Settlement Trough Due To Tunneling in Cohesive Ground
Settlement Trough Due To Tunneling in Cohesive Ground
Settlement Trough Due To Tunneling in Cohesive Ground
Key words Abstract: Surface settlements of soil due to tunneling are caused by stress relief and subsidence
Tunnel, clay, finite elements, due to movement of support by excavation. In this paper, the shape of settlement trough caused
by tunneling in cohesive ground is investigated by different approaches; namely, analytical
settlement, complex variable
solutions, empirical solutions and numerical solution by the finite element method. The width of
settlement trough was obtained through the finite element method by establishing the change in
the slope of the computed settlement profile. The results show that the finite element method
overpredicted the settlement trough width compared with the results of Peck for soft and stiff clay
but there is an excellent agreement with Rankin's estimation. The results show that there is a good
agreement between the complex variable analysis for z/D=1.5, while using z/D = 2 and 3, the
curve diverges in the region far away from the center of the tunnel.
γ D2 z 2
δz = − (2)
Analytical Solutions 4G (x2 + z 2 )
SETTTLEMENT TRO
OUGH DUE TO O TUNNELING IN I COHESIVE G GROUND
Mohammed Y. Fattah, Kaiis T. Shlash an
nd Nahla M. Sa
alim
The commplete solution n of the probleem consists of maximum vertical ssettlement, x is the transsverse
the
t sum of the e three partial solutions. In order
o to verify distance e from tunnel centerline, an nd i is the wid
dth of
the
t consistenccy and the accuracy of the solution, a settleme ent trough whiich is the dista
ance to the pooint of
computer proggram had been n developed by Verruijt and inflectio
on of the curve
e (corresponding to one standard
Booker (2000)). This program m, named MIND DLIN, enables deviatioon of the normal distribution curve), and is
the
t user to obtain
o numericcal results of stresses and determined by the e ground conditions.
c Va
arious
displacementss in each poin nt of the field,, to construct expressions have bee en proposed for calculatingg i In
different quanttities and to va
alidate the solu
ution. practicee, Rankin (19888) suggested that the relatioon will
be as foollows:
Empirical Solution
i = k zo (
(6)
Empirical Greenfield Settlem
ment Trough
w
where k is a dimensionless constant,
c depe
ending
Peck (1969)
( described settlemen nt data from on soil type
t (0.5 for clay
c & 0.25 forr cohesionless soils)
over twenty caase histories avvailable to him
m at that time. and z0 is the depth of the tunnell axis below gground
It was possible
e to deduce tha at the short-terrm transverse level.
settlement
s trrough in the e ‘Greenfield d’ could be
approximated by a normal distribution or Gaussian P
Peck (1969) and Cording an nd Hansmire (11975)
curve shown inn Figure 1. presente
ed a normalize ed relation of the width param
meter,
2i/D versus the tunne el depth, zo/D, for tunnels driven
d
The equation represe
enting the asssumed trough through different geolo
ogical conditionns. That is:
shape
s is as follows:
0.8
2i ⎛ z 0 ⎞
x2 =⎜ ⎟ (
(7)
δ = δ max expp( − 2 ) (5) D ⎝D⎠
2i
In which
h D is the diam
meter of the tunnel.
Where δ is the surfa
ace settlement, δmax is the
India
an Geotechniccal Journal, 41
1(2), 2011
Fujita (1982)
( statisttically analyzed the maximu um T method off predicting the
The e settlement trough
t
surface
s ment caused by shield tunnelling based on 96
settlem due to tunneling
t is investigated, an
nd the effect ofo rigid
cases in Japa an. A reasonab ble range of δmax for differe
ent boundaries is explorred. In using and develop ping a
types
t of shieeld machines driven througgh different soils model, it is necessarry to ensure that each tech hnique
conditions, with
w or withou ut additional measures was w used is correct and the ey are mutuallyy compatible. This
T is
suggested.
s achieved by testingg and validattion using exxisting
solutions. The problem studied by b Chow (199 94) is
A comparison of th he various em mpirical metho ods reanalyzzed in the follo
owing sections using the com mputer
discussed above was mad de on the assumption
a off a programm Modf-CRISP (Salim, 2006).
hypothetical foour meter diam meter tunnel loocated at a dep pth
of thirty meterrs which experrience a ground volume losss of Geommetry of the Problem and
a Soil
one percent (TTan and Ranjitth, 2003). From comparison of
various
v empiriical solutions for surface seettlement trough, Prope
erties
t maximum settlement ran
the nges from 3-5 mm whereas tthe
T tunnel analyzed by Chow has a diamete
The er of 5
trough
t width i varies between n 10 and 15 m.
m This shows th hat
m and different deptths. The eleme
ents chosen in this
there
t are siggnificant disccrepancies between empirical
applicattion are six node trianggular and 8-node
8
solutions
s to predict surface settlement tro ough because of
isoparammetric quadrilateral eleme
ents for the e two
different inte erpretations and
a databasee collection by
dimensional plane strain proble em. The ma aterial
different authoors (Tan and Raanjith, 2003).
properties which are e used by Gunn
G (1993),, and
representative of Lon
ndon clay as confirmed by Chow
Analysis
A by the Finite Element Method
d (1994), are presented in Table 1. The finite ele
ement
mesh is shown in Figure 2.
Most off the earlier nu
umerical modeels used the fin
nite
element meth hod. Other posssible numericcal methods a are
finite
f differencce and boundarry element tech
hniques.
The program Modf-CRISP is used in modelling tunnel axis. Values of z of 7.5 m, 10 m, and 15 m are
tunneling considering different soil models, as follows: used. The first analysis is carried out for short term
movement, undrained conditions; where the volumetric
1. Linear elastic. strain is zero. It follows that the value of undrained bulk
2. Nonlinear elastic. modulus must be infinite, and Poisson’s ratio equals
0.5. In finite element analysis, it is not possible to use
3. Elastic- perfectly plastic. an infinite undrained bulk modulus. To approximate
undrained conditions, a value of Poisson’s ratio of
Table 1 Soil properties for London clay (Chow, 1994)
0.495 is used. The soil stiffness is assumed to vary with
depth according to the following equation:
Property Value
E = Eo + m( zo − z ) (8)
Shear modulus, G (kPa) 33557
Where E is Young’s modulus, E0 is the initial
Poisson's ratio, ν 0.495
Young's modulus at datum elevation, m is the rate of
Unit weight of the soil, γ (kN/m3) 20 increase of Young's modulus with depth, z0 is the depth
of tunnel axis, and z is the depth to the point of interest.
Oteo and Sagaseta (1982) found that the bottom Mechanism of Short-Term Settlement Response
rigid base has the most significant effect on the
predicted settlements. When the depth of the rigid base Analytical approaches often require the
below tunnel axis, Hd, increases, the settlements identification of non-dimensional group of the important
decrease, resulting in surface heave for a value of parameters influencing a problem. One such group is
Hd/D>7. Hence in the present study, the depth of the the stability number N, which is defined by Broms and
bottom base of the mesh has been chosen accordingly. Bennermark (1967) as:
Computer Program (σ z − σ T )
N= (9)
Modf-CRISP allows elements to be removed to Su
simulate excavation and elements to be added to
simulate construction. The implied loading for both Where σz is the overburden pressure at the
these cases are automatically calculated by the tunnel axis, σΤ is the tunnel support pressure or internal
program. When performing a non-linear analysis pressure (if any), and Su is the undrained shear strength
involving excavation or construction, Modf-CRISP allows of clay. If σΤ =0 then N will be equal to σz /Su or γ z /Su.,
the effect of element removal or addition to be spread where z = C+D, C is the depth of tunnel’s crown. Three
over several increments in an "incremental block". An values of Su /γ z are used in the present analyses. The
increment block is just a series of ordinary increments values of Su /γ z and their corresponding Su for each z /D
grouped together in the input data for the program. are given in Table 2. These values are presented in
Element stiffness is always added or removed in the first Tables 3 and 4. The analyses were carried out using
increment of block, but the associated loads are homogeneous elastic analysis and homogeneous elastic
distributed over all the increments in the block. perfectly plastic analysis.
When a tunnel is excavated, unloading and stiffness
reduction occur simultaneously. The first results from Presentation of Results
removal of material having self weight. The second
occurs because the ground structure is altered. To
model the first of these processes, loads are applied to Linear Elastic Analysis
the mesh remaining after excavation. The second
process is modeled by excluding the stiffness of each The results obtained for three different depths of
element excavated from the calculation of structure the tunnel (z /D = 1.5, 2 and 3) using the program Modf-
stiffness. CRISP are shown in Figure 3, and presented in
dimensionless form as δ G / γ D2 versus x/D. Heave is
Linear Elastic and Elastic-Plastic Models noticed for z/D =1.5, 2 and 3. This is because the
movement of the soil is upward due to relief effect of
excavated soil in a purely elastic homogeneous medium.
The analyses are divided into two categories:
As x/D increases, the upward movement of the soil
elastic and elastic plastic. The analyses are carried out
decreases, this is because that the soil is remote from
for homogeneous soil and soil with properties varying
concentration of loading. These results conform with
with depth. For each soil type, an analysis is made for
Chow (1994) as illustrated in Figure 4.
three values of z, the depth below the surface of the
69
Table 3 Soil properties used for linear elastic model (from Chow, 1994)
Homogeneous Non homogenous
z (m) 7.5 10 15 7.5 10 15
1.5 2.0 3.0 1.5 2.0 3.0
z/D
Table 4 Soil properties used for elastic, perfectly plastic homogeneous analysis (from Chow, 1994)
Homogeneous
z (m)
7.5 10 15
z/D
1.5 2.0 3.0
G0 (kPa)
33557 33557 33557
m (kPa/m)
0 0 0
Su0 (kPa)
45 60 90 60 80 100 90 120 150
Reference* 1.5 ph3 1.5 ph4 1.5 ph5 2 ph3 2 ph4 2 ph5 3 ph3 3 ph4 3 ph5
0.25
1.5eh
Normalized Movement 0.2 2eh
3eh
0.15
0.1
0.05
0
-0.05
-0.1
0 5 10 15 20 25 30 35 40 45
X/D
Fig. 3 Surface displacement for elastic-homogeneous soil model predicted by the finite element method for different z/D ratios.
0.1
0.05
-0.05
-0.1
0 5 10 15 20 25 30 35 40 45
X/D
Fig. 4 Comparison of the analysis results with Chow's results for different z/D ratios using elastic homogeneous model
4
3
Fig. 5 Comparison between different approaches in predicting the surface settlement (z / D = 1.5)
1
Surface Settlement (mm)
-1
-2
Finite Element Analysis
Complex Variable Analysis
-3 Elastic Solution
Sagaseta Solution
-4
0 25 50 75 100 125 150 175 200 225
X (m)
Fig. 6 Comparison between different approaches in predicting the surface settlement (z/D = 2)
1
Normalized Movement
-1
-2
Fig. 7 Comparison between different approaches in predicting the surface settlement (z/D = 3)
72
-0.4
-0.6
-0.8
Finite Element Analysis
Complex Variable Solution
-1
0 50 100 150 200 250 300
x (m)
Fig. 8 Comparison of the surface settlement obtained by the finite element method with complex variable results (z/D =1.5)
0.2
Z/D=1.5en
Normalized Movement
0.15 Z/D=2.0en
Z/D=3.0en
0.1
0.05
0
-0.05
-0.1
0 5 10 15 20 25 30 35 40 45
X/D
Fig. 9 Surface displacement predicted by the finite element method using elastic nonhomogeneous soil model
for different z/D ratios and Su / γ z=0.3
0.2
Normalized Movement
0.1
0
-0.1
-0.2
-0.3 1.5ph3
2ph3
-0.4
3ph3
-0.5
0 4 8 12 16 20 24 28 32 36 40 44
X/D
Fig. 10 Surface displacement using elastic-plastic homogeneous soil model predicted by the finite element method
for different z/D ratios and Su / γ z=0.3
73
Although field investigations carried out by The results show that there is a good
Peck (1969) showed that surface settlements in agreement with Rankin (1988), O'Reilly and New
tunneling problems could fit with this type of (1982) but there is an approximately 36 %
probability function, computed settlements did not difference from Peck's estimation. This difference is
necessarily produce results that could be due to the fact that Peck used the probability
represented by this type of probability function. This function in the estimation of i which may not
is due to the fact that the data from field necessarily fit with the present results.
investigations were usually obtained in a region
Figure 13 shows the horizontal surface
comparatively close to the tunnel (e.g. < 20 m).
displacement obtained from the finite element
The predicted value of i becomes inaccurate results and empirical (Gaussian) solution (equation
since it is obtained from the gradient of ln δ/δmax 11). Here, the Gaussian model prediction is based
versus x2 plot, and is taken as the best fit line for the on the assumption that the vectors of ground
total horizontal width of the finite element mesh movement are oriented towards the tunnel axis
(200 m in this case).The second method was (O'Reilly and New, 1982). For the tunnel model, both
achieved by establishing the point where the change the form (allowing for the trough being too wide) and
the magnitude of maximum movement are similar to
in the slope of the settlement profile from positive to
negative occurs. the Gaussian model.
74
3.5
CRISP (pn3)
3 Stiff Clay (Peck)
2.5 Soft Clay (Peck)
O'Reilly and New (1982)
2
i/D
1
0.5
0
0 0.5 1 1.5 2 2.5 3 3.5
Z/D
Fig. 11 Plot of i/D against z / D using different methods in estimation of the settlement trough width (i)
X/D
-150 -100 -50 0 50 100 150
0
Surface Settlement (mm)
-1
-2
-3
-4
Finite elements (CRISP)
Peck, 1969
-5 Poulos and Davis, 1980
Sagaseta Solution
-6
Fig. 12 Surface settlement obtained from different approaches (deep tunnel) z 0 = 30 m , diameter = 4 m.
1
Horizontal Movement (mm)
0.8
0.6
0.4
0.2
0
-0.2
-0.4
-0.6 Finite Elem ents (CRISP)
-0.8 O'Reilly and New ,1982
-1
-125 -100 -75 -50 -25 0 25 50 75 100 125
X (m)
Fig. 13 Horizontal movement using different approaches for deep tunnel z 0 = 30 m , diameter = 4 m.
75
Table 5 Estimation of the settlement trough width (i) using different approaches.
i (m)
z (m) D (m) Finite Elements Peck O'Reilly and New Rankin
(Present study) (1969) (1982) (1988)
30 4 15 9.48 12.69 15