You are on page 1of 10

Centrifuge model studies of a

cantilever retaining wall in sand

G.J.W. KING, University of Liverpool and J.P. McLOUGHLIN,


City of Liverpool Community College, UK

Introduction
Flexible cantilever retaining walls resist the overturning moment, due to
the pressure from the retained soil, by developing a fixing moment due to
the earth pressures along the embedded portion of the wall. Appreciable
re-distribution of these pressures occurs, due to plastic yield in the soil, as the
height of retained soil is increased and the depth of embedment reduced by
excavation. For a satisfactory design the associated wall displacements and
the bending stresses developed in the wall must not be excessive. For a
specified depth of excavation in a particular soil, it is necessary to estimate
the required depth of embedment of the wall and its section properties.
Cantilever retaining walls are considered to be suitable for low retained
heights (<4.5m) and are mainly used for temporary works. According to
Head and Wynne (ref. 1), they account for up to 75% of all retaining walls
constructed at the present time. Steel sheet-piling is most commonly used
with the advantages of strength, variety of available sections, lightness,
re-usability, resistance to hard driving and the ability to be extended in
length.
Current design methods for cantilever walls are based on classical limit
state equilibrium approaches. These approaches, together with guidelines for
the measurement and selection of appropriate design parameters for perma­
nent and temporary walls in stiff clay, have been detailed in CIRIA Report
104 by Padfield and Mair (ref. 2). Comparative calculations, using the alter­
native design procedures suggested for the determination of embedment
depths to give certain factors of safety, were carried out by Symons and
Kotera (ref. 3). They concluded that pre-selection of a particular procedure
was not advisable. These approaches do not give any indications of wall
movements and do not consider the influence of wall flexibility on the earth
pressures which develop. Conventional laboratory tests on cantilever retain­
ing walls in cohesionless soils have been described by Rowe (ref. 4) and
Bransby and Milligan (ref. 5). While these have produced useful quantitative
information about stability and movement patterns, quantitative extrapola­
tion to prototype scale is of dubious value. Also bending stresses in the walls

Retaining structures. Thomas Telford, London, 1993 711

Downloaded by [ University of Liverpool] on [31/10/17]. Copyright © ICE Publishing, all rights reserved.
RETAINING STRUCTURES

Table 1. Dimensions and properties of model and prototype

Model Prototype

Wall height mm 120 11000

width mm 459 42077

thickness mm 2.02 185.2

second moment mm*/mm 6.90 x 1 0 " 1


4.93 x 1 0 5

of area (I)

modulus of kN/mm 2
193 207
elasticity (E)

flexural 2
kNm /m 1.33 x 10" 1
10.21 x 10*
rigidity

depth mm 229 21000

length mm 229 21000

width mm 460 42170

were not measured. Recent numerical studies by Fourie and Potts (ref. 6) have
indicated that limit state approaches over-predict wall bending moments
appreciably.
In this Paper centrifuge model studies on a flexible wall, with different
surface roughnesses, in both loose and dense sands are described. The results
are presented in terms of the behaviour of an 11 m high, Frodingham No. 5
section, sheet pile wall. The magnitudes of displacements are reported and
the observed bending moments and critical embedment depths compared to
those predicted by limit state analyses.

Centrifuge model tests


The cantilever retaining wall model was a flat stainless steel sheet. Ten 3
mm foil strain gauges were fixed centrally to each side of the wall at 10.9 mm
centres. This was used to model an 11 m high Frodingham No. 5 section
sheet-pile wall by spinning at 91.67g acceleration in the Liverpool University
centrifuge. This facility has been described in detail by King, Dickin and
Lyndon (ref. 7). The dimensions and properties of the prototype and model
are summarised in Table 1.
Three wall surfaces were used

712

Downloaded by [ University of Liverpool] on [31/10/17]. Copyright © ICE Publishing, all rights reserved.
KING AND MCLOUGHLEM

(a) the natural milled surface, subsequently referred to as having inter­


mediate roughness
(b) a smooth surface, achieved by coating the wall in silicone grease and
placing a single latex rubber sheet over both faces and
(c) a rough surface, achieved by applying double sided adhesive tape to
the faces of the wall and coating the outer surfaces with sand.
To minimise the influences of the boundary walls of the package, they
were coated with a layer of silicone grease and covered with latex rubber
sheet. Short greased membranes were also fixed to the sides of the wall to
prevent the sand passing between the ends of the wall and the sides of the
bucket.
The wall was placed centrally within the package and the sand was rained
into the bucket. The sand used was dry Erith sand which is a fine quartz sand
with an effective grain size of 0.16 mm and a uniformity coefficient of 1.25.
In order to obtain a dense packing, the sand was rained from a box with a
perforated base fixed 1.5 m above the bucket. A very repeatable average
3
density of 16.0 kNm" was achieved. A loose packing was obtained by raining
the sand from a box with a perforated base placed directly on top of the
3
bucket. Consistent average densities of 14.2 kNm" were obtained. Excess
sand was scraped off so that the sand surface was flat and level with the top
of the wall.
At the start of each test the centrifuge was spun up to the required speed
and dynamic datum readings taken for all strain gauges. The machine was
then brought to rest and an increment of excavation carried out using an
adjustable scraper and brush. The displacement transducer mounting was
then fixed in position and a transducer placed in contact with the centre-line
of the exposed face of the wall. The distance of the transducer from the top
of the wall was noted and a static datum reading taken. The machine was
again brought up to speed and strain gauge and transducer readings taken.
This procedure was repeated for each increment of excavation until catastro­
phic wall displacement occurred. The displacement transducer mounting
was adjusted and additional transducers introduced at greater depths as
excavation proceeded. Static zero datums were assumed to be appropriate
for new placed transducers in conformity with the observed behaviour of
existing ones. Ideally the excavations should have been carried out in flight
but no suitable mechanism for doing this has been developed. However, it is
felt that the unload-reload stress history applied did not have a significant
influence on the in-flight behaviour at any stage. In all tests, a further
increment of 0.5 m resulted in excessive walls movements before the test
speed was reached.

713

Downloaded by [ University of Liverpool] on [31/10/17]. Copyright © ICE Publishing, all rights reserved.
RETAINING STRUCTURES

Fig. 2: Typical wall displacements

714

Downloaded by [ University of Liverpool] on [31/10/17]. Copyright © ICE Publishing, all rights reserved.
Test results
The results of all tests have been recorded by McLoughlin (ref. 8). The
bending moment distributions and horizontal wall displacements at each
stage of excavation were typically as shown in Figs. 1 and 2. These are for an
intermediate rough wall in dense sand.
Oscillating bending moments, often with small negative peaks, were
recorded in the top part of the wall in all tests. When those moments were
negative, the corresponding displacement diagrams showed slight negative
curvature. The maximum bending moments occurred at depths of less than

Table 2. Maximum observed values close to failure

Soil Dense Loose

Surface Smooth Intermediate Rough Smooth Intermediate Rough

h(m) 7 8 8.5 6 7 8

M^CkNm/m) 152 168 259 137 166 248

s(m) 1 1 0.5 1.5 1 0.5

715

Downloaded by [ University of Liverpool] on [31/10/17]. Copyright © ICE Publishing, all rights reserved.
RETAINING STRUCTURES

2m below excavation level for all depths of excavation. These are consider­
ably shallower than the depths obtained, by numerical studies on a different
wall, by Fourie and Potts (ref. 6). It is interesting to note that these studies
also yielded negative bending moments in the top part of the wall.
The variations in maximum bending moment and maximum deflection
with height of excavation, for rough and smooth walls in loose and dense
sands is shown in Figs. 3 and 4. These curves also give a good indication of
the limiting heights of excavation. Clearly at any stable depth of excavation
both maximum bending moments and deflections reduce with increasing
surface roughness or soil density. However, because the limiting height of
excavation increases with increasing surface roughness and soil density, the
greatest bending moments can develop in rough surface piles in dense soil.
The maximum deflections observed without catastrophic failure were be­
tween 0.5 and 0.75m, for smooth surface piles and between 1.0 and 1.25 m for
rough surface piles.
A summary of the average maximum bending moments and their posi­
tions recorded at the maximum stable depths of excavation achieved, is given
in Table 2 for comparison with values predicted by limit state analyses.

716

Downloaded by [ University of Liverpool] on [31/10/17]. Copyright © ICE Publishing, all rights reserved.
KING AND MCLOUGHLDSf

Limit state analyses


The simplified active and passive earth pressure diagrams and equivalent
forces ususlly assumed are as shown in Fig. 5. The wall is considered to rotate
about point O close to the bottom of the wall. The piling below this point is
initially ignored and the pressures on this section replaced by the concen­
trated force R.
Moment equilibrium about O yields

= h / ( J K /K p A - 1 ) (1)

in which K A and Kp are appropriate earth pressure coefficients.


The actual penetration depth is then taken as

d = 1.2 d (2)
o

It is then checked that APp - A P A > R


The depth s below excavation level of zero shear and maximum bending
moment is given by

/ ( JK /K - 1 )P A
(3)

and the maximum moment by


I n - K A (h+s) 3
- K s )
p
3
7 / 6 (4)
Note that s and Mmax are not dependent on d . 0

For a fixed overall height of wall, H = h + 1.2d , Eq. 1 yields the limiting 0

excavation depth

h =H ( JK /K - 1 )/ ( JK /K + 0 . 2 )
P A P A (5)
Design with a factor of safety can be achieved by factoring the passive earth
pressure coefficient Kp.

R \
r P
0
)-2d f 0 AP A
IAFV
Fig. 5: Simplified pressure diagram

717

Downloaded by [ University of Liverpool] on [31/10/17]. Copyright © ICE Publishing, all rights reserved.
RETAINING STRUCTURES

Table 3. Earth pressure coefficients

Soli Dense (7 - 16 kNm" )


3
Loose (7 - 14.2 kNm - 3
)

Surface Smooth Intermediate Rough Smooth Intermediate Rough

49.5 49.5 40 40 40
*° 49.5

S ' 0 17.7 49.5 0 15.8 40

0.136 0.121 0.102 0.217 0.197 0.168


K A

K 7.35 13.77 48.06 4.61 8.61 13.41


p

Table 4. Limit state values

Soil Dense Loose

Surface Smooth Intermediate Rough Smooth Intermediate Rough

7.7 8.4 9.3 6.6 7.5 8.1

(kNm/m) 220 232 244 236 268 265

s(m) 1.2 0.9 0.5 1.8 1.3 1.0

Table 5. Design values (F -1.5)

Soil Dense Loose

Surface Smooth Intermediate Rough Smooth Intermediate Rough

h(m) 7.2 8.0 9.1 6.0 7.0 7.7

(kNm/m) 197 213 231 202 238 241

s(m) 1.4 1.0 0.5 2.2 1.6 1.2

Equations 3,4 and 5 were used to calculate the maximum excavation depth
and the position and magnitude of the maximum bending moment, both at
limiting equilibrium and with a factor of safety of 1.5, for each wall surface
embedded in dense and loose sand. The earth pressure coefficients used are
shown in Table 3.
The friction angles <>| and 8 were determined from plane strain and direct
sliding tests respectively and the coefficients from tables by Caquot and
Kerisel (ref. 9).

718

Downloaded by [ University of Liverpool] on [31/10/17]. Copyright © ICE Publishing, all rights reserved.
KING AND MCLOUGHLIN

The results are shown in Tables 4 and 5 for comparison with the figures in
Table 2.
Since all of the retaining walls failed when subjected to a further 0 . 5 m of
excavation, the limit state predictions of the critical heights are very good.
By chance the last stable heights of excavation achieved were close to the
design values calculated using as factor of safety of 1.5 on the passive earth
pressure. At this stage the pile displacements were quite large and their rate
of change with excavation depth was increasing rapidly. This does not inspire
confidence in this method of applying what would normally be considered
a reasonable factor of safety.
The design values of maximum bending moment agree reasonably well
with those observed for rough piling but are significantly larger than those
observed for intermediate rough and smooth piling.
The calculated depths to maximum bending moment agree with the values
observed for the rough pilings in dense sand but, otherwise, they are over­
estimated.

Conclusions
Centrifuge model studies on a flexible cantilever retaining wall with a
range of surface roughness, in both loose and dense sands have been de­
scribed. The results have been presented in terms of the behaviour of an 11
m high Frodingham No. 5 Section wall.
At stable depths of excavation both maximum bending moments and
deflections reduced with increases in surface roughness and in soil density.
However, because the limiting height of excavation increased with increasing
surface roughness and soil density, the greatest bending moments developed
in rough surface piles in dense soil. At all depths of excavation the maximum
bending moments occurred at depths of less than 2 m belowthe excavation
level.
Limit state predictions of critical heights were very close to those observed.
However observed maximum displacements, and rates of change of displace­
ment with heights of excavation, corresponding to design with a factor of
safety of 1.5 on passive pressure, were alarmingly high. The maximum
displacements were generally in excess of 0.5 m for smooth surface piles and
1.0m for rough surface piles. Design bending moments were significantly
larger than those observed for all but very rough piling and were predicted
to occur at greater depths then those observed for all but the rougher pilings
in dense sands.
The simple limit state approach was found to be adequate in many
respects. However, modifications to this approach are being considered
where it has been observed to be deficient.

719

Downloaded by [ University of Liverpool] on [31/10/17]. Copyright © ICE Publishing, all rights reserved.
RETAINING STRUCTURES

References
1. HEAD J.M. and WYNNE C P . Designing retaining walls embedded in stiff
clay. Ground Engineering, 1985,18, No. 3, pp 30-33.
2. PADFIELD C.J. and MAIR R.J. The design of propped and cantilever walls
embedded in stiff clays. CIRIA Report 104, London, 1984.
3. SYMONS I.F. and KOTERA H. A parametric study of the stability of embedded
cantilever retaining walls. TRRL Report RR116, Transport and Road Re­
search Laboratory, Crowthorne, Berks, 1987.
4. ROWE P.W. Cantilever sheet piling in cohesionless soil. Engineering, 1951,
September, pp 316-319.
5. BRANSBY P.L. and MILLIGAN G.W.E. Soil deformations near cantilever
sheet pile walls. Geotechnique, 1975,24, No. 2, ppl75-195.
6. FOURIE A.B. and POTTS D.M. Comparison of finite element and limiting
equilibrium analyses for an embedded cantilever retaining wall. Geotech­
nique, 1989,39, No. 2, pp 175-188.
7. KING G.J.W., DICKIN E.A. and LYNDON A. The development of a
medium sized centrifuge testing facility. Proc. Int. Symp. on the Application
of Centrifuge Modelling to Geotechnical Design, Manchester, 1984, pp 25-46.
8. McLOUGHLIN J.P. The behaviour and design of cantilever retaining walls in
sand. M.Sc.(Eng) Thesis, University of Liverpool, 1989.
9. CAQUOT A. and KERISEL J. Tables for the calculation of passive pressure,
active pressure and bearing pressure offoundations. Gauthier-Villars, Paris,
1948.

720

Downloaded by [ University of Liverpool] on [31/10/17]. Copyright © ICE Publishing, all rights reserved.

You might also like