Int. Workshop on Geotechnics of Soft Soils-Theory and Practice. Vermeer, Schweiger, Karstunen & Cudny (eds.

2003 VGE

The influence of tunnel boring on foundations and buildings in

urban areas - A numerical study

R.B.J. Brinkgreve
Geotechnical Laboratory, Delft University of Technology & Plaxis bv

W. Broere
Geotechnical Laboratory, Delft University of Technology & A.Broere bv

ABSTRACT: The advancement of a tunnel boring machine in the ground has been numerically
modelled using a phased excavation scheme. Special attention is given to the possibilities and
limitations of numerical modelling of the tunnel boring process and the influence of soil stiffness.
The method has been applied to a situation where tunnel boring may lead to damage in adjacent
buildings on wooden piles. It is concluded that soil stiffness plays an important role in predicting
the width of the settlement trough and consequently the influence on adjacent buildings.

1 INTRODUCTION
The increasing performance of desktop computers may eventually lead to the situation that
geotechnical engineers can apply non-linear three-dimensional finite element calculations on a
daily basis for ordinary design and consulting purposes. However, this situation is not yet reality
and requires more than just computer power. Finite element models for settlement analysis due to
tunnelling are often known to over-predict the width and to under-predict the gradient of the
settlement trough. To obtain realistic results it is, among other things, necessary to use advanced
soil models and to carefully select the corresponding model parameters. In this contribution
attention will be focused on the possibilities and limitations of 3D finite element models for the
analysis of the effects of shield tunnelling on old masonry buildings founded on wooden piles. The
situation considered is an arbitrary case based on a real large-scale project in Amsterdam, The
Netherlands, where a tunnel is constructed (bored) in the ancient city centre. This large-scale
project is known as the North-South (metro) line. The metro tunnels will be constructed near
historical masonry buildings, founded on wooden piles, as well as newer buildings, founded on
prefabricated concrete piles. Along the main part of the line the tunnels will run close to or beneath
the toes of the wooden foundation piles. Both the piles and the masonry structure have little margin
for deformation before damage to the construction will occur.
In recent years, several papers have been published with details about soil conditions
(Teunissen, 1998), design aspects and other issues related to the North-South line project.
Extensive tests and monitoring programmes have been carried out during the construction of the
Second Heinenoord Tunnel near Rotterdam, in view of the situation at the North-South line
(Bakker, 1996). Predictions and back-analyses have been made of the settlement profile that occurs
due to the tunnel boring process (Strack, 2000). The numerical modelling of this process itself has
also been subject of several papers (Vermeer, 2000, 2001; Van der Vliet, 2001).

In the next section of this contribution it will be evaluated to what extent 3D finite element
calculations can be used for geotechnical applications and which pitfalls should be taken into
account. In the third section the example situation and the corresponding finite element model will
be described in detail. Section four deals with the 3D modelling of the tunnel boring process. The
fifth section describes the results of the finite element analysis in terms of the influence of the
tunnel boring process on the surrounding buildings and the role of soil stiffness in the numerical
analysis. Finally, in Section six, the main conclusions of this research are summarised.

POSSIBILITIES AND LIMITATIONS OF 3D FEM MODELLING OF SHIELD TUNNELS

Apart from the computer power requirements, three-dimensional finite element calculations are
difficult to perform due to various reasons. Probably the most important reason is that it is much
more difficult to properly construct and interpret a three-dimensional computer model than a 2D
model. Modelling errors are easily overlooked. Of course, advanced 3D computer graphics and 3D
analysis tools can help, but it definitely requires spatial perception from the user. Particularly for an
analysis that involves multiple subsequent construction stages (some modellers call this a '4D
analysis', where Time is the fourth dimension), the use of animations can be of great help to
evaluate the results or to locate input errors.
Regarding the modelling of soil behaviour, a range of soil models is available nowadays in
modern finite element computer programs. Simple constitutive laws, like the standard MohrCoulomb model, lack aspects of soil behaviour that are vital to obtain accurate results.
Nevertheless, most finite element calculations are still performed using this first order model,
because it requires a limited amount of parameters and calculation times tend to be smaller than
when using advanced soil models. Regardless of the soil model, it should be realised that the strain
levels observed even close to a shield tunnel are usually quite small (except in extreme cases) and
the stress paths involve unloading rather than primary loading. As a result, the stiffness of the soil
tends to be very high and is generally underestimated. Advanced soil models involve stressdependent stiffness relations and distinguish between primary loading and unloading, but the input
of a small-strain stiffness is often not recognised. A real small-strain stiffness can only be
determined from dynamic soil testing, but this type of testing is not standard.
For the modelling of the tunnel boring process of a shield tunnel various methods were
developed over the years. Aspects that should be considered are the face pressure (for slurry or
earth pressure balance shields), the conicity of the TBM, the weight distribution of the TBM and
other equipment, the grout injection in the tail void between the TBM and the final lining, the
consolidation process of this grout layer, the segmentation and connections of the final lining, as
well as various other aspects. Most modern finite element programs provide facilities to apply
various types of loadings (pressure control), volumetric straining (strain control), tunnel lining
contraction and, last but not least, staged construction. However, it is usually a tremendous job to
correctly enter the necessary conditions for all calculation phases.
A special aspect that needs to be considered is the start condition. Finite element models
generally model the tunnel boring process on the route, but start from an undeformed mesh with
initial soil stresses. These start conditions definitely influence the results. A certain number of
construction steps are required to set the proper 'initial' conditions and to avoid influences of the
boundary where the tunnelling process is started. Something similar can be said about the face of
the TBM. Some soil deformation will already take place in front of the TBM. As soon as the TBM
is activated (usually modelled by means of shell elements), these elements are pre-deformed but
not properly stressed. However, if the analysis is performed according to the small deformation
theory (only first order deformations) rather than an Updated Lagrangian analysis, these initial
deformations do not really influence the stresses and forces, whereas in reality initial stresses and
forces do exist.
In reality, the tunnel boring process involves conditions that depend on the observations and
measurements that are made during the tunnelling process and on how the TBM operation crew
reacts on these observations. Examples of such conditions are the local pore pressure distribution,
the face pressure (depending on TBM advancement and cutting speed) and the grout injection

pressure and volume. Hence, when a finite element analysis is used as a true class A prediction, the
modelling may deviate significantly from the real process. Nevertheless, it can still be quite useful
to perform a finite element analysis to qualitatively analyse the effects of tunnel boring.

3 THE EXAMPLE SITUATION AND FINITE ELEMENT MODEL

In addition to an earlier analysis made for The Second Heinenoord Tunnel, an arbitrary but realistic
situation is considered here in which an 8.5 m diameter tunnel is bored with its axis at a depth of 20
m below the ground surface (15.75 m soil above the tunnel lining). The situation is modelled using
a 3D finite element model which is symmetric with respect to a vertical plane through the tunnel
axis. The bottom of the mesh is at a depth of 30 m (5.75 m below the tunnel lining). In addition to
the tunnel, a block of houses on piles is modelled adjacent to the tunnel at a distance of 10 m from
the tunnel axis (5.75 m from the lining). The block of houses is 18 m long, 10 m wide and 12 m
high (above the ground level). The piles are founded on the sand layer in which the tunnel is bored.
Above this sand layer (S) there is a 2 m thick clay layer (C), a 7 m thick peat layer (P) and a 3 m
thick fill (F). All layers are modelled with the Hardening Soil model (ref). Table 1 gives an
overview of the soil layers and their properties and parameters.
The stiffnesses are reference stiffnesses at a reference stress level of 100 kN/m2. The dilatancy
angle of the stiff sand layer is 6.5 degrees; the other layers have a zero dilatancy. The power in the
stress-dependent stiffness relation is 0.5 for all layers, except for the clay layer, where it is 1.0.
The hydrostatic pore pressure distribution in the model is derived from a phreatic level at 2 m
below the ground level.
The 3D finite element model used consists of 9125 quadratic volume elements, divided into 25
slices with identical cross section. However, in each slice, elements may be (de)activated or may
have different properties (see Figure 1). The thickness of most slices is 3.0 m, corresponding with
the advancement steps of the tunnel boring process. The 12 piles and the walls of the houses were
created using 0.40 m thick columns and slices of volume elements that were filled with 'pile' and
'wall' material (linear elastic model; pile= wall=24 kN/m3 ; Epile=3106 kN/m2 ; pile=0.0 ; Ewall=1106
kN/m2 ; wall=0.2). A small slice is followed by a 2.60 m thick slice to complete the advancement
step of 3.0 m. The TBM is modelled as a 9 m long cylinder of shell elements (three tunnel
advancement lengths). The flexural rigidity of the shell EI = 50103 kNm2/m and the normal
stiffness EA = 10106 kN/m. The weight of the shell is taken as w = 38.15 kN/m2, representing the
full weight of the TBM including equipment. The final concrete tunnel lining was modelled as a
0.35 m thick continuous ring, using volume elements with a volumetric weight = 24 kN/m3,
Young's modulus E = 24.6106 kN/m2 and Poisson's ratio = 0.2.

Figure 1. Finite element model in final situation

(some elements have been made invisible to show the pile foundation)

Table 1. Soil layers and their parameters (Hardening Soil model)

Top

Type

F
P
C
S

0.00
-3.00
-10.00
-12.00

Drain
Drain.
Undr.
Drain.

E50ref

Eoedref

[kN/m ]

[-]

19.5
12.0
18.5
20.5

0.2
0.2
0.2
0.2

sat

[kN/m ]

16.5
12.0
16.5
17.5

[m+NAP]

ur

unsat

Eurref

c'

'

[kN/m ]

[kN/m ]

[kN/m2]

[kN/m2]

[]

30000
5000
15000
40000

30000
5000
12000
39614

150000
25000
75000
200000

3.0
7.0
3.0
0.0

30.0
24.0
27.0
36.5

4 3D MODELLING OF THE TUNNEL BORING PROCESS

The modelling of the tunnel boring process consists of different aspects (see Figure 2), as
mentioned in Section 2, although not all aspects are considered here. Inside the tunnel the soil is
excavated by de-activating the corresponding volume elements and water pressures. At the face of
the TBM a face pressure is applied to support the soil during excavation. This face pressure is 234
kN/m2 at the top of the TBM and increases linearly to 336 kN/m2 at the bottom.
The TBM itself is slightly conical. The tail radius is 20 mm less than the front radius, which
effect has been modelled using a total contraction of 0.48% at the TBM tail, i.e. an incremental
contraction of 0.16% in each cross section at 3 m, 6 m and 9 m behind the TBM face.
contraction of shield

TBM

grout
pressure

final lining

face pressure

Figure 2. Various excavation phases modelled in the phased excavation procedure

The influence of the grout injection at the shield tail has been modelled using a distributed load,
acting over 6 m (two tunnel advancement lengths). Thereafter it is assumed that the grout has
settled and dewatered enough such that no additional deformations occur. The grout pressure has
been set to 186 kN/m2 at the top, increasing to 254 kN/m2 at the bottom. Behind the grout injection
zone the elements that form the tunnel lining are activated and are given concrete properties.
The initial stresses are calculated using a coefficient of neutral effective stress according to
Jaky. The full finite element analysis is divided into 20 calculation phases. In the first two phases,
the piles and the building are constructed, but the corresponding deformations are not taken into
account in further steps. In the following phases, the advancement of the tunnel boring process is
simulated by advancing the tunnel face 3 m in every phase, starting from the back-side of the
model and taking into account the previously described modelling aspects behind the tunnel face as
long as they fit in the model. The first tunnel excavation step takes place in the third calculation
phase. In the sixth phase the grout pressure becomes active for the first time (TBM shield is locally
de-activated). In the eighth phase, when the tunnel heading has advanced to the first row of piles,
the concrete lining is activated for the first time at the back-side of the model. In phase 14 the
tunnel heading has advanced to the last row of piles. In the final phase 20 the tunnel heading has
advanced 54 m, i.e. 18 m behind the last row of piles, such that most deformations of the buildings
have occurred. The results of the analysis are presented in Section 5.

5 RESULTS OF THE FINITE ELEMENT ANALYSIS

The influence of tunnelling close to pile foundations on the bearing capacity of those foundations
and the resulting deformations of the structures founded on them is topic of several research
projects (e.g. Teunissen, 1998). In this case it is the aim to investigate numerically and qualitatively

Figure 3. Total displacements at the end of Phase 20 (TBM well beyond houses)
(deformations enlarged 100 times)
to what extent pile foundations that are located primarily above the excavation level may be
influenced by the tunnel boring process. Moreover, the importance of soil stiffness in the finite
element model is shown.
Figure 3 shows the tilting of the houses due to the tunnelling process. At the left side the vertical
settlement is 17.8 mm and at the right side the vertical settlement is 5.7 mm, which gives a gradient
of 0.0012, i.e. less than 1:800. The gradient is slightly more than the gradient of the ground surface
at the back-side of the model. This is due to the fact that the soil deformation and gradient at the
pile tip level are higher than at the ground level.
A gradient by itself may not be most harmful for historical buildings, but what might be more
harmful is the torsion that occurs when the tunnel boring process passes the block of houses. The
torsion effect reaches a maximum when the TBM has advanced about half way the houses (see
Figure 4).

Figure 4. Displacements at the end of Phase 10 (TBM half way houses) showing torsion of the
houses (deformations enlarged 100 times)

From Figure 4 it can be seen that the wall at the back-side has settled and is inclined whilst the
front wall is still more or less undeformed. This torsion effect may lead to cracks in the masonry
walls, although the amount of torsion in this example case is quite limited.
It is known that the width of the settlement trough above a tunnel is generally overestimated in a
finite element analysis, and, as a result, the gradient of the trough is underestimated. Similar
observations have been made for the width of the settlement trough behind a sheet pile wall or a
diaphragm wall. It is the authors' opinion that the underestimation of the gradient is mainly due to
an underestimation of stiffness in the small-strain region.
To validate this statement an additional finite element analysis was performed for the situation
as described in Section 3, taking into account small-strain stiffness effects. The analysis was
performed by setting the stiffness moduli to values five times higher than used in the first analysis.
It is recognized that this approach does not fully replicate the small-strain stiffness effect, since that
effect is particularly pronounced at a somewhat larger distance from the tunnel. Nevertheless, the
corresponding aspects can, to some extent, still be observed. The deformations around the tunnel
are mainly strain-controlled and not stress controlled. Hence, the deformations immediately around
the tunnel are not influenced by the larger soil stiffness, so a comparison of deformations is still
valid. With the higher stiffnesses, the soil becomes plastic in a zone around the tunnel in an earlier
stage, which effectively reduces the stiffness in this zone. As a result, the stiffness is lower around
the tunnel and higher at a somewhat larger distance from the tunnel. This effect is similar (but not
equal) to the small-strain stiffness effect.
Figure 5 shows a comparison of results after Phase 20, where the model has been cut at the
back-side wall of the houses. Although the deformations of the tunnel are almost the same, the
settlement trough right above the tunnel in the stiff model (b) is indeed more concentrated and
steeper than in the soft model (a). Nevertheless, the gradient of the houses in the stiff model is less
than in the soft model. This is caused by the fact that in the stiff model the houses and their
foundation are located just outside the zone in which most settlements occur whereas in the soft
model the settlement area is wider and influences the pile foundation more than in the stiff model.
Also the settlements of the ground surface are less in the stiff model than in the soft model (see
Figure 6). Especially in the stiff model the inclination of the ground surface near the houses is less
than at the back-side of the model. For the soft model the situation is reversed, as mentioned
before.

Figure 5. Total displacements at the back-side wall and beyond at the end of Phase 20
(deformations enlarged 100 times)
a. Original analysis with stiffnesses according to Table 1
b. Modified analysis with five times higher stiffnesses

Figure 6. Comparison of settlements at the ground surface at the end of Phase 20

a. Original analysis with stiffnesses according to Table 1
b. Modified analysis with five times higher stiffnesses

6 CONCLUSIONS
A 3D finite element analysis can be used quite well to simulate the tunnel boring process and to
analyse the effects on adjacent structures. However, it is important to take sufficient modelling
aspects into account. Regarding soil behaviour, it is important to realise that the soil behaves quite
stiff due to unloading and small-strain stiffness effects. In the model, the selected soil stiffness
influences the width of the settlement trough and the effects of the tunnel boring process on
adjacent structures.

REFERENCES
Bakker K.J., van Scheldt W. Plekkenpol J.W. (1996) Predictions and a monitoring scheme with respect to
the boring of the Second Heinenoord Tunnel. Geotechnical aspects of underground construction in soft
ground. Balkema, Rotterdam, 459-464.
Brinkgreve R.B.J., Vermeer P.A. (2001) Plaxis 3D Tunnel (Validation manual). Balkema, Lisse.
Strack O.E., Verruijt A. (2000) A Complex Variable Solutions for the Ovalization of a Circular Tunnel in
an Elastic Half-Plane. GeoEng2000, An International Conference on Geotechnical & Geological
Engineering. Technomic.
Teunissen E.A.H., Hutteman M. (1998) Pile surface settlements at full scale tests North/South metro line.
Tunnels and Metropolises, WTC 98. Balkema, Rotterdam
Vermeer P.A., Ruse N. (2000) Face stability when tunneling in soil and homogeneous rock. Proc.
Developments in Theoretical Soil Mechanics The John Booker Memorial Symposium. Sydney, 123-138.
Vermeer P.A. (2001) On a smart use of 3D-FEM in tunnelling. Plaxis Bulletin (11), 2-7
Van der Vliet C., de Boer A., Blom C.B.M., Ros P.L.M. (2001) Strategy and results of calibration for
complex 3D finite element model for tunnel linings in soft soil. Modern Tunneling Science and Technology.
Swets & Zeitlinger, 835-838.