Professional Documents
Culture Documents
G. Zheng, Y.W. Lei, X.S. Cheng
G. Zheng, Y.W. Lei, X.S. Cheng
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Experimental study on the progressive collapse mechanism in the braced and tied-
a MOE Key Laboratory of Coast Civil Structure Safety, Tianjin University, Tianjin
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
300072, China
b Department of Civil Engineering, Tianjin University, Tianjin 300072, China
300072, China
E-mail: cheng_xuesong@163.com
E-mail: zhenggang1967@163.com
Abstract
influence of the failure of some retaining structure members on the overall safety
performance of a retaining system has not been studied. Model tests of failures of
retaining piles, struts or anchors were conducted in this study, and the load transfer
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
deformations occurred in certain piles, the increasing ratios of the bending moments in
adjacent piles were much larger in the braced retaining system than in the cantilever
system and more easily triggered progressive failure. When the strut elevation was lower
or the excavation depth was greater, the degree of influence and range of pile failures
became larger. When certain struts/anchors failed, their loads transferred to a few adjacent
of strut or anchor failure on piles were different from those of pile failure. As the number
of failed struts or anchors increases, the bending moments of the piles in the failure zone
first decrease and then increase to very high values. Therefore, the progressive failure
path extends from struts/anchors to piles and will lead to large-scale collapse.
1. Introduction
In practice, braced piles/walls and tied-back piles/walls are very common retaining
In recent years, many excavations using such retaining structures collapsed worldwide,
with most of these collapses being triggered by the partial failure of a retaining structure,
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
i.e., the failure of certain retaining structures within a relatively small and local range of
the excavation. For example, in the Singapore metro collapse accident (Artola 2005; COI.
2005; Corral and Whittle 2010; Davies 2007; Ishihara and Lee 2008; Puzrin et al. 2010;
Whittle and Davies 2006), the failure started at the ninth-level strutting connection.
Yielding of the connection allowed the diaphragm walls to deform, overloading the struts
in the levels above and causing them to buckle. This triggered a 100 m-long collapse of
the excavation. A Hangzhou metro accident (Chen et al. 2013; Li and Li 2010; Tan et al.
2018) began at the fourth-level struts that were not installed in a timely manner, and
finally led to a 70 m-long collapse. Moreover, in addition to the failures of struts and
retaining piles or walls, many collapses of excavations have also been triggered by the
breakage of anchors, such as the accidents in Cologne and Japan (Haack 2009; Itoh et al.
2011).
deformations and nonlinear and multifield coupling, because the failure can propagate
along the depth, width and length directions of the excavation. The traditional stability
analysis methods for excavations (overall stability failure, overturning failure, basal
heaving failure, etc. ( Kempfert and Gebreselassie 2006; Ou 2006; Lu and Tan 2019; Luo
et al. 2012)) are mostly based on the relatively simplified 2D instability failure mode of
an excavation profile (Bjerrum and Eide 2008; Chang 2000; Hsieh et al. 2008; Huang et
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 4 of 78
al. 2008; Osman and Bolton 2006; Ukritchon et al. 2003), which cannot reflect the
physical process of progressive failure in terms of time and space after the partial failure
are not permanent structures, they are usually designed with relatively low safety factors
and use member-based methods rather than system-based approaches. Therefore, the
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
retaining structures may lack integrity and sufficient redundancy, i.e., the system safety
performance of the retaining structures is not high, which may be a reason why
an excavation and found that when the maximum load transfer coefficient (the peak
increasing ratio of the maximum bending moment in the pile after failure over the
corresponding value before failure) is greater than the bearing capacity safety factor of
the piles, partial failure will lead to progressive collapse. However, these conclusions
were based on excavations retained by cantilever piles and few studies have examined
the system safety and progressive collapse of excavations retained by anchored or strutted
piles.
Stille and broms (1976) reported a field study of anchor failures at four sites in Sweden
and found that the maximum change in the anchor load expressed as a percentage of the
initial load of the failed anchor was 35% for walls with two or more rows of anchors.
Using numerical analysis, Goh and Wong (2009) found that the failure of one or two
struts due to an accident will not result in detrimental failure of the entire excavation
system if the struts are of adequate strength. Clause 3.7.4 of Technical Reference 26: 2010
(TR26. 2010) states that the design of deep excavations should accommodate for the
possible failure of any individual strut, tie rod, ground anchor, structural member or
Page 5 of 78
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
connection at each stage of construction. Itoh et al. (2011) designed a centrifuge test of
excavation collapse caused by the breakage of anchor heads. Zheng and Cheng et al.
(2011, 2014 and 2015) indicated that when the redundancy of retaining structures is not
sufficient, partial failure will easily evolve to progressive collapse. Pong et al. (2012)
indicated that one-strut failure (OSF) analysis using 2D plane strain analysis lead to more
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
conservative designs with heavier strut sections and proposed a simplified method using
evaluation index to analyse the structural safety of anchored concrete panels after anchor
ruptures through the calculation of the bending moments and punching forces in the
panels. Choosrithong and Schweiger (2018) and Goh et al. (2018) indicated that the load
from a failed strut may transfer to adjacent struts and therefore cause one or more struts
to fail if the adjacent struts are not of sufficient bearing capacity. Zhao et al. (2018) used
basement excavation in Shanghai has been investigated (Wang et al. 2019; Tan et al.
2020). The above literature review shows that most partial collapse analyses were
The focus of this paper is an experimental study of pile, strut and anchor failures to
investigate how loads from the failed retaining structures are transferred to the adjacent
retaining structures and lead to the collapse of the whole retaining system, i.e., the
progressive collapse mechanism. Three categories of model tests for strut-pile and
anchor-pile excavation were designed, conducted and compared in this paper. The tests
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 6 of 78
include the initial failures of certain retaining piles, struts or anchors. In these tests, the
changes of the earth pressure, pile internal forces, and strut or anchor axial forces caused
by different types of partial failures were monitored and analysed. The influences of
several crucial factors (partial failure type, excavation depth, etc.) on the progressive
To study the influence of partial retaining structure failure on the adjacent intact
retaining structures and the whole retaining system, 7 tests were conducted, including 5
tests (test 1-5) on a braced-pile retaining system and 2 tests (test 6-7) on a tie-back
retaining system. Tests 1-4 and test 5 simulated the failures of certain retaining piles and
struts in a braced-pile retaining system, respectively, and tests 6 and 7 simulated anchor
The soil used in this model test was dry fine sand, which was excavated in Tangshan
district, Hebei Province, China. The sand was the same as that in Cheng’s study (Cheng
et al. 2017). The main parameters of the sand are listed in Table 1. The critical-state
friction angle (Schofield 1968; Simoni and Houlsby 2012; Yang and Luo 2018) of the
sand sample at the mid-height of the model, as measured by the direct shear test, was 31°.
Altaee and Fellenius (1994) and Gibson (1997) 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 equal proximity
to the steady-state line (SSL). The scaling of the constitutive behaviours of the sand and
structures were the same as those in Cheng’s study (Cheng et al. 2017). The scaling
Page 7 of 78
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
relations in terms of the geometric scaling ratio n (ratio of prototype/model) are presented
in Table 2, and the geometric scaling ratio n was set to be 16 for all models in this study.
The information for the 7 model tests, including the depth of excavation, the elevation
of the strut or anchor, and the failure model of the retaining structure, are illustrated in
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
Tests 1-4 were model tests of excavation collapse caused by the failure of 4 retaining
piles (P20-P23), representing the retaining pile failure type. Test 1 is illustrated in Fig. 1
(b). The difference between tests 1 and 2 was the elevation of the strut. The difference
between tests 1 and 3 was the excavation depth. The influences of the strut elevation and
excavation depth on the load transfer during partial failure were studied in tests 1-3.
Furthermore, test 4 simulated large deformations rather than pile breakages (P20-P23) in
preset locations, as shown in Fig. 2, and the other model conditions were the same as
those of test 3. The difference between the influences of large localized deformation and
pile breakage on the progressive collapse mechanism was studied in this test.
excavation, test 3 in Cheng’s paper (Cheng et al. 2017) was introduced as the comparative
test and is labelled test 0 in this paper, as shown in Fig. 1 (a). In test 0, the boundary of
the window was considered to be a symmetrical plane. Two initiating failure piles (IFPs
P1 and P2) were used in test 0, which was equivalent to the four IFPs in tests 1-3. Test 0
was the model test of partial collapse caused by failures of certain cantilever retaining
piles, with the total excavation depth H of 750 mm and the corresponding prototype
excavation depth of 12 m.
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 8 of 78
In test 5, the initiating failure struts (IFSs) (S5-S10) were disabled one by one,
representing the strut failure type, as illustrated in Fig. 1(c). The sequence of strut failure
was S7, S6, S5, S8, S9 and S10. In tests 6 and 7, the initiating failure anchors (IFAs) (A6-
A14) were disabled sequentially, representing the anchor failure type, as illustrated in Fig.
1(d). The sequence of anchor failure was A10, A9, A11, A8, A12, A7, A13, A6 and A14.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
The difference between tests 6 and 7 was the excavation depth. Additionally, the
differences in load transfer between the strut failure and anchor failure of a retaining
For tests 1, 5 and 6, the total excavation depth H was 750 mm, and the corresponding
prototype excavation depth was 12 m. The elevation of the strut or anchor was the ground
surface (+0 mm). For test 2, the total excavation depth was also 750 mm, but the elevation
of the strut was 150 mm (0.2H) below the ground surface (GS). In these tests, when the
excavation depth was shallower than 600 mm, each excavation step was 100 mm. For
deeper excavation depths, the excavation step was 50 mm. For each excavation step, the
sand was removed manually by a steel scraper and the depth marks labeled on the piles
were used for the control of excavation depth in each step. After each excavation step,
when the measured pile head displacement became stable, next excavation step started.
For tests 3, 4 and 7, the total excavation depth was 900 mm (prototype excavation depth:
14.4 m), and the elevation of the strut or anchor was the GS (+0 mm). When the
excavation depth was shallower than 800 mm, each excavation step was 100 mm. For
The inner dimensions of the soil tank were 2.50 m × 2.46 m × 1.40 m (length × width
× height). A schematic model setup and photographic model example are shown in Fig.
Page 9 of 78
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
3 and Fig. 4, respectively. To observe the movements of the soil and retaining structure,
a tempered glass window was installed on one lateral side of the soil tank. A Teflon (PTFE)
film was attached to the internal wall surfaces of the tank (except the window) to reduce
any friction between the soil and the sidewalls. As shown in Fig. 3(a), sand-pouring
As shown in Fig. 4, 39 total contiguous piles were placed along the length of the soil
tank (2.46 m) to serve as the retaining structure, including 4 failure-initiating piles (IFPs),
9 monitoring IPs (MPs) and 26 normal IPs. In addition, the average pile spacing was 63
mm. Each model pile was composed of a rectangular polyvinyl chloride (PVC) tube with
of each model pile was 1.20 m. The long side of the pile cross section was perpendicular
to the cross section of excavation. Through the loading test of the simply supported model
pile, the flexural stiffness EI (where E is elastic modulus and I is moment of inertia) of
Because the retaining pile is mainly a bending member, the geometric scaling ratio n
was determined based on the equivalent flexural stiffness EI principle. In this study, the
prototype pile was set to be the C30 cast-in-place pile with a diameter of 0.8 m and a
length of 19.2 m (Cheng et al. 2017), which is commonly used in practice. The flexural
stiffness EI of the prototype pile was 603 MN·m2, which is approximately 165 times of
the flexural stiffness of the model pile. Therefore, based on the scaling relations in Table
2, the geometric scaling ratio n was set to be 16 for all models in this study. In addtion,
the compressive stiffness EA of the model pile and prototype pile were 2.2 MN and 9011
MN, respectively.
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 10 of 78
As shown in Fig. 3(b) and Fig. 4, the retaining piles are numbered from 1 to 39,
increasing with distance to the observation window. P4, P7, P10, P12, P14, P16, P18, P19
and P26 are MPs, and P20 to P23 are IFPs. Dial indicators were set on top of the MPs to
monitor the horizontal displacement of the MP heads, and earth-pressure cells were
arranged to monitor the soil pressure acting on the MPs at depths of 40 cm and 60 cm
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
below the GS. In addition, similar to the MP in Cheng’s experiments (Cheng et al. 2017),
the distribution of the pile body strain along the pile length could be measured to calculate
To simulate the partial failure of the retaining pile structure, certain piles remained
intact during the excavation stage but were broken automatically at a preset location when
they received a control signal. Such a pile is called an IFP in this paper, and the working
mechanism is shown in Fig. 5. An IFP has two parts (upper and lower) that are separated
by a preset inclined rupture plane (60 cm above the bottom of the pile). The results of the
numerical simulation and tests showed that the position of the maximum bending moment
was close to this preset inclined rupture plane for most of the tests performed in this paper.
The initiating-failure struts (IFSs) and monitoring struts (MSs) were modelled using
hollow circle PVC tubes with cross sections of 40 mm × 3.2 mm (external diameter ×
thickness). A total of 13 struts was set along the length of the excavation; that is, the
supporting range of each strut was 3 piles, as shown in Fig. 4. A screw rod with a round
head was set at both ends of the strut, and this design has two advantages. While the strut
length could be adjusted using the screw to create close contact between the struts and
the waler beam, because of the round head of the screw rod, the two ends of the strut were
in contact at a point, which ensured that the support was subjected to the axial force. The
Page 11 of 78
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
waler beam adopted the same PVC tube as the retaining pile, with a width of 40 mm, a
Among the 13 struts, 8 MSs had strain gauges installed on their outer surfaces to obtain
the axial force of the MS. According to the axial tensile calibration test, the compressive
stiffness EA of the model struts was 926 kN; the strain at the strain gauge was 1.08 under
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
the action of a unit axial force (1 N). Based on the scaling relations in Table 2, the
compressive stiffness EA of the corresponding prototype was 3793 MN, which was close
to the compressive stiffness EA of the steel pipe with a diameter of 609 mm and a wall
thickness of 8 mm.
To simulate the partial failure of the struts, some struts were designed to be IFSs that
could be broken manually. The working mechanism is shown in Fig. 6. During the
excavation stage, limit blocks were used to avoid the ends of struts sliding on the waler
surface. For the partial failure stage in which the struts needed to be broken, the upper
limit blocks were removed first. Then, the IFS was pulled manually by a steel strand to
make it fail.
The model initiating failure anchors (IFAs) and monitoring anchors (MAs) were 2.0-
mm-diameter steel strands installed at 15° from the horizontal direction. The ends of the
steel strands were connected to the load cells, as shown in Fig. 3(c). In total, 19 anchors
were set along the length of the excavation, i.e., the anchoring range of each anchor was
2 piles, as shown in Fig. 1(d). According to the loading test in the tension testing of the
anchors, the tensile stiffness EA of the model anchors was 262 kN; the strain value of the
strain gauge was 3.82 when the anchor was subjected to a unit axial tension load (1 N).
Based on the scaling relations in Table 2, the tensile stiffness EA of the prototype anchor
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 12 of 78
was 1073 MN. Before the excavation, all anchors were prestressed precisely by turning
the screws, as shown in Fig. 6 (c). After the excavation, to simulate anchor failure, the
steel strand of the designated IFA was cut manually with cutting pliers, as shown in Fig.
6 (d).
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
The model test process can be divided into two stages: a normal-excavation stage and
a partial-failure stage. This section will introduce the results of the normal-excavation
stage. To easily understand the model test results and associate them with practical
engineering implications, the results discussed in the following sections are all converted
Fig. 7 shows the induced bending moments and deflection along the pile at various
excavation depths in two typical tests, test 1 (braced pile) and test 6 (tied-back pile). The
pile deflection profile can be obtained by integrating the bending moment distribution
twice using the measured pile head displacement and rotation as the input boundary
conditions (Leung et al. 2000; Tan et al. 2015). The bending moment along the pile
increased with the excavation depth, and the location of the maximum bending moment
was observed to be approximately 9.6 m below the GS level. By considering this factor,
it was reasonable to choose this depth for the preset rupture plane of the IFP.
When the excavation depth was 12.0 m, the bending moments in piles of different
retaining systems differed. For the braced retaining piles in tests 1, 3 and 5, the maximum
bending moments were 710.41 kN·m, 625.21 kN·m and 591.13 kN·m (the tension at the
excavation side was positive), resulting in an average of 642.25 kN·m, when the
Page 13 of 78
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
excavation depth was 12.0 m. For the tied-back retaining structure, the maximum bending
moments in tests 6 and 7 were 389.3 kN·m and 444.9 kN·m, respectively. For the
cantilever pile in test 0, the maximum bending moment (Cheng et al. 2017) was 1652.82
kN·m, which was much larger than those of the braced and tied-back piles. In addition,
when the excavation depth was 14.4 m, the maximum bending moments of the piles in
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
tests 3 (braced piles) and 7 (tied-back piles) were 988.94 kN·m and 682.9 kN·m,
respectively. For the same excavation depth, the maximum bending moment in a braced
pile was larger than that in a tied-back pile, mainly because the strut could provide a
stronger lateral stiffness compared to the anchor (EA values of the strut and anchor were
In this paper, the increase of the axial forces of the struts was induced only by the
excavation depth. When the excavation depth was 12.0 m, the average axial compressive
forces of the struts were 169.98 kN, 218.32 kN, 153.19 kN, 145.82 kN, and 163.97 kN
for tests 1-5, respectively, while the average axial tensile forces of the anchors were 44.03
kN and 45.18 kN for tests 6-7. The average axial compressive forces of the struts in test
2 were approximately 28% greater than those in other tests, indicating that a reduction of
the elevation of the strut can improve the supporting effect and provide greater lateral
stiffness to the retaining pile. Additionally, the average axial compressive forces of the
struts were clearly greater than the average axial tensile forces of the anchors, which also
indicated that struts have a higher lateral stiffness than that of anchors.
When the excavation finished, the partial failure of the retaining structure was triggered
by controllable equipment. As shown in Fig. 8, after the IFPs failed, the soil behind the
IPs adjacent to the IFPs and inside the boundary of the crater became loose and gradually
slid into the excavation in test 1, forming a changing stress arching behind these piles
during collapse. After collapse, the bending moments of these piles (P14-19) increased,
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
and the axial forces of some struts (S7-S8) decreased. As the sand moved into the
excavation, a crater appeared outside the excavation, and the sand formed a slope inside
the excavation, as shown in Fig. 9. For tests 0-3, the sizes of the craters outside the
excavations, which represented the range of the unloading effect, are summarized in
Table 4. At the same excavation depth (tests 0-2), the crater sizes in different tests were
similar, and when the excavation depth increased (test 3), the size of the crater also
increased.
4.2 Changes in the pile head displacements induced by partial failure in each test
Fig. 10 shows the increase in the pile head displacement induced by the partial failure
of the retaining structure in different tests. For cases with pile failure, the pile heads were
slightly displaced toward the outside excavation in tests 0-3 when the 4 IFPs (P20-23)
were kicked out completely because the soil behind the IFPs flowed into the excavation,
which reduced the total horizontal earth pressure acting on the piles and thus reduced the
axial forces of certain piles near the local failure zone, as shown in Fig. 8. Generally, the
pile head displacements were consistent with the axial force reduction phenomenon. In
engineering practice, when the retaining piles (walls) move outside the excavation, the
joints between the horizontal struts and the retaining piles will be damaged, which will
For the large deformations of the 4 IFPs (P20-23) in test 4, the IFPs moved 24 cm
inside the excavation along the preset rupture plane, and the stress arching in the
horizontal and vertical planes (as shown in Fig. 2) induced by the soil deformation slightly
For the piles that failed during cantilever excavation in test 0, the pile heads clearly
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
moved toward the inside of the excavation (Cheng et al. 2017). For cases with strut or
anchor failure, the pile head also moved inside the excavation in tests 5-7, and the
increases in the pile head displacements were clearly greater than those in the other tests
(such as test 0) because the struts or anchors undertook a large amount of horizontal load
and limited the pile head displacements. Once the horizontal retaining structures failed,
5. Analysis and comparison of the load transfer mechanisms in the retaining pile
failure cases
To study the influence of partial retaining pile failure on the adjacent retaining
structures and the whole retaining system, four tests (tests 1-4) are introduced in this
section. The load transfer mechanisms, including the variations in the earth pressure, the
axial forces of the struts and the internal forces of adjacent IPs caused by the failure of
IFPs, were analysed in detail based on test 1. The influence of the struts, strut elevation,
excavation depth and the initial localized large deformation of piles on the load transfer
Fig. 11 shows the curves of the earth pressures acting on the piles at depths of 6.4 m
and 9.6 m after the failure of the 4 IFPs (P20-P23) in test 1. As shown in Fig. 11(a), the
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 16 of 78
increasing ratios of the earth pressure at different locations followed different loading and
unloading modes at the failure moment of the piles. The earth pressure acting on the
closest pile, i.e., P19, first increased and then decreased as the sand slid into the
excavation. The reason for the earth pressure increase was the loading of the horizontal
soil arch (Cheng et al. 2017) due to the stress redistribution of the soil pressure behind
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
the piles at the moment the piles failed, and the decrease was due to the loss of soil behind
P19 (the height of the sand loss was approximately 5.6 m). Although the sand behind P19
and P18 was missing at certain heights, the increasing ratios of the earth pressure
stabilized to approximately 2.09 and 1.96 after the collapse, respectively. Therefore, the
results proved that the loading effect of the soil arch was greater than the unloading effect
Slightly farther from the IFPs, the earth pressure acting on P16 and P14 continued to
increase until the sand was stable. The fluctuation phenomenon in these earth pressure
variation curves was also described in Cheng’s paper (Cheng et al. 2017) and can be
interpreted based on the coupling between the soil arch effect and the unloading effect.
In addition, the earth pressure acting on the distant piles (P10, P7 and P4) did not change
The variation in the earth pressure acting on the piles at a depth of 9.6 m was similar
to that at a depth of 6.4 m, as shown in Fig. 11 (b). Due to the deeper burial depth, the
sand loss had little influence. Therefore, the loading effect of the soil arch played a
dominant role compared with the unloading effect of the sand sliding into the excavation.
Fig. 12 shows the maximum increasing ratio of the earth pressure at a depth of 6.4 m
induced by partial failure in tests 0 and 1. The increasing ratio of the earth pressure acting
Page 17 of 78
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
on piles adjacent to the IFPs (P20-P23) in test 1 was greater than that in test 0, but the
influence range of the earth pressure in test 1 was significantly smaller than that in test 0.
This result is because the horizontal struts in test 1 increased the lateral stiffness of the
retaining structure and made the braced retaining system have a much higher lateral
stiffness than the cantilever retaining system. In test 0, the soil arching effect caused by
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
partial collapse forced the IPs to move toward the excavation (Cheng et al. 2017), which
caused the additional soil pressure acting on the IPs to be redistributed and transferred to
the distant IPs. In summary, the unloading effect induced by the pile body displacement
decreased the soil pressure acting on these piles. In contrast, because of the struts, the
displacement of the IPs caused by the soil arching effect in test 1 was much smaller than
that in test 0; thus, the unloading effect induced by pile body displacement was weaker.
Therefore, the final increase in the soil pressure acting on the IPs was much greater than
that acting on the cantilever piles. Furthermore, the pile heads of some IPs adjacent to the
collapse position moved several millimetres toward the outside of the excavation, as
shown in Fig. 7 and Fig. 10, which further resulted in a greater increase in the soil pressure
In conclusion, the increment of the soil pressure acting on the IPs adjacent to the
collapse position was influenced by three factors: the loading effect of the soil arching
effect, the unloading effects of soil movement and the displacement of the pile. When the
pile at the foot of the soil arch has a lower lateral stiffness, the arching effect will be
significantly weakened. In addition, the IPs adjacent to the failed piles will have larger
Consequently, the influence zone of the earth pressures in test 0 was larger than that in
test 1. Therefore, the lateral stiffness of the piles was one of the chief factors of the
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 18 of 78
different ranges of the influence of failure between the braced excavation and the
cantilever excavation.
Fig. 13 shows the curves of the changes in the strut axial forces in test 1. When the
IFPs (P20-P23) were broken at the preset positions, the strut axial forces of S7 and S8
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
within the failure zone decreased instantaneously. Because the waler beam could not
provide effective support for the failed piles within the failure area, the earth pressure
previously acting on the IFPs did not transfer to the waler beam. Therefore, the strut axial
forces of S7 and S8 decreased. The strut force of S6 also clearly decreased because the
load acting on the waler beam near the failure area decreased considerably. For the struts
far from the IFPs (S4 and S5), the soil arching effect increased the earth pressure acting
on the piles within the support range of S4 and S5 and thus slightly increased their strut
axial forces at first. Then, due to the unloading effect caused by the sand sliding into the
excavation, the strut forces of S4 and S5 decreased slightly. Additionally, because the
distances between struts S1-S3 and the IFPs were much larger than those between the
other struts, the strut forces increased slightly until reaching stability.
Based on the above analysis, when pile failure occurred in the retaining structure, it
caused the pile head and waler beam to move toward the active zone, as shown in Fig.
10. Therefore, the strut forces near the pile failure decreased considerably. If the
connection between the struts and the retaining structure is weak, then the struts could
drop, which is similar to the phenomenon in which the struts dropped from a waler beam
in the Xianghu station collapse of the Hangzhou metro (Zheng et al. 2011). The integrity
and redundancy of the retaining structure would be influenced, and the degree of the
The changes in the increasing ratios of the bending moments of pile sections 10.4 m
below the GS, which are close to the maximum bending moments of the piles, are
illustrated in Fig. 14. After partial failure of the IFPs (P20-23), piles P19-10 adjacent to
the collapse position were loaded by the soil arching effect, and the bending moments of
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
these piles increased immediately to the first peak. Then, the bending moments of P19
increased slowly to the maximum, and this process lasted approximately 9 s. Finally, due
to the unloading effect of the sand entering the excavation, a slight unloading
phenomenon was observed at P19 and P18. At piles far from the failure position, such as
P7-P4, the bending moments did not clearly change after partial failure. When the sand
behind the IFPs (P20-P23) moved inside the excavation and gradually stabilized, the
Under the same initial failure condition, the variations in the pile bending moment
between cantilever excavation and one-strut-retained excavation were almost the same,
as shown in Fig. 12 and Fig. 14. However, the maximum increasing ratio of the pile
closest to the partial collapse zone (P19) was the greatest at 1.20 in test 0, which was
considerably smaller than the corresponding result of 1.36 in test 1. Considering the trends
in the increase of the soil pressure acting on the IPs between the two tests, the strut-
retained excavation had a stronger lateral stiffness in test 1, and the loading effect caused
by partial failure was obvious. While cantilever excavation had a weaker lateral stiffness,
the retaining structure produced a greater displacement response for the same partial
failure condition. Therefore, the unloading effect of the greater displacement certainly
After the bending moment of the piles peaked in test 1, the unloading magnitude was
much smaller than that of the cantilever pile, as shown in Fig. 15, indicating that the
unloading phenomenon of the pile bending moment is very weak. This phenomenon was
also seen in tests 1-5 and represents one of the differences between cantilever excavation
and braced excavation after the failure of piles. Furthermore, this phenomenon can be
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
horizontal struts. The failure of piles clearly leads to the soil arching effect, and the
adjacent piles will be reloaded by the deformation coordination of the waler beam during
the spread of the soil arching effect. Therefore, it is difficult to observe obvious unloading
5.1.4 Load transfer coefficients of test 1 (braced pile) and test 0 (cantilever pile)
The peak increasing ratio of the bending moment in an adjacent pile induced by partial
failure was defined as the “load transfer coefficient (bending moment)” (Cheng et al.
2017), which was used to measure the change in the internal force of the piles outside the
partial failure zone. As shown in Fig. 12, the load transfer coefficients of piles P19-P16
in test 1 were higher than those in test 0 within the failure zone; however, they were lower
than those in test 0 far from the failure position because the decrease in the load transfer
coefficient was greater in test 1, i.e., the range of the influence of failure was smaller. The
reasons behind this phenomenon have been discussed in Section 5.1.1, as follows: the
braced retaining system has a much larger lateral stiffness than does the cantilever
retaining system; thus, the additional displacement of the piles induced by the soil arching
effect was smaller, and the additional soil pressure caused by the soil arching effect was
acting on only the piles close to the zone of influence of the failure, leading to a higher
5.2 Influence of the strut elevation on the load transfer mechanisms (comparison of
tests 2 and 1)
The difference between test 2 and test 1 was the installation elevation of the struts. In
test 2, the struts were 0.2H (H: 12 m) below the GS. When the IFPs (P20-P23) were
broken, the change in the horizontal earth pressure at a depth of 6.4 m and the strut force
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
in test 2 were similar to those in test 1 (and will not be introduced here). Fig. 16 shows
the increase ratios of the bending moments of the pile sections 10.4 m below the GS after
the failure of the IFPs in test 2. Fig. 17 shows the comparison of the load transfer
As shown in Fig. 17, the increasing ratio of the bending moment of P19 in test 2 was
1.46, which was greater than that in test 1 (1.36). The influence range in test 2 was also
larger than that in test 1. This result is because the lateral stiffness of the retaining
structure and the magnitude of the unloading caused by the partial failure were two
important factors that significantly influenced the value of the load transfer coefficient
and the range of the load transfer. In tests 1 and 2, the increments of maximum pile
deflection were 48 mm and 37 mm from excavation depth of 3.2 m (the depth when the
Fig. 7(a). Therefore, the lateral stiffness of the retaining piles with struts installed at a
certain depth under pile head in test 2 was greater than that with struts installed at the
level of pile head in test 1. In addition, although the excavation depths in test 1 and test 2
were the same, axial force on the strut in test 2 was greater than that in test 1, as discussed
in Section 3.2. Therefore, the unloading induced by the failure of piles P20-P23 in test 2
was higher than that in test 1, causing a greater load to be transferred to the adjacent piles.
Because of the above two factors, the load transfer coefficient (bending moment) and
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 22 of 78
influence range caused by partial failure in test 2 were significantly greater than those in
test 1.
5.3 Influence of the excavation depth on the load transfer mechanisms (comparison
of tests 3 and 1)
5.3.1 Analysis of the soil pressure and load transfer coefficient (bending moment)
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
The difference between test 3 (14.4 m) and test 1 (12.0 m) was the excavation depth.
After the IFPs (P20-P23) were broken, the change in the earth pressure over time at a
depth of 6.4 m below the GS was approximately the same as that in test 1. However, the
increasing ratios of the soil pressure caused by the partial failure in test 3 were generally
greater than those in test 1, except for that of P19, as shown in Fig. 18. In the strutted pile
retaining system, the horizontal soil load supported by each strut increased with the
excavation depth. Therefore, the strut unloading induced by the failure of piles P20-P23
in test 3 was significantly greater than that in test 1. Consequently, in test 3, the load
transferred to adjacent piles was greater, and the effect of the soil arching was stronger.
Similarly, the load transfer coefficient and influence range caused by partial failure in test
3 were clearly greater than those in test 1. For P19, because the crater caused by the failure
outside the excavation was deeper in test 3, as shown in Table 4, the missing soil behind
P19 reduced the stress arch acting on P19. Therefore, the increasing ratio of the soil
Fig. 19 shows the changes in the increasing ratios of the forces of the 8 MSs in test 3,
which were similar to those in test 1. However, compared to test 1 and test 2, the soil
arching effect caused by the unloading in test 3 was greater and had a greater loading
effect on the struts far from the local failure position (e.g., S4 and S3).
Page 23 of 78
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Near the IFPs (P20-P23), the strut axial forces of S7 and S8 first decreased, then
increased, and ultimately decreased again. The forces acting on S7 and S8 decreased
immediately after pile failure occurred. At the same time, the soil body behind the IFPs
moved into the excavation, developing the soil arch that acted on the IPs adjacent to the
IFPs. Then, the loading of the soil arch acted on S3-S5 through coordination with the
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
waler beam, causing their axial forces to increase quickly. Subsequently, as the soil
behind the IFPs slid into the excavation, the axial forces of S7 and S8 eventually
decreased. During this process, the deformation coordination of the waler beam, the effect
of the soil arch and the unloading of the local soil mass played important roles in changing
5.4 Influence of large pile deformation on the load transfer mechanisms (comparison
of tests 3 and 4)
The difference between test 4 and test 3 was the state of the IFPs (P20-P23) after
receiving the signal command for IFP failure. In test 4, the IFPs were broken in the preset
section, and the upper parts of the IFPs moved 24 cm toward the excavation in the
horizontal direction; however, the upper parts of the IFPs were not completely kicked out
and fell into the excavation. Therefore, the soil behind the IFPs did not flow into the
excavation, which is unlike what occurred in tests 0-3. The state of the IFPs in test 4 was
The changes in the earth pressure over time at a depth of 6.4 m below the GS are shown
in Fig. 20. The earth pressure acting on P19 first increased, then decreased, and ultimately
increased and stabilized. First, the loading effect of the horizontal soil arching effect led
to the immediate increase of the earth pressure when the large deformation of the IFPs
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 24 of 78
occurred. Then, the soil behind the upper part of the IFPs followed the movement of the
IFPs in the horizontal direction, which decreased the earth pressure due to the soil
unloading effect of the large deformation. Finally, because of the large deformation of
the IFPs, the loading of the vertical soil arching effect led to another increase in the earth
pressure. However, the maximum increasing ratio of the earth pressure acting on the other
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
piles was clearly lower than that in test 3, as shown in Fig. 18, mainly because the soil
arching effect caused by excessively large deformation of the IFPs was weak.
Fig. 21 shows the changes in the increasing ratios of the MS axial forces after the large
deformation of the IFPs occurred in test 4. The horizontal axial forces of S7 and S8
immediately decreased, then increased, and gradually stabilized. The horizontal axial
forces of S6-S3 first fluctuated and then increased to stable values. The axial forces of S7
and S8 first decreased because at the moment of IFP breakage, the earth pressure acting
on the IFPs could not effectively transfer to the waler beam and S7 and S8. Finally, the
axial force of each strut increased significantly compared with that before failure because
a large soil deformation near the preset section (depth of 9.6 m) resulted in soil arching
in both the horizontal and vertical planes in the surrounding soil, as shown in Fig. 2.
The soil arch in the horizontal plane transferred the loading released by the IFPs to the
adjacent IPs and further compressed all struts through deformation coordination with the
waler beam, immediately increasing the horizontal axial force of the struts. The upper
foot of the soil arch in the vertical plane was at the waler beam, causing the earth pressure
acting on the waler beam to increase. Therefore, the large deformation of the IFPs had a
loading effect on all struts, which also led to all piles deforming toward the excavation,
Page 25 of 78
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
as shown in Fig. 10. The influence of the large deformation of the IFPs on the struts was
As shown in Fig. 22, in test 4, although the upper and lower parts of the IFPs were not
completely separated, the partial damage led to loading due to the soil arching effect,
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
causing the bending moments of adjacent IPs to quickly peak and then stabilize. For such
a response, the retaining piles did not need to be completely broken. A large deformation
of the retaining pile body could lead to a significant soil arching effect behind the pile
wall, resulting in a redistribution of the internal forces of the retaining structure. Therefore,
because they are an important factor related to the progressive collapse of deep
excavations. The maximum load transfer coefficient (bending moment) caused by the
large deformation of the IFPs (P20-P23) was 1.54, while that induced by the IFPs (P20-
To study the influence of horizontal bracing structure failure, such as the failure of
struts and anchors, on the adjacent retaining structure, three tests (tests 5-7) are introduced
in this section. Test 5 included the failure of 6 struts that supported 18 piles (the sequence
of strut failure was S7, S6, S5, S8, S9 and S10), and tests 6 and 7 included the failure of
9 anchors that also supported 18 piles (the sequence of anchor failure was A10, A9, A11,
A8, A12, A7, A13, A6 and A14). The influence of the excavation depth on the load
As shown in Fig. 23, six struts (S5-S10) were broken in test 5. The failure of the six
IFSs also led to the redistribution of the soil pressure behind the piles. The six IFSs (S5-
S10) were located between P14 and P29. Each IFS could not function, reducing the
stiffness of the corresponding position of the retaining structure and leading to an increase
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
in the horizontal displacement of the pile top toward the excavation. Then, the soil arching
effect behind the piles loaded the piles beyond the range of influence of strut failure; thus,
the earth pressure acting on these piles increased. For example, the earth pressure acting
on P10 and P4 continuously increased as each strut broke. With the increase of the number
of failed struts, a pile in the range of the influence of strut failure underwent a greater
displacement in the horizontal direction; thus, the earth pressure acting on the pile
decreased. As shown in Fig. 23, the earth pressure acting on P19 and P16 decreased
For test 5, the changes in the strut axial force caused by the successive failure of the
six struts (S5-S10) are shown in Fig. 24. S7 was broken first, and the load that it supported
was transferred to the adjacent struts due to the load transfer of the waler beam. The axial
forces of S6 and S8 (i.e., the struts nearest to S7) increased clearly. When struts S7 and
S6 were broken, the axial forces of S5 and S8 increased the most. Likewise, when the
remaining struts were successively broken, each strut transferred most of its load to the
two nearest struts, with one on each side. After the failure of strut S10, the axial force of
strut S4 adjacent to the failure zone increased to 1015.8 kN (approximately 1.99 times the
strut axial force when the excavation was finished), while the axial force of strut S3,
which was one strut farther from the failure zone, increased much less than did that of
Page 27 of 78
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
adjacent strut S4. Therefore, each strut failure will increase the load on the nearest strut
on either side, which was similar to the findings in the numerical simulations conducted
by some researchers (Choosrithong and Schweiger 2018; Goh et al. 2018). This loading
pattern is referred to as the nearest phenomenon of load transfer caused by strut failure.
When a strut was broken, its load was not transferred to many other intact struts but was
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
concentrated on the adjacent struts, increasing the axial force of some struts and leading
to further failure. Therefore, wide-ranging progressive failure of the struts might occur in
this case.
According to the survey results of the collapse accident of Singapore's Nicoll metro
excavation, a ninth-level strut failed first, which resulted in the load supported by the
eighth-level struts increasing significantly and to the struts sustaining damage. Then, a
domino effect caused the progressive failure of the struts in the excavation profile (COI.
2005; Puzrin et al. 2010; Whittle and Davies 2006). Singapore's Nicoll metro accident
As shown in Fig. 25, in test 5, the failure of strut S7 caused the bending moment of
adjacent pile P18 to decrease slightly. Subsequently, when struts S6 and S5 were broken,
the bending moments of P4 and P10 increased slightly. However, the bending moments
of piles P14-19 adjacent to the strut-failure zone decreased. When S8, S9 and S10
successively failed, the bending moments of the piles adjacent to the strut-failure zone
decreased significantly, and those of the distant piles increased slightly, as shown Fig. 26.
After the failure of S5-S10, the bending moments of piles outside the failure zone, such
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 28 of 78
as P4, P7, P10 and P12, increased, mainly due to the arching effect, as discussed in
Section 6.1.1. However, the bending moments of piles in the failure zone decreased,
especially for the piles at the central area of the failure zone, such as P18. The changes of
the bending moment of the pile in the failure zone can be interpreted as follows.
Fig. 27 shows the changes in the bending moment of P26 within the strut-failure zone
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
as the number of failed struts increased. The bending moment distribution of pile P26 was
a typical single-row strutted pile distribution before strut failure. When the struts far from
P26, namely, struts S7, S6 and S5, were successively damaged, the bending moment of
P26 slightly increased due to the soil arching effect caused by the increase in the pile
displacement within the region of struts S7-S5. When the struts close to P26, namely,
struts S8, S9 and S10, were successively broken, the maximum bending moment of the
upper part of P26 gradually decreased, while the bending moment of the lower part of
P26 gradually increased. However, in general, the absolute value of the maximum
bending moment of a pile body decreased after the failure of the struts, as shown in Fig.
25 and Fig. 27. The resulting situation can be interpreted by the force analysis of an
isolation body of P26, as shown in Fig. 28. When x ≤ 12.0 m (x is the depth below the GS
and anchor head), i.e., when point C is above point B (excavation surface), the bending
by the earth pressure in the active zone). Since the earth pressure in the active zone above
the excavation bottom decreased slightly, as shown in Fig. 29, M1 also decreased slightly.
When the strut failed, the axial strut force decreased to 0 kN. Then, the shear force in the
pile head within the anchor-failure zone, which equals the support force F of the waler
beam acting on the pile head, decreased significantly. Therefore, the influence of F·x on
Mc was greater than that of M1, and the bending moment Mc in the retaining pile above
Page 29 of 78
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
the bottom of the excavation decreased. The above analysis is a major mechanism for the
decrease of the maximum bending moment of retaining piles in the anchor-failure zone.
The influences of strut failure and pile failure on adjacent piles were significantly
different.
6.2 Load transfer mechanisms under the conditions of anchor failure and the
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
As shown in Fig. 29, nine anchors (A6-A14) were broken in test 6. Each IFA failure
reduced the stiffness of the corresponding position of the retaining structure, increasing
the horizontal displacement of the pile top toward the excavation and leading to the
redistribution of the soil pressure behind the piles. Generally, the earth pressure acting on
the piles in the anchor-failure zone decreased gradually as each anchor failed, while the
earth pressure acting on the distant piles slightly increased. The location of the increase
in earth pressure was the soil arch foot. This phenomenon is similar to that of the strut
For test 6, the changes in the anchor axial forces caused by the successive failure of the
nine anchors (A6-A14) are shown in Fig. 30. After nine anchors were broken, their loads
were distributed to the adjacent anchors due to the load transfer of the waler beam and
the redistribution of the soil stress. The axial force of A15 and A5 (i.e., the anchors nearest
to the anchor-failure zone) increased clearly, as shown in Fig. 31(b). The resulting
situation was similar to that due to strut failure; that is, each anchor failure increased the
load of the nearest anchor on either side, but the load transfer coefficient (axial force) was
clearly smaller than that in the strut failure test. In particular, after 9 anchors were broken,
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 30 of 78
the load transfer coefficients (anchor forces) of A5, A4 and A3 were 1.49, 1.17, and 1.11,
respectively, as shown in Fig. 31(b). However, after 6 struts were broken, the load transfer
coefficients (strut forces) of S4 and S3 were 2.99 and 1.32, as shown in Fig. 31(a). In
addition, the range of influence of anchor failure was larger than that of strut failure, as
shown in Fig. 31. These phenomena can be interpreted by the lateral stiffness of the
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
struts/anchors and the effect of the waler beam. Because struts can provide a stronger
lateral stiffness compared to anchors, the two nearest struts will bear most of the load
released by the failure of a strut, and the distant struts will share a small percentage of the
load transferred due to the failure of the strut. However, anchors provide a weaker lateral
stiffness than do struts. Therefore, the load released by the failure of an anchor will
redistribute to more anchors through the waler beam. The higher the lateral stiffness of
the retaining structure, the larger will be the load transfer coefficient (axial force) and the
As shown in Fig. 32, in test 6, the failure of anchor A10 caused the bending moment
of adjacent piles P18 and P19 to decrease very slightly. Subsequently, when anchors A9
and A11 were broken, the bending moments of P4 and P10 increased. However, the
bending moments of the piles close to the anchor-failure zone (P14-P19) decreased.
Similarly, when A8, A12, A7, A13, A6 and A14 successively failed, the bending
moments of the piles close to the anchor-failure zone decreased significantly, and those
Fig. 33 shows the variation of the bending moment distribution in the pile body of P19
at the central area of the failure zone, which can also be interpreted by the mechanism of
test 5 in Section 6.2.3. Because P19 was located at the influence zone of anchor failure at
Page 31 of 78
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
all times, the strut effect of P19 provided by the anchors and waler beam gradually
decreased significantly, and then the bending moment distribution pattern changed from
the single-row strut type to the cantilever type. As the number of failed anchors increased,
the maximum bending moment, which is in the upper part of P19, gradually decreased,
while the bending moment of the lower part of P19 gradually increased. However, in
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
general, the absolute value of the maximum bending moment of a pile body decreased
In addition, based on Fig. 27 and Fig. 33, it can be speculated that when the failure
range of the strut or the anchor continues to increase, the bending moments in the lower
part of the piles in the failure range will increase and become the largest bending moments
in the entire pile. Simultaneously, the support force of the waler beam acting on the pile
head will decrease, which indicates that the mechanics that characterize the pile change
from single-strut mode to cantilever mode, and then the absolute value of the largest
bending moment may be larger than that before partial failure. Based on this analysis,
when the failure zone of anchors or struts is small, the load transfer coefficients (axial
forces) of adjacent anchors or struts are significantly larger than 1; however, the load
transfer coefficients (bending moment) of the piles are relatively small. Then, the
progressive failure path still follows the anchors or struts. When the failure zone of the
anchors or struts is large enough, the mechanics that characterize the piles in the failure
zone enter cantilever mode, and the load transfer coefficients (bending moments) of these
piles become very large. Then, the failure of the piles begins, which will lead to large-
6.3 Influence of the excavation depth on the load transfer mechanisms in the anchor
The difference between test 7 and test 6 was the excavation depth (14.4 m and 12 m,
respectively). After the IFAs (A6-A14) were broken, the changes in the anchor axial force
over time in test 7 were approximately the same as those in test 6, as shown in Fig. 31(b).
The difference in the results between test 6 and test 7 for anchor failure was in the load
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
transfer coefficients (axial forces) of adjacent anchors. For example, the load transfer
coefficients (axial forces) of A5 and A15 were 2.00 and 2.30 in test 7, while those of A5
and A15 in test 6 were 1.49 and 1.51. In the anchor-pile retaining system, the horizontal
soil load supported by each anchor increased with the excavation depth. Therefore, the
unloading effect of anchors induced by the failure of IFAs (A6-A14) in test 7 was
significantly stronger than that in test 6. Consequently, the load effect of the failed
anchors on adjacent anchors in test 7 was also stronger than that in test 6.
6.3.2 Comparison of the load transfer coefficients (bending moments) of tests 6 and 7
In test 7, the change in the pile moment caused by partial failure was similar to that in
test 6. When A6-A16 successively failed, the bending moments of the piles within and
close to the anchor-failure zone decreased significantly, and those of the distant piles
increased slightly, as shown in Fig. 32. The horizontal stiffness of the pile-anchor system
in test 7 was smaller than that in test 6, and the horizontal displacements of the waler
beam caused by the partial failure and its range in test 7 were much larger, as shown in
Fig. 10. Therefore, the range of decrease due to the bending moments in the piles
increased in test 7.
7. Conclusions
In this study, model tests of partial failure in the retaining structures of strut-pile and
anchor-pile excavations were designed and conducted. The load transfer mechanisms
Page 33 of 78
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
after different types of partial failures were analysed and compared. Based on the failure
type, the main conclusions of this research can be divided into two parts and are
summarized below.
(1) At the same excavation depth, when certain piles failed, the bending moment
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
increasing ratios of adjacent piles in braced-pile excavation were greater than those in
cantilever-pile excavation, because the braced retaining piles had a much larger lateral
stiffness. This indicates that the same number of failed piles would more easily trigger a
(2) In braced-pile excavation, the partial failure of the piles caused the axial forces of
the struts in the partial failure zone to decrease significantly because the load previously
sustained by the failed piles could not transfer to the waler beam. This change led the pile
head and waler beam in the pile failure zone to deform towards outside of the excavation,
resulting in a significant reduction of the axial compressive force in the horizontal strut
and possibly of the tensile force. If the connections between the struts and the retaining
structure are weak, the struts may drop and reduce the integrity and redundancy of the
entire retaining system and even cause large-scale progressive collapse, similar to case of
the Hangzhou metro accident (Chen et al. 2013). Therefore, the strength of the
(3) The magnitude of strut unloading caused by the partial failure of the retaining pile
and the lateral stiffness of the structure were two important factors that significantly
influenced the value of the load transfer coefficient and the range of the load transfer. For
braced-pile retaining systems, the factor of strut unloading is dominant. When the strut
elevation was lower, or the excavation depth was deeper, the magnitude of strut unloading
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 34 of 78
caused by the partial failure of the retaining pile was larger, and the load transfer
coefficients of the adjacent piles and the influence range were also greater. However, for
cantilever excavation, the lateral stiffness of the retaining structure may be the dominating
factor, because the load transfer coefficients decreased and the influence range increased
(4) When large deformations occurred in certain retaining piles, it also caused a
substantial increase in the bending moments of the adjacent piles, although the maximum
load transfer coefficient was smaller than that in the pile-rupture situation. Therefore,
large deformation of some pile bodies may lead to the progressive failure of the
excavation.
(1) For single-level strut-pile excavation, each strut failure increased the load on the
nearest (1-2) struts at two sides but had a very limited influence on farther struts. This
loading pattern is referred to as the nearest phenomenon of load transfer caused by strut
failure, which can easily cause the progressive failure of struts. In practical engineering,
when certain struts in an excavation are in a dangerous situation, they should be reinforced
(2) For single-level anchor-pile excavation, the load transfer coefficient (axial force)
(1.51 in test 6) was much lower than that in the strut-failure case (2.99 in test 5). However,
the influence range of the failed anchors (3-4 anchors) was larger than that of the strut-
failure case (1-2 struts). This was because anchors provide a weaker lateral stiffness than
do struts. In addition, when the excavation was deeper, the load transfer coefficients (axial
force) of the anchors were larger, which is similar to the case for braced excavation.
Page 35 of 78
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
(3) The influences of strut and anchor failure on piles were significantly different from
to the influence of pile failure on adjacent piles. When the failure zones of anchors or
struts were not very large, the bending moments of piles in the failure zone decreased.
The potential progressive failure path only follows anchors or struts. As the number of
failed anchors or struts increases, the piles in the failure zone gradually enter cantilever
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
mode, and their bending moments first decrease and then increase to much higher values
than those before failure. Therefore, the load transfer coefficients (bending moments) of
these piles become very large, and the progressive failure path extends to other piles and
Acknowledgments
This work was supported by the National Key Research and Development Program of
China under grant number 2017YFC0805407, the National Natural Science Foundation
of China under grant number 41630641 and the Natural Science Foundation of Tianjin
References
Altaee, A., and Fellenius, B. H. 1994. Physical modeling in sand. Canadian geotechnical
Artola. J. 2005. A solution to the braced excavation collapse in Singapore. M.Sc. thesis,
Bjerrum, L., and Eide, O. 1956. Stability of strutted excavations in clay. Geotechnique,
Chang, M. 2000. Basal stability analysis of braced cuts in clay. Journal of Geotechnical
0241.
Chen, R. P., Li, Z. C., Chen, Y. M., Ou, C. Y., Hu, Q., and Rao, M. 2013. Failure
10.1061/(ASCE)CF.1943-5509.0000557.
Cheng, X. S., Zheng, G., Diao, Y., Huang, T. M., Deng, C. H., Lei, Y. W., and Zhou, H.
10.1016/j.engfailanal.2016.06.011.
Cheng, X. S., Zheng, G., Diao, Y., Huang, T. M., Deng, C. H., Nie, D. Q., and Lei, Y. W.
10.1139/cgj-2016-0284.
COI. 2005. Report of the Committee of Inquiry into the incident at the MRT circle line
Manpower, Singapore.
Page 37 of 78
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Corral, G., and Whittle, A.J. 2010. Re-analysis of deep excavation collapse using a
Davies, R.V. 2007. Factors in the Nicoll highway collapse. Tunnels & Tunnelling
International, 2: 31–34.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
Goh, A.T. C., and Wong, K. S. 2009. Three-dimensional analysis of strut failure for
Goh, A. T. C., Fan, Z., Liu, H.L., Zhang, W. G., and Dong, Z. 2018. Numerical Analysis
6632-0_43.
Haack, A. 2009. Construction of the North-South-Metro Line in Cologne and the accident
Hsieh, P. G., Ou, C. Y., and Liu, H. T. 2008. Basal heave analysis of excavations with
Huang, M. S., Song, X. Y., and Qin, H. L. 2008. Basal stability of braced excavations in
Ishihara, K., and Lee, W.F. 2008. "Forensic diagnosis for sitespecific ground conditions
characterization. Edited by R.Q. Coutinho and P.W. Mayne. Taylor & Francis
Itoh, K., Kikkawa, N., Toyosawa, Y., Suemasa, N., and Katada, T. 2011. Failure
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
Mechanism of Anchored Retaining Wall due to the Anchor Head Itself Being
Kempfert, H. G., and Gebreselassie, B. 2006. Excavations and Foundations in Soft Soils.
Netherlands: Springer.
Leung, C. F., Chow, Y. K., and Shen, R. F. 2000. Behavior of pile subject to excavation-
Li, G. X., and Li, X. M. 2010. The shear strength in stability analysis of subway pit in
Lu, Y., and Tan, Y. 2019. Overview of Typical Excavation Failures in China, Geo-
Luo, Z., Atamturktur, S., Cai, Y., and Hsein juang, C. 2012. Simplified Approach for
Meymand, P.J. 1998. Shaking table scale model tests of nonlinear soil-pile-superstructure
Osman, A. S., and Bolton, M. D. 2006. Ground Movement Predictions for Braced
Ou, C. Y. 2006. Deep Excavation Theory and Practice. Taylor & Francis group, London,
UK.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
Pong, K. F., Foo, S. L., Chinnaswamy, C. G., Ng, C. C. D., and Chow, W. L. 2012. Design
practising engineers. The IES Journal Part A: Civil & Structural Engineering, 5(3),
Puzrin, A. M., Alonso, E. E., and Pinyol, N. M. 2010. Braced Excavation Collapse: Nicoll
Silva, A. P. D., Festugato, L., and Masuero, J. R. 2017. A new methodology to assess the
doi: 10.1016/j.tws.2017.04.022.
Simoni A., Houlsby, G. T. 2006. The Direct Shear Strength and Dilatancy of Sand–gravel
Stille, H., and Broms, B. B. 1976. Load redistribution caused by anchor failures in sheet
pile walls. In Proceedings of the 6th European Conference on Soil Mechanics and
Tan, Y., Jiang, W., Luo, W., Lu, Y., and Xu, C. 2018. Longitudinal Sliding Event during
doi: 10.1061/(ASCE)CF.1943-5509.0001181.
Tan, Y., Jiang W. Z., Rui H. S., Lu, Y., and Wang, D. L. 2020. Forensic Geotechnical
Tan, Y., Wei, B., Zhou, X., and Diao, Y. P. 2015. Lessons learned from construction of
TR26. 2010. Technical reference for deep excavation. Spring Singapore, Singapore.
Ukritchon, B., Whittle, A. J., and Sloan, S. W. 2003. Undrained Stability of Braced
Wang, W. D., Ng, C. W., Hong, Y., Hu, Y., and Li, Q. 2019. Forensic study on the
2006, Singapore.
Yang, J., Luo, and X. D. 2018. The critical state friction angle of granular materials: does
Zhao, W., Han, J. Y., Chen, Y., Jia, P., Li, S. G., Li, Y., and Zhao, Z. 2018. A numerical
10.1177/1687814018756775.
Zheng, G., Cheng, X. S., and Diao, Y. 2011. Concept and design methodology of
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
List of captions
Tables
Table 2 Scaling relations of 1-g physical modelling in sand (Altaee and Fellenius 1994;
Table 3 Parameters of the piles and excavations in the prototype and model
Figures
Fig. 1. Schematic model information, including the pile, strut and anchor arrangements,
instrument locations and excavation depths (plan view from the top): (a) test 0, (b) test
Fig. 2. Large deformation of piles in test 4: (a) vertical soil arch and (b) horizontal soil
arch.
Fig. 3. Schematic model setup (all dimensions are in mm): (a) side elevation for the
braced pile excavation in test 1, (b) side elevation for the tied-back pile excavation in
test 6, and (c) plan view from the top for the braced pile excavation in tests 1, 3 and 5.
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 42 of 78
Fig. 4. Example of an excavation model test (test 1) (IFP: initiating failure pile; IP:
Fig. 5. Device and working mechanism of the initiating failure pile (IFP).
Fig. 6. Device and working mechanism of the initiating failure strut (IFS) and initiating
failure strut (IFA): (a) excavation stage, (b) strut failure stage, (c) excavation stage, and
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
Fig. 7. Moment curves and pile deflection for tests 1 and 6: (a) test 1 and (b) test 6.
Fig. 8. Illustration of the movements of the IFPs that acted as a horizontal trapdoor and
Fig. 9. Collapse scenarios: (a) illustration of the slope inside the excavation and (b)
Fig. 10. Increases in the pile head displacement induced by partial failure in each test
Fig. 11. Change curves of the earth pressures acting on adjacent piles in test 1: (a)
increasing ratios of the earth pressures at a depth of 6.4 m and (b) increasing ratios of
Fig. 12. Comparison of the load transfer coefficients (earth pressure and bending
Fig. 13. Change in the axial forces of struts caused by the IFPs in test 1.
Fig. 15. Moment curves of the first pile adjacent to the IFPs before and after pile failure
Fig. 17. Comparison of the load transfer coefficients (bending moment) in test 1 and test
2.
Page 43 of 78
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Fig. 18. Comparison of the load transfer coefficients (earth pressure and bending
Fig. 19. Change in the axial forces of struts caused by the IFPs in test 3.
Fig. 20. Change in the earth pressure at a depth of 6.4 m over time in test 4.
Fig. 21. Changes in the axial forces of struts caused by the large deformation of IFPs in
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
test 4.
Fig. 23. Change in the earth pressure at a depth of 6.4 m versus the strut failure order in
test 5.
Fig. 24. Change in the axial forces versus the strut failure order in test 5.
Fig. 25. Change in IP bending moments versus the strut failure order in test 5.
Fig. 26. Comparison of the load transfer coefficients (bending moments) in tests 5-7.
Fig. 27. Profiles of the bending moment of IP P26 for each strut failure in test 5.
Fig. 28. Stress analysis diagram of the support pile isolation body.
Fig. 29. Changes in the earth pressure at a depth of 9.6 m versus the anchor failure order
in test 6.
Fig. 30. Changes in the anchor forces versus the anchor failure order in test 6.
Fig. 31. Comparison of the load transfer coefficients (axial force) in tests 5-7: (a) strut
axial forces in test 5 and (b) anchor axial forces in tests 6-7.
Fig. 32. Changes in IP bending moments versus the anchor failure order in test 6.
Fig. 33. Profiles of the bending moment of IP P19 versus anchor failure in tests 6-7.
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 44 of 78
Table 2 Scaling relations of 1-g physical modelling in sand (Altaee and Fellenius 1994;
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
Table 3 Parameters of the piles and excavations in the prototype and model
Excavation Prototype Elevation of
Retaining
Test no. depth H excavation strut or Failure mode
structure
(mm) depth (m) anchor BGS
Test 0
2 IFPs (P20-P21)
(Cheng et Cantilever pile 750 12
(Symmetry)
al. 2017)
Test 1 Braced pile 750 12 0H 4 IFPs (P20-P23)
Test 2 Braced pile 750 12 0.2H 4 IFPs (P20-P23)
Test 3 Braced pile 900 14.4 0H 4 IFPs (P20-P23)
4 IFPs (P20-P23)
Test 4 Braced pile 900 14.4 0H
(Large deformation)
Test 5 Braced pile 750 12 0H 6 IFSs (S5-S10)
Test 6 Tied-back pile 750 12 0H 9 IFAs (A6-A14)
Test 7 Tied-back pile 900 14.4 0H 9 IFAs (A6-A14)
Note: IFPs-initiating failure piles; P-pile; IFSs-initiating failure struts; S-strut; IFAs-initiating failure
anchors; A-anchor; BGS-below ground surface.
Page 45 of 78
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
*Half model.
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 46 of 78
(a)
0 Pile No.: P1 P39
P1-P2: IFPs
-750
1200
Symmetrical plane
39 piles: 2 IFPs + 37 IPs
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
-750
1200
-750
1200
15°, @130
-750
Anchor No.:
1560 A1 A19 A6-A14: IFAs
instrument locations and excavation depths (plan view from the top): (a) test 0; (b) test
A A
Preset
rupture Outside
plane Main stress excavation
direction of
Main stress soil arch
direction of Excavation
soil arch bottom Planform of preset rupture plane A-A
Fig. 2. Large deformation of piles in test 4: (a) vertical soil arch and (b) horizontal soil
arch.
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 48 of 78
Strut
600
Strain
750
Height of model: 1400
gauges
Preset
1200
rupture
plane
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
Excavation
Sand Earth bottom
pressure
cell
750
Masonry tank
Monitoring pile
(b) Screw Dial indicator
Anchor head
GS +0(mm)
Load
cell
Strain
gauge
Anchor
750
(Steel, ø2@130)
Pulley
Earth Excavation
Pressure bottom
Sand cell
1510
-1200
Masonry 650
-1400 tank
(c)
1710
Outside
63
Excavation
Monitoring pile for IP: MP
P20 - P23: IFPs
Instrumentation of MP:
Dial Intact
indicator pile
Earth Strain
pressure cell gauges
S13
P39
Strain gauge
Masonry tank
Fig. 3. Schematic model setup (all dimensions are in mm): (a) side elevation for the
braced pile excavation in test 1, (b) side elevation for the tied-back pile excavation in
test 6, and (c) plan view from the top for the braced pile excavation in tests 1, 3 and 5.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 49 of 78
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 50 of 78
Strut Strut
Upper part Upper part
of pile
60 cm
of pile
Steel rod kicked out
Linear
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
bearing
120 cm
Preset Partial
rupture collapse
plane signal
Excavation Steel rod is pulled
bottom into the lower part
Lower Steel wire rope of pile
part of
pile Air cylinder Piston rod push out
Air Air
compressor compressor
Air pipe
Solenoid valve: Switch OFF Solenoid valve: Switch ON
Fig. 5. Device and working mechanism of the initiating failure pile (IFP).
Page 51 of 78
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Crank Crank
(a) Pulley (b) Pulley
Waler beam Waler beam
Strut Steel
GS +0(mm) GS +0(mm) strand
Beam Beam
Pile Pile
-750 -750
Excavation Excavation
bottom bottom
(c) (d)
Screw Screw
Cutting
piler
Steel
strand 15° Pulley 15°
Anchor Anchor
Load
cell Pulley Load
cell
Sand Sand
Fig. 6. Device and working mechanism of the initiating failure strut (IFS) and initiating
failure anchor (IFA): (a) excavation stage, (b) strut failure stage, (c) excavation stage,
and (d) anchor failure stage.
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 52 of 78
-8 -8
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
Pile
-9.6 -9.6
-12.0 m -12 .0 m
-11.2 -11.2
-12.8 Excavation bottom -12.8
-3.2 m in test 2
-14.4 -14.4
-1.6 m -3.2 m -12 .0 m in test 2
-16 -4.8 m -6.4 m -16 -1.6 m -3.2 m
-4.8 m -6.4 m
-17.6 -8.0 m -9.6 m -17.6
-8.0 m -9.6 m
-19.2 -10.4 m -12.0 m -19.2 -10.4 m -12.0 m
(b) Bending moment (kN·m) Deflection (mm)
0 0
-300 0 300 600 -20 0 20 40 60
-1.6 -1.6
GS
-3.2 -3.2
Anchor -4.8
-4.8 - +
-6.4 -6.4
Burial depth (m)
Burial depth (m)
-8 -8
Pile
-9.6 -9.6
-12.0 m -12 .0 m
-11.2 -11.2
-12.8 Excavation -12.8 Excavation bottom
bottom -14.4
-14.4
-1.6 m -3.2 m -1.6 m -3.2 m
-16 -4.8 m -6.4 m -16
-4.8 m -6.4 m
-17.6 -8.0 m -9.6 m -17.6 -8.0 m -9.6 m
-19.2 -10.4 m -12.0 m -19.2 -10.4 m -12.0 m
Fig. 7. Moment curves and pile deflection for tests 1 and 6: (a) test 1 and (b) test 6.
Page 53 of 78
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Symmetrical plane
Boundary of crater at ground surface after the collapse
Peak increasing ratio of the max. bending
moment caused by the arching effect
Residual increasing ratio of the max. bending
moment when the collapse finished 1.35 D/2
1.32
1.08 Movements of IFPs:
1.01 P14 P19
GS +0 (mm) P1 P4 P7 Trapdoor
+0 P39
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
1200
-750
S1 S4 S7 S8 S13
0.09 Residual increasing ratio of
0.10
1.12 the max. axial force when
1.01 the collapse finished
1.10 1.01
1.22
Peak increasing ratio of the max. axial
force caused by the arching effect
Test 1 (Partial failure)
Fig. 8. Illustration of the movements of the IFPs that acted as a horizontal trapdoor and
some consequences of the collapse.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
(b)
(a)
Window
Fig. 9. Collapse scenarios: (a) illustration of the slope inside the excavation and (b)
Page 54 of 78
Page 55 of 78
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
S1 S2 S3 S4 S5 S6 S7 S8 S9 S10
-60 P1
P30
30 Anchor
- +
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
Test 0: 2 IFPs
60 Test 1: 4 IFPs bottom
Test 2: 4 IFPs
90 Test 3: 4 IFPs
Test 4: 4 IFPs
120 (Large deformation)
Test 5: 6 IFSs (S5-10)
150 Test 6: 9 IFAs (A6-14)
Test 7: 9 IFAs (A6-14)
180
0 4 8 12 16 20 24 28 32
Pile No.
Fig. 10. Increases in the pile head displacement induced by partial failure in each test.
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 56 of 78
2.0
Pile
failure
1.5 time
1.0
st
1 pile adjacent to partial failure in Test 0
0.5
-5 0 5 10 15 20 25 30 35 40
Time (s)
(b) 3.0
P1 P4 P10 P14 P19
9.6 m
Increase ratio of earth pressure
2.5
Earth pressure cell P16 P18 P20 P23
2.0
P19 P18 P16
P14 P10 P4
1.5
1.0
Pile failure time
0.5
0 5 10 15 20 25 30 35 40
Time (s)
Fig. 11. Change curves of the earth pressures acting on adjacent piles in test 1: (a)
increasing ratios of the earth pressures at a depth of 6.4 m and (b) increasing ratios of
the earth pressures at a depth of 9.6 m.
Page 57 of 78
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
3.0 1.5
P20 P23
GS Strut
Load transfer coefficient (earth pressure)
4 6 8 10 12 14 16 18 20 22 24
Pile No.
Fig. 12. Comparison of the load transfer coefficients (earth pressure and bending
-300
-250
-200
-150
-100
-50
0
50
100
150
-5
Pile
failure
0
P1
(P20-23)
S1
5
S2
10
S3
15
S4
S5
S1
S5
Time (s)
S6
20
S8
S7
S1
S5
S6
S2
S6
25
P19
S7
30
IFPs
S7
S3
S8
S4
35
P24
S2
S8
S4
40
Fig. 13. Change in the axial forces of struts caused by the IFPs in test 1.
Page 58 of 78
Page 59 of 78
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
1.2 P19
st
1.1 1 pile adjacent to the IFPs in test 0
1.0
Pile failure
P20~P23
0.9
-5 0 5 10 15 20 25 30 35 40
Time (s)
Fig. 14. Curves of IP bending moments versus time in test 1.
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 60 of 78
-9.6 -9.6
-12.0 m -12 .0 m
-11.2 -11.2
-12.8 -12.8 Excavation
-14.4 -14.4 bottom
-16 -16
-17.6 -17.6
-19.2 Test 0 -19.2 Test 1
Fig. 15. Moment curves of the first pile adjacent to the IFPs before and after pile failure
in tests 0 and 1: (a) test 0 and (b) test 1.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 61 of 78
1.0
1.2
1.4
1.6
1.8
0
10.4 m
P1
8
st
st
P4
P1
16
P12
P7
Time (s)
P4
24
P16
Bending moment guage P12
P14
P7
32
P18
40
P10
P16 P19 IFPs
P19
P18 P20 P23
Bending
1.2
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
moment
(-10.4 m)
1.1
IFPs
1.0 P19 P20~P23
st
1 pile adjacent to the IFPs
0.9
0 2 4 6 8 10 12 14 16 18 20 22 24
Pile No.
Fig. 17. Comparison of the load transfer coefficients (bending moment) in test 1 and test
2.
Page 63 of 78
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
3.0 2.0
P20 P23 GS Strut
pressure 1.4
Test 4: 4 IFPs (large deformation) (-6.4 m)
Bending
1.5 moment
(-10.4 m) 1.2
IFPs
P19 P20~P23
1.0 st 1.0
1 pile adjacent to the IFPs
0 2 4 6 8 10 12 14 16 18 20 22 24
Pile No.
Fig. 18. Comparison of the load transfer coefficients (earth pressure and bending
moment) in test 1 and tests 3-4.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
2.0
0
P1
8
S1
S2
16
S5
S1
S3
Time (s)
24
S4
S6
S2
S5
32
S6
S7
S3
P19
S7
40
IFPs
S8
S8
S4
P24
48
Fig. 19. Change in the axial forces of struts caused by the IFPs in test 3.
Page 64 of 78
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 65 of 78
0.6
0.8
1.0
1.2
1.4
1.6
-2 -4
6.4 m
P1
0
P4
of IFPs (P20-P23)
Large deformation
2
P7
P10
4
Time (s)
P12
P19
P14
6
P18
8
P10
P18
IFPs
10
P20 P23
P7
P14
12
Fig. 20. Change in the earth pressure at a depth of 6.4 m over time in test 4.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
0.0
0.3
0.6
0.9
1.2
1.5
1.8
2.1
2.4
0 -1
P1
S1
1
S2
2
S3
S4
S8
3
S4
test 4.
S5
4
Time (s)
S3
S7
S6
5
P19
S7
S2
S6
6
S8
7
P24
of IFPs (P20~P23)
Large deformation
S5
Fig. 21. Changes in the axial forces of struts caused by the large deformation of IFPs in
Page 66 of 78
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 67 of 78
0.8
1.0
1.2
1.4
1.6
1.8
-2
0
10.4 m
P1
P14
P1
2
P4
P4
P7
P16
4
Bending moment guage
P10
P7
Time (s)
P12
6
P18
P14
8
Large deformation of IFPs (P20-P23)
P10
P16 P19
P19
10
IFPs
P18 P20 P23
12
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
0.3
0.6
0.9
1.2
1.5
1.8
m
6.4
S7
0
P1
S1
P19
P4
P14
P19
S2
S6
Failure order
40
S3
S5
P10
P18
P10
80
S4
test 5.
S5
P16
S8
Time (s)
P4
120
P16
S6
S9
S7
160
S8
S10
P10
P14
200
S9
S10
P29
P4
P18
240
Fig. 23. Change in the earth pressure at a depth of 6.4 m versus the strut failure order in
Page 68 of 78
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 69 of 78
-800
-400
0
400
800
1200
P1
S7
0
S1
S8
S7
S2
S7
S4
S1
S6
40
Failure order
S3
S6
S5
S8
S2
S5
S4
80
S5
S5
S8
Time (s)
S3
S6
S6
120
S7
S9
160
S8
S4
S10
S9
200
S3
S10
S2 S1
P30
240
Fig. 24. Change in the axial forces versus the strut failure order in test 5.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.6
m
1.4 10.4
S7
0
Strut
S1
failure
P4
P1 P4
P12
P18
S6
S2
Failure order
40
P7
S3
P7
S5
P19
P16
80
P10
S4
P14
P10
S8
Time (s)
P26
P14
120
S6
S9
S7
P18 P20
160
S8
S10
200
S9
P26 P29
S10
240
Fig. 25. Change in IP bending moments versus the strut failure order in test 5.
Page 70 of 78
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 71 of 78
0.25
0.50
0.75
1.00
1.25
1.50
0
P1
A1
2
S1
S2
4
H: excavation depth
S3
6
H: 14.4 m
H: 12.0 m
H: 12.0 m
S4
A6
8
Test 7: 9IFAs (A6-A14),
Test 6: 9IFAs (A6-A14),
S5
Pile No.
10
S6
12
S7
14
S8
16
S9
A14
18
S10
P30
20
Fig. 26. Comparison of the load transfer coefficients (bending moments) in tests 5-7.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
-19.2
-16.0
-12.8
-9.6
-6.4
-3.2
0.0
-600
-
Outside
-400
+
GS Strut
S9
S8
S5
S6
S7
S10
bottom
-200
Depth 12m
order
Failure
0
200
Bending moment (kN·m)
400
600
bottom
Inside
Excavation
-12.0 m
800
Fig. 27. Profiles of the bending moment of IP P26 for each strut failure in test 5.
Page 72 of 78
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 73 of 78
Anchor
Excavation
Pile
A
B
C
1
A
C
F
Fig. 28. Stress analysis diagram of the support pile isolation body.
xm
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 74 of 78
-400
-300
-200
-100
0
200
300
P1
0
A10
A1 A2
40
A7
A1
A12
A9
100 Anchor failure
Failure order
A4
80
A8
A2
A10 A9 A11
A11
A13
A6
120
A8
A8
A9
A3
A8
160
A14
A12
A12
Time (s)
A4
A10
200
A10
A7
A7
A15
240
A5
A12
A13
A11
280
A6
A16
A13 A6
A6
A14
320
A14
A14
A16
P32
360
Fig. 30. Changes in the anchor forces versus the anchor failure order in test 6.
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 76 of 78
S5 - S7
2.0 S5 - S8
S5 - S9
1.5 S5 - S10
1.0
0.5
0 1 2 3 4 5 6 7 8 9
Strut No.
P1 P38
A1 A6 A8 A10 A12 A14 A19
2.00
Load transfer coefficient (anchor axial force)
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
0
m
10.4
A10
P4
40
P18
P12
A1 A2
A9
P1 P4
Failure order
Anchor failure
80
A4
A11
P7
P16
P19
120
P10
A8
A6
160
P10
P14
P14
Time (s)
A12
A8
200
A7
P19
A10
240
A13
A12
280
A6
320
A14
A14 A15
360
Fig. 32. Changes in IP bending moments versus the anchor failure order in test 6.
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 78 of 78
0.0
Failure
order Depth
-3.2 Depth 14.4 m
12.0 m
A10
Burial depth (m) -6.4 A9
A11
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by UPPSALA UNIVERSITET on 06/19/20
A8
-9.6 A12
A7
A13 -12.0 m
-12.8 A6
A14 -14.4 m
Excavation
bottom
-16.0 Excavation
bottom
-19.2
-300 0 300 600 0 300 600 900
Bending moment (kN·m) Bending moment (kN·m)
(a) Test 6 (b) Test 7
Fig. 33 Profiles of the bending moment of IP P19 versus anchor failure in tests 6-7.