Professional Documents
Culture Documents
www.elsevier.com/locate/compgeo
a
Civil Engineering Laboratory, Biskra University, BP 145 Biskra 07000, Algeria
b
URGC Géotechnique, INSA Lyon, 20 Avenue, Albert Einstein, 69621 Villeurbanne Cedex, France
Received 28 July 2004; received in revised form 2 March 2005; accepted 3 March 2005
Available online 10 May 2005
Abstract
Urban development often requires the construction of deep excavations. These might be for deep basements, cut-and-cover tun-
nels, underground parking or underground transportation systems. The design of deep excavations is often dominated by the water
flow around sheet piles or propped walls. The seepage flow influences the stability of the walls where bulk heave, piping or failure by
reduction of the earth pressure may occur. Several methods of calculating the stability against seepage failure of the soil have been
proposed in the literature, leading sometimes to great differences on the hydraulic head loss inducing failure. In this paper, the
FLAC-2D Code using an explicit finite difference method is used to analyse seepage failure of the sandy soil within a cofferdam sub-
jected to an upward seepage flow. Based on this study, the conditions for seepage failure occuring by boiling or heaving of the soil
behind sheet piles are clearly identified.
2005 Elsevier Ltd. All rights reserved.
0266-352X/$ - see front matter 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compgeo.2005.03.001
N. Benmebarek et al. / Computers and Geotechnics 32 (2005) 264–273 265
Nomenclature
the wall. He assumed, from experimental evidence, that General upheaval ‘‘heaving’’, which involves a
the body of sand, which is lifted by water, has the shape greater volume of soil.
of a rectangular prism (Fig. 1) with a width equal to the
half of the wall penetration D/2 and the horizontal base Piping or boiling occurs when a small prism of soil at
at some depth D0 below the surface (0 6 D0 6 D). It is the excavation level is not sufficiently heavy to resist up-
assumed that at the instant of failure, the effective hori- lift forces due to the upward flow. Terzaghi (1943) de-
zontal stresses on the prism vertical sides and the corre- fined the concept of critical hydraulic gradient (or
sponding frictional resistance are zero. flotation gradient) ic which controls the phenomenon
Therefore, the prism rises up and collapses as soon as of Piping or boiling.
the total excess water pressure Ue on the bottom of the McNamee defined the safety factor against boiling as
prism OA becomes equal to the submerged weight of the the ratio of the critical gradient to the exit gradient ie at
prism W 0 . the excavation level, hence
The safety factor against bulk heave is determined by F ¼ ic =ie . ð2Þ
the ratio of the submerged weight of the prism to the ex-
cess water force on the prism base: The author presented charts enabling the determina-
tion of the corresponding safety factor for a range of
F s ¼ W 0 =U e ¼ ic =im ; ð1Þ
dimensions of the excavation.
where im is the average hydraulic gradient between IJ Marsland [5] undertook extensive model tests using
and EF. both dense and loose homogeneous sands in an open
McNamee [4] identified two main types of failure: water excavation. He concluded that in loose sand, fail-
ure occurs when the pressure at the sheet pile tip is suf-
Local failure as ‘‘piping’’ or ‘‘boiling’’ is most likely ficient to lift the column of submerged sand near the
to begin at a point on the surface adjacent to the sheet wall of the cofferdam (the width of half the embedment
pile as it lies within the shortest seepage path. is not mentioned). In dense sand, failure occurs when the
exit gradient at the excavation surface reaches a critical
value.
By examining the equilibrium conditions with re-
spect to excess pore pressures, Davidenkoff [6] showed
that bulk heave of a rectangular prism is possible only
H if the shear forces at vertical faces are neglected and its
width is smaller than noted by Terzaghi [3]. He con-
D/2
cluded that the prism rises up as soon as the mean
E F gradient calculated over the embedment length of the
wall becomes equal to the critical gradient and the fail-
Wsat ure prism begins from the wall toe for homogeneous
D D0 soils.
Bazant [7] established a failure criterion with regard
I J to the soil shear strength by plotting the excess pore
pressure at the sheet pile tip divided by the embedment
depth of the sheet pile against the soil internal friction
angle.
U
Davidenkoff and Franke [8] proposed a diagram
Fig. 1. Failure by heaving [3]. based on model studies, which can be used to determine
266 N. Benmebarek et al. / Computers and Geotechnics 32 (2005) 264–273
the safety factor against piping for excavations in open 3. Numerical modeling procedure
water with different thickness of the pervious stratum.
The comparison between the safety factors derived 3.1. Case study
from these various approaches in the case of a sheet pile
wall driven in a semi infinite homogeneous and isotro- This paper deals with the numerical study of bottom
pic soil reveals significant variations reaching up to stability against seepage failure of sand within a coffer-
75% [9]. dam. A sheet pile with penetration depth equal to D in
Soubra et al. [10] published results of the passive homogeneous isotropic semi-infinite soil, which is sub-
earth pressure coefficients in the presence of hydraulic jected to hydraulic head H, is considered as shown in
gradients using the variational approach applied to Fig. 3.
the limit equilibrium method. Their results (Fig. 2) This problem has many parameters: excavation or
showed a quasi-linear decrease in the passive earth pres- cofferdam width, penetration of retaining wall in the
sure coefficients Kp for the hydraulic head loss (H/D) soil, thickness of permeable soil, wall flexibility, strut
values varying from 0 to 2.5, for the case of a single rigidity, wall translation and rotation . . . etc.
sheet pile wall driven into a homogeneous and isotropic The objective of this work is not to consider the influ-
semi infinite soil medium. The passive earth pressures ence of all these parameters but to check if a numerical
are sensitive to the soil–structure interface friction d analysis using the finite difference or the finite element
but vanish completely at the same value of H/D = 2.78 approach can describe correctly the various failure
for different interface friction angles. They concluded mechanisms observed due to upward seepage flow into
that the angle of friction at the soil–structure has no ef- a cofferdam. Therefore, the problem has been simplified
fect on the H/D value causing failure by heaving. For by referring to some model tests observations of failures
the same case, TerzaghiÕs approach gives a value of and to the classical approaches where the action of the
H/D against seepage failure by heaving equal to 2.82, wall on the soil is not directly considered, but the wall
while the boiling phenomenon which appears for a crit- is being considered as fixed.
ical hydraulic gradient at point E (Fig. 1), occurs with a An infinite rigid wall fixed by struts is considered in
theoretical value of the hydraulic head loss equal to this study. This is a simplification of the real problem
H/D = 3.14 = p. where the wall flexibility leads to a limited partial sup-
From these overviews, it appears that seepage failure port of the soil on the wall embedment. This allows to
at the excavation bottom can arise from different failure show the presence of the supposed fixed wall effect on
mechanisms. the failure mechanisms as well as the comparison of
Our aim in this paper is to propose a numerical pro- the results with the classical approaches where the wall
cedure in order to evaluate the head loss for seepage fail- is taken into account from the hydraulic point of view
ure at bottom excavation and to identify the different only.
failure mechanisms using the explicit finite difference The analysis is carried out using the computer code
method implemented in FLAC-2D code. FLAC-2D (Fast Lagrangian Analysis of Continua)
which is a commercially available finite difference expli-
cit program.
Fig. 2. Kp versus H/D for / = 30 and d// = 0,1/3,1/2 and 2/3 in the
case of a homogeneous isotropic semi-infinite medium [10]. Fig. 3. Case study.
N. Benmebarek et al. / Computers and Geotechnics 32 (2005) 264–273 267
The soil behavior is modeled by the elastic-perfectly The prediction of collapse conditions under steady
plastic nonassociative Mohr–Coulomb model encoded plastic flow conditions is one that can be difficult for a
in FLAC-2D code. All subsequent results are given for numerical model to simulate accurately [12]. Frydman
csat/cw = 2, elastic bulk modulus K = 30 MPa and shear and Burd [13], and Erickson and Dresher [14] operating
modulus G = 11.25 MPa. Four values of the angle of with viscoplastic or elastoplastic algorithms, have shown
internal friction u = 20, 30, 35, 40, four values of clearly the dependence of the plane strain bearing capac-
the friction angle at the soil/wall interface d/u = 0, 1/ ity factor Nc on the geometry of the mesh and indicate
3, 2/3, 1, and three values of the dilation angle w/ the reduction of this factor with decreasing value of soil
u = 0, 1/2, 1 are considered in the analysis. dilation angle.
Accordingly, in order to develop an acceptable anal-
ysis scheme for later computations, preliminary simula-
4. Modeling procedure tions have been carried out by testing the influence of
the mesh dimensions, the element size, the boundary
In the case of a rough wall, modeling the interface be- conditions and the earth pressure coefficient at rest K0
tween the soil and the wall is invariably an integral part as well.
of the analysis. The results show the difficulty in the capturing of the
The material properties, particularly stiffness assigned failure mechanisms for coarse meshes and prove the
to an interface, depend on the way in which the interface requirement of refined mesh to capture it clearly. As
is used. In the case of soil–structure interaction, the an illustration, Fig. 5 shows the failure mechanism pre-
interface is considered stiff compared to the surrounding sented by the displacement field and the corresponding
soil, but it can slip and may be open in response to the distribution of maximum shear strain rates at steady
loading. Joints with zero thickness are more suitable for state plastic flow for coarse mesh 40 · 20 (horizontal
simulating the frictional behaviour at the interface be- by vertical) and fine mesh 80 · 40 elements.
tween the wall and the soil. Also, the results confirm that variation in practical
The interface model shown in Fig. 4 has been used to range of the elastic soil parameters and earth pressure
simulate the soil/wall contact described by Coulomb coefficient at rest K0 do not have any significant influ-
law. The logic contact for either side of the interface is ence on the critical hydraulic pressure loss, as the
similar in nature to the interface used in the distinct ele- numerical estimation of the bearing foundation capacity
ment method [11]. factor Nc [13,14].
The spring in the tangential direction and the slider Fig. 6 shows the mesh (80 · 40 elements) and bound-
(Fig. 4) represent the Coulomb shear-strength criterion. ary conditions retained for this analysis. The mesh size is
The spring in the normal direction and the limit strength fine near the wall where deformations and flow gradients
represent the normal contact. The interface has a friction are concentrated. In order to minimize boundary effects,
angle d, a cohesion c = 0 kPa, a normal stiffness the length from the wall and the depth of the mesh are
Kn = 109 Pa/m, and a shear stiffness Ks = 109 Pa/m. These respectively located at six and five times the wall pene-
values of Kn and Ks are selected to approximate the results tration. As a general rule for the boundary conditions,
for the case where the wall is rigidly attached to the grid. the bottom boundary is assumed to be fixed, the right
(a)
0 2E 0
Boundary plot
0 2E 0
(b)
Boundary plot
0 2E 0
Max. shear strain-rate
0.00E+00
4.00E-06
8.00E-06
1.20E-05
1.60E-05
2.00E-05
2.40E-05
2.80E-05 80x40 elements
Contour interval= 2.00E-06
Displacement vectors
Max Vector = 3.164E-01
0 1E 0
Fig. 5. Capture failure mechanisms when / = 35, d// = 2/3, w// = 1/2, H/D = 3.00 for: (a) coarse mesh 40 · 20; (b) fine mesh 80 · 40.
and left lateral boundaries are fixed in the horizontal or boiling, the following three simulation procedure
directions. steps are adopted:
The sheet piles wall is modeled by structural beam
elements connected to the soil grid via interface elements 1. The geostatic stresses are computed assuming the
attached on both sides of the beam elements. The wall material to be elastic. The initial pore pressure and
thus acts as an impermeable member. effective stresses are established using the fish library
To identify the limiting cases corresponding to zero function assuming that the ratio of effective horizon-
passive earth pressures and seepage failure by heaving tal stress to effective vertical stress at rest is taken 0.5.
N. Benmebarek et al. / Computers and Geotechnics 32 (2005) 264–273 269
Wall
Z
H H=0
∂H
=0 ∂H
∂x =0
∂x
Impermeable
∂H
=0
∂z
the hydraulic head loss H/D, various governing parame- These results indicate clearly that the bottom stability
ters soil friction, soil dilation and interface friction. against seepage failure always corresponds to a bulk
heave except in the case of a dense sand / P 40, a dilat-
ing material w// C 1/2 and a rough wall d// P 2/3
5. Results and discussion where boiling would occur. It can be seen that boiling
starts from a hydraulic pressure loss of H/D = 3.16. This
Numerical studies are performed for different soil value is very close to the theoretical critical head loss va-
friction angles and the results are presented in Table 1 lue H/D = 3.14. In this case, the exit hydraulic gradient
for / = 20, 30, 35 and 40. Results for failure by attains the critical hydraulic gradient value. Fig. 8 shows
heaving are presented by one asterisk or two and failure the failure mechanism indicated by the displacement
by boiling by three. field and the corresponding distribution of maximum
Boundary plot
0 2E 0
Displacement vectors
Max Vector = 6.112E-01
0 2E 0
0 2E 0
Fig. 8. Displacement field and the corresponding distribution of maximum shear strain rates when / = 40, d// = 2/3, w// = 1, H/D = 3.2.
N. Benmebarek et al. / Computers and Geotechnics 32 (2005) 264–273 271
shear strain rates obtained in the case of / = 40, d// demonstrate that soil dilation angle has a significant
= 2/3, w// = 1 and H/D = 3.2 where boiling phenomena influence on the failure mechanism shape.
is indicated. From Table 1, for d// = 0, w// = 1 and / varies from
For w// = 0, a rectangular soil prism similar to that 20 to 40, the critical value of H/D lies in the range of
proposed by Terzaghi [4] is observed. Fig. 9 shows the 2.64–2.93 It is noted that for various values of / Ter-
failure mechanism in the case of / = 35, w// = 0, d// zaghiÕs solution is H/D = 2.82 whereas that of Soubra
= 2/3 and H/D = 3 where bulk heave of a rectangular et al. [11] solution is H/D = 2.78. Also, for / = 40,
soil prism with a width smaller than that of TerzaghiÕs w// = 1 and d// varies from 0 to 1, the critical value
method is observed. However, for dilating material of H/D lies in the range of 2.93–3.16.
w// P 1/2, a triangular soil prism is obtained and as These results show that the critical hydraulic pressure
an example, Fig. 10 shows the case of / = 35, w// = loss H/D corresponding to the zero effective passive
1/2, d// = 2/3, H/D = 3 where bulk heave of a triangle pressures depends on the soil friction angle and the
soil prism is observed. The present simulation procedure interface soil/wall friction.
Boundary plot
0 2E 0
Displacement vectors
Max Vector = 1.562E+00
0 5E 0
Boundary plot
0 2E 0
Max. shear strain-rate
0.00E+00
5.00E-05
1.00E-04
1.50E-04
2.00E-04
2.50E-04
3.00E-04
Fig. 9. Displacement field and the corresponding distribution of maximum shear strain rates when / = 35, d// = 2/3, w// = 0, H/D = 3.
272 N. Benmebarek et al. / Computers and Geotechnics 32 (2005) 264–273
Boundary plot
0 2E 0
Displacement vectors
Max Vector = 3.789E-01
0 1E 0
Boundary plot
0 2E 0
Max. shear strain-rate
0.00E+00
3.00E-06
6.00E-06
9.00E-06
1.20E-05
1.50E-05
1.80E-05
2.10E-05
2.40E-05
Fig. 10. Displacement field and the corresponding distribution of maximum shear strain rates when / = 35, d// = 2/3, w// = 1/2, H/D = 3.
This behaviour can be explained as follows: when the exit gradient becomes critical before the bulk heave of
soil adjacent to the wall expands by dilation at failure, the triangular prism and initiates the boiling. This case
the shear forces induced on the vertical faces of the may also occur for a flexible wall supported partially
prism block the rising of the rectangular prism. There- by the soil.
fore, a triangular failure prism appears instead. This
corresponds to a kinematically admissible mechanism
within the frame of limit analysis theory. This phenom- 6. Conclusions
enon has been experimentally observed by Kastner [9].
For high values of / and w, shear forces on the wall Several methods have been proposed to assess the
embedment and horizontal soil expansion by dilation risk of seepage failure at the excavation bottom of a
delay the triangular prism failure. Therefore, the surface braced cofferdam by heaving or boiling. Sometimes,
N. Benmebarek et al. / Computers and Geotechnics 32 (2005) 264–273 273