Effects of tunnel construction on nearby pile foundations 2

CHAPTER 1


INTRODUCTION





1.1 Background

Pile foundation subjected to passive loading from soil movement has been an important issue in
recent years especially with concern towards its performance. Such soil movement can be caused
by the construction of embankment (Stewart, 1992; Ellis & Springman, 2001), deep excavation
(Poulos & Chen, 1997; Goh et al., 2003), tunnelling (Loganathan, 1999; Jacobsz et al., 2002) or
slope movement (Fukuoka, 1977; Chen & Martin, 2002). In this thesis, the effects of tunnel
construction on nearby pile foundation are investigated.

The increased demand for underground systems in urban areas particularly for mass transportation
can be seen in big cities like Singapore, Hong Kong and London. This has led to many tunnels
being constructed in proximity to structures. Extensive research has been carried out in the United
Kingdom particularly on the Jubilee Line Extension (Burland et al., 2002) on the effects of
tunnelling on nearby structures. However, most of the structures in the studies are supported on
shallow foundations and very little work has been carried out on structures supported on pile
foundations. This is due to the fact that most structures were built long before the tunnels are
planned. As a result, instrumentation inside existing pile foundation is difficult to install for
further investigation.

1
1.1.1 Tunnel construction near pile foundation

Among all the passive loading caused by soil movement, the one caused by tunnelling is perhaps
by far the most complicated. This is due to the complex tunnelling processes particularly for shield
tunnelling, which comprises shield machine advancement, application of face pressure, tail void
grouting and lining installation. These caused significant disturbance to the surrounding soil with
shearing (during shield advancement), loading (application of face pressure and grout pressure)
and unloading (soil stress relief) mechanism. Since the structures usually exist in an urban
environment long before a tunnel is planned, engineers only have the choice of aligning tunnel
position relative to the nearby pile foundation. Figure 1.1 shows a schematic illustration of two
typical situations being encountered in practice: (a) tunnelling under pile foundation and (b)
tunnelling adjacent to pile foundation.

During tunnelling, stress relief will occur in the surrounding soil. When the tunnel is constructed
under a pile, it is likely that the pile base resistance will be first reduced and in turn leads to pile
settlement. To maintain equilibrium of load, the base resistance is transferred to the pile shaft. If
the tunnel stress relief is great enough to fully mobilise the pile shaft resistance, a larger pile
settlement would be anticipated. However, pile lateral response is unlikely to be of significance in
this situation.

In the second situation where the tunnel is constructed adjacent to a pile foundation, a different
mechanism is observed. The stress relief due to tunnelling would cause soil settlement above
tunnel. This in turn causes negative skin friction (NSF) to act along the pile shaft above the tunnel
level. In order for the force to be in equilibrium, the pile shaft below tunnel level (which is not
subjected to settlement) would support the dragload from NSF above the tunnel level. Only when
the positive shaft resistance and pile base are fully mobilised, settlement would become a problem.
2
This depends on the availability of pile length extension below tunnel level. It is also to be
checked that the dragload would not cause the pile to overstress. In addition, the lateral pile
response can be significant since horizontal soil movement is largest near tunnel.

Besides the two possible situations, other possibilities of tunnel alignments near pile foundation

were identified from reported case histories and can be grouped into four different categories as
shown in Figure 1.2. The categories are based on the relative position of tunnel to pile foundation.
The pile responses due to tunnelling in each category requires further investigation owing to the
varying observations by previous researchers as presented in Chapter 2.

The ignorance of not taking into account the additional loading caused by soil movement as
mentioned above could lead to excessive pile settlement or pile structural capacity being exceeded.
Under-designed pile foundations will be reflected on the superstructure such as cracks on beam,
column or wall and ultimately collapse if the damage is large. Aside from potential damage, poor
understanding of the mechanism will also lead to an expensive protective or mitigation work.
Therefore, further studies would be required to develop a better understanding of the problem and
contribute to an economic design.

1.1.2 Current design and construction approach

The current codes of practice do not provide guidance and basis for the design of piles subjected to
soil movement caused by tunnelling. The design requirements are usually stipulated by the local
authorities. To date, there are a few design approaches available to analyse the problem. Figure 1.3
summarises the approaches obtained from literature review. In the direct method, pile responses
are computed from some analytical and numerical analyses. These include finite element analysis,
boundary element analysis and soil-spring analysis. Besides, design charts are also available. In
3
the indirect method, pile responses are not computed. For example, the ‘risk of damage to
building’ assessment method categorises the overall building damage, whereas methods used by
Jacobsz et al. (2005) assumed the pile to response similarly to greenfield soil movement.
Furthermore, Nakajima et al. (1992) and Inose et al. (1992) check the factor of safety of pile
bearing capacity by hand calculation. More details of the methods mentioned are described in
Chapter 2.

Despite the design and assessment carried out, some local authorities also impose restrictions on

the tunnelling activities near critical structures. In Singapore, the Land Transport Authority (LTA)
imposes a criterion for the design of deep foundation to allow for an additional movement of
15mm in the short term and 25mm in the long term to account for future development (LTA,
2002). In addition, LTA defines an area of first reserve which is typically 6m away from the
tunnel extrados. Within this area, the activities should be dealt with care. In Japan, the provision of
tunnelling work adjacent to foundation can be classified into three zones (Figure 1.4) as according
to Fujita (1989). If the tunnel lies within Zone 1, there is no work restriction. In Zone 2, the
tunnelling work shall proceed with prudently selected methods and techniques. In Zone 3,
auxiliary measures and prediction should be carried out. Similar classification was also adopted by
Moroto et al. (1995).


1.2 Objectives of the study

With the current understanding of the effects of tunnelling near pile foundation based on the
limited available case histories, laboratory tests and numerical studies, the design of tunnel near
pile foundation is yet to be understood. Despite the increasing demand for underground systems,
the constraint in the congested area inevitably creates pressure for tunnelling engineers to make
4
sure that the construction of tunnel does not cause detrimental effects to adjacent pile foundations.
The thought and potentially wide scope have naturally stimulated the eagerness to undertake the
current study. The aim of the thesis is to develop a better understanding of the effects of tunnel
construction on nearby piles and hence contribute to the design and analysis of such piles. This
aim is achieved through the study using various methods as outlined below.

As mentioned above, placing instrumentation inside existing pile foundation is not feasible, which
restricts further understanding of pile responses caused by tunnelling. Therefore, it is one of the
objectives of this research to carry out field study with in-pile instrumentation. A Mass Rapid
Transit (MRT) line constructed in Singapore, North-East Line (NEL) C704 was identified for this
study. In the NEL C704, twelve working piles were instrumented with extensive strain gauges

which allow the bending moment and axial force developed during tunnelling to be monitored.
The case study will add to the existing knowledge of field monitoring of pile responses during
actual tunnelling.

The case study was only limited to specific range of tunnel-pile configurations, tunnelling process
and soil type. Further understanding of the problem outside the range is therefore unknown. To
overcome the limitation, three-dimensional (3-D) finite element (FE) studies were performed. The
FE model allows varying parameters and configurations to be studied. In order to do a general
study, a reliable finite element model has to be established first. Due to the extensive monitoring
data of the NEL C704, a 3-D finite element model was set-up to back-analyse the case study. The
aim of the back-analysis is also to verify the reliability of the field monitoring data. Subsequently,
a simplified 3-D model was used for parametric studies to identify the critical and non-critical
aspects in the analysis of pile responses due to tunnelling.

5
Although 3-D finite element model is a powerful tool for analysing such a complex interaction
problem, 3-D software is usually not adopted in practice for design due to the limitation of
computational resources available and also the time consuming effort. Therefore, the problem is
usually idealised in plane strain condition in 2-D analysis. The plane strain idealisation is just an
approximation to the actual 3-D nature of the problem. The idealisation techniques available are
not well understood and could be misused since it is difficult to select adequate parameters to
represent the 3-D effect. Besides, there are very few studies being carried out on 2-D idealisation
of tunnel-pile interaction. Therefore, the current research would also provide greater insight into
the use of 2-D finite element model in analysing the problem.


1.3 Organisation of thesis

Firstly, a literature review on the current research topic is presented in Chapter 2. The review
consists of past observations from case histories, laboratory tests and field studies. This is

followed by a review on the various prediction and design methods available in practice. The
current understanding and outstanding issues were also discussed.

Chapter 3 presents a unique case history in Singapore on the monitoring of the effects of tunnel
construction adjacent to full-scale working piles. The field instrumentation data were studied and
presented together with the project overview, ground condition and construction details. Also
presented is the initial analysis carried out using existing design charts.

Chapter 4 presents the development of a three-dimensional finite element model to back-analyse
the case history reported in Chapter 3. Full tunnel construction process was simulated. The
analysis results for single tunnel advancement, twin tunnel advancement, pile response to single
6
tunnel and twin tunnels were discussed. Sensitivity studies were also carried out, investigating
various factors affecting the 3-D model results.

Chapter 5 gives a general overview on some of the considerations required in finite element
analysis of piles due to tunnelling. Parametric studies were carried out to cover a wider range of
tunnel-pile configurations beyond the case study examined in Chapter 4.

Chapter 6 presents the study of tunnel-pile interaction problem using plane strain idealisation
technique. Some of the techniques available for modelling pile foundation and tunnelling were
reviewed followed by thorough comparative study between 2-D and 3-D models. Sensitivity
studies were carried out and calibration charts were produced to serve as a guide for analysing the
tunnel-pile problem in plane strain model. Three case studies were analysed using 2-D FE model,
engaging the technique and charts presented earlier.

Chapter 7 lists the summary of findings and conclusions from the research work carried out in this
thesis. Recommendations for future research areas are identified.

7
CHAPTER 2


LITERATURE REVIEW





2.1 Introduction

Research on the effects of tunnel construction on nearby pile foundation began since the 1970s
notably by Morton & King (1979). Since then, research on this topic has been discontinued all the
way until the 90s. More recently, the number of researchers carrying out such study has increase
tremendously owing to the demand such as the construction of Channel Tunnel Rail Link 2
(CTRL2) in UK, Circle Line (CCL) in Singapore and North-South Line (NSL) in Amsterdam. In
these projects, tunnel alignment was designed inevitably close to pile foundation and some even
have to cut through piles. Physical observations from case histories as well as laboratory tests and
full scale pile tests are presented in Section 2.2.

Section 2.3 discusses the design and prediction methods that have been used by engineers to

analyse the problem. The advantages and shortcomings of each method are highlighted. Based on
the limited publications available, the current understanding and outstanding issues are identified
and discussed in Section 2.4.



8
2.2 Pile responses caused by tunnelling: Physical observations

2.2.1 Case histories

Engineers have reported evidence of case histories where tunnels were constructed close to pile
foundation. Table 2.1 summarises some of the case histories and its details including the distance
between tunnel axis and pile centre (X
pile
), tunnel depth (H
tun
), tunnel diameter (D
tun
), pile diameter
(D
pile
), pile length (L
p
) and volume loss (V
L
).

In Singapore, construction of the MRT North-East Line (NEL) C704 was a unique case for
studying the responses of pile subjected to tunnelling (Coutts & Wang, 2000). As part of the

contract C704, a viaduct bridge was planned in conjunction with the tunnel advancement. The
bridge which consists of 2 abutments and 39 piers was constructed in parallel alignment with the
new twin tunnels configuration (Figure 2.1a). The piers were supported by groups of four to six
1.2m and 1.8m diameter bored piles. Along the alignment, 6.5m diameter tunnels were located
very closely with 1.6m clear distance to the pile foundation. The piles were founded at relatively
greater depth to tunnel level with L
p
/H
tun
ratio of 3 (i.e. long pile). A total of twelve piles were
installed with strain gauges at various levels where the axial force and bending moment were
obtained. Only a summary of the monitoring results for six piles were reported. Substantial
dragload and bending moment were observed in the piles, as high as 91% and 59% of the design
working load and bending moment respectively. Unfortunately, many important details such as
tunnel volume loss, pile-tunnel configurations, construction sequence and soil data were not
mentioned by the authors, which lead to limited interpretation. Despite that, further collection of
data from the contractor and the authority was carried out as part of the works in this thesis and
formed the study as presented in Chapter 3.

9
Tham & Deutscher (2000) reported another stretch of the MRT NEL tunnel (Contract C705)
passing by a 4-storey workers’ quarters supported on 0.45m diameter bored piles with tunnel-pile
clear spacing of approximately 1.85m (Figure 2.1b). The tunnel was located at a depth of
14m.b.g.l. which correspond to the same length of the piles (i.e. L
p
/H
tun
of 1). Comprehensive
monitoring by building settlement markers showed that the building did not suffer detrimental
effect. Settlement of up to 7mm was measured at the edge of the building. The success of the

operation was due to the good tunnelling control where volume loss was only 0.4%.

In Hong Kong, construction of the 7.9m diameter twin tunnels for Mass Transit Railway (MTR)
Island Line posed the same concern to adjacent piled-building which was located approximately
3m to the tunnel extrados (Forth & Thorley, 1996). The 31-storey building was supported on 2m
diameter bored pile with pile group consisting of up to 4 piles. The L
p
/H
tun
ratio was between 1.6
and 2.5 which were considered as long pile (Figure 2.2). Again, the performance of piles could
only be judged indirectly from building movement at ground surface. Settlement of only 5mm was
observed on the near side of the building.

In London, Mair (1993) and Lee et al. (1994) reported the construction of a hand-dug escalator
tunnel at the Angel Underground Station where the tunnel was constructed very close to pile
foundation (i.e. 1m clear distance between tunnel and pile). The new 7-storey building was
supported by 1.2m diameter under-reamed bored piles (Figure 2.3) which were installed below the
tunnel depth hence, a long pile condition. These piles were de-bonded down to 4m above the pile
tip to reduce negative skin friction above the tunnel level. Owing to the early planning and co-
operation between tunnel owner and building developer, both in-ground and in-pile
instrumentations were possible prior to tunnelling. Measurements of the in-pile inclinometers
showed that the nearest pile was only subjected to maximum lateral deflection of 8mm for volume
loss up to 2%. Besides, both the in-ground and in-pile inclinometers results were very similar. The
10
authors concluded that tunnel could be constructed very close to pile foundations in London Clay
and would only cause small horizontal deflection.

Powderham et al. (1999) reported construction for the Jubilee Line Extension (JLE) station
beneath piled-supported buildings. Figure 2.4a shows sectional view of the tunnels and buildings.

Generally, the tunnels were driven below the 20m long bored piles. Prior to tunnelling, permeation
grouting was carried out around the tunnels and one of the buildings was underpinned with
concrete raft. During tunnelling, compensation grouting was applied to minimise settlement.
Maximum settlement of only 18mm was observed after the two tunnels were driven. The risk
management approach used was able to limit the damage of buildings to acceptable levels. For the
same tunnels, Selemetas et al. (2002) reported the settlement of New London Bridge House
(NLBH) which is adjacent to the buildings mentioned by Powderham et al. (!999). The 1.4m
diameter under-reamed pile supporting a corner of the NLBH was located 2m above the tunnels
but not directly under the NLBH (Figure 2.4b). Monitoring of NLBH showed a settlement of up to
24mm after the tunnels were constructed. The authors concluded that no adverse effects were
encountered on the piled-structure.

Recent tunnelling works for the CTRL2 in London also exposed piled-structures to potential
damage. Jacobsz et al. (2005) reported the monitoring of three piled-bridge foundations with one
on end bearing piles (Figure 2.5a) and two on friction piles (Figures 2.5b and c). The authors
recommended that re-assessment of pile capacity to be carried out as large factor of safety can
often be found in piles and redistribution of loading is possible.

In Japan, construction of a tunnel near a highway bridge posed the same concern (Moroto et al.,
1995). Piles supporting the bridge piers are 15.5m away from the tunnel axis (Figure 2.6). No
settlement or tilt was observed at the piers while maximum surface settlement of 6mm was
11
measured. For the construction of Nanboku Line in Tokyo, Nakajima et al. (1992) and Inose et al.
(1992) described the planning of tunnels below piled bridges (Figure 2.7) and a large piles
supported dome stadium (Figure 2.8) respectively. Pile settlement and bearing capacity were of
major concern during tunnel construction due to its relative position directly below tunnels. In
similar condition, Ikeda et al. (1996) made use of compensation grouting between tunnels and pile
foundations to restrict settlement of piles. More recently, Takahashi et al. (2004) reported the
construction of Rinkai Line in Tokyo where twin tunnels were driven underneath pile foundations
with minimum clearance of 3.4m. Figure 2.9 shows the relative location of tunnels and pile

foundations. With the use of mitigation measure such as grout injection and good control of tunnel
construction which limit the volume loss to 0.5%, settlement of bridge pier supported by the piles
was monitored to be less than 4mm.

2.2.2 Laboratory and centrifuge tests

As described in Section 2.2.1, only two out of the many case histories were installed with in-pile
instrumentation. The pile responses from the remaining case histories could only be examined
indirectly from the monitoring of the structures. Therefore, some researchers resorted to
experimental simulation where pile can be instrumented to obtain valuable data.

One of the earliest studies on pile responses due to tunnelling was reported by Morton & King
(1979). In their study, four 1-g static model tests were carried out in medium dense to dense dry
sand. The tunnel was excavated with jacking liner tube by concurrently rotating a close fitting full-
face head. Wooden piles with sustained working load were located at three levels above the tunnel
crown (Figure 2.10) in separate tests. Despite the violation of scaling law, the authors have
concluded that the settlement of friction pile could be large as the pile tip is closer to the tunnel.
Besides, the prime factor of pile failure is due to dilatancy of soil within a zone above the tunnel.
12
The model indicated that even small dilatancy can induced immediate failure of the piles located
within the critical zone. However, the results were questionable due to the 1-g representation of
real problem where the low confining stress in sand (which is stress-dependent) caused an abrupt
failure in pile without gradual increase in settlement (Mair, 1979).

With the availability of centrifuge modelling technique, a prototype problem can be simulated in a
scaled laboratory model hence overcoming the limitation of 1-g model. Some of these centrifuge
tests on tunnel-pile interaction have been carried out in clay overlying dense sand (Bezuijen &
Schrier, 1994; Hergarden et al., 1996), stiff clay (Loganathan, 1999), soft clay (Ran et al., 2003),
dense dry sand (Jacobsz et al., 2002; Feng et al., 2002) and dense saturated sand (Lee & Chiang,
2004). Table 2.2 summarises all the centrifuge tests and its details. It should be noted that all the

tests considered a plane strain tunnel in the simulation. Besides, most of the tests were simulating
volume loss as small as 1% and up to 20% to investigate the pile responses over more detrimental
tunnelling situation.

The three tests carried out by Bezuijen & Schrier (1994) (see also Hergarden et al., 1996)
simulated a prototype tunnel of 7m diameter at a depth of 14.5m, 18m and 23m below ground
surface. Six end-bearing piles of 0.4m diameter were installed to a depth of 18m below ground
surface (Figure 2.11). The six piles corresponded to four different tunnel-pile distance (X
pile
) of
4.9m, 6.5m, 9.7m and 12.9m. The authors concluded that pile settlement can be significant when
volume loss is 1% or more and the clear distance between tunnel and pile is less than one tunnel
diameter. Furthermore, settlement and bearing capacity of pile is not affected at all when the clear
distance is more than two times the tunnel diameter.

Loganathan (1999) carried out three centrifuge tests which have quite similar tunnel-pile relative
vertical position except that the clear distance between tunnel and pile was fixed at 2.1m. Besides,
13
both single pile and 2x2 pile group were investigated in stiff clay (see Figure F.1). In the tests,
both the axial and bending responses of the piles were measured. The author concluded that
bending moment (BM) and lateral deflection of a pile is critical when tunnel springline is located
at or near the pile base. The axial force is found to be critical when the tunnel springline is below
the pile base. Besides, the BM and lateral deflection of both single pile and pile in the group at the
same distance from tunnel are almost identical.

Studies by Jacobsz et al. (2002) focused on the investigation of settlement and load distribution of
single driven pile in homogeneous dense sand. All the tests were carried out with the pile base
above the tunnel. The zone of influence to determine the pile’s critical position was identified as
shown in Figure 2.12 (shaded area). As volume loss increases, load transfer from pile base to shaft
changes gradually and once full mobilisation of shaft resistance has been achieved, large

settlement occurs. Pile base located within the zone would be subjected to large settlement if
volume loss increases beyond 1.5%.

Lee & Chiang (2004) also simulated the problem in sand but concentrated on pile base located
outside the zone of influence. A total of twelve tests were carried out on single piles with varying
pile length and tunnel depth. From the observation of unit skin friction development during
volume loss, load transfer mechanism was identified for long pile (i.e. L
p
/H
tun
> 1.0) and mid-
length pile (i.e. L
p
/H
tun
= 1.0) as shown in Figure 2.13.

Preliminary centrifuge tests were also carried out in National University of Singapore by Feng et
al. (2002) and Ran et al. (2003). However, the test set up uncovered a problem in simulating actual
tunnel stress relief. Instead, an ovalisation shape of tunnel deformation was observed which
caused unusual soil movement around tunnel and could not be verified further.
14
2.2.3 Full scale pile tests

Despite the usefulness of laboratory tests as mentioned above, the simplification (i.e. plane strain
tunnel) and boundary problems (i.e. boundary effect and drainage) of such tests raised
uncertainties in the actual pile responses.

In view of the construction for North/South Line in Amsterdam where tunnels have to be driven
under more than 250,000 wooden and 2000 concrete piles, full scale pilot test was first carried out.

A total of 43 timber piles and 20 concrete piles were purposely installed and loaded at the Second
Heinenoord tunnel site (Teunissen & Hutteman, 1998; Kaalberg et al., 2005). Figure 2.14a shows
the piles and tunnel layout. Pile and soil settlement were measured and the results were concluded
in the zone of influence as shown in Figure 2.14b. Basically, pile settlement would be greater than
soil surface settlement if the pile base is founded in Zone A. If the pile base is in Zone B, both the
pile and soil surface settlement would be equal. Where as if it is in Zone C, pile settlement would
be significantly lesser than the soil surface settlement.

Selemetas et al. (2005) reported another full scale trial pile test for CTRL Contract 250 in the U.K.
Four purposely installed piles (2 end bearing and 2 friction piles) were instrumented and loaded
with kentledge. These 0.48m diameter piles of 13m length were located strategically to investigate
the zone of influence as categorised by Jacobsz et al. (2002). Figure 2.15 shows the categorisation
of influence zone from the field data. The categorisation is slightly different from Jacobsz et al.
(2002) and Kaalberg et al. (2005) mainly by the angle of zone.



15
2.3 Pile responses caused by tunnelling: Prediction and design methods

Given the complexity of pile responses due to tunnelling, prediction and design would not be
straightforward. For simplicity, empirical methods are adopted (Tham & Deutscher, 2000;
Nakajima et al., 1992; Jacobsz et al., 2005). However in recent years, various methods have been
used to analyse the problem such as the 2-D finite element method (Vermeer & Bonnier, 1991;
Lee et al., 1994), 3-D finite element method (Mroueh & Shahrour, 2002; Cheng et al., 2004; Lee
& Ng, 2005), numerical & analytical methods (Broms & Pandey, 1987; Chen et al., 1999;
Sawatparnich & Kulhawy, 2004; Kitiyodom et al., 2004) and design charts (Chen et al, 1999).
Table 2.3 summarises some of the reported studies and their details.

2.3.1 Empirical method

In the tunnelling near Woodleigh Workers’ Quarters (Figure 2.1b) as described in Section 2.2.1,
the pile responses were not assessed. Instead, an indirect approach was adopted to assess the risk
of damage to building according to Mair et al. (1996). In this method, the tensile strain is
calculated and a greenfield soil settlement is assumed to imposed on the building. However, it
should be noted that the method was originally developed for tunnelling near building founded on
shallow foundation. The use in piled-structure is therefore not justifiable.

Jacobsz et al. (2005) described the empirical methods used to assess pile foundation supporting
three bridges (Figure 2.5) near the CTRL2 tunnels. The methods used to assess pile settlement and
pile overstress are summarised in Figures 2.16a and b respectively. A few assumptions were made
in the assessment based on observations from centrifuge tests by Jacobsz et al. (2005) and field
16
study by Selemetas et al. (2005). The method is applicable to pile with its base located in the zone
of influence. Besides, only pile axial response can be assessed.

Nakajima et al. (1992) reported some of the methods used to assess pile bearing capacity during
shield advancement and due to tail void grouting. The procedure is summarised in Figures 2.16c
and d respectively. Besides, Inose et al. (1992) adopted an alternative approach to assess the pile
bearing capacity based on an imaginative cone around the pile base (see Figure 2.16e). These
methods can only be used for pile base located above tunnel.

2.3.2 Finite element method

The empirical methods as described above can only be used to assess pile response individually.
There is a need for unified approach to analyse both the pile axial and lateral responses
simultaneously such as finite element (FE) method.

Vermeer & Bonnier (1991) carried out 2-D FE analyses using PLAXIS to simulate a typical pile
settlement problem due to tunnelling in Amsterdam. A row of piles were modelled as sheet pile

wall using line elements whereas the pile-soil interface was modelled with joint elements. The
method used to reduce the pile stiffness and skin friction was described. However, the accuracy of
the method is unknown and not investigated. It was concluded that the pile settlement followed
exactly the settlement of bearing layer where the pile base were founded.

Lee et al. (1994) made a prediction for the Angel Underground Development as described in
Section 2.2.1 using 2-D FE program “OASYS SAFE”. Linear elastic soil model and undrained
condition were assumed. The piles subjected to tunnelling were not modelled directly, but were
17
assumed to act as slender members that deform with soil. The authors claimed that the FE
prediction of pile lateral deflection is similar to the measured and provided an upper bound value.

Owing to the limitations of 2-D analysis as described, 3-D FE analyses were performed. Mroueh
& Shahrour (2002) carried out simulation of 3-D shield tunnel advancement on adjacent pile
foundation. Instead of the full procedure for shield advancement, a simplified stress release zone
was simulated to represent the effect due to shield, over-cut and tail void grouting. No physical
data was back-analysed and a reference case was set up for parametric studies. The results show
that tunnelling induced significant deflection and axial force in piles depending on the location of
pile base. Positive effect with reduction of axial force in rear piles was observed. However, there
was no significant reduction in bending moment. Besides, the presence of pile cap only affects the
upper portion of the piles and analysis can be performed assuming free head for pile group.

Lee & Ng (2005) also carried out a 3-D FE analysis where the responses of single pile due to an
open face tunnel (i.e. no face pressure and unsupported length of 3m in front of tunnel) was
studied. The model was developed to mimic one of the centrifuge tests by Loganathan (1999)
where the geometry and tunnel-pile configuration were same. However, the ground was modelled
as London Clay which does not give any basis for comparison. Besides, the simplified model does
not represent a shield tunnel nor a typical NATM tunnel and certainly not a plane strain tunnel as
modelled in the actual centrifuge test. Despite that, some importance observations can be drawn
from the analysis. Firstly, the pile factor of safety was reduced from 3.0 to 1.5 after the tunnel had

advanced past the pile (Figure 2.17). Secondly, a significant zone of one tunnel diameter ahead
and behind the tunnel was identified where pile settlement was greater than soil surface settlement.
Thirdly, the transverse bending moment was three times larger than the longitudinal bending
moment. Finally, the induced axial force and bending moment were not significant.
18
Cheng et al. (2004) developed a Displacement Control Method to simulate tunnelling adjacent to
single pile in 3-D FE model. In the model, plane strain tunnel was simulated instead of tunnel
advancement. Parametric studies were carried out and revealed that bending moment is negligible
when X
pile
/D
tun
> 2. Furthermore, pile cracked moment is exceeded when X
pile
/D
tun
< 1. The axial
force was found to be dependent on pile base position, soil stiffness and volume loss. The field
data from Coutts & Wang (2000) was back-analysed and good agreement of bending moment and
axial force between measurement and FE analysis was obtained.

2.3.3 Numerical and analytical methods

Broms & Pandey (1987) reported one of the earliest numerical studies dedicated to pile response
due to tunnelling. A 2-D FE model was adopted. Generally, the FE model was first used to
compute the greenfield lateral soil movement at the pile position. Subsequently, the lateral soil
deflection was converted to an equivalent lateral load and was imposed on the beam on elastic
foundation. The beam was assumed to be supported on a series of Winkler springs. In the
approach, the pile was considered to be infinitely stiff and pile-soil-tunnel interaction was not
taken into account. A series of design charts were prepared for the assessment of pile bending

moment. Maximum bending moment was observed near pile cap whereas the bending moment
was small at tunnel springline level.

Chen et al. (1999) also carried out a similar analysis as Broms & Pandey (1987). The lateral and
axial responses of a single pile were analysed in two-stage approach. In the first stage, vertical and
lateral greenfield soil movements were computed from an analytical method. In the second stage,
the computed soil movements were imposed on boundary element analyses (i.e. PIES and
PALLAS) to compute the pile responses. The authors noted that the lateral and axial pile
19
responses computed separately would lead to a lower predicted bending moment. Parametric
studies were also carried out and found that pile responses are influenced by tunnel-pile geometry,
volume loss, soil stiffness and strength. The same study was extended to investigate the influence
of pile head condition in Chen & Poulos (1997). Pile head condition has a significant effect on
bending moment in pile especially when the pile head is fixed from translation and restrained from
rotation. Besides, a rigorous 3-D boundary element program GEPAN was further developed to
couple the pile lateral and axial responses. The above studies were then extended to pile group and
was reported in Loganathan et al. (2001).

Surjadinata et al. (2005) continued to adopt the exactly same procedure as described by Chen et al.
(1999). However, in the current two-stage approach, 3-D FE analysis was used to compute the
greenfield soil movements instead of using analytical method.

Kitiyodom et al. (2004) developed their own numerical program called PRAB to model the effect
of tunnelling on single pile, pile group and pile raft. The approach used was similar to Loganathan
et al. (2001), i.e. 2-stage approach. In the program, piles, soil and raft were modelled as elastic
beam, springs and thin plate respectively. The program was verified by comparison of results from
GEPAN (Loganathan et al., 2001) and FLAC (Matsumoto et al., 2005). Parametric studies were
also carried out and the authors concluded that single pile analysis can be used to represent the
results of piles in pile group (i.e. bending moment, lateral deflection and settlement) except pile
axial force. The author also confirmed that this is only true for large pile slenderness ratio.

Lastly, Sawatparnich & Kulhawy (2004) presented the similar two-stage approach used by Broms
& Pandey (1987). The greenfield soil lateral deflection was first computed using analytical method
by Loganathan & Poulos (1998). Then the soil lateral deflection was imposed on the pile modelled
as beam element and soil-pile interaction modelled as springs. However, a more complex
20
hyperbolic force-displacement curve was developed to represent the springs. The authors carried
out a set of parametric studies on pile bending moments and concluded that L
p
/H
tun
and tunnel
radius have great influence on the results.

2.3.4 Design charts

From the two-stage approach mentioned above, Chen et al. (1999) carried out a thorough
parametric study and developed a set of design charts for use to estimate the maximum pile
responses due to tunnelling such as the induced bending moment, axial forces, settlement and
lateral deflection. Corrections for soil undrained shear strength, pile diameter, pile length to tunnel
depth ratio were considered.


2.4 Current understanding and outstanding issues

Research works carried out to date had shown some similarities and variations in their results. The
following sections present consecutively the comparison of all the reported field observations,
results of pile settlement, axial force, lateral deflection, bending moment and pile group effect.
Outstanding issues were identified and discussed.

2.4.1 General

The case histories with various tunnel-pile configurations as shown in Section 2.2.1 have indicated
all possible situations encountered in practice. As noted, only two case histories were reported
with in-pile instrumentation, i.e. Mair (1993) and Coutts & Wang (2000). These data are
21
invaluable and serve as references to others. All the cases showed some common observations as
follow:-

• Tunnelling with well controlled volume loss (typically 0.5% to 1%) caused no significant
effect on the piled-structure particularly the settlement.
• In some cases, measures such as compensation grouting (Ikeda et al., 1996; Takahashi et
al., 2004), mass concrete underpinning (Powderham et al., 1999), pile shaft grouting
(Jacobsz et al., 2005) and slip coating on pile (Mair, 1993) were adopted to minimise the
effects on piles or directly on structures.
• Other similar cases reported by Coutts & Wang (2000), Tham & Deutscher (2000) and
Forth & Thorley (1996) showed that piles were able to withstand the tunnelling induced
movement without any mitigation measure and no detrimental effect was observed.

The varying opinions on similar problems have raised questions on the confidence of tunnelling
near to pile foundations and the necessity for expensive mitigation works. More field monitoring
data are therefore required.

2.4.2 Pile settlement

Centrifuge tests by Loganathan (1999) showed that the largest pile settlement occurs when L
p
/H
tun

is equal to 1.0 and smallest when L
p
/H
tun
is greater than 1.0. Interestingly, settlement for L
p
/H
tun
<
1.0 is less than settlement for L
p
/H
tun
= 1.0 (Figure 2.18a). Centrifuge tests by Bezuijen & Shrier
(1994) and Hergarden et al. (1996) also observed similar trend. However, in Hergarden et al.
(1996), the trend reversed when X
pile
/D
tun
is increased. This is believed to be due to the change of
pile base position relative to the zone of influence. Besides, it is found that pile settlement is
significant if the pile-tunnel distance normalised with the tunnel diameter, X
pile
/D
tun
is less than 1.0
22
and volume loss is equal or more than 1%. For pile with L
p
/H

tun
< 1.0, Jacobsz et al. (2002),
Kaalberg et al. (2005) and Selemetas et al. (2005) have defined their own zone of influence as
shown in Figures 2.12, 2.14 and 2.15 respectively. The defined zones are similar. Furthermore,
Cheng et al. (2004) showed in the 3-D FE model that pile head settlement is in a decreasing trend
when L
p
/H
tun
is increased (see Figure 2.18a).

2.4.3 Pile axial force

Loganathan (1999) in the centrifuge tests shows that pile axial force is maximum at tunnel
springline level when L
p
/H
tun
is greater than 1 and at pile base level when L
p
/H
tun
is equal to or less
than 1. This trend agrees with those reported in numerical analyses where the axial force is
increasing with depth due to the downdrag caused by tunnelling. For L
p
/H
tun
greater than 1, the
pile length below tunnel springline would be subjected to positive skin friction to support the

downdrag (i.e. negative skin friction). In terms of magnitude, the induced axial force is greatest for
L
p
/H
tun
< 1 (Figure 2.18b). The maximum axial force for volume loss of 1% is 180kN. This
corresponds to only 7% of the ultimate pile capacity. However, the trend is totally different when
compared to studies by others. For example, Mroueh & Shahrour (2002) showed no clear trend of
variation with L
p
/H
tun
whereas Cheng et al. (2004) showed that the axial force is consistently
increasing with increasing L
p
/H
tun
(see also Figure 2.18b). Besides, it can be observed that the
magnitude of axial force is significantly small in the centrifuge tests for the similar details of
tunnel and pile. Furthermore, a pile could be subjected to tensile force particularly near the pile
head. This is likely to be caused by the type of restraint at the pile cap.




23
2.4.4 Pile lateral deflection

Centrifuge tests by Loganathan (1999) showed that maximum lateral deflection of a pile occurs at
the pile base regardless of L

p
/H
tun
. This contradicts field observation by Lee et al. (1994) where the
maximum occurs at the tunnel springline level for pile with L
p
/H
tun
greater than 1. In terms of
magnitude, the lateral deflection is largest when L
p
/H
tun
is equal to 1 (Figure 2.18c). Besides, the
magnitude of greenfield soil movement at pile position is similar to pile lateral deflection. A
magnitude of 7.5mm was observed for volume loss of 1%. However, Lee et al. (1994) only
observed a magnitude of 7mm for volume loss of 2%. The difference could be due to many other
different factors such as pile diameter, pile-tunnel distance and soil condition. Cheng et al. (2004)
also observed a similar magnitude of lateral deflection.

2.4.5 Pile bending moment

In Loganathan (1999), bending moment was observed to be the maximum near pile base for
varying L
p
/H
tun
. Besides, the maximum bending moment occurred for L
p
/H

tun
of 1 (Figure 2.18d).
For volume loss of 1%, the maximum bending moment is observed to be 90kNm. This
corresponds to only 12% of the ultimate moment capacity (M
ult
). The author also showed that for
volume loss of up to 10%, the bending moment only reached 75% of M
ult
. Mroueh & Shahrour
(2002) in their FE analyses shows that the maximum bending moment varies similarly as the
centrifuge tests by Loganathan (1999) where the maximum occurred for L
p
/H
tun
at approximately
1. Despite that, Cheng et al. (2004) showed that the maximum bending moment increases with
L
p
/H
tun
(see also Figure 2.18d). The authors also concluded that bending moment would be
negligible when X
pile
/D
tun
is greater than 2.

24
In actual 3-D tunnel advancement, pile would be subjected to bending in transverse as well as
longitudinal direction. Most of the previous studies have been focusing on transverse bending due

to the simulation of plane strain tunnel. However, Mroueh & Shahrour (2002) have carried out 3-
D tunnel advancement simulation in FE and showed that the transverse bending moment is
approximately three times the longitudinal bending moment.

2.4.6 Pile group effect

To-date, only a few studies have been carried out on effect on pile group due to tunnelling.
Loganathan (1999) simulated a 2x2 pile group in the centrifuge tests. From comparison with a
single pile of the same distance to tunnel, the front pile was subjected to a reduction of
approximately 16% and 22% of bending moment for pile with L
p
/H
tun
of 0.86 and 1.00
respectively. However, the bending moment increased by 40% for pile with L
p
/H
tun
of 1.20. This
inconsistency was not observed in the boundary element analysis carried out by the same authors
reported in Loganathan et al. (2001). In the front pile of the pile group, reduction of 12%, 29%,
6% and 15% were observed in pile head settlement, axial force, lateral deflection and bending
moment respectively when compared to a single pile of the same distance to tunnel. Similarly to
the rear pile in the same pile group, reduction of 10%, 43%, 0% and 21% were observed
respectively. To be noted, lateral deflection for the rear pile was subjected to no reduction.

Besides, Mroueh & Shahrour (2002) also reported in their 3-D FE analyses that reduction of 20%
and 3% were observed for maximum axial force and bending moment in the front pile of a 2x2
pile group. For the rear pile, much higher reduction were observed; 60% and 45% respectively.
Contrary to the above, Kitiyodom et al. (2004) in their numerical analyses showed that almost

negligible reduction was noted when the pile slenderness ratio (L
p
/D
pile
) was large (i.e. 25).
25