You are on page 1of 16

115

Numerical analysis of a multipropped


excavation in stiff clay
Charles W.W. Ng, B. Simpson, M.L. Lings, and D.F.T. Nash

Abstract: This paper presents the procedures and the results of numerical analyses of a multipropped deep excavation at Lion
Yard, Cambridge, using the nonlinear Brick model. The computed results are compared with the comprehensive case record.
The observed small deflections and bending moments of the wall, low prop forces, and relatively small ground movements
during the main excavation have been taken into account. Shear strains which developed around the site during the main
excavation were generally less than 0.3%. Significant reduction of lateral stress in the ground during wall installation and the
highly nonlinear stress–strain characteristics of the Gault Clay are the chief reasons for the observed unusual behaviour. The
Gault Clay exhibits first yield at a threshold shear strain of about 0.001%, beyond which the stiffness deteriorates significantly
from an initial very high value. This high stiffness at very small strains may be due to cementation bonding between soil
particles, as a result of the presence of calcium carbonate. Simple drainage assumptions for the stiff fissured clay on both sides
of the diaphragm wall appear to be inadequate for design analyses.
Key words: numerical analysis, multipropped excavation, Gault Clay, nonlinear brick model, small strain stiffness.

Résumé : Cet article présente les procédures et les résultats d’analyses numériques effectuées pour une excavation à étais
multiples à Lion Yard Cambridge, grâce au modèle non-linéaire Brick. Les résultats du calcul sont comparés aux observations
complètes qui ont pu être faites pour ce cas. On a pris en compte les déflections et moments fléchissants faibles observés sur le
mur, les forces réduites dans les étais et les mouvements du sol relativement peu importants pendant l’excavation principale.
Les déformations de cisaillement qui se sont développées autour du site pendant cette phase étaient en général inférieures à
0,3%. Une réduction importante de la contrainte latérale dans les sols pendant l’installation du mur et les caractéristiques
fortement non-linéaire en contraintes–déformations de l’argile de Gault sont les principales causes du comportement
inhabituel observé. L’argile de Gault montre un premier écoulement lorsqu’on dépasse un seuil de déformation de
cisaillement de l’ordre de 0,001% à partir duquel la raideur se détériore sensiblement. Cette forte raideur à faible déformation
pourrait être attribuée aux liens de cimentation qui se développent entre les particules de sol par suite de la présence de
carbonate de calcium. Les hypothèses simples que l’on fait pour le drainage de l’argile raide fissurée des deux côtés du mur
diaphragme paraissent inadéquate lors des études.
Mots clés : analyse numérique, excavation à étais multiples, argile de Gault, modèle non-linéaire Brick, raideur à petite
déformation.
[Traduit par la rédaction]

Introduction and relatively simple soil model which can reflect the most
Field monitoring provides a means by which the geotechnical important features of the actual soil behaviour. Ideally, a good
engineer can verify the design assumptions and the contractor case record should provide sufficient information for checking
can execute the work with safety and economy. More impor- the results obtained from the back-analysis procedure. Other-
tantly, the field data may also be assembled into a comprehen- wise, misleading conclusions may be drawn.
sive case record that is then available for use when checking The assumption of a linear stress–strain relationship of soil
the validity of any analytical and numerical models. The ongo- inside the state boundary surface has been applied to almost
ing process of back-analysis can help to refine and improve our all geotechnical engineering practice for years. Only recently
understanding, which in turn provides guidance for future de- has it been widely accepted that the stress–strain relationship
signs. is highly nonlinear even at very small strains. The stiffness of
To achieve the tasks of back-analysis, it is essential to have most soils decreases as strain increases, and the stiffness at
a reliable and comprehensive case record and an appropriate small strains is strongly dependent on the recent stress or strain
history of each soil element (Atkinson et al. 1990; Atkinson
and Sällfors 1991). For heavily overconsolidated clays such as
Received February 6, 1997. Accepted September 18, 1997. London Clay and Gault Clay, the influence of mean effective
stress and void ratio on soil stiffness is relatively unimportant
C.W.W. Ng. Department of Civil and Structural
Engineering, Hong Kong University of Science and
compared with the effect of recent stress history and level of
Technology, Clear Water Bay, Kowloon, Hong Kong. strain developed. This new knowledge has emerged from ad-
B. Simpson. Arup Geotechnics, 13 Fitzroy Street, London, vances in both laboratory and field measuring techniques (At-
W1P 6BQ, U.K. kinson and Sällfors 1991; Jardine et al. 1991; Simpson 1992).
M.L. Lings and D.F.T. Nash. Department of Civil This paper presents the results of numerical analyses of the
Engineering, University of Bristol, Bristol, U.K. 10 m deep multipropped excavation at Lion Yard, Cambridge,

Can. Geotech. J. 35: 115–130 (1998). © 1998 NRC Canada


116 Can. Geotech. J. Vol. 35, 1998

Fig. 1. Site plan, showing the location of cross section X–X (see Fig. 2).

using the nonlinear Brick model (Simpson 1992). The Brick (Fig. 1). The cross section of the site is shown in Fig. 2. The
model provides an elastic–plastic constitutive relationship for existing ground level is at approximately +10 m above ordi-
soils in which very high stiffnesses occur at changes in the nary datam (OD). The ground conditions at the site comprise
direction of stress paths, gradually reducing as strains proceed. 3–4 m of made ground and gravel above 38 m of Gault Clay
The model is expressed in effective stress and it has the ability which overlies the Lower Greensand. The soil profile obtained
to capture a wide range of soil behaviour including shear fail- during site investigation has been discussed by Lings et al.
ure and the effects of void ratio, mean effective stress, over- (1991).
consolidation, and recent stress history on soil stiffness. The Gault Clay was laid down under marine conditions
Descriptions and discussion of the soil parameters used, simu- during the Cretaceous period and was probably overlain by up
lation procedures adopted, and analyses undertaken are given, to 400 m of chalk, which has since been eroded (Worssam and
and the computed results are compared with the comprehen- Taylor 1969). The actual thickness of chalk is not known for
sive case record of the Lion Yard excavation. The objective of certain. Since erosion, the clay has been reloaded by 3–4 m of
the comparison is to acquire a better understanding of gravel. The Gault Clay in its natural state is heavily overcon-
soil–structure interaction associated with deep excavations solidated, having natural water contents close to the plastic
and to improve class A predictions for similar types of con- limit (31%). It consists of stiff to hard silty grey clay of high
struction in stiff fissured clays in the future. plasticity and is laminated with closely spaced fissures and
joints. Worssam and Taylor (1969) found that the Gault Clay
The Lion Yard development in the Cambridge area is highly calcareous, containing up to
30% by weight of calcium carbonate. A similar proportion
Soil profile and properties (27.5%) was found in samples taken from Lion Yard (Ng
The Lion Yard site is situated in the city centre of Cambridge, 1992). Prior to construction, groundwater conditions were ap-
United Kingdom, and is approximately 65 m by 45 m on plan proximately hydrostatic below a level of about +7.0 m OD.

© 1998 NRC Canada


Ng et al. 117

Fig. 2. Cross section X–X.

usual pore water pressure response and clay swelling were


The structure recorded in the centre of the site (Ng 1992; Nash et al. 1996).
The structure consists of a three-level underground carpark
beneath a five-storey hotel and is supported on large-diameter
underreamed bored piles founded 30 m below ground level in Finite element analyses undertaken
the Gault Clay. The 10 m deep excavation is retained by a 17 m
deep perimeter concrete diaphragm wall 0.6 m thick, con- The multipropped excavation was simulated using the non-
structed under bentonite in panels typically 8.5 m in length. linear finite element Brick model, which has the ability to
The development has been constructed top-down. The bored capture soil behaviour at small strains, and a linear elastic –
piles and the diaphragm wall were installed before the excava- perfectly plastic model with the Mohr–Coulomb failure crite-
tion of the underground carpark. rion. The former is adopted for Gault Clay and the latter is used
On two opposite sides of the site 3.5 m wide and 19 m long to model gravel and made ground. Plane strain deformation
rectangular openings were left in the slabs adjacent to the dia- was assumed. The finite element mesh adopted in the analyses
phragm wall (see Fig. 1). These were at the positions of the is shown in Fig. 4. Eight-node linear strain finite elements with
sloping vehicle ramps that allow access to the underground four Gauss points were used. The intersection between the
carpark levels. Temporary steel props were installed across Gault Clay and the Lower Greensand was taken as a rigid
these 3.5 m wide openings at each level to support the dia- boundary (BC). The other fixed soil boundary (CD) was as-
phragm wall during excavation. sumed to be 230 m away from the centre of the site. A vertical
sliding boundary (AB) was given to the axis of symmetry in
Instrumentation the centre of the site.
Some of the instrumentation at Lion Yard is shown in
Figs. 1–3. Full details of the structure, construction sequence, Drainage assumptions
and instrumentation together with data were given by Ng Idealization of the excavation is an essential step for any nu-
(1992). A preliminary appraisal of the observed behaviour of merical modelling. A coupled consolidation was not carried
the deep excavation has been reported by Lings et al. (1991). out because (i) the observed pore-pressure response during the
In the context of this paper, the key findings from the field main excavation stages and the swelling behaviour of the clay
monitoring were as follows: (i) there has been a marked reduc- did not follow the conventional Terzaghi consolidation theory
tion in lateral stresses accompanied by small ground move- (Lings et al. 1991; Ng 1992; Nash et al. 1996), since substan-
ments due to diaphragm wall construction (Ng et al. 1995); tial swelling has taken place in the upper part and in the centre
(ii) during the main excavation, wall deflections, strut loads, of the clay stratum; (ii) the heterogeneous nature of soil perme-
and bending moments observed on site were substantially ability is difficult to quantify precisely; and (iii) some of the
lower than the conventional design predictions; and (iii) un- boundary drainage conditions are not known for certain.

© 1998 NRC Canada


118 Can. Geotech. J. Vol. 35, 1998

Fig. 3. Details of the instrumentation at panel 22. on both sides of the wall. This assumption is valid if the exca-
vation is rapid and little drainage occurs within the construc-
tion period.
(3) Type III assumption—Soil is assumed to be undrained
inside the site and outside the site, except near the soil–wall
interface where local drainage in a zone 2.9 m wide is allowed.
This assumption has been made in an attempt to model the
observed significant and rapid fall of pore pressures at the
soil–wall interface shortly after each stage of excavation, prob-
ably due to the presence of “bleed channels” observed on the
exposed faces of the diaphragm wall (Lings et al. 1991). The
selection of the width of the local drainage zone is based on
some parametric analyses using the type I assumption together
with varying widths of the local drainage zone (i.e., 0, 1.4, and
2.9 m wide). The purpose of the parametric analyses is to study
the sensitivity of the ground surface settlement to the assumed
width of the local drainage zones behind the wall. The meth-
odology used to specify the measured pore-water pressure dis-
tribution within each local drainage zone behind the wall is
identical to that adopted in the type I drainage assumption
described above. Selection of the most appropriate drainage
assumption for the Lion Yard site is discussed later.
All undrained analyses presented in this paper are based on
the effective stress approach. The bulk compressibility of satu-
rated soil is due to the bulk compressibility of the water phase
alone. This is based on the assumption that the elastic bulk
modulus of soil grains is much larger than that of water. In the
finite element calculations, the elastic bulk modulus of water
is added to the appropriate element stiffness matrices. Pore-
water pressures are derived such that the change of volume
within each element will be zero, and smoothing is carried out
across element boundaries to give a continuous distribution of
pore-water pressure (Simpson et al. 1979). The undrained
analysis produces no change in the effective stress, and the
Detailed discussions on the measured changes of pore-water external loading is entirely carried by the excess pore-water
pressure in response to the excavation are given by Lings et al. pressure.
(1991) and Ng (1992).
Apart from the idealization of the Lion Yard site geometry
as shown in Fig. 4, some known or simple drainage conditions Soil parameters
have been specified for soil on both sides of the wall in this
paper as follows: Selection of soil stiffness
(1) Type I assumption—Soil is assumed to be undrained The knowledge of the stress–strain behaviour for Gault Clay is
inside the site, whereas all soil elements behind the wall are rather poor, particularly at small strains. Very few laboratory
set with a pore-pressure distribution based on the pneumatic or field tests have been performed and reported. Powell and
piezometer readings taken at the soil–wall interface on the Butcher (1991) summarized field tests carried out by the
retained side (see Fig. 3). For ease of specifying different Building Research Establishment (BRE) at Madingley, United
measured pore-water pressures at various depths, soil elements Kingdom. However, there is considerable scatter in the data,
behind the wall are grouped into different horizontal zones depending on what type of in situ measuring device was used.
(even if they have the same soil parameters). Within each hori- A factor of four can be found between the measured maximum
zontal zone, a single value of the measured pore-water pres- and minimum soil stiffnesses at very small strains.
sure at each stage of the main excavation is specified at the For the nonlinear Brick model, an S-shaped curve which
depth corresponding to the elevation of the pneumatic pie- defines the way that shear stiffness varies with shear strain is
zometer located in this zone. For other soil elements located required. As far as the authors are aware, no laboratory meas-
in the same zone, pore-water pressures are extrapolated with urements of soil stiffness at small strains of Gault Clay had
a static hydraulic gradient of 10 kPa/m depth. This type I drain- been published at the time when the analyses were carried out
age assumption is mainly intended to verify whether the ob- in 1992. Thus the field measurements of soil stiffness at very
served rapid response of pore pressures at the soil–wall small strains on Gault Clay at Madingley (Powell and Butcher
interface (Lings et al. 1991; Ng 1992) is just a localized phe- 1991) were used in deriving appropriate S-shaped curves for
nomenon or if this was extended horizontally to the entire parametric studies.
region on the retained side of the wall. Since the field-measured soil stiffness at very small strains
(2) Type II assumption—Soil is assumed to be undrained shows substantial variations, parametric studies were carried

© 1998 NRC Canada


Ng et al. 119

Fig. 4. Finite element mesh for simulating the main excavation.

Fig. 5. (a) Deduced normalized stiffness–strain relationship for Gault clay. (b) Comparison of the deduced and recent laboratory-measured
S-shaped curves.

out by varying the maximum shear modulus (Gmax) value and sin φcs′ by a factor of about [1 + β(λ – κ) ln(OCR)], where
within the measured upper and lower bounds. Throughout this λ is the slope of the one-dimensional compression line in volu-
series of parametric studies, type III drainage conditions were metric strain (v) – ln s′ space, κ is the average slope of the
assumed. unload–load line in v – ln s′ space, and OCR is the overcon-
For a chosen Gmax value, a best-fit S-shaped curve has been solidation ratio. This approach implicitly assumes that the
drawn through the Gmax value and the laboratory measured Gmax value determined from the geophysical method (refrac-
shear stiffness (Ng 1992) on reconstituted, normally consoli- tion technique) is the same as the value for reconstituted clay
dated Gault Clay specimens at constant mean effective stress at very small strains. The derived upper and lower bound S-
(p′) with 180° of rotation of the stress path, following the con- shaped curves for the parametric studies are shown in Fig. 5a.
cept of the Brick model (Simpson 1992). The area under this As a comparison, test results on an intact specimen taken from
selected curve (Gt/s′ vs. γ, where Gt is the tangent shear modu- the Lion Yard site are also included in the figure. No internal
lus, s′ is mean effective stress in plane strain, and γ is the linear small strain measuring devices were available for these tests,
shear strain) was kept equal to sin φcs′, where φcs′ is the critical but they nevertheless show significantly higher stiffnesses than
state angle of friction. Although the S-shaped curve is a de- that from the reconstituted specimens.
scription of stiffness, the area under the curve dictates the an- The results of the parametric studies suggested that analysis
gle of shearing resistance at failure. Details and an explanation with the lower bound S-shaped curve substantially overpre-
of the derivation are given by Simpson (1992). To allow for dicted the measured wall deformation by a factor of 3 and 1.5
the effects of overconsolidation, the material constant β was at the end of the first and the last stages of the main excavation,
set to 3, which has the effect of increasing both the stiffness respectively. In contrast, analysis with the upper bound

© 1998 NRC Canada


120 Can. Geotech. J. Vol. 35, 1998

S-shaped curve computed wall displacements which were in Table 1. Soil parameters used in the analyses.
reasonably good agreement with field observations at all three
Description Parameter
stages of the main excavation. More detailed discussion is
given later in this paper. Gault Clay
Recently, some independent laboratory triaxial stress path Slope of one-dimensional compression line
tests on intact Gault Clay specimens taken from Madingley in v – ln s′ space λ = 0.072
have been conducted by a research student at Cambridge using Slope of unload–reload in v – ln s′ space κ = 0.018
some internal measuring devices to determine the small-strain Slope of initial unload–reload in v – ln s′
stiffness of the clay (Dasari et al. 1995). Some of the test re- space ι = 0.0006
sults are compared with the deduced upper bound S-shaped Poisson’s ratio for effective stress analysis µ′ = 0.2 (assumed)
curve shown in Fig. 5b. A number of points can be drawn from Material parameter β=3
the comparisons as follows: Preconsolidation pressure s′max = 3200 kN/m2
(1) For constant p′ drained tests on intact specimens ob- Unit weight of Gault Clay γs = 20 kN/m3
tained from Lion Yard and Madingley, the two sets of test Angle of friction φ′cs = 24°, φ′p = 34°
results are consistent with each other at moderately large Stress–strain characteristics As shown in Fig. 5a
strains (see Figs. 5a, 5b). Gravel and made ground
(2) As expected, the deduced upper bound S-shaped curve Angle of friction φ′ = 40°
based on measurements on normally consolidated reconsti- Angle of dilation (not a key parameter) 0° (assumed)
tuted specimens lies below the recent test results obtained from Drained Young’s modulus E′ = 50 000 kN/m2
overconsolidated intact samples at strains greater than 2 × Poisson’s ratio for effective stress analysis µ′ = 0.2 (assumed)
10–5. The use of the material parameter, β = 3, in the Brick Unit weight of gravel and made ground γs = 18 kN/m3
model would have the effect of increasing both the stiffness Coefficient of earth pressure at rest K0 = 0.35
and sin φcs′ by a factor of about [1 + β(λ – κ) ln(OCR)].
(3) Gault Clay possesses highly nonlinear stress–strain
characteristics and exhibits the first yield at a threshold shear laboratory tests are given by Ng (1992). A summary of the soil
strain of about 0.001%, beyond which the stiffness deteriorates parameters used in the analyses is given in Table 1.
significantly from an initial very high value. After modest The elastic parameter iota (ι) used in the analysis has been
straining, its shear stiffness reduces to values comparable to calculated from the following equation:
those for London Clay (Jardine et al. 1991). This high stiffness  s′ 
ι= (1 − 2µ′)
G 
response at very small strains is probably attributed to the pres- [1]
ence of about 30% of calcium carbonate (Ng 1992), which  max
results in some weakly cemented bonding in the soil. Atkinson where µ′ is Poisson’s ratio and was assumed to be 0.2, and
et al. (1990) clearly demonstrated that an artificially cemented Gmax was deduced from the refraction test results (Powell and
sand is about four times stiffer than an uncemented sand at Butcher 1991).
small strains. High stiffnesses at small strains have also been
measured on carbonate clays by Georgiannou et al. (1991). Finite element simulation procedures
(4) The deduced shear stiffness at very small strains (less
than 0.0001%) from the in situ geophysical tests using the Simulation of the geological history at Lion Yard
refraction technique corresponds reasonably well with the Correct modelling of the initial stress conditions in the ground
laboratory tests on intact specimens using internal measuring and wall installation effects is important for the back-analysis
devices mounted on soil samples. The slight discrepancy be- of the main excavation at Lion Yard (Ng et al. 1995). Thus,
tween the laboratory and field measurements could be as a numerical modelling of the main excavation was started by
result of some degree of sample disturbance or inaccurate simulating the geological history of the site.
measurements introduced both in the field and in the labora- The geological history at Lion Yard is complex. To model
tory, or a combination of both. the initial K0 condition for this site, an additional stage was set
(5) It is evident that the recent laboratory measurements up to represent the geological history of the site by consolidat-
give strong support to the deduced upper bound S-shaped ing the Gault Clay one-dimensionally to its estimated precon-
curve. Based on the current knowledge and the results of the solidation pressure, followed by erosion, and then reloading
parametric studies, the upper bound S-shaped curve was con- by made ground and gravel. Since the actual thickness of the
sidered as the most appropriate for the Gault Clay. chalk cannot be determined with great confidence, a sensitivity
analysis was carried out by assuming a different thickness of
Other soil parameters eroded chalk ranging from 200 to 400 m. There was no sig-
The Brick model does not require the coefficient of earth pres- nificant difference in the computed initial lateral stresses. It is,
sure at rest (K0) as an input parameter. Instead, an estimated therefore, believed that, even though there are errors in esti-
preconsolidated pressure for the site is required so that the mating the thickness, it is unlikely to alter the conclusions
geological history may be simulated, including calculation of arising from the present investigations. Figure 6 shows the
the initial lateral stresses in the ground. The actual simulation computed and measured K0 values at Lion Yard using the self-
procedure to derive the initial stress conditions is given later. boring pressuremeter (SBPM), together with predictions using
The angle of friction (φ′) and the compressibility indices Schmidt’s (1966) semiempirical rules with and without allow-
(λ, κ) have been determined from laboratory tests using stress ing the effect of reloading of the overlaying gravel, and the
path triaxial and oedometer cells, respectively. Details of the theoretical predictions given by the Brick model. It should be

© 1998 NRC Canada


Ng et al. 121

noted that the effects of redeposition or reloading are to in- Fig. 6. Comparison of computed and measured K0 values. SBPM,
crease vertical effective stress with relatively little increase of self-boring pressuremeter.
horizontal effective stress (Burland et al. 1979). Thus, the
value of K0 will decrease if the effects of redeposition are
included, particularly at shallow depths. For the first 10 m of
the clay, the measured K0 values using SBPM at Lion Yard are
not believed to be very reliable. First, after allowing for the
effects of redeposition of 3–4 m of gravel on top of Gault Clay
at Lion Yard, the measured values are not consistent with the
SBPM measurements on Gault Clay at Madingley (Powell
1990), which is about 5 km away. The measured values at Lion
Yard are substantially higher than those obtained from Mad-
ingley. Second, if the reloading of gravel is considered when
applying Schmidt’s (1966) semiempirical rule to predict K0
values at Lion Yard, the measured values are significantly
higher than those from the empirical predictions (see Fig. 6).
Third, the measured values at shallow depths are substantially
higher than the theoretical predictions given by the Brick
model. Fourth, the measured values are higher than the maxi-
mum possible K0 value deduced from the analysis of the three-
dimensional effects of diaphragm wall installation (Ng et al.
1995). The analysis was based on field measurements of earth
pressure at the soil–wall interface.

Modelling stress reduction due to diaphragm walling


For simulating the stress reduction due to wall installation,
measured earth pressures (after wall installation) were im-
posed inside an excavated trench and the measured pore pres-
sures were also specified in the surrounding soil. The
reliability of the earth pressure measurements has been as-
sessed in the light of strut forces, wall deflection, and rebar
stresses (Lings et al. 1993). Reasonable force and moment
equilibrium could be achieved. Subsequently, the excavated
trench was replaced by concrete elements and the imposed
pressures were removed. Detailed analysis of the changes of
lateral earth pressure due to the diaphragm construction is
given in the discussion section later.
During this stage of analysis, a maximum of 15 mm hori- Main excavation simulation
zontal soil displacement was computed on the ground surface,
as compared with the measured value of 2 mm, which has an Diaphragm wall stiffness
accuracy of ±2 mm. The large overprediction of ground move- The 600 mm concrete diaphragm wall was modelled by linear
ment during wall installation resulted from neglecting horizon- elastic elements with Young’s modulus of 31 kN/mm2, which
tal arching in the plane strain analysis. Ng et al.(1995) have was determined from the test results of concrete cores taken
illustrated that the wall installation process is a truly three- from the wall itself. No cracking of the wall was modelled.
dimensional problem. Correct computation of ground move- This approach may not be strictly correct in small parts of the
ments may only be achieved if horizontal arching is taken into wall at the final stage of excavation. However, it was consid-
account. The current method used for modelling wall installa- ered to be a reasonable assumption, as the wall was probably
tion effects is just a two-dimensional representation of a three- slightly cracked only locally just above the final excavated
dimensional problem. The imposed average stress changes level (Ng et al. 1992). The use of an uncracked section in this
could be overestimated, as the final stress distribution was analysis will not alter any conclusions arising from the present
nonuniform behind the wall as a result of soil arching (Ng et al. investigations because slight local cracking can only insignifi-
1995). However, this approach is a simple but reasonable way cantly affect the computation of the maximum bending mo-
to incorporate stress and pore-pressure changes due to wall ment in the wall.
installation with the objective of creating appropriate initial
stress conditions for modelling the subsequent main excava- Prop stiffness
tion stages. Another similar approach was that adopted by Five temporary steel struts at 1.7 m centre to centre were in-
Powrie and Li (1991), who modelled the installation effects by stalled at each level inside the rectangular vehicle opening left
reducing the lateral earth pressure coefficient in all the soil in the concrete slab (see Fig. 1) to support panel 22 during the
layers above the toe of the wall. The problems of their ap- main excavation. These struts were simply supported by angle
proach have been discussed by Ng et al. (1995). brackets, and the axial forces required to support the

© 1998 NRC Canada


122 Can. Geotech. J. Vol. 35, 1998

diaphragm wall were resisted by the concrete slab. Thus, very in good agreement with field observations. However, some
little rational restrain was provided by the concrete slab for the excessively large settlements were computed behind the wall
section considered in the analysis. By considering the force as compared with field measurements. This was probably be-
equilibrium between the concrete slab and the steel props, a cause the use of measured pore-water pressures at the soil–wall
simple equivalent stiffness for the whole supporting system interface for specifying the pore-water regime for the entire
can be derived as follows: soil mass behind the wall gave an overestimate of the vertical
1 1 1 effective stress in most areas, i.e., the reduction of pore-water
[2] = + pressure due to the main excavation was not uniform through-
Keq Ksb Ksp
out but decreased with increased distance from the wall. A
where Keq is the overall equivalent prop stiffness, Ksb is the possible explanation is that the measured pore-pressure re-
concrete slab stiffness, and Ksp is the steel prop stiffness. sponse only represented a local phenomenon probably due to
Based on eq. [2], the estimated Keq was 36 600 kN⋅m–1⋅m–1 the existence of the “bleed channels” formed at the soil–wall
run for the level 4 support system, and 57 600 kN⋅m–1⋅m–1 run interface (Lings et al. 1991; Ng 1992). The majority of the soil
for the rest of the propping systems at levels 2 and 3. The behind the wall remained undrained during the excavation
values of Ksp were calculated using the actual length, cross- stages. Subsequently, an analysis assuming the soil to be un-
sectional area, and Young’s modulus of the steel strut. Simi- drained on both sides of the wall (type II) was carried out as a
larly, Ksb was estimated according to the thickness of the comparison. This analysis underpredicted the maximum settle-
concrete slab, half width of the site’s dimension minus the ment behind the wall and the maximum wall deflection by a
length of the struts, and the Young’s modulus of concrete. factor of 2 and 1.5, respectively (see Figs. 7a, 13), but the
Concrete creep and shrinkage were allowed for in estimating shape of the settlement distribution curve corresponded well
the slab stiffness in accordance with guidance provided by the with measurements.
British Standard Institution (1985). In view of this analysis, and the evidence of bleed channels,
it is believed that during the main excavation only a local zone
Simulation of construction of soil adjacent to the wall behaved as drained material,
The actual construction sequence was simulated by the follow- whereas the majority of soil behind the wall actually behaved
ing steps: more like an undrained material. This hypothesis is consistent
(1) model the effects of geological history and wall instal- with the observed variation of the pore-water pressure with
lation as described earlier; distance behind the wall (Ng 1992). As the purpose of back-
(2) install temporary prop at +9.3 m OD by using elastic analysis is to model as closely as possible what has actually
spring element with reduced stiffness (1/20th) to account for happened, a type III assumption with the upper bound
some missing props; S-shaped curve has been presented.
(3) excavate to level 3 (+6.4 m OD) by replacing relevant
soil elements with air elements, which have zero density and
stiffness, to model the effects of stress relief due to excavation; Comparison of the observed and
stresses (including body forces) that originally exist in the ex- computed performance
cavated soil elements are used to compute equivalent nodal
forces acting on the faces of the excavation, thus the potential Soil–structure interaction at panel 22
errors that arise from modelling excavations as discussed by
Brown and Booker (1985) should not appear in the analysis; Diaphragm wall deformation
for simulating the effects of stress relief due to excavation, As discussed previously, the soil stiffness of the Gault Clay at
negative (opposite sign) equivalent nodal forces are imposed small shear strains was not known with confidence at the time
by the finite element program SAFE when the relevant soil of these computations. A sensitivity analysis has been carried
elements are declared to be air elements; out and the calculated wall deflections using the upper and
(4) install temporary prop at +6.4 m OD by using an elastic lower bound S-shaped stiffness–strain curves (see Fig. 5a) are
spring element; compared with the field observations made at the first and last
(5) continue to excavate to level 2 (+3.4 m OD) by further stages of the main excavation as shown in Fig. 7b. The abso-
replacing soil elements with air elements and install temporary lute displacements of the wall were determined from the hori-
props at +3.4 m OD ready for the last stage of excavation; and zontal alignment monitoring of the ground surface station,
(6) excavate to the final basement level (+0.0 m OD). using a theodelite, and taping from the surface station to the top
Cavitation of the fissured clay has been modelled by limit- of the wall. It can be seen that using the lower bound stiffness
ing negative pore-water pressures to –60 kPa measured inside results in overprediction of the wall deformation at the first
the site at the final stage of excavation. This is an additional stage of excavation by a factor of 2. The accumulative wall
analysis step allowing some limited local drainage to occur deformation computed at the last stage of excavation was less
near the soil surface, where in situ suctions were measured by influenced by the variation of soil stiffness at small strains than
tensiometers installed about 1 m below the exposed surface the first stage of excavation. This seems to suggest that the soil
after each stage of the main excavation. was not operating at very small strains during the last stage of
excavation, as an average accumulative shear strain of about
Selection of drainage assumptions 0.3% was developed adjacent to the wall as shown in Fig. 15,
Analysis was first undertaken using the construction sequence hence high stiffness at very small strains was not very relevant
described above with the upper bound S-shaped curve and type for this stage. The difference between the upper and lower
I drainage assumption. The computed wall displacements are bounds of shear modulus may not be so important for the

© 1998 NRC Canada


Ng et al. 123

Fig. 7. (a) Sensitivity of the wall deformation to drainage assumption on the retained side. (b) Sensitivity of the wall deformation to soil
stiffness.

incremental deformations between the penultimate and the last During the subsequent two stages of excavation, the de-
stages of excavation. flected profiles of the wall were typical for a multipropped wall
Hereafter, all computed results presented in this paper are constructed in stiff clay, i.e., deep-seated inward movements
based on the upper bound S-shaped stiffness–strain curve. occurred (Burland et al. 1979). Generally the computed and
Figure 8 shows the comparison between the computed and measured deflections of the wall are in good agreement, al-
measured displacements of the wall for the three stages of the though the Brick model overpredicted the lateral inward move-
main excavation. At the first stage of excavation, the wall be- ment in the region of the toe, which resulted in smaller
haved essentially as a cantilever, with only small propping computed wall curvature than the measured value.
effects at the top. This was because the installation of some of Inward toe movement as a result of vertical stress relief is
the top level props had not been completed before the start of clearly depicted in Fig. 8 by the field measurements and numeri-
the excavation. An allowance was made in the analysis by cal predictions. The conventional approach of deducing wall dis-
reducing the prop stiffness by a factor of 20 at this stage of placements from field data by assuming fixity at the toe would
excavation. have underestimated the true wall deformation slightly.

© 1998 NRC Canada


124 Can. Geotech. J. Vol. 35, 1998

Fig. 8. Comparison of the measured and computed displacements Fig. 9. Comparison of the computed and deduced bending moments
of the wall during excavation. and shear forces.

Bending moments and shear forces Fig. 10. Comparison of the computed and measured propping
Figure 9 shows the computed and deduced bending moments forces.
and shear forces at the end of excavation. The deduced bending
moments and shear forces were derived from the measured
wall rotation and strut load data (Lings et al. 1993). To obtain
satisfactory agreement with the measured rotation data, Lings
et al. (1993) assumed a continuity bending moment at the top
of the wall. This is considered to be logical, as the wall sup-
ports one of the major concrete columns for the superstructure,
which has a full moment connection. However, this additional
moment was difficult to include in the finite element analysis
because the degree of the fixity provided by the superstructure
was not known.
Both the field observations and numerical predictions indi-
cate that the maximum bending moment occurred at a level
just above the final excavated level. The computed maximum
bending moment was smaller than the value deduced from
field observations. This was due to the overprediction of in-
ward toe movement which reduced the maximum curvature of
Fig. 11. Total horizontal stresses during construction.
the wall (see Fig. 8).

Prop forces
A comparison of the measured mean and computed prop forces
at the final stage of excavation is shown in Fig. 10. The com-
puted values appear to correspond reasonably well with the
field observations, despite the large overprediction of prop
force at L3. This is probably because the fixed end moment
provided by the concrete column at the top of the wall was
ignored in the analysis. The introduction of a negative moment
at the top of the wall would have resulted in a smaller com-
puted prop force at L3.
Good agreement between the measured and computed prop
forces is important, as this increases the confidence in the com-
puted earth pressures, which are in force equilibrium with the
computed prop forces.

Total horizontal stresses


Figure 11 shows the calculated and inferred earth pressures on
both sides of the wall. The inferred earth pressures at the end 1993), agree reasonably well with the finite element results. It
of the L1 excavation, which were derived in the light of the is important to note that during wall installation there was a
measured prop forces and rotations of the wall (Lings et al. marked increase in computed lateral stress below the toe of the

© 1998 NRC Canada


Ng et al. 125

Fig. 12. Contours of total horizontal stresses around the site at the end of excavation.

wall as a result of the significant reduction in lateral stress ble using a linear elastic model by a number of researchers
above the toe of the wall. This marked reduction in lateral (Burland and Hancock 1977).
stress was considered to be the chief reason for the measured The computed distribution of lateral inward movements at
low prop forces and bending moment of the wall. The reduc- ground surface indicates a maximum value at about 9 m away
tion of lateral stress due to the construction of the diaphragm from the wall, which coincides with the only two measured
wall was substantially greater than the stress reduction during data points. Although the radius of influence of ground move-
the subsequent 10 m deep excavation. ment extended well over three times the maximum depth of
Figure 12 shows the total horizontal stress contours around excavation (10 m), as for the observations made at New Palace
the site at the end of excavation together with the inferred earth Yard in London Clay (Burland and Hancock 1977), surface
pressures. Generally the inferred earth pressures are consistent ground movements are more concentrated within a distance
with the predictions. It can be seen that the major stress redis- approximately equal to the full depth of the wall, i.e., 17 m.
tribution occurred within a distance approximately equal to the Figures 14 and 15 show the computed general ground de-
full depth of the wall (17 m) as illustrated by the gradual flat- formation pattern and shear strain contours, respectively,
tening of the contours beyond this zone. Thus the use of a around the site at the end of the main excavation. Due to ver-
single reduced K0 value for all the soil above the toe of the tical stress relief inside the site, soil on the retained side moved
wall to model wall installation effects (Powrie and Li 1991) towards the excavation, whereas the soil underneath the exca-
may underestimate the stress reduction close to the wall, vation heaved upwards as depicted in Fig. 14.
whereas it will overestimate the stress reduction farther away It can be seen in Fig. 15 that the shear strains developed
from the wall. around the excavation were small and generally less than 0.3%
except in front of the wall, just below the final excavated level
General ground deformation due to main excavation where the soil was yielding. A similar pattern and magnitude
The measured ground movements due to wall installation and of shear strains were deduced from field measurements at the
piling are excluded because the initial reference data set for the multipropped excavation in London Clay at the British Library
surveying stations was taken at different times and in many in Euston (Atkinson and Sällfors 1991). These indicate that the
cases after the start of diaphragm walling. To make consistent moduli adopted for any calculations based on the theory of
comparisons, the surveying results just before the main exca- linear elasticity should be appropriate for the average small
vation have been chosen as a single reference datum for all strain level anticipated during the main excavation. By study-
stations. For consistency, the computed ground movements ing the strains developed around excavations, an appropriate
presented are only those arising from the main excavation. soil stiffness corresponding to the average small strains could
Figure 13 shows the comparison of the computed and meas- perhaps be selected for design from laboratory test results (Ng
ured distributions of ground movements behind the wall at the 1992).
end of excavation. The computed distribution of settlements
behind the wall shows a “settlement trough” at about 8 m from Pore-water pressure distribution
the wall and the settlement gradually reduces to zero at ap- Figure 16 shows the computed contours of pore pressure
proximately 70 m away from the wall. Computation of a deep around the site at the end of excavation, marked with the meas-
settlement trough and a correct curvature of the settlement ured values at appropriate locations, both inside and outside
profile behind a retaining wall has been shown to be impossi- the site. The changes in pore-water pressures were

© 1998 NRC Canada


126 Can. Geotech. J. Vol. 35, 1998

Fig. 13. Comparison of observed and computed ground movements behind the wall. PD, partially drained; UD, undrained.

overpredicted within the site, resulting in larger suctions in the Fig. 14. Displacement vectors at the end of excavation.
upper levels of clay than were measured. However, it is con-
sidered that the computed values of pore-water pressures are
generally consistent with the measured values, given that some
drainage will have occurred. Outside the site, the computed
pore-water pressures correspond reasonably well with the two
pore-pressure measurements in BH1 (for location of BH1 see
Fig. 1). This gives confidence to the specified pore-water pres-
sure regime within the local drainage zone behind the wall.
Detailed discussions on the measured changes of pore-water
pressure in response to the excavation are given by Lings et al.
(1991) and Ng (1992).
Two major points can be noted from Fig. 16. First, negative
pore pressures were computed for the top few metres of soil
below the final excavated level virtually across the whole site.
This is consistent with suction measurements using tensiome-
ters inside the site. Second, the zone where pore-water pres-
sures behind the wall fell due to the unloading inside the site
extends a distance about equal to the full depth of the wall. of the clay stratum and thus caused some general swelling in
Beyond this zone, pore-water pressure contours remained ap- the upper layer of the clay. Detailed studies of the swelling
proximately horizontal. This probably implies that unloading behaviour and pore-pressure response of the Gault Clay in both
effects due to lateral stress relief are unlikely to extend far short term and long term are given by Nash et al. (1996).
beyond this distance.
Insights into the soil behaviour around the
Soil heave and swelling in the centre of the site
excavation
A comparison between the observed vertical subsoil move-
ments and the calculated heave and swelling with depth at the It is well known that the ground movement around an excava-
end of excavation is shown in Fig. 17. Using the undrained tion is attributable to vertical and horizontal stress relief (Bur-
assumption inside the site, the analysis substantially underpre- land et al. 1979). The magnitude of the ground movement not
dicted vertical subsurface movements beneath the final exca- only depends on the change of stress but also relies on the
vated level. This is probably because water permeated rapidly current state of stiffness. To understand how the soil responds
through the fissure networks to the upper part and in the centre to stress relief due to excavation, it is useful to see how the

© 1998 NRC Canada


Ng et al. 127

Fig. 15. Shear strains around the site at the end of excavation.

Fig. 16. Contours of pore pressure (kPa) at the end of excavation.

stiffness varies at different locations as the construction pro- iour. Through an understanding of the ground behaviour, an
ceeds. Results from the numerical analysis have been extracted appropriate average soil stiffness can be selected for some sim-
to produce the “stiffness” paths for some typical soil elements ple analyses such as one using the Mohr–Coulomb model.
around the excavation to give insights into the ground behav- Figure 18 shows the computed variation of normalized

© 1998 NRC Canada


128 Can. Geotech. J. Vol. 35, 1998

Fig. 17. Vertical soil movements in the centre of the site. operated at a normalized secant stiffness ratio (G/cu) ranging
from 1300 to 600 at the wall installation stage and 500 to 80
during the three stages of the main excavation. The magnitude
of the current normalized stiffness ratio depends on the loca-
tion of the soil and the current state of stress of the soil.

Discussion

Drainage assumptions
One possible stability analyses of embedded retaining walls in
stiff clay could be done by assuming that soil to be fully
drained on the retained side and to be an undrained material
inside the excavation during construction (i.e., type I assump-
tion). For Gault Clay in Cambridge, however, the extension of
this approach to predict ground movements appears to result in
significant overpredictions in magnitude and extent of settle-
ments behind the diaphragm wall due to the overpredictions of
vertical effective stress, if a reduced uniform water table due
to excavation is assumed behind the wall. On the other hand,
secant shear stiffness (G/cu, where G is the shear modulus, and when the type II assumption was used, the analysis underesti-
cu is the undrained shear strength) with stress change ratio mated the maximum wall deflection and settlement behind the
(∆σh/σho, where ∆σh is the change of total horizontal stress, wall by factors of 1.5 and 2, respectively. The type III drainage
and σho is the initial total horizontal stress) for some typical assumption is useful for understanding the sensitivity of pre-
soil elements at locations around the excavation. The lateral dictions by allowing local drainage to take place at the
stress change ratio is defined as the difference between the soil–wall interface.
initial and final total horizontal stresses of a particular soil In view of these results, it seems reasonable to suggest that
element over its initial total horizontal stress before construc- analyses with types I and II assumptions can provide two ex-
tion. As the computed wall deflections and surface ground treme cases for modelling an embedded retaining wall in fis-
settlements are in reasonable agreement with field observa- sured stiff clay. However, for practising engineers in their
tions (see Figs. 8 and 13), the stiffness path effectively illus- design analyses, an improvement in predictions can be made
trates the deduced operative soil stiffness at various locations if effects of some local drainage and the formation of “bleed
during construction. channels” at the soil–wall interface are considered. A coupled
It can be seen that the soil stiffness adjacent to the wall consolidation analysis is unlikely to account for these effects
reduced substantially after the wall installation process, automatically.
whereas the stiffness of soil elements farther away from the
wall, such as P5, was not affected by this operation.
Class A estimations of wall installation effects
As the main excavation proceeded, the normalized stiffness
ratio (G/cu) of the majority of soil elements around the exca- It is rather difficult to derive an analytical, closed-form solu-
tion to account for stress changes in the ground after wall
vation dropped below 500. This implies that it would be rea-
sonable to use an average normalized stiffness ratio (G/cu) for installation. This is mainly because of the complex vertical and
a simple linear elastic – perfectly plastic model to represent lateral stress transfer mechanisms (Ng et al. 1995). However,
some of the main excavation operations. if the length of panel (on plan) is relatively large compared
The soil element A107 (about 7 m behind the wall) behaved with its width (on plan), the average resulting lateral pressure
differently from the other elements considered, as the lateral in the ground after wall installation may be approximated by
stress reduction ratio decreased during the main excavation the following theoretical bilinear wet concrete pressure en-
stages. A decrease in the lateral stress change ratio of course velop (Ng 1992):
means an increase in lateral stress during the main excavation. γ z z ≤ hc
This phenomenon can be explained by two mechanisms. One [3] σh =  c
possible mechanism is that after installing the props at the top (γc − γb)hc + γbz z > hc
two levels, the upper part of the wall was not free to move. As where γb and γc are the unit weights of bentonite and fluid
the last two stages of the main excavation progressed, the soil concrete, respectively; hc is the critical depth; and z is the
close to the wall was “squeezed” by the soil farther away from depth. The critical height, hc, is governed by a number of fac-
the wall. This was because uniform lateral movement of the tors such as rate of placing concrete (Ng 1992; Lings et al.
ground was prevented by the wall and the props (see Fig. 13). 1994). The above theoretical equation has been verified by
The soil farther away from the wall than soil element A107 field measurements of lateral pressure after wall installation at
would have moved more towards the wall and caused the in- three different sites. Very good agreements have been found.
crease of lateral stress at A107. The other equivalent mecha- As a rule of thumb, hc could be assumed to be equal to one-
nism is vertical soil arching. Once the top prop had been third of the wall depth. Within the critical depth, the wet con-
installed, lateral stress was transferred (redistributed) upward crete applies the full fluid pressure to the surrounding soil. At
below it as the last two stages of excavation proceeded. depths greater than the critical depth, the wet concrete pressure
The stiffness paths plotted in Fig. 18 show that the soil increases with depth at a rate given by the unit weight of the

© 1998 NRC Canada


Ng et al. 129

Fig. 18. Normalized stiffness paths for soil elements around the site.

bentonite. For equilibrium, the lateral earth pressure must model wall installation effects is likely to underestimate stress
equal the wet concrete pressure. Therefore, the equation can be reduction close to the wall, whereas it will overestimate the
used as the first approximation for class A predictions to take reduction farther away from the wall.
into account the effects of wall installation. (3) Shear strains that developed around the site during the
main excavation were small and generally less than 0.3%. This
Conclusions was similar to the multipropped excavation in London Clay at
the British Library in Euston. The study of shear strains around
(1) The performance of the multipropped excavation at excavations can increase understanding soil behaviour at
Lion Yard has been evaluated using the nonlinear Brick model. working stresses and can help to determine an appropriate
The results computed by the model are in reasonably good mean soil stiffness for design if a simple linear soil model is
agreement with the field observations. The observed small de- used.
flection and bending moments of the wall, low prop forces,
and relatively small ground movements during the main exca-
vation have been accounted for. The Gault Clay in the centre Acknowledgments
of Cambridge possesses highly nonlinear stress–strain charac-
teristics with a threshold shear strain value of 0.001%. Broadly The authors gratefully acknowledge financial support from JT
speaking, the Gault Clay exhibited a normalized secant stiff- Design Build and the Science and Engineering Research
ness ratio (G/cu) ranging from 1300 to 600 at the wall instal- Council, and the permission given by Arup Geotechnics for
lation stage and from 500 to 80 during the three stages of main using their finite element program SAFE.
excavation. The magnitude of the current normalized stiffness
ratio depends on the location of the soil and the current state References
of stress and strain history of the soil. The deduced initial high
stiffness at very small strains may be due to cementation bond- Atkinson, J.H., and Sällfors, G. 1991. Experimental determination of
ing between soil particles as a result of the presence of calcium stress–strain–time characteristics in laboratory and in-situ tests.
carbonate. General report. In Proceedings of the 10th European Conference
(2) The resulting low horizontal stresses in the ground due on Soil Mechanics and Foundation Engineering, Florence, Italy,
to wall installation are the chief reason for the observed low Vol. 3, pp. 915–956.
Atkinson, J.H., Coop, M.R., Stallebrass, S.E., and Viggiani, G. 1990.
prop forces, bending moments, and wall displacements during Measurement of stiffness of soils and weak rocks in laboratory
the main excavation. The significant destressed zone behind tests. In Proceedings of the 25th Annual Conference of the Engi-
the wall as a result of wall installation and main excavation neering Geology Group, Leeds, U.K. British Geology Society,
seems to be confined within a distance approximately equal to pp. 21–27.
the full depth of the wall. Thus, the use of an average single British Standard Institution. 1985. Structural use of concrete.
reduced K0 value for all the soil above the toe of the wall to BS8110: Part 1 and Part 2. British Standard Institution, London.

© 1998 NRC Canada


130 Can. Geotech. J. Vol. 35, 1998

Brown, P.T., and Booker, J.R. 1985. Finite element analysis of exca- European Conference on Soil Mechanics and Foundation Engi-
vation. Computers and Geotechnics, 1: 207–220. neering, Florence, Italy, Vol. 1, pp. 153–156.
Burland, J.B., and Hancock, R.J.R. 1977. Underground carpark at the Powrie, W., and Li, E.S.F. 1991. Finite element analyses of an in situ
House of Commons, London: geotechnical aspects. Structural En- wall propped at formation level. Géotechnique, 41(4): 499–514.
gineer, 55(2): 87–100. Schmidt, R. 1966. Discussions. Canadian Geotechnical Journal, 3:
Burland, J.B., Simpson, B., and St John, H.D. 1979. Movements 239–242.
around excavations in London clay. In Proceedings of the 7th Simpson, B. 1992. Thirty-second Rankine lecture: retaining struc-
European Conference on Soil Mechanics and Foundation Engi- tures: displacement and design. Géotechnique, 42(4): 541–576.
neering, Brighton, U.K., Vol. 1, pp. 13–19. Simpson, B., O’Riordan, N.J., and Croft, D.D. 1979. A computer
Dasari, G.R., Bolton, M.D., and Ng, C.W.W. 1995. Small strain model for the analysis of ground movements in London Clay.
measurement using modified LDTs. Cambridge University Engi- Géotechnique, 29(2): 149–175.
neering Department, Technical Report CUED/D-SOILS/TR275. Worssam, B.C., and Taylor, J.H. 1969. Geology of the country around
Georgiannou, V.N., Rampello, S., and Silvestri, F. 1991. Static and Cambridge. 2nd ed. Her Majesty’s Stationary Office, London.
dynamic measurements of undrained stiffness on natural overcon-
solidated clays. In Proceedings of the 10th European Conference
on Soil Mechanics and Foundation Engineering, Florence, Italy, List of symbols
Vol. 1, pp. 91–95.
Jardine, R.J., Potts, D.M., St. John, H.D., and High, D.W. 1991. Some cu undrained shear strength
practical application of a non-linear ground model. In Proceedings E′ drained Young’s modulus
of the 10th European Conference on Soil Mechanics and Founda- G shear modulus
tion Engineering, Florence, Italy, Vol. 1, pp. 223–228. Gmax maximum shear modulus
Lings, M.L., Nash, D.F.T., Ng, C.W.W., and Boyce, M.D. 1991. Ob- Gt tangent shear modulus
served behaviour of a deep excavation in Gault Clay: a preliminary hc critical depth
appraisal. In Proceedings of the 10th European Conference on Soil K0 coefficient of earth pressure at rest
Mechanics and Foundation Engineering, Florence, Italy, Vol. 2,
pp. 467–470.
Ksb concrete slab stiffness
Lings, M.L., Nash, D.F.T., and Ng, C.W.W. 1993. Reliability of earth Ksp steel prop stiffness
pressure measurements adjacent to a multipropped diaphragm Keq overall equivalent prop stiffness
wall. In Proceedings of an International Conference on Retaining OCR overconsolidation ratio
Structures, Cambridge, U.K. Thomas Telford, London. pp. p′ mean effective stress, (σ1′ + 2σ3′)/3
258–269. q deviator stress, σ1′ – σ3′
Lings, M.L., Ng, C.W.W., and Nash, D.F.T. 1994. The pressure of s′ mean effective stress in plane strain, (σ1′ + σ3′)/2
wet concrete in diaphragm wall panels cast under bentonite.
smax′ preconsolidation pressure
Geotechnical Engineering, Proceedings of the Institution of Civil
Engineers, 107: 163–172. v volumetric strain
Nash, D.F.T., Lings, M.L., and Ng, C.W.W. 1996. Observed heave z depth
and swelling beneath a deep excavation in Gault clay. In Proceed- β a material constant allowing for overconsolidation effects
ings of the International Symposium on Geotechnical Aspects of γ shear strain (ε1 – ε3)
Underground Construction in Soft Ground, April, City University, γb unit weight of bentonite
London. Edited by R.J. Mair and R.N. Taylor. A.A. Balkema, γc unit weight of fluid concrete
Rotterdam, pp. 191–196. γs unit weight of soil
Ng, C.W.W. 1992. An evaluation of soil–structure interaction associ-
∆ increment or change
ated with a multi-propped excavation. Ph.D. thesis, University of
Bristol, Bristol, U.K. ε1, ε3 principal strains
Ng, C.W.W., Lings, M.L., and Nash, D.F.T. 1992. Back-analysing the ι initial slope of unload–load line in v – ln s′ space
bending moment in a concrete diaphragm wall. Structural Engi- κ average slope of unload–load line in v – ln s′ space
neer, 70(23 and 24): 421–426. λ slope of one-dimensional compression line in v – ln s′
Ng, C.W.W., Lings, M.L., Simpson, B., and Nash, D.F.T. 1995. An space
approximate analysis of the three-dimensional effects of dia- µ′ Poisson’s ratio
phragm wall installation. Géotechnique, 45(3): 497–507. σh total horizontal stress
Powell, J.J.M. 1990. A comparison of four different pressuremeters
σho initial total horizontal stress
and their methods of interpretation in a stiff heavily overconsoli-
dated clay. In Proceedings of the 3rd International Symposium on σ1′, σ3′ major or minor principal effective normal stresses
Pressuremeter Testing, Oxford, U.K., pp. 287–298. φ′ angle of friction
Powell, J.J.M., and Butcher, A.P. 1991. Assessment of ground stiff- φcs′ critical state angle of friction
ness from field and laboratory tests. In Proceedings of the 10th φp′ peak angle of friction

© 1998 NRC Canada

You might also like