Professional Documents
Culture Documents
PII: S1350-6307(16)31072-X
DOI: doi: 10.1016/j.engfailanal.2017.04.030
Reference: EFA 3110
To appear in: Engineering Failure Analysis
Received date: 7 November 2016
Accepted date: 25 April 2017
Please cite this article as: Gang Zheng, Tao Cui, Xuesong Cheng, Yu Diao, Tianqi Zhang,
Jibin Sun, Longbo Ge , Study of the collapse mechanism of shield tunnels due to the
failure of segments in sandy ground. The address for the corresponding author was
captured as affiliation for all authors. Please check if appropriate. Efa(2017), doi: 10.1016/
j.engfailanal.2017.04.030
This is a PDF file of an unedited manuscript that has been accepted for publication. As
a service to our customers we are providing this early version of the manuscript. The
manuscript will undergo copyediting, typesetting, and review of the resulting proof before
it is published in its final form. Please note that during the production process errors may
be discovered which could affect the content, and all legal disclaimers that apply to the
journal pertain.
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
Study of the collapse mechanism of shield tunnels due to the failure of segments in
sandy ground
Gang Zheng1,2, Tao Cui2, Xuesong Cheng1,2,*, Yu Diao1,2, Tianqi Zhang2, Jibin Sun2, Longbo Ge2
1
MOE Key Laboratory of Coast Civil Structure Safety, Tianjin University, Tianjin 300072, China
2
Department of Civil Engineering, Tianjin University, Tianjin 300072, China
Abstract
PT
Catastrophic tunnel collapses usually start with the partial failure of tunnel linings.
RI
However, research on the mechanism of such progressive collapses is limited. Among
SC
such collapses, a considerable number of accidents have arisen from the construction of
mechanism underlying tunnel collapse due to this operation. The test results showed
MA
that the inrush of the sand due to the removal of a certain segment at the springline may
lead to the loss of soil resistance around the openings, resulting in a sudden increase in
D
the bending moments of the adjacent linings. In particular, an increase in the cover-
E
PT
depth-to-diameter ratio (C/D) can aggravate the development of the maximum bending
moment, although it has little influence on the affected length of the tunnel. To address
CE
the load sustained by the segments and joints during collapse that the model tests fail to
show, 3D numerical analyses based on the load-structure model were performed. Via
AC
comparison with existing failure criteria for both segments and joints, the safety of the
Keywords: Tunnel safety; partial failure; model tests; progressive collapse; failure
criterion
*
Xuesong Cheng. Tel.: +86 13821807355. E-mail address: cheng_xuesong@163.com
1
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
1. Introduction
The shield-driven tunneling method is widely used for the construction of urban
ground traffic and surface structures [1,2]. Because the tunnel is composed of thousands
of segments connected by relatively weak joints, there exist a large number of potential
PT
risks during both the construction and operation of the tunnel. Disasters involving large-
scale tunnel collapses induced by the partial failure of the tunnel linings can cause
RI
enormous craters. Such disasters have already occurred multiple times throughout the
SC
world [3-5] with catastrophic consequences, especially when located in densely built-up
areas.
NU
In recent years, a large number of shield tunnels have collapsed with severe
MA
consequences, such as the accidents in Shanghai [3], Nanjing [4], and Kaohsiung [5]. In
these cases, the collapse process exhibited certain typical processes of a domino effect,
D
catastrophes usually started with a partial failure of the lining but soon developed into a
PT
instance, the accident records of the Shanghai Metro Line 4 [3] indicated that the 274-
CE
m-long collapse began from the weakest part of the lining structures, where a
AC
connecting passage was being built. Due to the failure of the freezing method applied
around the connecting passage construction area, soil and water abruptly rushed into the
tunnel via the opening, where the segment had been moved. Afterwards, the tunnel
rings adjacent to the connecting passage began to fail progressively in the form of a
domino effect until approximately 274 m of tunnel was ultimately destroyed. In the
collapse, the ground sank and the buildings also inclined severely, as shown in Fig. 1.
2
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
PT
(a)
RI
SC
NU
MA
(b)
Depth from ground surface
32 m
Average depth of the upper part Original depth of the tunnel
33 m of the lining is 33.5 m
D
center is 33.5 m
34 m
E
35 m
Average depth of the upper part
36 m of the lining is 36.4 m
PT
39 m
part of the lining is 38.1 m
40 m
41 m
42 m
(c)
Fig. 1 The Shanghai Metro Line 4 accident: (a) schematic profile of the tunnel, (b)
inclined buildings and damaged ground, and (c) cross-sections of damaged tunnel
linings
systematic and intensive [6-10]. Design codes and guidelines related to resisting
3
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
progressive collapse in superstructures have been issued in many countries and regions
also been proposed [15,16], including the alternative path method [17], enhanced local
However, most investigations of tunnel safety problems have been performed in the
PT
context of forensic investigations of tunnel failures [18-20], analyses of the stability of
RI
linings during disasters [27-29]. Studies on the progressive collapse of a tunnel
SC
beginning with the partial failure of segments appear to be scarce.
In this study, the collapse mechanisms of a shield tunnel caused by the partial failure
NU
of segments at the springline in sandy ground were investigated using model tests and
MA
numerical simulations. In the model tests, the stiffness of the segments and joints and
the relative density of sand in the model were determined by strictly following 1-g
D
scaling laws [30-35]. Based on the test results, the influence of the partial collapse on
E
the adjacent tunnel rings was analyzed. The earth pressures measured around the tunnel
PT
were then entered into 3D numerical models to calculate the load that the segments and
joints sustained. Via comparison with failure criteria, the safety of the adjacent linings
CE
Altaee and Fellenius [30] and Gibson [31] proposed a scaling law for small-scale
testing under the 1-g condition in sand. They noted that the chief condition for
agreement between the model and prototype is that their initial soil states must have
4
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
equal proximity to the steady-state line (SSL), as shown in Fig. 2. When the stresses are
normalized to the initial mean stress, the model can be expected to behave similarly in
all aspects to the prototype. The scaling relations in terms of the geometric scaling ratio
n and stress scaling ratio N (ratios of the prototype/model) are presented in Table 1. Due
to the difference in the void ratios between the prototype and model, the densities at the
PT
homologous points in the prototype and model are slightly different. Therefore, the
prototype/model ratio for the geostatic stress at the homologous points does not equal
RI
the geometric scaling ratio n. However, the density difference between the model and
SC
prototype is small and can be ignored [31,32], making the stress-scaling ratio N equal to
the geometric scaling ratio n. The corresponding scaling relations under this condition
NU
are also listed in Table 1. The geometric scaling ratio n was set at 20 for all the models
MA
in the tests.
D
Line of similar
PT
normalized constitutive
Δem behavior
Void ratio e
CE
Model
Δep
Prototype
Δem = Δep
AC
5
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
PT
Displacement n n 20
Force Nn2 n3 8,000
Moment Nn3 n4 160,000
RI
em = ep + em = ep - 0.0389 ×
Void ratio em = ep + λln(1/n)
λln(1/N) ln(1/20)
SC
Young’s modulus of
N n 20
segments
Flexural stiffness of
Nn4 n5 3,200,000
segments
NU
Compressional
Nn2 N3 8,000
stiffness of segments
Time n1/2 n1/2 4.47
MA
To extrapolate the tunnel behavior in the 1-g model tests to the field conditions, the
E
scaling laws for both the segment and the joint of the tunnel should be deduced. The
PT
scaling rules for the segment can be easily determined according to the π theorem
CE
[34,35] and the dimensional analysis method, as listed in Table 1. The scaling rules for
Based on the similarity analysis methods of the joint recommended by Ye et al. [36],
a new similarity method was proposed by considering the profile and the deformation
deformation Δlb when the tunnel lining is subjected to an external load. Therefore, the
6
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
relative rotation angle of the two adjacent segments under small strain conditions can be
expressed as follows:
PT
where Eb is the Young’s modulus of the bolt, Ab is the cross-sectional area of the bolt, lb
RI
is the length of the bolt, and n and n' are the numbers of bolts in the circumferential joint
SC
in the prototype and the model, respectively.
NU
σ0(ε0)
MA
t d x
θ
D
Tensile member
E
By substituting Eq. (1) into Eq. (2), the following equation can be derived:
CE
Fb lb
(3)
n Eb Ab d x
AC
Because the similarity ratio of the rotation angle Cθ is equal to 1, the relationship can
be obtained as follows:
Fb lb Fb lb
nE A d x nE A d x (4)
b b
p b b m
According to the relationships of the similarity ratios, the following equations exist:
7
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
F
CF CE CL
b p 2
(5)
F b m
s
d x
CL
p
(6)
d x m
where CL is the geometric similarity ratio, CEs is the similarity ratio of the Young’s
PT
Therefore, Eq. (4) can be converted to Eq. (7) as follows:
RI
n Eb Ab n Eb Ab
l CE CL (7)
p lb m
s
SC
b
Similarly, the similarity relation for the longitudinal joints can also be expressed in
NU
the form of Eq. (7). The material and dimension of the model bolts will be determined
according to Eq. (7) and will be discussed in detail in the next section.
MA
This model tunnel is designed to simulate the universal subway tunnel in China [38],
E
which is typically composed of C50 reinforced concrete (RC) segments and M30 steel
PT
The materials commonly used to simulate the shield tunnel segment include
and aluminum alloys [39]. In this study, PMMA, polyoxymethylene (POM) and
polypropylene (PP) were selected as the candidate materials for the model segment, and
their tensile strengths and elastic moduli were derived from tensile tests, as listed in
Table 2.
8
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
PT
Aluminum wire (type 1060) 44.00 31.68
RI
As listed in Table 1, the similarity constant of the Young’s modulus CEs should be set
SC
at 20. The Young’s modulus of the concrete in the prototype is 34.50 GPa, and the
The material of the bolt model must be suitable for secondary processing to ensure
E
PT
that the bolt can satisfy the scaling law and meet the requirements of the segment
assembly. Therefore, aluminum welding wire (type HS301), steel welding wire (type
CE
JL-50), lead wire and aluminum wire (type 1060) were selected as candidate materials.
E A
Based on CEs = 20.4 and CL = 20, the value of b b can be calculated using Eq.
lb m
(7). Then, the optimal sizes of the circumferential joints constructed from those
materials can be determined, as listed in Table 3. Note that the size of the bolt
constructed from aluminum wire (type 1060) is relatively small and matches the size of
9
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
convenient. Therefore, this type of wire was selected for the circumferential joint in the
model test; it was also selected for the longitudinal joints using the sizes calculated from
Eq. (7). The mechanical and geometric properties of the tunnel model are provided in
PT
Table 3 Optimal sizes of the bolts constructed from four types of materials to produce
RI
Eb Ab Young’s Diameter
modulus E
SC
Length of
Material b m
l b of bolt db
bolt lb (m)
(N/m) (GPa) (m)
Aluminum welding wire (type
NU
3.711e8 35.60 0.078 0.0016
HS301)
Steel welding wire (type JL-50) 3.711e8 103.51 0.128 0.0012
Lead wire 3.711e8 14.95 0.116 0.0030
MA
Young’s Young’s
Poisson’s ratio Poisson’s ratio
E
Lining Lining
Lining Ring Length Diameter
outer inner Number of
Type thickness width of bolt of bolt
diameter diameter bolts
(m) (m) (m) (m)
(m) (m)
Prototype 6.200 5.500 0.350 1.200 0.400 0.030 12/16
0.026 0.001
Model 0.310 0.275 0.0175 0.060 12/8
/0.033 /0.0016
Note: The first and second values of the bolt properties represent the circumferential
10
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
apply the same structure and assembly method used in the prototype tunnel to the model
tunnel, due to its small size. Therefore, certain adjustments of the tunnel structure were
made in the model tunnel without affecting its structural stress or deformation
PT
characteristics, as follows. (1) For more convenient installation and formation of the
entire ring, straight bolts were used in the circumferential joints, although curved bolts
RI
are often used in prototype segments, as indicated in Fig. 4 (b). However, for the
SC
longitudinal joints, curved bolts were used to avoid using the slots, as indicated in Fig. 5
(a), and pre-tightening forces (25 N, corresponding to 100 kN in the prototype) [40]
NU
were applied to the longitudinal joint bolts; (2) for more convenient installation of the
MA
bolts for the circumferential joint, access holes were placed in the outside of the
segment without changing the bolt location; and (3) for simplicity, the tunnel segment
D
grooves and gaskets at the joints were omitted in the model tunnel. Fig. 4 and Fig. 5
E
illustrate the structure and the distribution of the circumferential and longitudinal bolts
PT
in the model segmental ring, respectively, and Fig. 6 illustrates the distribution of the
hand holes and the bolt holes in the model tunnel. After manufacturing the segments
CE
and the bolts, the tunnel model with stagger-jointed rings could be assembled, as shown
AC
in Fig. 7 (b).
11
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
PT
RI
(a)
SC
NU
MA
(b)
Fig. 4 Segment partitioning scheme and circumferential bolt distribution (unit: mm): (a)
E D
segment partitioning scheme and (b) circumferential bolt distribution and detail of a
PT
circumferential bolt
CE
AC
(a)
12
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
PT
(b)
Fig. 5 Illustration of the longitudinal bolt distribution (unit: mm): (a) distribution and
RI
detail of the longitudinal joint bolt and (b) longitudinal joint bolt
Access hole
SC
The longitudinal joint bolt hole
NU
MA
segment models: (a) top segment, (b) adjacent segment and (c) standard segment
PT
sandy soil, 4 sets of model tests with different cover depths of the tunnel were
conducted.
A rigid steel tank with inner dimensions of 1.80 m × 1.08 m × 1.90 m (length × width
× height) was designed to conduct the model tests, as illustrated in Fig. 7 (a). Two 20
mm thick Plexiglas windows were installed at the front and back sides of the tank to
13
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
observe the test. Except for the windows, the internal surfaces of the steel tank were
lined with a two-layer Teflon (PTFE) film to reduce the friction between the sand and
the sidewalls of the tank [41]. A circular opening of 180 mm in diameter was drilled in
each of the front and back windows. The size of the opening was smaller than the inner
diameter (275 mm) of the tunnel. A sheet of sponge covered by a thin film was installed
PT
on each end of the tunnel to minimize the interaction of the tunnel and the Plexiglas and
to prevent the sand from leaking through the gap between them.
RI
A sand pluvial device was installed above the soil tank. During the sand dropping
SC
process, the hopper controlled by a servo device could move upward automatically to
maintain a constant height (300 mm in the test) above the sand surface.
NU
As shown in Fig. 7(a), the tunnel was located in the middle of the length of the strong
MA
box. The clear distance between the lateral boundaries (inside walls of the soil tank) and
the tunnel was 2.4 times the tunnel diameter (2.4D and D = 310 mm) to reduce the
D
boundary effect on the tunnel deformation. The thickness of the sand layer below the
E
3.2 Instrumentation
CE
For the tests, the tunnel model was assembled using eighteen rings, as illustrated in
AC
Fig. 7(c). The partial failure was set in the ninth ring, as will be described in detail in
Section 3.4. The bending moments of the springline, invert and crown in the fifth to
ninth rings were monitored by strain gauges. In particular, the monitoring sections of
the eighth ring were equidistantly distributed at every 45° from the crown. For each
monitoring section, two strain gauges were installed on the inside and outside of the
segment.
14
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
Plexiglas face
Plastic membrane
Centric horizontal shaft
Handle
PT
Strain gauge wire
RI
SC
NU
(a)
Partial failure
Load cell location Strain gauge
MA
E D
Ninth ring
PT
(b)
Fig. 7 Model test devices: (a) strong box and (b) model tunnel
AC
In addition to the bending moment, the earth pressures acting on the tunnel were also
monitored. The earth pressures were monitored by earth pressure cells (EPCs), which
had an accuracy of ±0.02% and a diameter of 25 mm. Eight EPCs were mounted on the
segment of the tenth ring, equidistantly distributed at every 45° from the crown. Other
EPCs were mounted on the crowns and springlines of the eleventh to fifteenth rings, as
15
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
illustrated in Fig. 7 (c). The strain and earth pressure measurements were collected by a
To ensure that the EPCs could measure the correct pressures, they were tested using
hydrostatic pressure loading, and the measured data were recorded and compared with
the expected pressure values [42]. The results presented a pressure that linearly
PT
increased with the water depth. At a depth of 1.2 m below the water surface, the
hydrostatic pressure was measured to be 11.6 kPa, which was consistent with the
RI
expected value.
shown in Fig. 8. The mean grain size, D50, of the sand was 0.23 mm, and the coefficient
of nonuniformity, Cu, was 2.25. Other parameters of the sand are listed in Table 6.
D
A series of consolidation undrained (CU) triaxial tests were performed to estimate the
E
steady-state behavior of the sand [31]. Sand specimens with different initial relative
PT
densities were tested. The SSL determined from the CU tests is illustrated in Fig. 9, and
the slope of the SSL λ in the e - lnp plane is -0.0389. For the falling height of 300 mm
CE
used in this study, the void ratio of the sand is approximately 0.67, and the
AC
corresponding density is approximately 1,550 kg/m3. The critical state friction angle of
the sand measured using the direct shear test was 30.96°.
Based on the scaling relation em = ep + λln(1/n), the void ratio in the prototype is 0.55,
which lies within the range defined by the maximum and minimum void ratios (0.85
and 0.43, respectively) of the sand, indicating that the void ratios are reasonable.
16
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
PT
100
80
RI
Percent passing
60
SC
40
20
NU
0
MA
0.9
E
0.8
PT
0.7
0.6
CE
Steady-State Line
Void ratio
0.5
y = -0.0389 ln(x) + 0.8490
0.4
AC
0.3
0.2
0.1
0
1 10 100 1000
Mean effective stress p' (kPa)
Fig. 9 SSL of the sand
17
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
To simulate the partial failure of the tunnel structure, the tunnel ring remains intact
during the sand raining stage and fails (thus initiating partial failure) when it receives
the control signal. Such a ring is called a partial failure ring (PFR) in this paper, and its
working mechanism is shown in Fig. 10. The segment at the springline of the ninth ring
PT
(PFR) was pulled into the tunnel to create an opening of the tunnel and initiate partial
failure, which is similar to the opening of the connecting passage in the Shanghai Line 4
RI
accident [3]. This initial failed segment (IFS) was installed on the interlock, which
SC
could be controlled by the handle. NU
MA
(b)
Fig. 10 Device and working mechanism of the PFR: (a) device and (b) working
mechanism
The cover depths for Tests 1-4 were 1D, 2D, 3D and 4D (D = 310 mm), respectively.
18
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
(1) From the tank base up to the level of the tunnel invert, the sand was prepared
using the pluvial deposition method, and for each 200 mm thick layer, the sand was
tamped on the surface using a steel plate to achieve a relatively dense condition.
(2) The model tunnel was placed at the pre-set position. Then, the centric horizontal
shaft was installed, and the IFS was connected to the interlock.
PT
(3) From the level of the tunnel invert to the tunnel crown, the sand was rained with a
constant falling height of 300 mm. In this stage, the monitoring devices were switched
RI
on to record the earth pressure and the strain.
SC
(4) Above the tunnel crown, the sand was continuously dropped with a constant
falling height of 300 mm to a level of 1.0D above the tunnel crown. After this layer was
NU
finished, the model was left to stand until the sensor readings reached a stable state.
MA
(5) The fourth step was repeated layer by layer until the sand surface reached the
(6) When the sensor readings reached a stable state, the handle was turned rapidly to
E
initiate the partial collapse of the tunnel. After the sensor readings reached a stable state
PT
To easily understand the model test results and associate them with practical
engineering implications, the results discussed in the following sections are all
Fig. 11 shows the bending moments in the eighth ring at different cover depths for
Test 1 to Test 4. When the cover depth is 1D, the maximum bending moments were
19
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
72.96 kN•m, 74.88 kN•m, 72.8 kN•m and 74.4 kN•m for Test 1 to Test 4, respectively.
These results were similar, which indicates that the repeatability of this series of tests
was satisfactory.
Fig. 12 shows the bending moments in the third to eighth rings in Test 4. The
difference between the bending moments in the same location of different rings was
PT
minimal, indicating that the boundary effect in the model tests was relatively slight.
120
0° CR
RI
Cover depth of 1.0D +M
θ
90° SL
SC
80
180° IN
Moment M (kN· m)
NU
40
0
MA
-80
(a)
PT
200
0° CR
Cover depth of 1.0D
+M
Cover depth of 2.0D
CE
150 θ
90° SL
100
180° IN
AC Moment M (kN· m)
50
0
0 45 90 135 180 225 270 315 360
Angle from
-50 crown θ (°)
-100
-150
(b)
20
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
250
0° CR
Cover depth of 1.0D
+M
200 Cover depth of 2.0D
Cover depth of 3.0D θ
90° SL
150
100 180° IN
Moment M (kN· m)
50
0
0 45 90 135 180 225 270 315 360
PT
-50 Angle from
crown θ (°)
-100
RI
-150
SC
-200
(c)
350
NU
Cover depth of 1.0D 0° CR
300 +M
Cover depth of 2.0D
250 Cover depth of 3.0D θ
90° SL
200 Cover depth of 4.0D
MA
150
180° IN
Moment M (kN· m)
100
50
D
0
0 45 90 135 180 225 270 315 360
-50
E
Angle from
-100 crown θ (°)
PT
-150
-200
CE
-250
-300
(d)
AC
Fig. 11 Bending moments in the eighth ring for different tests: (a) Test 1, (b) Test 2, (c)
21
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
350
Monitoring location
CR PFR
Cover depth of 1.0D
300 Cover depth of 2.0D +M
Cover depth of 3.0D
# # # # # # # # # #
Cover depth of 4.0D 3 4 5 6 7 8 9 10 11 12
250
Ring number
Moment M (kN· m)
200
150
PT
100
RI
50
SC
0
3 4 5 6 7 8 9
Ring number
NU
(a)
0
3 4 5 6 7 8 9
Ring number
-50
MA
-100
-150
Moment M (kN· m)
-200
E
-250
PT
PFR
-300
Cover depth of 1.0D +M
CE
(b)
22
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
300
PFR
Cover depth of 1.0D +M
250 Cover depth of 2.0D
# # # # # # # # # #
Cover depth of 3.0D 3 4 5 6 7 8 9 10 11 12
Cover depth of 4.0D IN
Monitoring location Ring number
200
Moment M (kN· m)
150
PT
100
RI
50
SC
0
3 4 5 6 7 8 9
Ring number
NU
(c)
0
3 4 5 6 7 8 9
Ring number
-50
MA
-100
-150
Moment M (kN· m)
-200
E
-250
PT
PFR
-300 Cover depth of 1.0D +M
Cover depth of 2.0D
CE
L-SL
-350 Cover depth of 3.0D # # # # # # # # # #
3 4 5 6 7 8 9 10 11 12
Cover depth of 4.0D
Monitoring location Ring number
-400
AC
(d)
Fig. 12 Bending moments in different rings for Test 4: (a) on the crown, (b) at the right
The earth pressures recorded by the EPCs at the crown and the springline are depicted
in Fig. 13. The pressure acting on the crown is larger than that at the springline, which
23
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
can also be observed from Fig. 14, in which the earth pressures around the
circumference of the tenth ring are presented. The earth pressures were not exactly
symmetrical with respect to the vertical axis of the tunnel section due to the
asymmetrical arrangement of the segments. The earth pressures in the other tests were
PT
320
EPC Monitoring location
Cover depth of 1.0D
280 Cover depth of 2.0D
RI
Cover depth of 3.0D
# # # # # # # # # #
Cover depth of 4.0D 6 7 8 9 10 11121314 15
240
PFR Ring number
SC
200
Pressure (kPa)
160
NU
120
MA
80
40
0
D
8 9 10 11 12 13 14 15 16
Ring number
E
(a)
240
PT
# # # # # # # # # #
Cover depth of 4.0D 6 7 8 9 10 11121314 15
120
80
40
0
8 9 10 11 12 13 14 15 16
Ring number
(b)
24
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
Fig. 13 Earth pressures acting on the tunnel crowns and at the springline for different
rings in Test 4: (a) on the tunnel crowns and (b) at the springlines.
PT
RI
SC
NU
MA
Fig. 14 Earth pressures acting at different locations around the 10th ring in Test 4.
E D
In Tests 1–4, after the failure of the PFR, the adjacent rings remained intact because
the strength of the tunnel in the model tests was large, as it was not scaled based on the
AC
similarity laws. Test 4 was adopted as a typical case to discuss the general process of
the collapse, as shown in Fig. 15; the collapse processes in the other tests were similar
to that in Test 4. When the PFR failed, the collapse developed quickly within a very
In the horizontal cutting plane of the model at the elevation of the center of the IFS,
the horizontal movements of the IFS were similar to those of the trapdoor test, as shown
25
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
in Fig. 15 (b). In the conventional trapdoor test, the trapdoor usually moves downwards
in the vertical direction, whereas the IFS served as a trapdoor that moved in the
horizontal direction. Therefore, the movements of the IFS caused a horizontal stress
arching in the soil around the initial failure area. However, when the IFS became
entirely detached from the surrounding segments, a large amount of sand began to slide
PT
into the tunnel quickly, as shown in Fig. 15 (c). The inrush of the sand lasted for more
RI
SC
NU
MA
(a) (b)
E D
PT
CE
(c) (d)
AC
(e) (f)
Fig. 15 Inside view of the tunnel before and after partial collapse (the time is in the
prototype scale): (a) moment of partial collapse, (b) 0.18 s after partial collapse, (c) 0.54
26
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
s after partial collapse, (d) 0.89 s after partial collapse, (e) 1.34 s after partial collapse,
PT
Fig. 16 (a) shows the changing curves of the earth pressures acting on the monitoring
rings. After the failure, the earth pressures acting on the IFS, i.e., the springline of the
RI
FPR (#5 EPC), decreased almost to zero instantaneously because of the sudden inward
SC
movement of the IFS, which is similar to the earth pressure acting on a trapdoor. The
earth pressures acting on the crown of the FPR (#1 EPC) also decreased significantly
NU
when the sand sliding evolved to this point; however, its decrease lagged behind that of
MA
the #5 EPC.
The earth pressures acting on the springline of the adjacent ring (#6 EPC) increased
D
rapidly for a short period of time due to the horizontal arching effect discussed in
E
Section 5.1. However, when the sand surrounding the IFS began to slide into the tunnel
PT
after the IFS was pulled into the tunnel, the earth pressure on the #6 EPC sharply
decreased. The springing of the stress arching then moved from the 10th ring to the 11th
CE
ring, causing the earth pressure of the #7 EPC to increase slightly. However, the stress
AC
arching became considerably weaker. Simultaneously, the earth pressures acting on the
crown of the 10th rings (#2 EPC) began to decrease as the sand at this area began to
slide. The decrease of the earth pressure of the #2 EPC was much smaller than that of
the #1 EPC. The earth pressures acting on the crown of the 11th rings (#3 EPC) changed
only minimally, indicating that it was not influenced by the sand sliding. After 2.68 s,
27
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
Based on above analysis, in the horizontal cutting plane, the width of the trapdoor in
this test was B, i.e., the width of the ring, and the influenced range of the trapdoor was
approximately 2B on each side, i.e., the width of two rings. This range of influence is
consistent with the results (approximately 2–3B) derived in the conventional trapdoor
tests conducted by other researchers [43,44]. However, because the sand slid rapidly
PT
into the tunnel after the IFS was entirely pulled into the tunnel, the arching effect in this
test existed for only a short period of time, and the changes of the earth pressures were
RI
primarily affected by the sand sliding surrounding the IFS.
2.0
SC
NU
IFS EPC
#1 EPC #5 EPC #1 #2 #3 #4 CR
#2 EPC #6 EPC #5 #6 #7 #8
SP
#3 EPC #7 EPC
1.5
MA
#4 EPC #8 EPC
Increase ratio of earth pressure
# # # # # #
7 9 11 13 15 16
PFR Ring number
The moment that the sand began to slide into the tunnel
1.0
E D
0.5
PT
CE
0.0
Partial collapse time
0 1 2 3 4 5
AC
Time (s)
(a)
4d
28
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
3.0
Monitoring location IFS
R9-CR CR
R8-CR R8-SP
Increase ratio of bending moment 2.5 R7-CR R7-SP SP
# # # # #
R5-CR R5-SP 3 4 5 6 7 8# 9#1011
# # #
12
1.5
PT
RI
1.0
SC
0.5
0 1 2 3 4 5 6
Time (s)
NU
(b)
4d
Fig. 16 Curves of the soil pressures acting on different rings and of the moments in
different rings in Test 4: (a) soil pressures and (b) bending moments
MA
The variation curves of the increase ratios of the bending moments, which are equal
E
to the ratio of the bending moments after collapse over the moments before partial
PT
failure, are illustrated in Fig. 16 (b). Because the PFR was not intact after partial failure,
CE
the upper part of this ring was similar to a cantilever beam. Before the partial failure, the
outer surface of its crown sustained compression, which transformed to tension after the
AC
partial failure. The bending moment in the crown of the PFR ultimately increased by a
factor of 1.95.
As discussed in the last section, in a short period of time after the partial failure, the
earth pressure acting on the springline of the adjacent ring increased slightly due to the
horizontal arching effect, which alleviated the convergence deformation of this ring and
caused the bending moment in the springline of the 8th ring (R8-SP) to decrease slightly,
29
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
as shown in Fig. 16 (b). Afterwards, the earth pressures acting on the crown and
springing of the adjacent ring decreased to different levels within approximately 0.89 s
after the partial failure. However, the decrease of the earth pressure at the crown was
much smaller than that at the springline. Therefore, the convergence deformation of the
ring became larger, and simultaneously the bending moments in this ring became larger,
PT
as shown in Fig. 16 (b), where the bending moments in the crown and the springline in
RI
The peak increase ratio of the bending moments in the 8th ring was 1.72. As the
SC
distance between the ring and partial failure became larger, the increase ratio became
smaller. For the rings that are more distant than 3 ring widths, the bending moments
NU
were almost unchanged.
MA
As shown in Fig. 17 and Fig. 18, the changes of the earth pressures and bending
E
moments in Tests 1–3 were similar to those in Test 4. After the failure, the earth
PT
pressures acting on the #5 EPC decreased to almost zero instantaneously. The earth
pressures acting on the adjacent #6 EPC first increased due to the horizontal arching
CE
effect and then decreased significantly. The maximum increase ratios of the earth
AC
pressure of the #6 EPC in Tests 1–4 were 1.02, 1.06, 1.13 and 1.32, respectively,
indicating that the horizontal arching effect increased when the cover depth was larger.
The earth pressures of the #1 EPC and #2 EPC also decreased, but the decreases were
In Tests 1–3, the bending moment in the springline of the 8th ring (the first adjacent
ring) decreased slightly first due to the horizontal arching effect and then increased
30
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
significantly due to the redistribution of the earth pressures caused by the sand inrush,
and the bending moment in the crown of the 8th ring also increased sharply after the
partial collapse. The changes of the bending moments in Tests 1–3 were similar to that
in Test 4.
However, as shown in Fig. 19, which plots the increase ratios of the bending
PT
moments in different locations in the 4 tests, when the initial cover depth was larger, the
peak increase ratio was larger, but the influenced extent was negligible. This finding
RI
indicates that a progressive collapse might occur more easily in a larger cover depth
SC
condition. NU
2.0
IFS EPC
#1 EPC #5 EPC #1 #2 #3 #4 CR
MA
#2 EPC #6 EPC
#5 #6 #7 #8
#3 EPC #7 EPC SP
1.5 #4 EPC #8 EPC
Increase ratio of earth pressure
# # # # # #
7 9 11 13 15 16
PFR Ring number
D
1.0
E
PT
0.5
CE
0.0
Partial collapse time
AC
0 1 2 3 4 5
Time (s)
(a)
1d
31
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
2.0
IFS EPC
#1 EPC #5 EPC #1 #2 #3 #4 CR
#2 EPC #6 EPC
#5 #6 #7 #8
#3 EPC #7 EPC SP
Increase ratio of earth pressure 1.5 #4 EPC #8 EPC
# # # # # #
7 9 11 13 15 16
PFR Ring number
1.0
PT
0.5
RI
0.0
SC
Partial collapse time
0 1 2 3 4 5
Time (s)
NU
(b)
2.0 2d
IFS EPC
#1 EPC #5 EPC #1 #2 #3 #4 CR
#2 EPC #6 EPC
MA
#5 #6 #7 #8
#3 EPC #7 EPC SP
1.5 #4 EPC #8 EPC
Increase ratio of earth pressure
# # # # # #
7 9 11 13 15 16
PFR Ring number
D
1.0
E
PT
0.5
CE
0.0
Partial collapse time
AC
0 1 2 3 4 5
Time (s)
(c)
Fig. 17 Curves of the soil pressures acting on 3d
adjacent rings in Tests 1–3: (a) Test 1, (b)
32
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
2.25
Monitoring location IFS
CR
R9-CR
2.00 R8-CR R8-SP
SP
Increase ratio of bending moment R7-CR R7-SP # # # # #
3 4 5 6 7 8# 9#1011
# # #
12
R5-CR R5-SP
PFR Ring number
1.75
Partial collapse time
1.50
PT
1.25
RI
1.00
SC
0.75
0 1 2 3 4 5 6
Time (s)
NU
(a)
3.0 1d
Monitoring location IFS
R9-CR CR
R8-CR R8-SP
MA
# # # # # # # # # #
3 4 5 6 7 8 9 1011 12
R5-CR R5-SP
PFR Ring number
2.0
Partial collapse time
E D
1.5
PT
1.0
CE
0.5
AC
0 1 2 3 4 5 6
Time (s)
(b)
2d
33
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
3.0
Monitoring location IFS
R9-CR CR
R8-CR R8-SP
R7-CR R7-SP SP
Increase ratio of bending moment 2.5 # # # # #
3 4 5 6 7 8# 9#1011
# # #
12
R5-CR R5-SP
PFR Ring number
1.5
PT
RI
1.0
SC
0.5
0 1 2 3 4 5 6
Time (s)
NU
(c)
3drings in Tests 1–3: (a) Test 1, (b) Test 2,
Fig. 18 Curves of the moments of the adjacent
2.25
R9-CR R8-CR R7-CR R5-CR
D
Peak increase ratio of the bending moment
1.75
CE
1.50
1.25
AC
1.00
0.75
1 2 3 4
Cover depth (D)
Fig. 19 Peak increase ratios of the bending moments in these 4 tests
34
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
In the model tests, the bending moments in only a few sections of the tunnel linings
were monitored, which cannot address the overall characteristics of the internal forces
in the segments and the joints. Therefore, it is difficult to determine whether the tunnel
PT
lining failed when partial collapse occurred. In addition, the joints are the most
damageable parts in the tunnel lining; it is at their locations where the tunnel collapse
RI
usually commences [45]). To obtain all the internal forces in the tunnel linings, the
SC
tunnels in the model tests were simulated by a numerical model using FLAC 3D [46].
NU
6.1.1 Numerical tunnel model
MA
calculate the internal forces in the segmental lining using the load-structure method, part
D
of which is shown in Fig. 20 (a). The geometric and mechanical parameters of the
E
numerical tunnel were the same as those in the model tests, i.e., on the model scale.
PT
Both the circumferential and longitudinal joints were considered in the 3D model. The
tunnel segments were simulated using a linear elastic linear element based on the
CE
classical Kirchhoff plate theory [46]. This type of lining element provides two links on
AC
each node, one of which can connect the segments into a tunnel [47-49], as shown in
Fig. 20 (b). These links have six degrees of freedom, including three translational
components and three rotational components, and they are all simulated by springs. In
this study, three springs in a link were considered, as they govern the deformation of the
joint, i.e., a rotational spring, an axial spring and a radial spring, as shown in Fig. 20 (c),
and their stiffnesses were represented by Kh, KA and KR, respectively [47-49]. The
35
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
stiffnesses of these three springs were determined using the simplified procedures
presented by Thienert and Pulsfort [50], Zeng and He [51] and Do et al. [47-49], as
listed in Table 7.
6.1.2 Earth pressures acting on the tunnel before and after partial collapse
PT
In the numerical simulation, two typical conditions in each model test were calculated:
the condition before the partial collapse and the (most dangerous) condition after the
RI
partial collapse. The pressures acting on the tunnel under these two conditions were
SC
determined and simplified based on the model test results. Taking Test 4 as an example,
according to Fig. 16 (b), the bending moments in the adjacent rings were at their
NU
maximum values 5.58 s after the partial failure, which represents the most dangerous
MA
situation of these rings in the model test. Therefore, the pressures 5.58 s after the partial
failure were used in the numerical simulation. Fig. 21 shows the earth pressures used in
D
L2
F
AC
B3 L1
B2 B1
(a)
36
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
PT
(b)
Segment A KR KA Segment B
RI
Kθ
SC
(c)
Fig. 20 Schematic diagram of the tunnel and the joint: (a) 3D numerical tunnel model,
NU
(b) links between the segments and (c) springs in a link
MA
Fig. 21 Earth pressures in prototype scale used in the numerical simulation of Test 4
37
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
Similar to the results discussed in the model tests, the results derived in the numerical
simulation were all converted to the prototype scale. As discussed in Section 5.2.2, the
peak increase ratio of the bending moment in the adjacent ring (the 8th or 10th ring) was
PT
largest. Therefore, we focus on the variation characteristics of the internal forces in the
8th ring. The internal forces, including the bending moments and axial forces, in the 8th
RI
ring before and after the partial failure obtained from the 3D numerical results are
SC
presented in Fig. 22.
As shown in Fig. 22 (a), several peak bending moment points in the segmental ring
NU
were located near the crown, springlines and invert. Additionally, the values calculated
MA
in the numerical simulation are close to those derived in the model test, indicating that
both the results of the model test and the numerical simulation were relatively reliable.
D
After the partial failure of the PFR, the bending moments in the 8th ring increased,
E
especially at the crown, springlines and invert, as well as the joint positions. The axial
PT
forces in the 8th ring also changed, as shown in Fig. 22 (b). In the circumferential
direction, the tunnel lining sustained both bending and compression loads. The
CE
combinations of internal forces in the 8th ring at the key positions (including the crown,
AC
springlines, invert and certain joints) before and after the partial collapse are listed in
38
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
PT
RI
SC
NU
(a)
MA
E D
PT
CE
AC
(b)
Fig. 22 Internal forces of the 8th ring before the partial failure and at 5.58 s after the
partial failure in numerical model 4: (a) bending moments and (b) axial forces
39
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
Table 8 Combinations of internal forces in the 8th ring at key positions in numerical
model 4
PT
6.2.2 Failure criteria of the segment and the joint
RI
Assume that the strength design of the prototype tunnel in this study was the same as
SC
that of a real project in Tianjin [38], in which the tunnel segment is in the form of
NU
symmetrical reinforcement. The ultimate strength envelope curve of the segment can be
calculated according to the Code for Design of Concrete Structures (GB 50010-2010)
MA
[52]. The ultimate strength envelope curve of the circumferential joint was derived
based on the mechanical model of the joint, as shown in Fig. 23, and the derivation
D
procedure will be described in detail below. For the joints under negative and positive
E
fck
h0
CE
h0 x
σ0(ε0)
t d x
d
N
AC
θ t
M Bolt T
Tensile member
The failure modes in the joint area include the crushing of the concrete and the failure
of the bolt. Some assumptions are made to simplify the derivation of the failure criteria:
40
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
(a) the deformation complies with the plane section assumption, and the stress is equal
to zero in the open part of the joint; and (b) the influence depth of the compression
According to the balance of force in the joint area and the assumptions, the following
PT
equations can be derived:
x h0
N T Ec c b (8)
RI
2
SC
h x h0 x 2h h
M Ec c b h0 0 N T d (9)
2 2 NU3 2
2 Eb A cu
T T0 d x (10)
L
MA
c x h0
(11)
cu x
where Ec is the Young’s modulus of the concrete, Eb is the Young’s modulus of the bolt,
D
T0 is the pre-tightening force of the bolt, T is the tensile force of the bolt, εc is the yield
E
PT
strain of the concrete, εcu is the ultimate strain of the concrete, b is the width of the
segment, h is the thickness of the segment, h0 is the yield height of the segment, L is the
CE
M G E 2 N 2 2 G E F H E 0.5 h N G F 2 H F K (12)
2 L cu c
E (13)
4 Eb A cu2 f ck b L 2 cu c
2 cu c L T0 2 Eb A cu d
F (14)
4 Eb A cu2 f ck b L 2 cu c
41
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
f ck b c 3 cu 2 c
G 0.5 f ck b (15)
6 cu c
2
2 Eb A cu2 d
H= (16)
L cu c
2 Eb A cu d 2
K T0 d (17)
L
PT
(2) Failure of the bolt:
RI
According to the balance of force in the joint area and the assumptions, the following
SC
x h0
N T f ck b (18)
2
NU
h x h0 x 2h0 h
M f ck b h0 0 N T d (19)
2 2 3 2
MA
T f s ,u A (20)
D
s ,u
A Eb T0 L x d x
E
(21)
c A E1 x h0
PT
where fck is the compressive strength value of the concrete, fs,u is the ultimate tensile
CE
strength of the bolt, and εs,u is the ultimate tensile strain of the bolt.
E
s ,u
A Eb T0 L
(22)
c A E1
2 N T
F (23)
f ck b
d 2 E 2 E d 4 E F
2
x (24)
2
42
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
By substituting the axial force N into Eq. (23) and substituting Eq. (22) and Eq. (23)
into Eq. (24), the equation of the envelope curve can be derived:
M f ck b F 2 F x x 2 N T d
1 h
(25)
6 2
6.2.3 Failure analysis of the tunnel after the partial collapse in Test 4
PT
The safety of the adjacent tunnel lining after the partial collapse can be evaluated
RI
based on the failure criteria and the combinations of the internal forces in the tunnel
SC
lining. If the internal force combinations of the segments and joints all fall within their
ultimate strength envelope curves, the tunnel lining is safe. Otherwise, the tunnel lining
NU
adjacent to the partial collapse will fail, which indicates that progressive collapse occurs.
The ultimate strength envelope curves and the internal force combinations in the 8th
MA
ring before the partial failure and at 5.58 s after the partial failure are illustrated in Fig.
24.
D
Before the partial failure, the internal forces of both the segments and the joints were
E
PT
within the corresponding envelope curves. After the partial failure, the internal forces
moved from the beginnings to the ends of the red arrows. For the segments, the internal
CE
force combination was inside the envelope curve, indicating that the segment was safe.
However, the internal force combination in joint #3 was beyond the ultimate strength,
AC
which indicated that joint #3 would fail. Progressive collapse under the condition of
partial failure would occur in this situation. However, in Test 4, because the strength of
the joint was not scaled based on the scaling law and was larger than that corresponding
to the prototype tunnel, the adjacent tunnel linings remained intact after the partial
collapse. Therefore, the tunnel model needs to be improved to capture the progressive
collapse in the longitudinal direction via scaling the tunnel strength precisely.
43
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
1200 Internal force at the #3 joint (under the negative bending moment)
Internal force at the #4 joint (under the positive bending moment)
1000
PT
800
600
RI
400
Changing path of the combined
SC
200 internal forces
0
0 1000 2000 3000 4000 5000 6000
NU
Axial force N (kN)
Fig. 24 Changing paths of the combinations of the internal forces derived in numerical
MA
model 4
To investigate the influence of the cover depth on the change of the internal forces of
PT
the segments and the joints, the pressures acting on the tunnel before and after the
partial collapse in Tests 1, 2 and 3 were also determined based on the model test results,
CE
and the tunnels sustained these pressures were calculated. The internal forces in the 8th
AC
ring before and after the partial failure derived from the numerical simulations of Tests
1, 2 and 3 are presented in Fig. 25. The variation patterns of the internal forces due to
partial failure in Tests 1–3 are similar to that in Test 4. However, the increases of the
internal forces become smaller when the burial depth is smaller, which is consistent
with Fig. 19. According to the changing paths of the internal force combinations in
Tests 1–3 and the failure criteria of the tunnel, the safety of the tunnel was also judged.
44
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
In Test 1 and Test 2, the tunnels were safe after the partial collapse. The internal force
combinations in the 8th ring at the key positions in Test 3 derived in the numerical
simulation are listed in Table 9, and their changing paths are illustrated in Fig. 26. As
shown in Fig. 26, the segment was safe. The internal force combination in joint #3 was
PT
Table 9 Combined internal forces in the 8th ring at the key positions derived in the
RI
numerical simulation in numerical models 1, 2 and 3.
SC
Numerical Internal Right Left Joint
Position Crown Invert Joint #1 Joint #4
model forces springline springline #3
Bending 80.00 -59.20 67.20 -72.00 35.20 -20.80 22.40
Before
NU
moment
the
(kN·m)
partial
Axial force 234.56 552.16 174.16 586.32 623.36 264.8 628.24
failure
(kN)
Model 1
At 5.58 Bending 84.80 -70.40 89.60 -91.20 49.60 -48.00 49.60
MA
s after moment
the (kN·m)
partial Axial force 198.72 396.24 363.12 641.20 399.76 242.64 608.32
failure (kN)
Bending 195.20 -144.00 160.00 -174.40 84.80 -49.60 54.40
Before
moment
D
the
(kN·m)
partial
Axial force 297.76 1063.76 146.56 1147.68 1283.68 407.68 1285.04
E
failure
(kN)
Model 2
At 5.58 Bending 233.60 -184.00 225.60 -236.80 132.80 -115.2 118.40
PT
s after moment
the (kN·m)
partial Axial force 208.96 853.04 746.80 1374.48 792.24 511.12 1391.12
failure (kN)
CE
(kN)
Model 3
At 5.58 Bending 393.60 -308.80 368.00 -398.40 220.80 -187.2 192.00
s after moment
the (kN·m)
partial Axial force 203.68 1354.48 1211.44 2182.24 1204.64 825.20 2269.52
failure (kN)
45
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
PT
RI
SC
NU
(a)
MA
E D
PT
CE
AC
(b)
Fig. 25 Internal forces of the 8th ring before the partial failure and 5.58 s after the
partial failure in numerical models 1, 2, 3 and 4: (a) bending moments and (b) axial
forces
46
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
1200 Internal force at the #3 joint (under the negative bending moment)
Internal force at the #4 joint (under the positive bending moment)
1000
PT
800
600
RI
400
Changing path of the combined
SC
200 internal forces
0
0 1000 2000 3000 4000 5000 6000
NU
Axial force N (kN)
Fig. 26 Changing path of the combined internal forces corresponding to the prototype
MA
7. Conclusions
E
This study carried out a series of model tests investigating the tunnel collapse
PT
mechanism due to the removal of segments at the springline in sandy ground. The
influences of partial failure on the adjacent linings due to the arching effect and the
CE
inrush of the soil were discussed. Via 3D numerical analyses, the distributions of the
AC
bending moments and axial forces in both the segments and joints of the adjacent
linings were obtained. Based on the failure criteria, the safety of the adjacent linings
(1) After the removal of the initially failed segment, the movements of soil around the
failed segment were similar to a horizontal trapdoor, which created a horizontal arching
zone. Afterwards, the soil outside the openings became loosened and slid into the tunnel,
47
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
which caused a sharp increase in the bending moments of the adjacent linings due to the
resistance loss of the soils around the openings. The adverse effect of the partial failure
on the adjacent linings was primarily caused by the inrush of the sand into the tunnel
(2) The affected tunnel length of the horizontal arching zone caused by the movement
PT
of the initially failed segment was approximately 2B (where B is the width of one ring)
on each side, which is almost the same as those in the conventional trapdoor tests. The
RI
affected length of the entire partial failure process was also approximately 2–3B.
SC
(3) The influence of the cover depth on the collapse mechanism was also investigated.
(4) A 3D load-structure model was used to obtain the internal forces in both the
D
segments and bolts of the tunnel. This procedure enabled us to compare the loads
E
sustained by both the segments and bolts to the relevant failure criteria so that the safety
PT
of the tunnel could be evaluated. The results indicated that when C/D is larger than 3.0,
the joints in the adjacent rings may fail, causing further collapse.
CE
This study aimed at providing feasible access to explore the influence of segment
AC
failure on adjacent linings and to evaluate the possibility of further damage to the tunnel.
collapse, more sophisticated model tests are still needed in future studies, such as model
tests considering the scaling of the structure strength in addition to stiffness, which can
simulate actual progressive collapse processes and provide more information on the
48
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
Acknowledgements
This work was supported by the China Postdoctoral Science Foundation (Grant No.
51508382) and the Key Project of the National Natural Science Foundation of China
PT
RI
References
SC
[1] K.M. Lee, X.Y. Hou, X.W. Ge, Y. Tang, An analytical solution for a jointed
shield-driven tunnel lining, Int. J. Numer. Anal. Meth. Geomech. 25 (2001) 365–
NU
390. doi: 10.1002/nag.134.
MA
[2] International Tunnelling Association Working Group No. 2, Guidelines for the
design of shield tunnel lining. Tunn. Undergr. Sp. Technol. 15 (2000) 303–331.
D
recovering a collapsed tunnel in Shanghai metro No.4 line, China Civ. Eng. J. 42
PT
[4] Y. Lei, The repair construction technology for a tunnel accident in a city.
CE
Home and Abroad. China Architecture and Building Press, 2012. (in Chinese)
[6] D.M. Frangopol, J.P. Curley, Effects of damage and redundancy on structural
9445(1987)113:7(1533).
49
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
3828(2004)18:2(79).
PT
[9] U. Starossek, Progressive collapse of structures: nomenclature and procedures,
RI
[10] E. Masoero, F.K. Wittel, H.J. Herrmann, B.M. Chiaia, DEM simulations of the
SC
progressive collapse of framed structures.
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.385.1819&rep=rep1&
NU
type=pdf.
MA
[11] GSA, Progressive Collapse Analysis and Design Guidelines for New Federal
[13] BS8110-1 Structural Use of Concrete, Part 1: Code of Practice for Design and
Construction, 1997.
CE
50
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
9445(1988)114:9(1978).
[18] W.F. Lee, K. Ishihara, Forensic diagnosis of a shield tunnel failure, Eng. Struct.
PT
[19] C. Chen, W.F. Lee, Forensic investigation of a disastrous failure at the arrival
end of a station of the Kaohsiung mass rapid transit system, in: Proceedings of
RI
International Conference on Forensic Engineering, 2007.
SC
[20] W.F. Lee, C.L. Hung, K.J. Kuo, J.I.-W. Chen, S.-M. Woo, Forensic analysis of
the cases of failure at the arrival end of O2 Station, Sino-Geotechnics 105 (2005)
NU
35–46.
MA
[21] Z.X. Zhang, X.Y. Hu, K.D. Scott, A discrete numerical approach for modeling
[22] R.P. Chen, L.J. Tang, D.S. Ling, Y.M. Chen, Y.M. Chen. FACE stability
PT
analysis of shallow shield tunnels in dry sandy ground using the discrete element
10.1016/j.compgeo.2010.11.003.
AC
[23] E. Leca, L. Dormieux, Upper and lower bound solutions for the face stability of
doi: 10.1680/geot.1990.40.4.581.
51
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
[25] G. Mollon, D. Dias, A. Soubra, Face stability analysis of circular tunnels driven
10.1061/(ASCE)GT.1943-5606.0000194.
sensitivity analysis of tunnel face stability, Struct. Saf. 32 (2010) 402–410. doi:
PT
10.1016/j.strusafe.2010.06.001.
[27] J.S. Qin, W. Zhu, J. Chen, Study of dislocation of duct pieces and crack
RI
problems caused by shield attitude control, Construction Technology 33 (2004)
SC
25–27 [in Chinese].
[28] H.H. Mo, J.S. Chen, Study on inner force and dislocation of segments caused by
NU
shield machine attitude, Tunnelling Undergr. Space Technol. 23 (2008) 281-291.
MA
doi: 10.1016/j.tust.2007.06.007.
[29] Z. Wang, L. Wang, L. Li, J. Wang, Failure mechanism of tunnel lining joints
D
[32] R.F. Scott, Centrifuge modeling and technology: a survey, Rev. Franc. Geotech.
48 (1989) 15-34.
[33] P.J. Meymand, Shaking Table Scale Model Tests of Nonlinear Soil-Pile-
52
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
[34] A.A. Sonin, A generalization of the Π-theorem and dimensional analysis, Proc.
[35] L.I. Sedov, Similarity and Dimensional Analysis in Mechanics. Academic Press,
PT
Model test study on effective ratio of segment transverse bending rigidity of
RI
[37] Y. Koyama, Study on the Improvement of Design Method of Segments for
SC
Shield-Driven Tunnels. RTRI Report: Special No. 33, RTRI, 2000, pp. 114 (in
Japanese).
NU
[38] G. Zheng, T. Zhang, Y. Diao, Mechanism and countermeasures of preceding
MA
[39] W. Van Empel, R.G.A. De Waal, C. Van der Veen, Segmental tunnel lining
E
[40] W.Q. Ding, Y.C. Peng, Z.G. Yan, B. Shen, H. Zhu, X. Wei, Full-scale testing
CE
and modeling of the mechanical behavior of shield TBM tunnel joints, Struct.
AC
[42] C. Leung, M.A. Meguid, An experimental study of the effect of local contact
loss on the earth pressure distribution on existing tunnel linings. Tunn. Undergr.
53
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
[43] G.S. Pardo, E. Sáez, Experimental and numerical study of arching soil effect in
10.1016/j.compgeo.2014.01.005.
PT
Tunn. Undergr. Space Technol. 18 (2003) 171-183. doi: 10.1016/S0886-
7798(03)00025-7.
RI
[45] R. Wang, D. Zhang Mechanism of transverse deformation and assessment index
SC
for shield tunnels in soft clay under surface surcharge. Chin. J. Geotech. Eng. 35
115–127.
PT
[48] N.A. Do, D. Dias, P.P. Oreste, I. Djeran-Maigre, 3D modelling for mechanized
tunnelling in soft ground - influence of the constitutive model. Am. J. Appl. Sci.
CE
10 (2013) 863–875
AC
simulation for mechanized tunnelling in soft ground: the influence of the joint
54
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
metro shield tunnel. J. Southwest Jiaotong Univ. 39 (2004) 744-748. (in Chinese)
PT
RI
SC
NU
MA
E D
PT
CE
AC
55
ACCEPTED MANUSCRIPT
Collapse mechanism of shield tunnels
Highlights
The failure criteria for both segments and joints were derived.
The evaluation of the tunnel safety under partial collapse was discussed.
PT
RI
SC
NU
MA
E D
PT
CE
AC
56