to start dropping to residual. Little change
risk of collapse. Final collapse is quite abrupt, has occurred at è'. After 9 years the strength at
without signi®cant warning from acclerating dis- `b' has reached residual. The strength at è'
placement. reaches peak after 12 years. The strength at both
The complex behaviour is shown further in Fig. points then drops due to increasing pore water
19, where the stress paths predicted for two points pressure until collapse occurs at 14´5 years; at `b'
on the eventual rupture surface are plotted. Point the strength is at residual at è' it has just begun to
`b' is on the horizontal sliding surface formed at soften from peak.
an early stage. Point è' is on the rupture surface The predicted deformations are consistent with
at the back of the slip, which forms subsequently. the limited precedents from ®eld observation of
After excavation there is a small suction at `b', London Clay slopes. The best recorded data are of
which is quite close to failure. There is a positive the Kensal Green retaining wall (Skempton, 1977),
pore pressure at è', which is some way from which moved nearly 0´5 m before failing some 29
failure. After 3 years point `b' reaches peak years after its construction. The movements were
strength. Point è' has changed little. After 7 years about 100 mm after 26 years.
Pore pressure kPa
Q Q Q Q Q Q 0
Q
Q P P P 25
P
After excavation Q P
P
Q P
O O O 50
Q O
After 9 years
Q
Q O N 75
Q P O
Q O N
O
Q P O N 100
N N M M
Q P O N M
Q Q N M
Q Q P O N
P M
M L 125
P O N M L
P M L
O N
P P
P M L K 150
O M L
O N L K
L K
O
Change since excavation
Pore pressure kPa
Q Q Q Q Q Q 0
Q
Q P P 25
Q P
P
After excavation Q
P
P
O O 50
Q O O
Q After 14.5 years O
Q O
P O N N 75
N N N
Q P O N
Q O Long term
Q P M 100
Q O M M M
Q Q Q P N
O M
P O N M
O L L 125
P N M L L
P L
P P O N
P P M L
O N L 150
O N M L K K K
O N K
O O M L K
O O N
0 Scale 25 m
Pressure scale
0 100 200 kPa
Fig. 17. Typical analysis (S3): 3:1 slope, 10 m high, K 0 1´5, surface suction 10 kPa. Contours of pore water
pressure
Effect of coef®cient of earth pressure at rest the same steady state seepage solution, but differ
Analyses S1-6 were all performed for a 10 m because the rupture surface for each analysis is
high 3:1 slope but had values of K 0 increasing different. This plot shows that the time to collapse
from 1´0 to 2´15 (the last value is approximately is strongly dependent on K 0 , and reaches a mini-
equal to Kp , the passive coef®cient of earth pres- mum of about 13 years when K 0 is about 1´25.
sure). The variations of average pore water pressure Maximum pore water pressure ratios at collapse
ratio and mid-height horizontal displacement with are similar, ranging from ru 0:225 to ru
time from these analyses are plotted in Fig. 20. It 0:275. However, two types of behaviour may be
may be noted that the ùltimate equilibrium pore seen. First, when K 0 < 1:25, pore water pressures
water pressure' ratios shown on this plot are from reach close to their maximum values soon after
0.5
(a)
0.25
0
5 10 15 20
0 5 10 15 20
0
0.1
0.2 Base rupture surface

starts to form
0.3
0.4 (b)
0.5
Fig. 18. Typical analysis (S3): 3:1 slope, 10 m high, K 0 1´5, surface suction 10 kPa. (a) Development of
average pore pressure ratio on the rupture surface, and (b) mid-slope horizontal displacement with time
excavation. Second, when K 0 > 1:5, there is a The shapes of the rupture surface and the resi-
rapid equilibration of pore water pressures over dual factor, R, for each analysis are shown in Fig.
about 5 years, followed by a much slower increase 21. The rupture surfaces are strongly in¯uenced by
until collapse. This slow increase is accompanied the value of K 0 . They become deeper as K0 in-
by an initially increasing rate of displacement, creases from 1´0 to 1´75. As K0 increases further
which then slows down. Rates of displacement the back of the slip moves out towards the slope.
only accelerate again a few months before collapse. In all cases only a short length of the rupture
There is little warning of collapse from observa- surface is at a strength between peak and residual.
tions of either pore water pressure or movement. The rupture surface at the crest of the slip forms
Collapse is predicted when pore water pressures after the collapse event when K 0 is low. The rup-
are signi®cantly lower than ®nal long term equili- ture surfaces are compared with each other and
brium values. with ®eld surfaces in Fig. 22. The ®eld surfaces
(a) s 5 [σ1 1 σ3]/2, s 5 [σ91 1 σ93]/2 kPa
0 100 200 300
0
t 5 [σ1 1 σ3]/2 kPa
14.5
20 9
b9
40 E Resid
E ual
7
60 3 Pea
k
(b) s 5 [σ1 1 σ3]/2, s9 5 [σ91 1 σ93]/2 kPa

0 100 200 300
0
t 5 [σ1 1 σ3]/2 kPa
E E
20 e
e9
40 14.5 Resid
12 ual
60 Pea
k
b e
Legend:
Total stress paths Numbers denote time after
excavation in years.
Effective stress paths 'E' – End of excavation
Fig. 19. Typical analysis (S3): 3:1 slope, 10 m high, K 0 1´5, surface suction 10 kPa. Typical stress
paths for points (a) `b' and (b) è'
have been geometrically scaled to the same slope collapse, and the amount of progressive failure is
and slope height as the analyses. Observations reduced.
from two cross-sections are Fareham which had a
height of 19 m and a slope of 3:1 are given.
Northolt had a height of 8 m and a slope of 2´6:1. Effect of surface boundary suction
The predicted and observed rupture surfaces are in Analysis S7 was of a 3:1 slope, 10 m high, with
broad agreement. a K 0 2 and with a surface suction raised to
The amount of progressive failure as expressed ub ÿ20 kPa. Such an increase in suction might
by the residual factor, R, predicted from all the be produced by the management of slope vegeta-
analyses are similar. However, the in¯uence of tion. The results may be compared directly with
progressive failure changes as K0 exceeds 1´75, as analysis S5 which had ub ÿ10 kPa. Fig. 24
shown by the values of R in Fig. 23. Initially the shows how the average pore water pressure ratio
in¯uence increases with K0 , then it decreases, on the rupture surface and the horizontal displace-
®nally it increases again. The reason is shown in ment at mid-height develop in the two analyses.
Fig. 21. The length of the horizontal rupture sur- Collapse with the increased boundary suction oc-
face which develops from the toe of the slope curs after 100 years. Analysis S7 was re-run with-
increases with K 0 . With high values of K 0 it out strain softening (Fig. 24). No collapse
extends beyond the crest of the slope. Then the occurred, and the changes in behaviour as the base
back of the ®nal rupture surface develops subse- shear develops about 5 years after excavation were
quently closer to the slope, and the inner part of eliminated.
the horizontal rupture surface is not incorporated The rupture surface for analysis S7 with
in the ®nal rupture surface. As a consequence part ub ÿ20 kPa was slightly deeper than that for
of the horizontal stress in the zone in which the analysis S5. The average pore water pressure ratio
back of the slip ®nally develops is relieved before at collapse on the rupture surface of analysis S7
0.5 at equilibrium:
r u* 5 0.375 r u* 5 0.385 r u* 5 0.38 r u* 5 0.38

.
r * 5 0 34
u K0 5 1.75
r u* 5 0.315 K052.15
K0 5 2.0
K0 5 1.5 K0 5 1.0
K0 5 1.25
0.25
0
10 20 30 40 50
(a)
20.25

0 10 20 30 40 50
0
K0 5 1.0 (S1)
K0 5 1.25 (S2)
0.5 K0 5 1.75 (S4)

K0 5 1.5 (S3)
K0 5 2.0 (S5)
(b)
1 K0 5 2.15 (S6)
Fig. 20. Analyses (S1), (S2), (S3), (S4), (S5) and (S6): 3:1 slope, 10 m high, with varying K 0 . (a) Develop-
ment of average pore pressure ratio on the rupture surface, and (b) mid-slope horizontal displacement with
time
was ru 0:243, compared with ru 0:264 for Effect of slope geometry

analysis S5. The amount of progressive failure Flatter and deeper slopes. Analysis S14 was
increases to compensate, from R 0:5 for analysis made for a 3:1 slope, 15 m high, and can be
S5 to R 0:693 for analysis S7, which is a drop compared with analysis S5 for a similar slope
in average strength of some 12%. 10 m high. Fig. 25 shows the development of the
Although there is more progressive failure, the average pore water pressure ratio and mid-height
small increase in boundary suction (from ub horizontal displacement with time. Perhaps surpris-
ÿ10 kPa to ub ÿ20 kPa) improves stability sub- ingly, the higher slope takes longer to collapse
stantially. Extrapolation indicates that a further in- (51´25 years compared to 31´6 years). There are
crease in boundary suction to ub ÿ30 kPa would two opposing effects which control the time to
give an ultimate (steady state) pore water pressure collapse. Firstly, there is the magnitude of the pore
equivalent to ru 0:22 on the rupture surface, water pressure change on the rupture surface that
which would be just suf®cient to prevent collapse. is necessary for collapse to occur. Secondly, there
K0 5 1.0 (S1) [R 5 0.436]
K0 5 1.25 (S2) [R 5 0.583]
K0 5 1.5 (S3) [R 5 0.623]
K0 5 1.75 (S4) [R 5 0.643]
K0 5 2.0 (S5) [R 5 0.500]
K0 5 2.15 (S6) [R 5 0.508]
Rupture surface not formed
Rupture surface at peak
Rupture surface between peak and residual
Rupture surface at residual
Fig. 21. Rupture surfaces predicted by the analyses on 3:1 slopes, 10 m high,
with surface suction 10 kPa and varying K 0
is the average length of the drainage path from brium values whereas for the 15 m high slope
the slope boundary to the rupture surface. For collapse occurs after a change of 55%. However,
the 10 m high slope collapse occurs when the the rupture surface for the higher slope is much
pore water pressures on the rupture surface have deeper and penetrates further back into the slope
changed 73% of the way towards their ®nal equili- than the rupture surface for the 10 m high slope.
3:1
(a)
1.0(S1)
2.15(S6)
K0 1.25(S2)
2.0(S5)
1.5(S3)
1.75(S4)
0 20
Scale m
(b) 3:1
Northolt
Fareham 1
Fareham 2
Range of predictions from (a)
Fig. 22. (a) Summary of the rupture surfaces predicted by the analyses on 3:1 slopes, 10 m
high with surface suction 10 kPa and varying K 0 . (b) Comparison of prediction with some
®eld observations of slips in London Clay (James, 1970)
1
Residual factor: R
0.5
0
0.5 1 1.5 2 2.5
Coefficient of earth pressure at rest: K0
Fig. 23. Change in residual factor with the initial coef®cient of

earth pressure at rest
The drainage paths are longer, and the effect of value of ru 0:425. There was more progressive
this dominates. failure than in the 3:1 slope, with a reduction in
Two analyses were made of a 4:1 slope 10 m operational strength of 11%. The 15 m high slope
high (S11) and 15 m high (S15) with a surface collapsed after 145 years, with ru 0:361 (some
suction, ub ÿ10 kPa, and K 0 2. Fig. 25 shows way from ®nal equilibrium, ru 0:445) and with
the development of the average pore water pressure slightly less progressive failure (R 0:606). The
ratio and mid-height horizontal displacement with shorter time to collapse predicted for the higher
time. The 10 m high slope collapsed after 203 slope arises because the smaller degree of pore
years, with ru 0:396 and R 0:68 (compared water pressure equilibrium required (79% com-
with tf 31:6 years, ru 0:264 and R 0:5 for pared to 92% for the 10 m slope) compensates for
analysis S5 of the equivalent 3:1 slope). The pore the longer drainage paths, in contrast to the effect
water pressures were close to the ®nal equilibrium predicted for 3:1 slopes.
0.5
at equilibrium:
r * 5 0.38
u
steady seepage after 250 years
r u* 5 0.30
Surface suction 10 kPa Suction 20 kPa – no softening
Average pore pressure ratio: r u*
0.25
Surface suction 20 kPa
0
10 20 30 40 50 60 70 80 90 100
(a)
20.25

0 10 20 30 40 50 60 70 10 90 100
0
Suction 20 kPa – no softening

0.5 Surface suction 20 kPa
Surface suction 10 kPa

(b)
Fig. 24. Analyses (S5) and (S7): 3:1 slope, 10 m high, K 0 2´0. Effect of increasing surface suction from
10 kPa to 20 kPa on (a) development of average pore pressure ratio on the rupture surface, and (b) mid-
slope horizontal displacement with time
In other respects the analyses predict that a 4:1 geologic features such as sand or silt layers within
slope behaves in a similar manner to a 3:1 slope. the slope lead to more rapid swelling.
There is a rapid development of a horizontal shear
zone from the toe of the slope, accompanied by Steeper and lower slopes. Analyses S9 and S10
discontinuities in the plots of the average pore were performed to examine the stability of 2:1 and
water pressure ratio and displacement against time 2´5:1 slopes, 10 m high, with K 0 2. Results may
(Fig. 25). They indicate that a 4:1 slope is nearly be compared with analysis S5 for the 3:1 slope. The
stable when 10 m high, even in the very long term, development of pore water pressure and deforma-
but a 15 m high slope fails with pore water pres- tion with time are shown in Fig. 26. Both slopes fail
sures some way from predicted equilibrium values. in less than 10 years, with relatively low average
While predicted times to collapse are very long, pore water pressure ratios on their rupture surfaces
such that the eventual collapse may be considered (ru 0:07 and ru 0:087). There was slightly
rather academic, such slopes may be vulnerable if more progressive failure than for the 3:1 slope
0.5 at equilibrium:
r u* 5 0.437 r u* 5 0.445
r u* 5 0.425
r u* 5 0.38
Slope 4:1, H 5 10 m (S11)
Slope 4:1, H 5 15 m (S15)
0.25 Slope 3:1, H 5 15 m (S14)
Slope 3:1, H 5 10 m (S5)
0
50 100 150 200
(a)
20.25

0 50 100 150 200
0
Slope 3:1, H 5 10 m (S5)
0.5
Displacement: δh (m)
1.5 Slope 4:1, H 5 10 m (S11)

Slope 3:1, H 5 15 m (S14)
(b)
2 Slope 4:1, H 5 15 m (S15)
Fig. 25. Analyses (S5), (S11), (S14) and (S15): 3:1 and 4:1 slopes, 10 m and 15 m high, K 0 2´0. (a)
Development of average pore pressure ratio on the rupture surface, and (b) mid-slope horizontal
displacement with time
(R 0:5), with R 0:686 and R 0:636 respec- and 3:1. The 3:1 slope had not collapsed when
tively. Analysis S8 was of a 2:1 slope, 10 m high, pore water pressures reached equilibrium after 150
with K 0 1:5 and a surface suction ub ÿ10 kPa. years. However, inspection of the results show that
The results may be compared with analysis S3 for this slope was on the verge of collapse. The 2´5:1
the 3:1 slope. Again the steeper slope failed earlier slope collapsed after 36 years, with an average
(3´4 years compared to 14´5 years) with a lower pore water pressure ratio on the rupture surface of
average pore water pressure ratio on the rupture ru 0:236, quite close to the equilibrium value
surface (ru 0:027 compared to 0´238). The of 0´285, and with R 0:591. There is slightly less
amount of progressive failure was less than for progressive failure than for the 2´5:1 slope, 10 m
the 3:1 slope (R 0:434 compared to 0´623). high (analysis S10). The 2:1 slope collapsed after
Three analyses of lower slopes 6 m high were 12´6 years, with an average pore water pressure
made (S16, S17 and S18), with K 0 2, a surface ratio on the rupture surface of ru 0:107 and
suction of ÿ10 kPa, and with slopes of 2:1, 2´5:1 with R 0:393.
0.5
Slope 3:1 (S5)

0.25
Slope 2.5:1 (S10)

Slope 2:1 (S9)
0
5 10 15 20 25 30 35
(a)
20.25

0 5 10 15 20 25 30 35
0
Displacement: δh (m)
Slope 2:1 (S9)

0.5 Slope 2.5:1 (S10)
(b) Slope 3:1 (S5)
Fig. 26. Analyses (S5), (S9) and (S10): 3:1, 2:1 and 2´5:1 slopes, 10 m, K 0 2´0. (a) Development of average
pore pressure ratio on the rupture surface, and (b) mid-slope horizontal displacement with time
DISCUSSION coef®cient of permeability, k, on which the time to

The delay predicted for collapse and the average collapse directly depends, the times to collapse
pore water pressure ratio's, ru , on the rupture predicted by the analyses are consistent with ®eld
surface at collapse for all analyses are shown on observations, particularly when higher K 0 values
Fig. 27. Also shown on this ®gure are ®eld ob- (which are likely to be more representative of stiff
servations (Chandler, 1984a, 1984b). Typically, the clay) are assumed. As the present analyses show,
®eld values of ru are derived by averaging results time to collapse varies signi®cantly with factors
from 3±4 piezometers and assuming that this is other than permeability, slope angle and slope
representative for the slope. In some cases the height (Kovacevic, 1994).
value of ru along a hypothetical (or observed) slip The average strength at collapse predicted by all
surface was calculated (Skempton, 1977; Chandler, the analyses is shown on Fig. 28, together with the
1984a, 1984b). The agreement between the ®nite peak and residual strengths. Also shown are the
element predictions and ®eld measurements is operational strengths deduced from back analysis
good. Considering the dif®culties in estimating the of slips by Chandler & Skempton (1974). The
0.5
0.4
Average pore pressure ratio at collapse: r u*
0.3
Legend:
0.2 Slope 2.0:1 – predicted
Slope 2.5:1 – predicted
0.1 Slope 4.0:1 – predicted
Field observations
0
50 100 150 200
Time of failure, t f (years)
20.1
20.2
Fig. 27. Predicted average pore pressure ratio on the rupture surface against time to collapse, and ®eld
observations (Chandler, 1984a, 1984b)
100
Legend:
Average shear stress at collapse, τ: kPa
Peak strength
Residual strength
Field back-analysis c p′ 5 7 kPa, φ′P 5 20°
50
c r′ 5 2 kPa, φ′R 5 13°
c ′ 5 1 kPa, φ′ 5 20°
0
0 50 100 150
Average normal effective stress σ′n: kPa
Fig. 28. Average operational strength at collapse predicted by ®nite element analyses and calculated by limit
equilibrium analysis from ®eld cases (Chandler & Skempton, 1974)
strengths from the ®nite element analyses are gen- just above the `collapse' range (Fig. 29). This
erally lower than those of Chandler & Skempton. implies that although the time to collapse would
However, the ®nite element analyses typically pre- increase collapse would still occur (this can be
dict deeper rupture surfaces than the critical sur- veri®ed by comparing results from analysis S5 and
faces predicted by limit equilibrium analyses. S7). For collapse to be prevented then the surface
Typically, as the rupture surface becomes deeper, suction would have to be increased still further so
the amount of progressive failure increases, and the that the long term value of ru fell below the
average strength, expressed as a stress ratio, de- `collapse' range indicated on Fig. 29.
creases. At low values of K0 the relatively shallow Figure 16 shows a sharp discontinuity in volu-
rupture surfaces approximate to the critical surfaces metric strain across the horizontal rupture surface
predicted by limit equilibrium analyses. Then the of 2±3%, equivalent to a change of water content
average strengths predicted by the ®nite element of 1±2%. Such a discontinuity could be detected
analyses agree with those obtained by limit equili- by site investigation. This would be a potential way
brium back-analysis. of examining the stability of old slopes. As can be
Figure 29 shows values of ru on the eventual seen from Figs 14(a), 15(a) and 16(a) the basal
rupture surface, after excavation, at collapse and part of the rupture surface forms quite early on. Its
after full equilibrium (steady state), for all the formation does not indicate that collapse will even-
analyses, plotted against slope inclination. The pore tually occur, but its presence with continuing
water pressures required for collapse increase with movement on it (which could be monitored) indi-
decreasing slope angle, but so do the equilibrium cates decreasing stability. The absence of move-
pore water pressures. The analysis show that col- ment would indicate stability.
lapse is sensitive to the magnitude of the pore Figure 30 shows a section through a slip on the
water pressures in the slope, and thus to the M11 near Loughton in Essex, in a slope 18 years
equilibrium pore water pressures to which these old. Results from tests on samples taken from just
pressures tend. This plot could be used to assess above and below the rupture surface are given on
the effectiveness of any remedial drainage meas- the ®gure. There is nearly a 4% change in water
ures. For example the effect of increasing the content and two-fold change in remoulded un-
surface boundary suction from ub ÿ10 kPa to drained strength across the rupture surface. This
ub ÿ20 kPa for a 3:1 slope, 10 m high, is to observation is consistent with the numerical ana-
reduce the long term steady state value of ru to lysis.
0.6
Change in steady state value

0.5 due to increasing surface suction
from 10 kPa to 20 kPa
Range of values
0.4 at steady state
0.3
0.2 Range of values

at collapse
0.1
0
1.5 2.5 3.5 4.5
Slope (cotâ)
20.1
20.2
Legend:
Range of values
after excavation after excavation
20.3
at collapse
steady state
20.4
Fig. 29. The variation of average pore pressure ratio on rupture surface after excavation, at collapse and at steady
state with slope inclination
Age of slope at collapse 18 years

4
9.5 m
B
Brown London Clay
Water Liquid Plastic Remoulded

content limit limit quick undrained
shear strength
W: % WL: % WP: % Su: kPa
Sample A 33.2 70 33 26
Sample B 29.4 75 32 62
Fig. 30. Slip on the east side of the M11 near Loughton, Essex (1993)
CONCLUDING REMARKS strain-softening to start; and (iv), when K 0 is low,

No softening analyses an inclined section near the original ground surface
The analyses of 2:1 slopes, 10 m high, in which on which rupture only occurs post-collapse.
the soil did not soften all show well de®ned rup- The deep-seated nature of slips predicted at the
ture surfaces at collapse. The average pore water high values of K0 usually present in in situ London
pressures on the rupture surface at collapse, ru , Clay is typical of ®eld experience. The analyses
were independent of K0 . They are in¯uenced show that, at the higher values of K 0 , the rupture
slightly by any dilation that occurs when the soil surface is developed by progressive failure at a
fails. The depression in pore water pressures due greater depth than the critical surface which would
to dilation acts as a partial brake on the rate of be determined by limit equilibrium analysis. As a
post-collapse movement on the slip surface. consequence the predicted operational strength on
The values of ru at collapse predicted by the the rupture surface is less than that which would
analyses with zero dilation are in good agreement be determined for the same slope with the same
with those obtained from limit equilibrium analysis pore water pressures by limit equilibrium analysis
assuming a circular rupture surface. However the using a search technique to ®nd the critical sur-
position of the rupture surface predicted by the face.
numerical analysis differs from the critical slip The two analyses of a 4:1 slope with high K 0
circle. showed similar behaviour to the 3:1 slopes. Both
The time to collapse increases with K 0 . failed after more than 100 years. The 10 m high
slope was near pore pressure equilibration. The
15 m high slope was not, but time to collapse was
Softening analyses long because of the long drainage paths. However,
The parametric studies of the 3:1 slope, 10 m swelling times in real slopes could be shortened by
high, show that the probability of collapse is rela- geologic features such as sand layers within the
tively insensitive to the value of K0 assumed. The clay.
value of K 0 strongly in¯uences the location of the Six metre high slopes were more stable. While
shear surface and the time to collapse. There is a the 3:1 slope was just stable in the long term, the
slight increase in the amount of progressive failure behaviour of the 2´5:1 slope was similar to that of
with increasing K0 up to a critical value, after the 3:1 slope, 10 m high. Equilibrium long term
which the base rupture surface extends beyond the pore water pressures were decreased, and the
®nal inclined rupture surface, which relieves the amount of progressive failure was also slightly
stress in the slip prior to its ®nal formation and reduced.
reduces the amount of progressive failure. There is considerable variability in the predicted
The rupture surface has four parts, namely: (i) time to collapse of the slopes analysed, bearing
the horizontal base shear, which forms early and in mind that a constant permeability is assumed.
reaches residual well before collapse; (ii) a short The parametric studies on the 3:1 slope gives
transition length between peak and residual; (iii) an 11 , tf , 45 years. The steeper slopes fail in less
inclined back surface, in which rupture has oc- than a decade. The 15 m high, 4:1 slope failed
curred but where strains are not suf®cient for after 145 years, and the 10 m high, 3:1 slope with
increased surface suction after 100 years. The ®eld gations of slips in motorway slopes and their repair
experience also shows considerable variability in conducted by TRRL. The writers are particularly
time to collapse. grateful to Dr Miles O'Reilly and Dr John Perry of
There will be seasonal variation in super®cial TRRL for their help and support. These studies
pore water pressures, not represented in the ana- build on many years of work on this topic at
lyses, superimposed on the general trend (Vaughan Imperial College, London. Professor A. W. Skemp-
& Walbancke, 1973; Walbancke, 1976; Chandler, ton, who was responsible for most of this work,
1984b). Thus collapse will tend to occur in winter has taken a close interest throughout, and has made
when surface suction is zero. many helpful contributions.
The surface hydraulic boundary condition has a
strong effect on stability. The analyses show that
an increase in surface suction from 10 kPa to REFERENCES
20 kPa goes more than half way to stabilize the Apted, J. P. (1977). Effects of weathering on some
3:1 slope. Surface pore water pressures can be geotechnical properties of London Clay. Ph.D. thesis,
reduced by increasing evapo-transpiration through University of London.
the controlled use of vegetation, or by surface Bishop, A. W. (1955). The use of the slip circle in the
drainage which improves pore water pressures be- stability analysis of slopes. GeÂotechnique 5, No. 1,
low the depth of the drains. Farrar (1990) presents 7±17.
Bishop, A. W. & Bjerrum, L. (1960). The relevance of
observations of the effects of counterfort drains in the triaxial test to the solution of stability problems.
a slope in London Clay. These observations suggest ASCE research conference on shear strength of co-
a reduction in pore pressure at depth midway hesive soils, Boulder, CO, pp. 437±451.
between drains of about 11 kPa, and a consequent Bromhead, E. N. (1978). Large landslides in London Clay
average reduction of about 20 kPa. The develop- at Herne Bay, Kent. Q. J. Engng Geol. 11, 291±304.
ment of suitable drilled drainage systems could Burland, J. B. (1990). On the compressibility and shear
also reduce pore water pressures suf®ciently to strength of natural Clays. GeÂotechnique 40, No. 3,
eliminate the risk of long-term deep-seated sliding. 329±378.
These analyses show, beyond reasonable doubt, Burland, J. B., Longworth, T. I. & Moore, J. F. A. (1977).
A study of ground movement and progressive failure
that the mechanism causing deep-seated delayed caused by a deep excavation in Oxford Clay.
slips in these slopes is progressive failure promoted GeÂotechnique 27, No. 4, 557±591.
by swelling. They enable the importance of the Chandler, R. J. (1974). Lias Clay: the long-term stability
controlling variables to be established, and the of cutting slopes. GeÂotechnique 24, No. 1, 24±38.
effect of possible stabilizing measures to be evalu- Chandler, R. J. (1984a). Recent European experience of
ated. The new and sophisticated techniques of landslides in over-consolidated clays and soft rocks.
numerical analysis have an encouraging ability to Proc. 4th Int. Conf. Landslides, Toronto 1, 61±81.
reproduce observed behaviour, provided realistic Chandler, R. J. (1984b). Delayed failure and observed
input data is used. No assumptions need then be strengths of ®rst-time slides in stiff clays: a review.
Proc. 4th Int. Conf. Landslides, Toronto 2, 19±25.
made about behaviour mechanisms, as these are Chandler, R. J. & Skempton, A. W. (1974). The design of
established by the analysis. permanent cutting slopes in stiff ®ssured clays.
No shallow slips were predicted in the analysis GeÂotechnique 24, No. 4, 457±464.
presented here, in which it was assumed that the Chen, Z., Morgenstern, N. R. & Chan, D. H. (1992).
slopes analysed were of homogeneous clay without Progressive failure of the Carsington Dam: a numer-
seasonal pore pressure variation. Such slips can ical study. Can. Geotech. J. 29, Dec., 971±988.
only be explained if the effects of weathering and DeLory, F. A. (1957). Long-term stability of slopes in
pore pressure ¯uctuation near the slope surface are over-consolidated clays. Ph.D. thesis, University of
considered (Vaughan, 1994). The results of ana- London.
Dounias, G. T. (1987). Progressive failure in embankment
lyses in which this is done will be the subject of a dams. Ph.D. thesis, University of London.
further publication, which will consider also the Dounias, G. T., Potts, D. M. & Vaughan, P. R. (1988).
stability of clay embankments. Finite element analysis of progressive failure: two
case studies. Comput. Geotechnics 6, 155±175.
Dounias, G. T., Potts, D. M. & Vaughan, P. R. (1989).
ACKNOWLEDGEMENTS Numerical stress analysis of progressive failure and
The work reported here was performed as part cracking in embankment dams. Report to the Depart-
of a research programme into the design of slopes ment of the Environment, Contract No. PECD
for highway cuttings and embankments sponsored 7=7=222, Building Research Establishment.
Farrar, D. M. (1990). Observations of slope drainage at
by the Transport & Road Research Laboratory Romford, Essex. Research Report 302, Transport &
(TRRL) on behalf of the Ministry of Transport. Road Research Laboratory, Crowthorne.
The writers are grateful for permission to publish Garga, V. K. (1970). Residual strength under large strains
this paper. This research programme was a and the effect of sample size on the consolidation of
development of the extensive surveys and investi- ®ssured clay. Ph.D. thesis, University of London.
Henkel, D. J. (1957). Investigations of two long-term Skempton, A. W. (1970). First-time slides in over-
failures in London Clay slopes at Wood Green and consolidated clays. GeÂotechnique 20, No. 3, 320±324.
Northolt. Proc. 4th Int. Conf. Soil Mech. Found. Skempton, A. W. (1977). Slope stability of cuttings in
Engng, London 2, 378±381. brown London Clay. Proc. 9th Int. Conf. Soil Mech.
James, P. M. (1970). Time effects and progressive failure Found. Engng, Tokyo 3, 261±270.
in clay slopes. Ph.D. thesis, University of London. Skempton, A. W. & Henkel, D. J. (1957). Tests on
Kovacevic, N. (1994). Numerical analyses of rock®ll London clay from deep borings at Paddington,
dams, cut slopes and road embankments. Ph.D. thesis, Victoria and the South Bank. Proc. 4th Int. Conf.
University of London. Soil Mech. Found. Engng, London 1, 100±106.
Lupini, J. F., Skinner, A. E. & Vaughan, P. R. (1981). The Taylor, D. W. (1948). Fundamentals of soil mechanics.
drained residual strength of cohesive soils. GeÂotech- Wiley, New York.
nique 31, No. 2, 181±213. Terzaghi, K. & Peck, R. B. (1948). Soil mechanics in
Perry, J. (1989). A survey of slope condition on motorway engineering practice, 1st edn. Wiley, New York.
earthworks in England and Wales. Research Report Vaughan, P. R. (1994). Assumption, prediction and reality
199, Transport & Road Research Laboratory, Crow- in geotechnical engineering. GeÂotechnique 44, No. 4,
thorne. 573±603.
Potts, D. M., Dounias, G. T. & Vaughan, P. R. (1990). Vaughan, P. R., Kennard, R. M. & Greenwood, D. A.
Finite element analysis of progressive failure of (1983). Squeeze grouting of stiff ®ssured clay after a
Carsington embankment. GeÂotechnique 40, No. 1, tunnel collapse. Proc. 5th Eur. Conf. Soil Mech.
79±101. Found. Engng, Helsinki 1, 171±176.
Potts, D. M. & Ganendra, D. (1994). An evaluation of Vaughan, P. R., Dounias, G. T. & Potts, D. M. (1989).
substepping and implicit stress point algorithms. Advances in analytical techniques and the in¯uence of
Comput. Meth. Appl. Mech. Engng 119, 341±354. core geometry on behaviour. Proceedings of the
Sandroni, S. S. (1977). The strength of London Clay in conference on clay barriers for embankment dams,
total and effective stress terms. Ph.D. thesis, Uni- London, pp. 35±56.
versity of London. Vaughan, P. R. & Walbancke, H. J. (1973). Pore pressure
Skempton, A. W. (1948). The rate of softening in stiff changes and the delayed failure of cutting slopes in
®ssured clay, with special reference to London Clay. over-consolidated clay. GeÂotechnique 23, No. 4, 531±
Proc. 2nd Int. Conf. Soil Mech. Found. Engng, 539.
Rotterdam 2, 50±53. Walbancke, H. J. (1976). Pore pressure in clay embank-
Skempton, A. W. (1964). Long term stability of clay ments and cuttings. Ph.D. thesis, University of
slopes. GeÂotechnique 14, No. 2, 77±101. London.

Potts Et Al 2015 Delayed Collapse of Cut Slopes in Stiff Clay

Uploaded by

Copyright:

Available Formats

Potts Et Al 2015 Delayed Collapse of Cut Slopes in Stiff Clay

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Potts Et Al 2015 Delayed Collapse of Cut Slopes in Stiff Clay

Uploaded by

Copyright:

Available Formats

Potts, D. M., Kovacevic, N. & Vaughen, P. R. (1997). GeÂotechnique 47, No.

Delayed collapse of cut slopes in stiff clay

0 εPDP εPDR εPD

Fig. 1. Variation of the angle of shearing resistance ö9, and the

Table 1. Material properties assumed in the analyses

Possible discontinuity strength

(Chandler, 1984a; Bromhead, 1978) it e e le

Fig. 3. Drained residual strength of Brown London Clay

10 Field swelling from

Finite element predictions

Fig. 4. Observed and predicted swelling of Brown London Clay

Brecciated Upper Lias clay, Empingham.

London Clay, Wraysbury.

London Clay, Wraysbury.

Assumption made in analyses

Shear stress, τ (kPa)

0 0.05 0.1 0.15

Fig. 6. Assumed stress±strain and stress±displacement relationships

Axial strain, εa (%)

350 Finite element

Fig. 7. Measured and predicted behaviour in undrained triaxial tests on Brown

POTTS, KOVACEVIC AND VAUGHAN

DELAYED COLLAPSE OF CUT SLOPES IN STIFF CLAY

Fixed boundary: impermeable

Fig. 8. Typical ®nite element mesh

0.25 K0 5 1.0 (NS1) K0 5 1.5 (NS2) K0 5 2.0 (NS3)

Time since excavation, t (years)

Time since excavation, t (years)

Fig. 11. Comparison of predicted rupture surfaces from non-softening analyses

(a) 9 years after excavation

Rupture surface at peak

Rupture surface between peak and residual

Rupture surface at residual

(b) 14.5 years after excavation – just before collapse

Note: Accumulated deviatoric plastic strains

Note: Strain-softening starts when εDp 5 5%.

(a) 9 years after excavation

(b) 14.5 years after excavation – just before collapse

(a) 9 years after excavation

(b) 14.5 years after excavation – just before collapse

(a) 9 years after excavation

Note: In a saturated clay of γ 5 18.8 kN/m3, water content 32%, a volumetric

(b) 14.5 years after excavation – just before collapse

(a) 9 years after excavation

Change since excavation

Pore pressure kPa

(b) 14.5 years after excavation – just before collapse

Time since excavation, t (years)

0.2 Base rupture surface

(b) s 5 [σ1 1 σ3]/2, s9 5 [σ91 1 σ93]/2 kPa

r u* 5 0.375 r u* 5 0.385 r u* 5 0.38 r u* 5 0.38

Time since excavation, t (years)

Time since excavation, t (years)

0.5 K0 5 1.75 (S4)

was ru  0:243, compared with ru  0:264 for Effect of slope geometry

K0 5 1.25 (S2) [R 5 0.583]

K0 5 1.5 (S3) [R 5 0.623]

K0 5 1.75 (S4) [R 5 0.643]

K0 5 2.0 (S5) [R 5 0.500]

was ru 0:243, compared with ru 0:264 for Effect of slope geometry