Professional Documents
Culture Documents
Keywords: diaphragm wall friction, vertical stability of deep excavation, probabilistic analysis
INTRODUCTION
The excavation of deep building pits often requires a check against failure by uplift of low
permeability ground layers below excavation level. Whenever the weight of these soil layers is less
than the pore-water pressure underneath, measures to resist buoyancy are to be considered. The
measures most commonly adopted include: decreasing the water pressure by drainage, anchoring an
underwater concrete slab in the underlying strata or executing the excavation in a pressurised
working chamber.
This paper discusses a case study in which it was shown successfully that side wall friction too can
contribute significantly to the vertical stability of a diaphragm walled deep excavation. The safety
against failure by uplift is demonstrated through probabilistic analysis of all relevant parameters.
The project presented is the top-down constructed, up to 31 m deep building pit of Ceintuurbaan
Station, Amsterdam, The Netherlands. In the deepest excavation stage of this pit the soil weight and
uplift force would not balance. As dewatering or other measures were deemed not feasible, the
deepest part of the excavation and the construction of the bottom slab were initially designed to take
place under compressed-air. However, a considerable reduction of cost and construction time was
achieved after extensive analysis of the friction between soil and diaphragm walls proved that the
pressurised excavation could be limited.
Value engineering
Air
Ground+Air
Pore-pressure
Ground+Friction
Pore-pressure
Scope
Following an introduction to the project and case study, this paper will comprise in subsequent
sections:
verification of failure by uplift for a deep excavation
analysis of diaphragm wall friction, focusing on the calculation of horizontal stresses after
excavation
value engineering: assessment of the probability of failure by means of Monte Carlo analysis
observational method: verification of design assumptions through field and laboratory tests
conclusion.
Construction method
To minimize the duration of impact on city and infrastructure, the top-down method was selected
for construction. The construction phases as adopted in the design can be summarised as follows:
installation of diaphragm walls
installation of jet grout strut below deepest excavation level
construction of roof and restoration of street level
sub-roof, multiple stage, excavation and installation of pre-stressed struts
construction of intermediate floor
sub-floor excavation, installation of pre-stressed struts and construction of base slab working
under up to +1.6 bar compressed-air
TBM passage
final construction
2
Geometry
The geometry of the building pit in the deepest excavation phase is depicted in figure 3.
At first, an assessment into the settlement risk of adjacent historic buildings resulted in strict
requirements regarding deformations of the building pit. Typically, the historic buildings in
Amsterdam are founded on wooden piles driven into the so-called 1st sand layer. It is not
uncommon for these pile foundations to have barely sufficient bearing capacity, if verified along
modern-day standards. Hence, the retaining walls, roof and floor slabs, several layers of prestressed struts and a jet grout strut have been designed to resist all actions and forces, and to limit
differential settlements of adjacent foundations.
Secondly, the verification of the vertical stability of the building pit resulted in the requirement to
extend the diaphragm walls below the so-called intermediate sand layer. This layer at NAP -37 m is
water bearing with a piezometric head of approximately NAP -3 m. Practically speaking, lowering
this head to a secure level to avoid instability of the remaining Eem clay layer during excavation
could only be achieved by confining the intermediate sand layer. The D-walls have therefore been
designed to function as cut-off walls too, resulting in lengths of up to 45 m.
Geotechnical conditions
The geology, depicted in figure 3, is characteristic for the centre of Amsterdam (Wit, 1999). The
stratigraphy is formed by a glacial basin filled with sediments. The 3rd sand layer, relevant as
highly permeable aquifer, is at its base. For this papers case study, especially the Eem and Drenthe
clay layers and the intermediate sand layer are of interest. The latter are a glacial clay and
fluvioglacial sand deposited in the Saalian period. These glacial deposits are overlain by marine
clays of Eemian age. Above, the 1st and 2nd sand layer, often separated by the more silty Allerd
layer, have been combined into one layer here. These two medium to dense, aeolian (1st) and
fluvial (2nd) sand layers are of Weichselian origin. On top, the Holocene deposits have been
condensed into one layer too. This unit consists of a tidal sand and mainly of soft clay and peat
layers.
The sand layers are permeable, water bearing strata. For this case study it is assumed that all have a
head of circa NAP -3 m. The freatic level in the Holocene layers differs, and is circa NAP -0.5 m.
A summary of geotechnical parameters for all layers is included in figure 3. For further reference,
table 1 comprises additional data on the marine Eem clay and glacial Drenthe clay.
Table 1 - Geotechnical parameters Eem and Drenthe clay, mean values
Parameter
Eem clay
Drenthe clay
water content
w
[%]
36
23
liquid limit
wL
[%]
42
28
plastic limit
wP
[%]
23
18
liquidity index
IP
[%]
19
10
undrained shear strength
cu
[kPa]
150
180
compression index
Cc
[-]
0.358
0.143
secondary compression
[-]
0.0044
0.0018
C
swelling index
Csw
[-]
0.033
0.014
consolidation coefficient
cv
[m2/s]
1*10-6
2*10-6
permeability
k
[m/s]
2*10-9
1*10-9
sat
Description
'
3
[kN/m ]
[]
E'50;ref OCR
c'
[kPa] [kPa] [-]
0m
width = 11m
0m
Roof, t = 0.9m
-5m
Struts, pre-stressed
-10m
Struts, pre-stressed
-15m
Struts, pre-stressed
(temporary)
-19m
Floor, t = 0.9m
-25m
Struts, pre-stressed
(temporary)
-31m
-33m
-45m
D-wall, t = 1.2m
-3m
15
28
8000
n/a
-12m
19
34
34000 n/a
-25m
18
32
15
13000 2,0
-37m
Intermediate layer,
medium SAND
19.5
19.3
33
25000 n/a
-40m
33
11
15000 1,5
-45m
19.5
35
35000 n/a
Figure 3 - Soil parameters and geometry (modified for the purpose of the paper)
(1)
where Gd = design value of total weight of Ground (excavation level - bottom of clay layer)
Fd = design value of total side wall Friction
Ad = design Air pressure above atmospheric pressure
Pd = design value of total Pore-pressure (in 3rd sand layer applied under clay layer)
In the remainder of this paper the excess air pressure is presumed to be zero (i.e. atmospheric), the
excavation level is taken at NAP -26 m (i.e. just below the lowest temporary strut level) and the
weight density of water is 10 kN/m3.
The total weight of the soil and sum of the water pressures seem well defined parameters. Both can
be deduced from sampling and piezometer readings respectively. In this example:
Gd = 353 kN/m2 * W / m
Pd = 420 kN/m2 * W
W = 11 m width
m = 1.1 according to Dutch code NEN6740:1997
Clearly, without air-pressure or wall friction equation (1) would not be met.
The friction Fd between soil and diaphragm wall requires closer attention. The wall friction cannot
be tested or monitored in-situ directly. The next section therefore deals with the question how to
predict the maximum side wall friction.
It should be noted that the strength of the jet grout strut in the Eem layer is being neglected in the
analysis. Yet, its stiffness is taken into account implicitly. This approach has been chosen to design
on the safe side. Cause, despite of a perhaps vertically stabilising effect of the grout strut, its
primary function has been determined to reduce deformations. It has not been designed to take
lateral load. On the other hand, its presence reduces a (favourable) horizontal pre-stressing of the
soil. This effect is similarly being neglected in the analytical analysis below.
(2)
(3)
Below, first the distribution of normal stresses will be evaluated, and next the interface friction.
Normal stress distribution
Several methods are available for the estimation of effective stress normal to a retaining wall.
Essential in this case is the notion that the deep clay layers are in a state of over consolidation after
excavation. The ratio between horizontal and vertical stress changes during excavation and might
differ considerably from K0;nc. For comparison, the finite element code Plaxis and an analytical
model have been applied. In both the decrease in horizontal effective stress is deduced from the
reduction in vertical effective stress.
As illustrated in figure 4, the stress path for elastic unloading in Plaxis Hardening-Soil model can
be described with:
h = v * ur / ( 1 - ur )
(4)
Though, the ratio of horizontal and vertical stress is limited by the Mohr-Coulomb failure criterion.
Plastic failure would occur if:
h;excavated = v;excavated * Kp + 2 c Kp
(5)
'h [kPa]
Kp = ( 1 + sin ) / ( 1 - sin )
(6)
300
K_0;nc
v_ur / (1-v_ur)
200
s'v*Kp + 2c'sqrt(Kp)
Hardening-Soil
100
K0 - OCR (Mayne)
0
0
100
200
300
'v [kPa]
Figure 4 - Stress paths over consolidated Eem clay, unloading from v;initial = 255 kPa
Alternatively, application of the well known K0 - OCR relationship (Schmidt, 1966) suggests a
somewhat different stress path during unloading:
h = v * K0;nc * OCRsin = v * ( 1 - sin ) * OCRsin
(7)
where
OCR = v;max / v = OCRinitial * v;initial / v;excavated
6
(8)
The difference between the Hardening-Soil model and the K0 - OCR relationship becomes apparent
for small vertical stresses. Equations (4) and (5) show that the minimum value found for the
horizontal stress h when using the Hardening-Soil model for unloading would still equal 2cKp,
even for zero vertical stress. On the contrary, the K0 - OCR relationship results in zero horizontal
stress for zero vertical stress. The relationship (7) finds sound support in the statistical analysis of
ample laboratory test results on clay and sand samples (Mayne, 1982).
Concluding, the analysis of normal stresses after unloading by means of the K0 - OCR relationship
will result in a lower design value of total side wall friction. The latter is therefore regarded a safer
design approach and has been adopted in this case study.
Interface friction
The friction capacity of the soil - diaphragm wall interface might in two ways be affected by the
construction method itself. First, the slurry trench phase is likely to result in an alteration of initial
lateral stresses. Second, the bentonite support slurry is known to plaster the trench wall, forming a
filter cake, which might not be expelled during concreting of the panel. Especially in granular soils,
such cake of soft clay has been observed to decrease the shaft friction capacity.
To take the latter effect into account, a range of /-ratios can be found in literature. The possible
reduction of lateral stress has been expressed in a Ks/K0-ratio. With reference made to research data
on slurry supported bored piles (Kulhawy, 1991), the following values were initially recommended
for use within the framework of this case study:
/ = 0.8 - 1.0
Ks/K0 = 0.6 - 0.7
or combined into a Plaxis-type interface strength reduction factor R:
R = 0.5 (lower boundary value)
R = 0.7 (mean value)
Main considerations in selecting these values were: the sedimentation and presence of a bentonite
cake was regarded less likely at the low permeability marine and glacial clay layers. Besides,
former research had shown incremental outward horizontal displacements in the deeper layers
during concreting of a panel (Wit, 2002). Nonetheless, being a key parameter, it was determined
that validation of the interface friction by testing prior to the critical excavation stage would be
decisive. The results of these tests are summarised at the end of this paper.
Vertical eff. stress [kN/m2]
-100
-25
100
200
300
-100
-25
100
200
300
-100
-25
-30
-30
-30
-35
-35
-35
-40
-40
-40
initial
-45
excav.
initial
-45
excav.
-45
100
200
300
400
initial
excav. HS-model
excav. K0-OCR
Figure 5 - Interface stresses prior to and after excavation down to NAP -26m (mean values, 1D-analytical)
7
(9)
where
Z = Resistance - Sollicitation = (Gk + Fk) - Pk
(10)
Table 2 - Stochastic variables in Monte Carlo analysis (refer to figure 3 for mean values)
Parameter
Distribution
Stand. deviation
[kN/m3] standard normal
weight density
0.4 to 1.5
sat
angle of shearing resistance
standard normal
2 to 4
[]
cohesion intercept
c
[kPa]
n/a
0
interface reduction factor
R
[-]
uniform
0.2
over consolidation ratio
OCR
[-]
standard normal
0.25
length of D-wall
L
[m]
standard normal
0.1
piezometric level
[m]
uniform
0.2
- at excavation level
hexcav.
[m]
standard normal
0.5
- at intermediate sand layer
hint.sand
h3rd sand
[m]
standard normal
0.05
- at 3rd sand layer
Results
The Monte Carlo analyses provided clear and quantitative data on the probability of failure by
uplift, and on how this failure probability relates to the contributive factors: soil weight, wall
friction and uplift. Their probability density functions are shown in figure 6, together with a
histogram of the reliability function Z.
Reliability function; failure = Z<0
100
5000
F (1358)
G (3923)
P (4620)
3000
4000
75
50
4000
25
0
-100 -50
50 100
3000
2000
1000
0
-100
200
500
800
1100
1400
1000
2000
[kN/m']
5000
For the case study presented, the following results have been computed:
total number of calculations = 50,000
number of calculations with Z<0 = 8
probability of failure Pf = 8 / 50,000 = 1.6 * 10-4
result reliability index = 3.6 (assuming a normal distribution for reliability function Z)
required reliability index 3.6 (NEN6700:1997)
Concluding, the vertical stability of the deep excavation down to NAP -26 m meets the
requirements regarding reliability, also without pressurising the working chamber. The remainder of
the section shall be excavated under compressed-air. Although the frictions relatively wide
probability density function could be said to increase the uncertainty (the tail) of the reliability
function, its contribution to the safety is significant.
The presence or absence of a bentonite filter cake can, presumably, be explained by the difference
in permeability of the ground layers. It seems that where (nearly) no bleeding or consolidation of
the bentonite slurry could take place, such as in the low permeability clay layers, the slurry has been
effectively expelled during concreting of the D-wall panel. In contrast, the thickness of bentonite
cake in the sand layers exceeded the specific roughness of the diaphragm wall in several cases
observed. A possible correlation between cake thickness and duration of the slurry trench phase
could not be investigated.
Direct shear tests
To verify previously assumed values for shearing resistance direct shear tests on samples including
and excluding bentonite cake have been performed. For reference, also laboratory-made samples
with and without filter cake were tested. For the bentonite-cake-in-sand samples only those results
were considered where the shear plane clearly cut the filter cake (sand - sand shear disregarded).
The test results confirmed a /-ratio of 0.5 (lower boundary) to 0.7 (mean value) for diaphragm
wall friction in sand. In Eem clay it was suggested to calculate the maximum shear in connection
with the test results as:
max = N * tanshear test (note: c=0), with
mean = 35
st.dev. = 4.9
/ 1.
Since the possible decrease of horizontal stress due to installation effects could not be deduced from
the tests, it was recommended to apply the aforementioned Ks/K0-ratio of circa 0.7 undiminished.
Design verification
A re-run of the Monte Carlo analysis, implementing the direct shear test parameters, proved a
minimal difference with earlier calculations. Hence, the design parameters could be confirmed.
Note: the results depicted in figure 6 are based on the latter analysis.
CONCLUSION
Side wall friction can contribute significantly to the vertical stability of a diaphragm walled deep
excavation. In the case study presented, the safety against failure by uplift was demonstrated
through probabilistic analysis of all relevant parameters.
The approach to calculate the horizontal stress distribution after excavation by implementation of
the K0 - OCR relationship leads to a safer design than the application of the Hardening-Soil model.
In contrast to some literature, a bentonite cake on the diaphragm wall could be observed in sand
layers, but was virtually absent in low permeability clay layers.
Furthermore, the results lead to practical recommendations regarding groundwater management,
monitoring and lower acceptable levels for pressurised air during the deepest excavation stage than
earlier anticipated. Based on the analysis, the client was able to significantly reduce the application
of compressed air, without bearing higher risk.
10
REFERENCES
Kulhawy, F.H. (1991), Drilled shaft foundations, in Fang, H.Y. (Ed.), Foundation Engineering Handbook, Van
Nostrand Reinhold, pp.537-552
Mayne, P.W. and Kulhawy, F.H. (1982), K0-OCR Relationships in soil, Journal of the Geotechnical Engineering
Division, Proceedings of the American Society of Civil Engineers, Vol.108, No.GT6, pp.851-872
Schmidt, B. (1966), Earth pressures at rest related to stress history, Canadian Geotechnical Journal, National
Research Council, Ottawa, Vol.3, No.4, pp.239-242
Wit, J.C.W.M. de and Lengkeek, H.J. (2002), Full scale test on environmental impact of diaphragm wall trench
installation in Amsterdam, Proceedings of International Symposium on Geotechnical Aspects of Underground
Construction in Soft Ground, Toulouse
Wit, J.C.W.M. de, Roelands, J.C.S. and Schiphouwer, R.A. (1999), Geotechnical design aspects of the deep
underground stations in the North/South Line in Amsterdam, in Barends, F.B.J. et al. (Ed.), Geotechnical engineering
for transportation infrastructure, Taylor&Francis, pp.211-220
http://www.northsouthline.com/ (2009)
11