Bolton, M. D. & Stewart, D. I. (1994). Gtkmxhnique 44, No.

1, Ill-127

The effect on propped diaphragm walls

of rising groundwater in stiff clay

M. D. BOLTON* and D. I. STEWART?

The Paper investigates the stability and service- L’article etudie la stabilite et la capacite de service
ability of propped diaphragm walls constructed in de murs diaphragmes Ctayb, construits dans une
situ in stiff clay, first when an excavation is formed argile raide, premierement lorsqu’une excavation
in front, second when the groundwater level recov- est realisee devant le mur, deuxiemement lorsque
ers, and third when it is raised. Three centrifuge le niveau de la nappe se retablit et troisiemement
model tests were conducted to observe the behav- lorsque ce niveau est depasse. Trois essais de
iour of these structures. Simplified behavioural modelisation en centrifugeuse ont ete mis en oeuvre
mechanisms are shown to provide a method of pour etudier le comportement de ces structures.
analysis for a propped diaphragm wall which would Des mecanismes simplifies de comportement sont
permit a designer to calculate the soil and structur- proposes afin d’obtenir une methode d’analyse,
al deformations without recourse to expensive and pour un mur diaphragme etaye, qui permettrait au
complicated computer-based analyses. The essen- constructeur de calculer les deformations du sol et
tial conditions are the satisfaction of equilibrium de la structure sans avoir recours a des analyses
and compatibility requirements, and the ability to sur ordinateur onereuses et compliquees. Les condi-
use stress-path specific stress-strain data. Where tions majeures portent sur des conditions
there was little structural displacement, the earth d’equilibre et de compatibilite ainsi que sur la pos-
pressures were calculated using a bilinear approx- sibilite d’utiliser les don&es specifiques contrainte-
imation to the measured stress path of kaolin in deformation du chemin de contrainte. Pour les
one-dimensional swelling. deplacements structuraux faibles, la pression des
KEYWORDS: centrifuge modelling; clays; deforma- terres a et6 calculee a partir d’une approximation
tion; diaphragm and in situ walls; earth pressure; bilineaire du chemin de contrainte obtenur sur un
groundwater. kaolin lors d’un gonflement unidimensionnel.

INTRODUCTION roof beam in a cut-and-cover tunnel, or even part

The high value of land in many urban areas of the carriageway in a road cutting.
makes it desirable that the land-use of any given Many urban areas in the UK are situated on
site is maximized. One solution is to use a heavily overconsolidated clays, posing a number
propped diaphragm wall to support a deep base- of potential difficulties for the designer of a
ment. A reinforced concrete diaphragm wall can propped diaphragm wall. CIRIA Report 104
be cast in situ into a trench that is temporarily (Padfield & Mair, 1984) which is often used for
held open by bentonite slurry. Once the concrete the design of diphragm walls in the UK, offers
has gained sufficient strength, the main excava- design guidance in terms of ultimate limit states
tion can proceed with minimal disturbance to the and suitable safety factors, but could not specify
retained soil. When the proposed use of the exca- pressure distributions applicable to serviceability
vation permits, it is common practice to prop a limit states. The deformations required for stiff
diaphragm wall at one or more levels. This has clay to reach an ultimate limit state may be too
the advantage of reducing both the wall penetrat- large for the retaining wall to be considered ser-
ion required for stability and the displacement of viceable, certainly when the wall is well propped.
the retained soil caused by excavation. The props Also, the factors of safety employed against col-
can often form part of the permanent structure; lapse may mean that the stresses in service differ
possibly the floor slabs in a deep basement, the greatly from those assumed.
It has been observed that the groundwater
levels in many urban areas and notably in
Manuscript received 14 May 1992; revised manuscript
accepted 10 May 1993.
London (Simpson, Blower, Craig & Wilkinson,
Discussion on this Paper closes 1 July 1994; for further 1989) have been rising because of reduced indus-
details see p. ii. trial extraction of water from underlying aquifers.
* University of Cambridge. Increases in pore-water pressures in stiff clay near
t University of Leeds. a retaining wall may also be caused by water

111
112 BOLTON AND STEWART

leaking from damaged water mains or sewers. increased from 2 m in DWC20 to 5 m in DWC21
This has led to concern about the performance of and to 10 m in SSIOl. The wall bending stiffness
retaining structures m stiff clay subject to swell- in DWC20 was made very large in order to verify
ing. Clay swelling has the potential to cause the mode of deformation as one of rotation about
undesired wall movements or, if such movements the prop. The walls in DWC21 and SSIOI were
are resisted by a retaining wall, structural distress. made to model more typical diaphragm wall
Centrifuge tests that modelled diaphragm walls stiffnesses. The selected prop stiffnesses were suff~-
propped at excavation level are reported in this cient to ensure that their axial compression
Paper. A shallow penetration model (DWC20) remained much smaller than the maximum soil
was used to investigate the effects of wall move- displacements observed in the tests.
ment: on excavation the wall rotated about the It was desired that the soil models should
prop, the top of the wall moving towards the deform in conditions of plane strain, so ideally
excavation. The other two centrifuge tests the boundaries parallel to the plane of movement
(DWC21 and SSIOl) modelled deeper penetration of the model (the front window and the back
walls subject to long-term swelling. These walls plate) should be rigid and frictionless. The strong-
were unable to rotate, and therefore were repre- box used for these mode1 tests was developed and
sentative of any well-propped diaphragm wall described by Powrie (1986). The front is made
subject to rising groundwater. from 80 mm thick Perspex so that the centrifuge
model can be viewed and photographed during
testing; the back is made from 20 mm thick
THE CENTRIFUGE MODEL TESTS Dural stiffened by cross-beams. The front window
It can be shown that when a model geometri- and the backplate were well lubricated before the
cally similar to a field-scale structure is made at a soil models were inserted (Stewart, 1990). Powrie
scale 1: n and tested at a constant acceleration (1986) reported coefficients of friction in the range
equivalent to n gravities, the self-weight stress dis- 04-0.02 for these kaolin-grease-solid interfaces
tribution will be correctly modelled if the bound- tested in direct shear.
ary conditions are also similar (Schofield, 1980). A liner of the type first employed by Avgh-
An advantage of centrifuge modelling is that the erinos (1969) is used with this strong-box. This
time taken for both transient and steady-state liner (of internal dimensions 736 x 257 x 152
seepage is reduced by a factor of l/n2 from full mm) is common to a consolidation press, so the
scale; one day testing at 1259 represents nearly 50 soil can be pre-consolidated before centrifuging.
years in the field. Thus, a centrifuge model is the The liner seals against the front and back of the
only practical way to observe the long-term centrifuge strong-box, preventing leakage of
behaviour of structures in stiff clay. water from the soil model.
The dimensions and stiffnesses of the retaining A clay slurry was prepared by mixing kaolin
walls and supports were chosen to be broadly powder with deionized water under a partial
consistent with full-scale structures when appro- vacuum, at a moisture content of 120%. It was
priate scaling factors are applied. Table 1 sets out then placed in the liner within the consolidation
these details at both model and prototype scale press, and one-dimensionally consolidated to a
for tests DWC20 and 21 and SSIOl. Each proto- vertical effective stress of 1200 kPa in increments.
type wall retained a 10 m face of clay, and was The sample was then unloaded in vertical stress
propped at the final level of the excavation. The decrements of about 100 kPa. The final vertical
prototype wall penetration beneath that level effective stress in the consolidometer was selected

Table 1. Wall and support details

Excavated height: Penetration: Bending stiffness EI : Prop stiffness I:

m m kNm’/m MN/m per m

DWCZO
Model 0.08 0.016 5 158
Prototype 10 2 107 158

DWC21
Model 0.08 0.04 0.6 158
Prototype 10 5 1.2 x 10” 158

SSIOl
Model 0.08 0.08 0.6 79
Prototype 10 10 1.2 x 106 19
114 BOLTON AND STEWART

Control standplpe transducer

water
I
SUPPlY I Surface water
transducer
Zinc chlonde transducer
Excavation
transducer

Clay sample

1
i
txcavanon waw
control valve

Fig. 3. Fluid control system

the boundary condition imposed on the model tial transformers (LVDTs) were used to measure
walls before excavation would have been approx- movements of the clay surface, three triangulated
imately equivalent to an earth pressure coefficient LVDTs measured wall movements, six full strain
of unity in the soil it was supposed to represent gauge bridges (equally spaced on the centre line
(Bolton, Britto, Powrie & White, 1989). of the wall) measured wall bending moments, and
Prior to any construction activity the earth a full strain gauge bridge measured the force on
pressure coefficient in a stiff clay is usually greater each prop. Each strain bridge output was ampli-
than 1 and may be 3 or greater, but the slurry fied by a factor of 100 on the centrifuge arm.
trench phase of diaphragm wall construction Figure 3 shows how the fluids were controlled
reduces the coefficient to approximately unity during centrifuging. Before testing, zinc chloride
(Powrie, 1985; Tedd, Chard, Charles & Symons, solution was placed in the rubber bag, and in a
1984). Thus, the start of the centrifuge tests rep- top-up tank which was used to control the level
resented the stage of diaphragm wall construction during reconsolidation in the centrifuge. Water
after the slurry trench phase. levels at the model boundaries were controlled by
Figure 2 shows the position of the instrumen- standpipes. When a single water level was
tation within the centrifuge model. Eleven Druck required throughout a model test, a standpipe
PDCR81 miniature pressure transducers were with an overflow at the required level was used.
used to measure the pore-water pressures within Otherwise, a standpipe with a very small bleed-
the centrifuge models, five linear variable differen- hole was used; the height of the water in the

Table 2. Water supply conditions: RS = water supplied at a bead equivalent to the

height of the retained surface; EX = water supplied at a bead equivalent to the height
of the excavation

Head of water supply

DWC20 DWC21 SSIOl

Pre-excavation Base aquifer RS EX EX

Excavation RS EX No supply
Retained surface RS No supply No supply

Post-excavation Base aquifer No supply EX EX

Excavation EX EX No supply
Retained surface RS No supply No supply

Further changes Base aquifer Increased Increased

Excavation No supply No supply
Retained surface No supply No supply
Fig. 4. DWCZO flow net: no curvature correction; numbers shown are potentials in
m relative to excavation level

Prototype scale: years

0 10 20 30 40 50
40 I 1 I I I I 1 1 1

Excavation Centrifuge pause

t t
0 ,,,,,11~I,IIII,~~I~,~~‘~,~“’ 0
0 10 20 30
Model scale: h

(a)

Prototype scale years

0 10 20 30 40 50
I I I I I I I

30 -

0 10 20 30
Model scale: h

(b)

Fig. 5. Rising aquifer potential in models (a) DWC21, (b) SSIOl:

datum for potential is base of clay, 22 m (prototype) below excavation
116 BOLTON AND STEWART

standpipe could then be adjusted during a test by SSIOL Without a direct water supply, there was a
alteration of the water supply rate. In general, significant evaporation loss from both the
water could be supplied to three model bound- retained surface and the excavation. This was
aries: the retained surface, the excavation, and the minimized by covering these boundaries with thin
aquifer underlying the model. Table 2 details how rubber sheets, but it still resulted in steady-state
the water was supplied in each test. upward seepage. This is illustrated in Fig. 6,
The initial water supply conditions in test which shows the long-term seepage pattern after
DWC20 resulted (after a period of reconsolid- excavation for models DWC21 and SSIOl.
ation) in a water table at the retained soil surface.
Immediately before excavation was simulated, the
height of the water in the region to be excavated ONE-DIMENSIONAL CLAY SWELLING
was reduced to the excavation level, and the base Figure 7 shows the anticipated stress path in
aquifer was isolated from the water supply to act both rrv’+,’ and t-s’ plots for clay subjected to a
as an internal drain. The resulting long-term loading-unloading cycle in one-dimensional
seepage pattern after excavation is shown in Fig. strain (Schmidt, 1966; Wroth, 1972). At the start
4, the internal drain having been observed to of unloading, Wroth (1975) noted an approximate
achieve a constant potential 3 m above the level proportionality between the initial changes in
of the excavation. vertical and horizontal effective stress in an oedo-
The water supply conditions in tests DWC21 meter, consistent with isotropic elastic rebound at
and SSIOl were intended to represent a clay layer constant Poisson’s ratio. As an isotropic stress
overlying a recharging aquifer. During reconsoli- state was approached, however, the horizontal
dation, excavation and subsequent equilibration, effective stress started to decrease more sharply,
water was supplied to the base aquifer at a head until the decrements of eV’ and uh’ were roughly
equal to excavation base level. The water pressure equal and the stress path ran at constant t
in the base aquifer was then increased stepwise towards failure at small s’. Wroth (1972) intro-
several times, the models being allowed to duced the passive failure criterion as a limit on
approach a steady state between increases. This is the stress ratio in one-dimensional swelling.
illustrated in Fig. 5, which shows the pore-water Brooker & Ireland’s (1965) data indicate that very
potential in the base aquifer and at discrete loca- high stress ratios CT,,‘/O,’ + co, t/s’ + - 1 may be
tions within the model during tests DWC21 and achieved at low effective stress levels as a clay

‘(a)

Fig. 6. (a) DWC21 and (b) SSIOl long-term flow nets: no curvature correction;
potentials in m (prototype scale) relative to base aquifer level, 22 m below excava-
tion
 PROPPED DIAPHRAGM WALLS 117

l-

0.8 -

“0 0.2 0.4 0.6 0.8 1

o’da’, wax

0 0.2 0.4 0.6 0.8 1

S’/O’” Inax

Fig. 7. Stress paths for onedimensional consolidation

and swelling

sample is one-dimensionally unloaded. Burland & decrements, sampled, and then set up in the
Fourie (1985) reported a true cohesion intercept stress-path triaxial apparatus under an isotropic
(c’ = a,‘/2) in the ultimate failure states of undis- effective stress s’ = 200 kPa. A bilinear elastic-
turbed samples of London Clay subject to com- plastic relation has been fitted to the data of sub-
plete passive stress relief (0,’ = 0), and stress sequent one-dimensional swelling. The initial
ratios in excess of ah’/a,’ z 5, t/s’ z -0.68 phase of unloading is bounded by a line of slope
(corresponding to a secant 4 of 43”) for reconsti- 1 on the (t, s’) diagram. This is, by definition, a
tuted specimens dilating strongly at failure. line of constant horizontal effective stress. An
Data from one-dimensional swelling tests elastic interpretation of this response would be of
(Stewart, 1990) on kaolin, conducted in a strain- drained unloading with zero Poisson’s ratio.
path-controlled triaxial apparatus, are typified in Similarly, if this path were the result of a pore-
Fig. 8. The clay was pre-compressed one- pressure increase under constant vertical total
dimensionally to c&,,~~ = 1200 kPa, unloaded in stress, the horizontal effective stress would

t .
Test Vl DSWELL (Stewart, 1990)

50 100 150 200

0 b
S’

Fig. 8. Stress path data for one-dimensional swelling of kaolin

118 BOLTON AND STEWART

equally remain constant, so that the horizontal ANALYSIS OF A SHALLOW

total stress would increase by an amount equal to PENETRATION WALL PROPPED AT
the increase in pore pressure. EXCAVATION LEVEL
The subsequent plastic behaviour is bounded Test DWC20 indicated that a shallow penetrat-
by a line t = -c,. A line of zero slope on the (t, ion diaphragm wall propped at excavation level
s’) diagram corresponds to zero change in shear (penetration 20% of excavation depth) would
stress. If a clay deposit were swelling one- exhibit significant rotation about the prop imme-
dimensionally due to an increase in the pore- diately on excavation, and would become even
water pressure without a change in the vertical more unserviceable in the long term. Bolton &
total stress, the horizontal total stress would not Powrie (1988) have shown that statically admis-
be changing in this second stage. The data indi- sible plastic stress fields and kinematically admis-
cate that samples of overconsolidated kaolin clay sible plastic deformation mechanisms can be used
continue to support a significant deviatoric stress to check the status of walls rotating about either
when the vertical effective stress reduces mono- their crest or their toe. An analysis is now per-
tonically to zero, at which tensile cracking must formed which uses a similar geostructural mecha-
be expected. After monotonic swelling, the magni- nism for a wall rotating about a prop at
tude of shear stress eventually developed could be excavation level.
expressed as c, z 0.040:,,, An extra shearing Statically admissible stress fields for the col-
cycle, as experienced by the centrifuge samples lapse of a structure usually assume a constant soil
during model preparation, was found to lead to a strength (characterized by either an undrained
reduction to c, z 0.020:,,, (see Fig. 8). When strength c, or an angle of shearing resistance 4).
samples were continually subjected to cycles of Bolton & Powrie (1988) propose that similar
stress-reversal while swelling, however, failure stress fields are applicable before collapse, but
occurred by shearing at a critical state stress ratio based on mobilization of a constant proportion
(t/s’ = sin 4cri, = 0.38). of the ultimate strength. This, they suggest, will
Fitting of the bilinear approximation on the be a significant error only if the stress field before
lower bound of the data in Fig. 8 overestimates failure varies substantially from that at collapse.
the lateral swelling pressure. This overestimate is They propose that serviceability criteria can be
conservative for the purposes of checking struc- checked by the linking of mobilized shear
tural serviceability. Stewart (1990) showed that strength and shear strain using data of plastic
clay swelling in conditions where the radial strain hardening obtained from appropriate element
rate was small compared with the axial strain rate tests. The compatible wall displacement can then
followed an unloading stress path very similar to be inferred from a kinematically admissible strain
the one-dimensional swelling stress path. There- field.
fore it is thought that the idealized bilinear stress
path is applicable to clay swelling near sub-
structures when structural movements are posi-
Rotation hfl
tively restrained.

Fig. 9. Equilibrium stress field for wall propped at exca- Fig. 10. Kinematically admissible strain field for wall
vation base: undrained soil propped at excavation base: undrained soil
 PROPPED DIAPHRAGM WAILS 119

Undrained analysis
The statically admissible stress field and the
kinematically admissible strain mechanism,
shown in Figs 9 and 10 respectively, form a geo-
structural mechanism for the analysis of a
shallow penetration diaphragm wall propped at
excavation level. In regions of deformation, the
principal stress and strain directions are horizon-
tal and vertical, and the major stress and strain
directions everywhere coincide. Strictly, this
applies only to a frictionless wall. The mechanism
assumes that there is a single shear strain in all
regions of (undrained) deformation as the sup-
posedly rigid wall rotates about the prop, into the + Shear vane datum
excavation. If the mobilized strength before exca- . Triaxial datum

vation is zero (corresponding to K, = l), and

variations in soil stiffness are ignored, a unique 0.1 L------b
1 10 100
shear strain should mobilize a unique shear OCR on 0”’
strength.
Wroth (1984) suggested an approximate rela- Fig. 11. Undrained strength correlation with OCR: data
from Amerasinge (1973), Phillips (1986), Al Tabbaa
tion of undrained strength ratio (~,/a,‘) and over-
(1988), Stewart (1990)
consolidation ratio (OCR = cr:,,J~~‘)
c,/g,’ = a OCRb (I)
undrained strength profile for model DWC20
where a and b are constants. Fig. 11 shows based on equation (2). The proposed geo-
undrained strength data for kaolin tested in two structural mechanism assumes that undrained
different apparatus, plotted on logarithmic axes strength is constant with varying depth, therefore
as undrained strength ratio against OCR. Amer- the undrained strength profile is averaged over
asinge’s (1973) data are for Spestone kaolin, the the depth of the wall, to give an undrained
other data are for Speswhite kaolin; but this is strength c, ay of 46 kPa within predicted bounds
not considered significant (Mair, 1979). A linear 42 kPa < ci, By< 50 kPa.
regression analysis was performed on all the data Equilibrium of the stress field shown in Fig. 9
in Fig. 11. The intercept and slope of the regres- for model DWC20 would require a mobilized soil
sion line gives values for a and b in equation (1) strength of 38 kPa provided the tension crack at
the soil-wall interface remained dry. If sufficient
(C&T,‘) = 0.29 OCRo.405 (2) water were available to fill this tension crack, a
Fig. 11 also shows the 95% confidence limits for mobilized soil strength of about 300 kPa would
the central tendency. Fig. 12 shows the estimated be required for wall stability; this would have

c,: kPa

50 100
I A

Fig. 12. Undrained strength profile for DWCZO

120 BOLTON AND STEWART

i.Or

-
0

0
0
-- --
*
ii I
I
I
I/

l!L
0
0
I
l Datum
I
_ _ A”a,ySiS ““drains

0 I
Datum Drained

y: percentage
SC&S
Fig. 13. Plane-strain compression data for kaolin at Model 0 40 0 4mm
OCR = 10 +‘:‘I ,‘:‘I
Field 0 5.0 0 0.5 m
space Displacement
resulted in wall collapse. As the model was stable
Fig. 14. Immediate and long-term boundary displace-
immediately after excavation, it is inferred that
ments for DWCZO
any tension crack was dry.
Also, equilibrium of the stress field shown in
Fig. 9 would require a prop force of 0.8 MN/m at
DWC20 indicated that retained surface settle-
field scale (6.4 N/mm at model scale). A prop
ment became negligible at a distance of about 16
force of 1.1 MN/m at field scale (8.5 N/mm at
m prototype scale (130 mm model scale) from the
model scale) was measured immediately after
10 m (80 mm) high wall. The simple plastic defor-
excavation during the centrifuge model test.
mation mechanism would have underpredicted
Therefore, the proposed geostructural mechanism
the width of the settlement trough, setting the
estimates the wall propping force to be about
limit of influence at 10 m from the wall. It is gen-
75% of that measured. This agreement is satisfac-
erally found that linear elastic finite element pre-
tory considering the assumptions that are made the other hand, significantly
dictions, on
about the stress field.
overestimate the zone of influence (Hubbard,
The stress-path data shown in Fig. 13 were
Potts, Miller & Burland, 1984; Jardine, Potts,
used in the deformation analysis of centrifuge
Fourie & Burland, 1986).
model DWC20. This relation, measured by
Powrie (1986) in an undrained plane-strain test, is
representative of the soil just below wall toe level. Drained analysis
The proportion of undrained strength (c,,,,/c,) In the long term, the wall came into equi-
required for wall equilibrium is 0.83, with predic- librium (see Fig. 14) with pore-water pressures
tion bounds 0.76 < c,,~/c, < 0.91. Therefore, the dictated by the pattern of seepage shown in Fig.
estimated soil shear strain y is 2.4%, with predic- 4: these pressures were linearized against depth
tion bounds y = 1.6% and 4.3%. The shear-strain for calculation purposes. The consequent increase
mechanism shown in Fig. 10 would estimate a in effective stress throughout the depth of the clay
wall rotation of 0.012, with prediction bounds of was partly responsible for the relatively large
0.008 and 0.022. The measured wall rotation long-term settlement on the retained side. The
immediately after excavation was 0.023. There- sharp changes of horizontal effective stress near
fore, the best estimate of wall rotation is an the toe, due to the reversal of strain direction
underestimate by a factor of 2, but the small about the prop and to the reduction of vertical
uncertainty on undrained strength could explain effective stress owing to upward seepage beneath
the shortfall: the stress-strain curve has become the excavation, lead to difficulties in analysis.
quite flat at this high degree of strength mobi- Milligan & Bransby (1976), dealing with a similar
lization. configuration in dense sand, showed that the
Figure 14 compares the undrained boundary passive support on the retained side beneath a
displacements calculated for model DWC20 with pivot could be very large, especially when wall
those measured immediately after excavation friction could be developed. They showed experi-
during the centrifuge model test. As noted above, mentally that the passive resultant acted close to
the calculated wall rotation is about half of that the base of the wall. Fig. 15 shows the predicted
measured; calculated and measured settlements state of drained equilibrium for DWC20 based on
immediately behind the wall show a similar rela- the achievement of fully active conditions above
tion. Photographic film measurement for model the prop (L., = 29.5” at OCR = 20) (Stewart,
 PROPPED DIAPHRAGM WALLS 121

props could be constructed. In such circum-

stances, the wall would either require temporary
shoring or have to rely on the undrained strength
of the soil for support: even unsupported excava-
tions in stiff clay can be stable before drainage
occurs. The centrifuge model tests were represen-
tative of the situation.

Prop P Construction of the excavation

The movements of an unpropped rigid wall
immediately on excavation are thought to be
characterized well but the shear-strain mechanism
shown in Fig. 16 (Bolton, Powrie & Symons,
1990), in which all regions of soil deformation
Fig. 15. Drained equilibrium stresses and forces for
DWCZO undergo the same shear strain. If the original
earth pressure coefficient in the wall vicinity was
approximately unity and accepting that the pore-
water pressure boundary conditions of models
1990). The effective stress component of the DWC21 and SSIOl resulted in an approximately
passive support force is taken at the base; the uniform OCR throughout, a single shear strain in
effective active pressure beneath the excavation is regions of soil deformation might be assumed-as
ignored. Moment equilibrium about the base now a first approximation-to mobilize a single soil
dictates that the propping force should be 1.36 strength c,,,,~ = ci due to excavation.
MN/m (10.9 N/mm at model scale). This com- Bolton, Powrie & Stewart (1987) presented the
pares well to the value of 1.2 MN/m (9.2 N/mm) stress field shown in Fig. 17 for an unpropped
obtained from the model test. rigid diaphragm wall embedded in stiff clay. The
equilibrium diagram assumes conditions of fixed
earth support: this is accurate enough since the
ANALYSIS OF DEEP PENETRATION WALLS point of rotation can be shown to lie very close to
PROPPED AT EXCAVATION LEVEL the base (Bolton et al., 1990). The mobilized soil
When excavation was performed in model tests strength and the force Q are then found by con-
DWC21 and SSIOl, the immediate wall move- sidering moment and horizontal force equi-
ments were very small. These wall movements librium.
were insufficient even to bring the walls into effec- Before excavation, the stress state at a general
tive contact with the wall props, but load cells in point R in the retained soil behind a diaphragm
the wall props indicated that the walls were effec- wall can be represented by points R, and R,’ in
tively propped very shortly after excavation. If a Fig. 18 (the pore-water pressure at point R has
field-scale diaphragm wall were to be propped at arbitrarily been taken as negative). If it is
excavation base level, the excavation would have assumed that the surface of the retaining wall is
to proceed (at least partially) before the wall frictionless, on excavation there will be no change

-de

7++de

Fig. 16. Kinematically admissible strain field for a free

cantilever in undrained soil
9

Fig. 17. Approximate equilibrium for undrained excava-

tion against a free cantilever
122 BOLTON AND STEWART

Fig. 18. Stress path at and after excavation: retained soil

in the vertical total stress in the retained soil. the end of excavation is
Therefore the total stress state immediately after
excavation will be indicated by point Ri mobi- Uh = Cr”+ 2ci (4)
lizing a shear stress t = ci, where R, Ri is a The total stress path for point E during excava-
straight line of gradient - 1 in t-s space. The tion can be deduced from the stability of a par-
horizontal total stress acting on the retained side tially formed excavation in the same way.
of the wall at the end of excavation will therefore
be Duriny transient drainage
crh = 0” - 2ci (3) At the end of excavation there will be a nega-
tive excess pore-water pressure in the retained
The effective stress path R,‘R,‘, however, will be clay (Au, = -ci in Fig. 18), because during exca-
approximately perpendicular to the s‘ axis, as vation there is a reduction in the average total
shown, because the clay will deform quasi- stress s without a reduction in the average effec-
elastically at constant volume. tive stress s’. As already stated, this excess pore-
Before excavation, the stress state at a general water pressure must dissipate in conditions of
point E in the soil beneath the excavation can virtually no horizontal strain. Therefore, the effec-
be represented by points E, and E,’ in Fig. 19. tive stress path is assumed to follow the idealized
On excavation, the effective stress path E,’ Er’ will one-dimensional elastic swelling stress path, and
also be approximately perpendicular to the s’ the excess pore-water pressure will dissipate
axis, and will also reach a shear stress equal in without change in the horizontal effective stress.
magnitude to ci. Therefore the horizontal total Thus, the idealized effective stress path will follow
stress acting on the excavation side of the wall at a line of slope + 1 in Fig. 18, while the total stress

t Notension

I/ --IL l E

I \ Notenso”
Fig. 19. Stress path at and after excavation: soil beneath excavation
 PROPPED DIAPHRAGM WALLS 123
* the constant t portion of the idealized swelling
stress path. Therefore the horizontal total stress
acting on the excavation side of the wall after the
transient pore-water pressures have dissipated is
h
Crh= U” + 2c, (6)
If the shear stress ci required for undrained wall
stability immediately after excavation is greater
than c, , then it is assumed that the effective stress
path will move back on to the idealized one-
dimensional swelling stress path during pore-
water pressure dissipation.
The idealized earth pressure distribution in the
Fig. 20. Approximate drained earth pressure distribution long term after excavation is shown in Fig. 20. A
negative prop force P or a large toe force Q
path must have a slope of - 1 since vertical total would indicate that the one-dimensional swelling
stress is constant. Therefore both the total and assumption made above is inappropriate to the
effective stress states (Rd and R,’ in Fig. 18) can case under study due to a tendency towards wall
rotation.
be obtained, and it will be seen that the shear
stress t, = ~42. The horizontal total stress acting
on the retained side of the wall at this stage is
Swelling due to rising groundwater level
therefore
Increasing the groundwater level in the centri-
crh = 0” - 2t, = 0” - ci (5) fuge tests did not result in any accumulation of
surface water, either at the retained surface or in
A similar argument can be used to deduce the the excavation. Therefore, assuming no wall fric-
long-term stress state at the general point E in the tion, there will be no change in the vertical total
soil beneath the excavation. Once the excavation stress. So, at the general point R in the retained
is complete, there will be no further change in the soil, the total stress path will remain on the line of
vertical total stress (assuming no wall friction). gradient - 1, shown in Fig. 21. The effective stress
Thus the total stress path must remain on a line path will follow the idealized one-dimensional
in t-s space of gradient - 1. Excavation will have swelling path until the difference between the
induced a negative excess pore-water pressure at average total stress s and the average effective
point E (Aur in Fig. 19). The effective stress path stress s’ has reduced by an amount equal to the
is assumed to follow the idealized one- pore pressure rise Au, (see Fig. 21), and hence the
dimensional swelling stress path as this excess idealized total and effective stress paths for point
pore-water pressure dissipates, and so the total R will reach R, and R,’ respectively. At this stage
and effective stress states after dissipation of tran- the horizontal total stress acting on the retained
sient pore-water pressures will be E, and E,‘. side of the wall will be
Any reasonable excavation depth will result in
sufficient stress relief to ensure that point E,’ is on cr,, = uV - ci + Au, (7)

Fig. 21. Stress path for rising groundwater: retained soil

124 BOLTON AND STEWART

I \ Noterwon
Fig. 22. Stress path for rising groundwater: soil beneath excavation

but the assumption of the bilinear one- (c) at the end of the observation period after the
dimensional swelling stress path will impose a final increase in base aquifer potential: the
limit on the horizontal total stress of increase in pore-water pressure in the vicinity
of the model walls was less than the increase
fJh = 0” + 2c, (8) in the base aquifer water pressure because of
Similar reasoning can be used to deduce the surface evaporation, therefore transducer
idealized total and effective stress paths for PPT3 was used to estimate the pore-pressure
the general point E beneath the excavation. As increase near the model wall (see Figs 2 and
the effective stress path is already following the 5).
constant shear stress portion of the idealized one-
dimensional swelling response, the horizontal The undrained strength profiles of models
total stress in the soil beneath the excavation will DWC21 and SSIOl have been estimated by use of
not change, so equation (8) also applies. Therefore equation (2). In both models, the undrained
the average effective stress s’ will reduce by an strength immediately before excavation was
amount equal to Au,, and the total and effective approximately constant with depth; c,, TVwas 122
stresses at a general point E in the soil beneath kPa in test DWC21 and 135 kPa in test SSIOl.
the excavation will be represented by points E, On excavation, it is calculated that model
and E,’ in Fig. 22. DWC21 mobilized a strength of 57 kPa, and
The idealized earth pressure distribution after model SSIOl 52 kPa. Once a steady state was
an increase in groundwater level is shown in Fig. reached after excavation, the pore-water pressure
23. Again, a negative prop force P or a large toe in the wall vicinity was increased by a total of
force Q would indicate that the assumptions were -65 kPa during test DWC21, and by 80 kPa
inappropriate in the specific case. during test SSIOl. The value of c, was taken to
be 23 kPa, using the Fig. 8 data for a swelling test
with a single cycle of stress reversal.
Comparison of the analysis with Figure 24 compares the wall bending moments
the centrljuge model tests
During the centrifuge tests, the water pressure
in the base aquifer was increased incrementally,
and each water-pressure increment was followed
by a period of observation to study the influence
of that particular increment on the model behav-
iour. The analysis does not consider transient
flow, therefore the points of comparison of the
analysis and the centrifuge model tests are
(a) immediately after excavation: this analysis is
undrained; excavation occurred slowly in
model test SSIOl so no comparison is avail-
able for that test
(b) at the end of the observation period of tran-
sient flow after excavation Fig. 23. Approximate swelling pressure distribution
 PROPPED DIAPHRAGM WALLS 125

kNm/m kNm/m
2000 3000 4000 0 1000 2000 3000 4000
1

---- - - “,z:zis long-term after excavation

1
~ Analysis aftermcrease
I” groundwater level
after increase in groundwater level
. Datum I

Fig. 24. Bending moments for DWC21 Fig. 25. Bending moments for SSIOl

Table 3. Prop forces in DWC21

Prop force: kN/m

Additional force Q required
Measured Calculated for equilibrium in analysis: kN/m

Immediately on 0 157
excavation

Table 4. Prop forces in SSIOl

Prop force: kN/m

Additional force Q required
Measured Calculated for equilibrium in analysis: kN/m

Immediately on 0 185
excavation

Long-term 872 914 - 284

after
excavation

After 80 kPa 1714 2372 -353

increase in
pore-water pressure
 126 BOLTON AND STEWART

Model 0 20 40 60 80mm 0 20 mm
t I : : t l-A-7-l
Prototype 0 5 10 m 0 2.5 m
Sp?NX Displacement

Fig. 26. Displacements of model DWC21 when the pore-water pressure in the base aquifer had been increased by 120
kPa

calculated for model DWC21 with those mea- The accurate prediction of model behaviour
sured during the centrifuge model test; Table 3 depends entirely on the selection of appropriate
compares the calculated and measured prop stress path tests for soil elements. This work has
forces. Likewise, Fig. 25 and Table 4 compare the featured behaviour in swelling clays, and a large
analysis with measurements made during centri- programme of swelling strain path tests was con-
fuge model test SSIOL As expected, the analysis ducted. These showed that characterization of
gives a moderate overestimate of the prop force this behaviour as elastic is inapproptiate, even
and the wall bending moments following swelling. inside what is normally assumed to be the plastic
The two tests do suggest, however, that a design yield surface.
approach based on these principles will be suffi- The stress path followed by kaolin in one-
ciently accurate for practical purposes. Fig. 26 dimensional unloading was idealized as bilinear.
shows the final soil displacements following the This facilitated hand calculations for the horizon-
raising of groundwater pressures in DWC21, con- tal effective stress acting on a deep penetration
firming that one-dimensional swelling is a sulli- diaphragm wall propped at excavation level. By
ciently good representation of long-term soil fitting of the bilinear approximation below the
movements around this well-supported wall. test data on a t : s’ diagram, the horizontal lateral
stress will be overestimated: this procedure is
conservative in checking for structural service-
CONCLUSIONS ability. There is evidence from element tests that
Centrifuge model tests allowed the long-term cyclic straining reduces or eliminates the swelling
behaviour of a propped diaphragm wall in stiff pressures that can otherwise be retained at zero
clay to be observed in a few hours: this would effective vertical stress.
take many years at prototype scale.
Observation of centrifuge model tests was ACKNOWLEDGEMENT
essential in the creating of simplified behavioural The Authors are grateful for financial support
mechanisms for analysis. These simplified plastic of the Transport Research Laboratory, effected
mechanisms obeyed the conditions of equilibrium through a research contract placed with Cam-
and compatibility, with coincidence of principal bridge University. The opinions expressed are,
axes. however, solely those of the Authors.
 PROPPED DIAPHRAGM WALLS 127

NOTATION Brooker, E. W. & Ireland, H. 0. (1965). Earth pressures

a, b equation (1) constants at rest related to stress history. Can. Geotech. J. 2,
c’ true cohesion intercept No. 1, 1-15.
c,,,,~ mobilized shear strength (also denoted by t) Burland, J. B. & Fourie, A. B. (1985). Testing of soils
c, maximum shear stress t following strain under conditions of passive stress relief.
c, undrained strength Gtotechnique 35, No. 2, 193-198.
E general point in the soil beneath the excavation Hubbard, H. W., Potts, D. M., Miller, D. & Burland,
g gravitational acceleration J. B. (1984). Design of the retaining walls for the
n model scale factor M25 cut and cover tunnel at Bell Common.
OCR overconsolidation ratio ($, ,,,.Jo,‘) Giotechnique 34, No. 4,495-512.
prop force Jardine, R. J., Potts, D. M., Fourie, A. B. & Burland,
toe force J. B. (1988). Studies of the influence of stress-strain
general point in the retained soil characteristics in soil structure interaction.
stress in plane strain GCotechnique 36, No. 3, 377-396.
effective stress in plane strain Mair, R. J. (1979). Centrifuge modelling of tunnel con-
shear stress in plane strain struction in soft clay. PhD dissertation, University of
pore-water pressure Cambridge.
soil shear strain Milligan, G. W. E. & Bransby, P. L. (1976). Combined
excess pore-water pressure active and passive rotational failure of a retaining
effective horizontal stress wall in sand. G&technique 26, No. 3,473-494.
effective vertical stress Padfield, C. J. & Mair, R. J. (1984). Design of retaining
secant angle of effective shearing resistance walls embedded in stifl clay. Report 104. London:
Construction Industry Research and Information
Association.
Subscripts Phillips, R. (1986). Ground deformation in the vicinity of
av average a trench heading. PhD dissertation, University of
crit at a critical state Cambridge.
d after drainage Powrie, W. (1985). Discussion on Performance of
E relating to point beneath excavation propped and cantilevered rigid walls. GPotechnique
i initial, after excavation 35, No. 4, 346-348.
max maximum Powrie, W. (1986). The behaviour ofdiaphragmwalls in
R relating to point in retained soil clay. PhD dissertation, University of Cambridge.
s after swelling Schmidt, B. (1966). Discussion on Earth pressures at
rest related to stress history, by Brooker and
Ireland. Can. Geotech. J. 3, No. 4,239%242.
REFERENCES Schofield, A. N. (1980). The Rankine Lecture: Cam-
Al Tabbaa, A. (1988). Permeability and stress-strain bridge geotechnical centrifuge operations.
response of Speswhite kaolin. PhD dissertation, Uni- GCotechnique 30, No. 3, 227-267.
versity of Cambridge. Simpson, B., Blower, T., Craig, R. N. & Wilkinson,
Amerasinge, S. F. (1973). The stress-strain behaviour of W. B. (1989). The engineering implications of rising
clay at low stress levels and high over-consolidation groundwater levels in the deep aquifer beneath
ratios. PhD dissertation, University of Cambridge. London. Special publication 69. London : Construc-
Avgherinos, P. J. (1969). Centrijuge testing of models tion Industry Research and Information Associ-
made of soil. PhD dissertation, University of Cam- ation.
bridge. Stewart, D. I. (1990). Groundwater effects on in-situ walls
Bolton, M. D., Britto, A. M., Powrie, W. & White, T. P. in sti# clay. PhD dissertation, University of Cam-
(1989). Finite element analysis of a centrifuge model bridge.
of a retaining wall embedded in overconsolidated Tedd, P., Chard, B. M., Charles, J. A. & Symons, I. F.
clay. Comput. Geotech. 7, 289-318. (1984). Behaviour of a propped embedded retaining
Bolton, M. D. & Powrie, W. (1988). The behaviour of wall in stiff clay at Bell Common tunnel.
diaphragm walls in clay prior to collapse. Gtotechnique 34, No. 4, 513-532.
GCotechnique 38, No. 2, 167-189. Wroth, C. P. (1972). General theories of earth pressure
Bolton, M. D., Powrie, W. & Stewart, D. I. (1987). and deformations. Proc. 5th Eur. Con& Soil Mech.,
Effects on diaphragm walls of groundwater pressure Madrid 2, 33-52.
rising in clays. Proc. 9th Eur. Conf. Soil Mech., Wroth, C. P. (1975). In-situ measurement of initial
Dublin, 759-762. stresses and deformation characteristics. Proc. Spe-
Bolton, M. D., Powrie, W. & Symons, I. F. (1990). The cialty Conf: In-situ Measurement of Soil Properties,
design of stiff in-situ walls retaining overconsolidated Raleigh, 181-230.
clay. Contractor report 199, Crowthorne: Transport Wroth, C. P. (1984). The interpretation of in-situ soil
and Road Research Laboratory. tests. Gdotechnique 34, No. 4,449-489.