Three-dimensional Structural Analyses of the

Shield-Driven "Green Heart" Tunnel of the

High-Speed Line South
C. B. M. Blom, E. J. van der Horst and P. S. Jovanovic

A b s t r a c t --The tunnel lining generates a significant part of the bore tunnel project costs. This tunnel
structure is one of the most important components of the whole tunneling process. The tunnel structure has
to fulfill all necessary functional requirements during its lifetime. Because of this it is essential for engineers
first to understand the realistic tunnel-lining behaviour and then to design a tunnel structure in a proper way.
The design of the lining structure is actually quite simple because of the wide range easy-to-use models now
available. In contrast, predicting realistic tunnel-lining behavior is very difficult. The available numerical
models for a segmented concrete lining cannot predict realistic structure behavior at all stages of excavation
and during the tunnel lifetime. Conventional models ignore the influences of assembling processes,
imperfections of segments, type of joints and variation in stress distributions in the concrete sections. This
paper deals with three-dimensional finite element analyses of the tunnel structure, observations during the
construction phase and in-situ measurements on the Second Heinenoord Tunnel (Bakker 1999) applied on
the structural design of the shield-driven "Green Heart" Tunnel of the High Speed Line-South in the
Netherlands. 1999Elsevier Science Ltd. All rights reserved.

1.0 Introduction
he design of any tunnel structure has to fulfill some
basic requirements concerning structure stability,
durability, watertightness and reliability, low costs
and low-risk profile. To achieve these goals it is necessary
that engineers have an opportunity to predict unwanted
effects. Although it is difficult to translate the reality of
tunnel lining behaviour into a mechanical mathematical
engineering model, it is absolutely essential that this be
done.
In most cases, conventional modelling ignores the influences of segment and assembling imperfections, the construction method, type of joint material, jack forces, nonlinear effects, ground freezing, grouting, etc. As a result of
these effects, the three-dimensional stress distribution in
concrete segments develops with corresponding unwanted
deformations. Furthermore, this non-uniform stress distribution, with stress peaks and totally unexpected stress
paths in the segments, are the reason for the appearance of
damages and cracks, which lead to a decrease in structural
durability and an increase in maintenance costs, and thereby
a higher risk profile.

Present address: Ir. C. B. M. Blom,Ir. E. J. van der Horst, and Ir. P.

S. Jovanovic, Project Organization High-Speed Line South, Bore
Tunnel "Groene Hart" Project Office , Postbus 20000, 3502 LA,
Utrecht. All authors are on loan to this project from Holland
Railconsult.

Most ofthe conventional structural models assume planestrain conditions for the lining and the ground. The stiffness
of the lining is considered a constant value. Complete or
restrained structural hinges are sometimes considered, and
the active soil pressures on the lining are assumed to be
equal to the primary stresses in undisturbed ground. Furthermore, between the lining and ground exists a bond, both
for radial and tangential deformations, which results in
reaction stresses in the ground. Material properties and the
behaviour of the soiland lining are generally assumed to be
elastic. Grout pressure and grout hardening, as well as the
type of joints and packing material, are not taken into
account when modelling.

2.0 Necessity of Three-dimensional Analyses

In comparson with conventional models, three-dimensional finite element analyses of a tunnel structure, combined with observations during the construction phase and
in-situ measurements, show totally different stress distributions in concrete lining sections. With three-dimensional
models it is possible to predict the influences of the assembly
process and imperfections of segments on lining behaviour,
as well as the influence of joint type with various packing
materials.
Structure behaviour concerning strength and stability
has been analysed applying different sorts of available
methods, e.g. analytical, empirical and numerical. These
methods should fulfill necessary modelling requirements.
The models should be easy to use and should include all

www.elsevier.com/]ocate/tust
~n~elli~ and UndergroundSpace Technology,Vol. 14, No. 2, pp. 217-224, 1999
0886-7798/99/$ - see front matter 1999 Elsevier Science Ltd.
All rightB reserved.
P[I:S0886-7798(99)00035-8

Pergamon

Cv =

7H

O'B = K r U
N
=

Kl.=

const.
M

h = Ko

O v e"

radial
displacement

grpund reaction

pressure

normal
forces

bending
moment

Figure 1. Conventional models of the bedded beam type.

significant data related to the geometry, loads and groundstructure interaction. The wide range of available structural design models may be categorised as follows:
models based on subgrade reaction (conventional
models): continuum models, bedded beam models
(see Fig. 1).
models based on relative stiffness solution.
convergence-confinement models.
observational methods.
finite element method (2-D, 3-D).

1or. The philosophy of these conventional models is based on

the following assumptions:
It is sufficient to assume plane-strain conditions for
the lining and the ground. This leads to avoiding
three-dimensional effects.
Stiffness of the lining is considered constant. Complete or restrained structural hinges may or may not
be considered.
Active soil pressures on the lining are assumed to be
equal to the primary stresses in undisturbed ground.
Between the lining and ground there exists a bond,
both for radial and tangential deformations.
Owing to the bond between the lining and ground,

For some shield-driven tunnel projects now underway in

The Netherlands, conventional models have been used to
provide necessary information about tunnel lining behav-

h
2t~

2~

I d I } I I I l lil

Ill

I I I d

2~

(~D

'

Jilllltl

il

~iiIi

- -

llil

'

-2ID

'

-2~

1
'

i
~.

~!

'~
V

,,L~ ~.-~!

Figure 2. Stresses caused by the assembling process, measured in the ring within the shield.

218 TUNNELLINGANDUNDERGROUND SPACETECHNOLOGY

Volume 14, Number 2, 1999

TAN 29 igg~
13:11:!['
PLOT ~2.
1

TI~E~
RSY$-O

$,',~ =-598~
=lOl9
&Eta=4796

~ ,',1~' =24,3
ZZ5~5 = ~ 5 Ii,Ll
-: 315
_ :\:v
i

-4~30

~:2<

36~

-2095
-~]16
-~97.992
24C.32~
1019

Figure 3. Normal stresses are not uniformly distributed in radial, axial and tangential directions (stress paths around the
key segment).

o
i

10

15

20

-500

f~ce level after asse~ly

-1000
,o...o=.=,ooo. 0o.o ..ooo . . . . . . . . .

o....o.o..o.=,~o.ooo,..o==o,ooo=oo..oooo.oooo.o.oooooooo.ooo.oo.ooo.oooo.o

URl

-1500

The long-term ring forces remain

close to the level that was obtained
during the assembly (Fig. 4).
Eccentricity of the axial normal
forces is obviously present, as can be
seen from the bending moments in
the longitudinal direction ofthe tunnels (Fig. 5).
Maximum measured sectional forces
and moments are twice the highest
predicted with conventional models
(Fig. 6).

The appearance of assembly stresses

in the lining, which were measured, could
be explained by a hypothesis consider-3OOO
ing the combination of loads and construction effects. The presence of grout
Figure 4. Long-term ring forces remain close to the load that was obtained
hardening behind the TBM taft, together
with jack forces, assembly imperfections
during the assembly.
and tolerances, could lead to unexpected
stress development in the concrete segmented lining. The non-uniform jack forces, which are
deformations of the lining result in reaction stresses
supported by the segments to provide TBM driving, are
in the ground.
used at the same time for segment placing and fixing
Material properties and the behaviour of the soil and
towards the neighboring ring. Placing of the key segment
lining are generally assumed to be elastic.
between the other segments is definitely dependent on the
The serviceability stage is essential for the lining
available space that can be smaller or bigger than the size
design.
of the key segment.
The dynamic process of tunnel driving can be seen
After the completion of the first shield-driven tunnel in
perfectly by observing a ring inside the shield. When the
the Netherlands, The Second Heinenoord Tunnel (see Bakker
ring is completely assembled, the jack forces move the TBM
et al. 1999 in this issue of T&UST), it was concluded that
forward, allowing grout injection behind the TBM tail. At
some of the assumptions stated above are not applicable to
that moment the ring behind the last one in the TBM, which
Dutch soft soil conditions. As a result of measuring, obseris loaded by grout load, is going to "hang" on the ring within
vations and three-dimensional finite element analyses,
TBM. It is obvious that assembling a ring with a perfect
some important facts have emerged that should be considcircle shape on a deformed ring, in addition to the absence
ered carefully:
of full contact in the lateral and ring joints, could lead to
Stresses caused by the assembly process were meathree-dimensional stress distribution in all directions
sured in the ring within the shield (Fig. 2).
through the concrete segments.
Stresses are not uniformly distributed (stress peaks)
A combination of these effects could cause deformation of
in radial, axial and tangential directions (stress paths
the tunnel in the longitudinal direction, behind the TBM in
around the key segment) (Fig. 3).
the construction stage, resulting in bending moments and

Volume 14, Number 2, 1999

TUNNELLINGANDUNDERGROUNDSPACETECHNOLOGY219

JAN 29 1 9 9 9
13+15-~n
pLOT NO1
NODAL SOLUTION
STEP=!
S U B =I
TIME:~I
UZ
RSYSmO
DMX =,0047~9
S E P C ~ I 5. 517
S~
~-, 3 5 D E - 0 3
SMX ~, 6 2 7 E - 0 4
.350E-03
- . 3DIE+ 03
-. 2 5 8 E - 0 3
- .212E-03
-. 166E-0]
]
,121E-01
-.
718E-04
- .290E-04
IBIBII
.169E-0~
.627E+04

and tolerances under specific non-uniform jack forces and

overburden. Furthermore, the assembly within the shield
concerning sequential placing of the segments (load history)
and special loads (freezing, dynamic, etc.) can also be
analysed.
Nevertheless, the three-dimensional finite element models are not the models of common practice, but they could
be essential in predicting the realistic tunnel-lining
behaviour. Research that is now taking place within the
Project Office of the Green H e a r t Tunnel, deals with the
following objectives:
grout behavior (hardening);
packing materials;
explicit domination of relevant parameters;
avoidance of stress peaks;
verification of calculated and measured stresses;
reliability of three-dimensional finite element models
with applicability to risk analyses and test requirements.

3.0 Development of 3D-FEM model for shielddriven tunnels

Figure 5. Eccentricity of the axial normal forces is

obviously available.

400

300

Measured

200

100

- immlt/

Predicted

10

Model

Figure 6. Maximum measured moments are twice the

highest predicted.
definitely in an unforeseen three-dimensional stress distribution.
It is obvious t h a t conventional models are not able to
predict realistic structure behaviour in the construction
stage and operating stage. The observations and measurements on the Second Heinenoord Tunnel project, together
with 3-D finite element analyses of the tunnel structure,
gave a totally different stress distribution picture in the
lining, in comparison to the results from conventional
models. The assembly process, grouting process, measure
tolerances, and inaccurate placing of segments definitely
play a crucial role in three-dimensional stress development. Because of this it was concluded t h a t a threedimensional finite element model (block elements) should
be applied in the structural analyses and design of the
lining of the shield-driven Green Heart Tunnel.
This type of modelling can provide full information about
the influence of the key segment on the complete system of
rings. It is also possible to observe the realistic behaviour of
joints, packing materials, grout hardening, imperfections

220 TUNNELLINGANDUNDERGROUNDSPACETECHNOLOGY

As a result of the considerations mentioned above (in

relation to measurement results and 3D-research), Holland
Railconsult (participating in the Project Office of the Green
H e a r t Tunnel) developed a three-dimensional finite element model in order to be able to research the construction
behaviour of shield-driven tunnels. Using this model, which
is closer to reality than the available models, it is possible to
predict the so-called assembly stresses more realistically.
At the same time, the three-dimensional influences, such as
grout hardening, can be analysed much more accurately.
A few aspects concerning the influence of axial forces and
unequally loaded tunnel rings have already been shown
(Blom et al. 1998). Because of the continuous development
and enhancement of the finite element models, the possibilities for researching tunnel lining behaviour are constantly
increasing. For example, it is now possible to model a
significant part of the tunnel (more than 10 rings; see Fig.
8), including the grout at different stages of stiffness development. The stress distribution through the segments, as
they are assembled, also can be followed in the future.
This paper describes the background with the applied
assumptions to the model and their relation to the sitemeasurements. Furthermore, an example is given of a
twelve-ring model, which was used for researching the
influence of hardening grout behind the TBM. Finally, an
overview is given of some future plans.

3.2 Mode/Description
The 3-D model shown in Figure 7 was made (1996) with
the ANSYS finite element sot%ware package. The concrete
segments were modelled using solid volume elements. The
first models were designed for three rings (seven segments
and the key segment each) with, in total, 8100, each consisting of eight nodes, solid brick elements (concrete quality
B45). Each node had three degrees of freedom. Ground
property was represented by applying spring-damper elements (1418 in total) at three-quarters of the perimeter
(after Duddeck 1980), with Young's modulus of elasticity
around 45,0 MPa/m 2.
The interaction between the segments (in all directions)
was realised by applying interface or contact elements.
These "point-to-point" elements were used because of their
capacity to simulate cold interface conditions transmitting
only compression in the direction normal to the surfaces and
shear in the tangential direction (Fig. 8).
The packing material, the only contact between the rings
at ring joint, was represented by four contact elements each,
behaving as linear springs until sliding occurs. The stiffness

Volume 14, Number 2, 1999

Figure 7. Three-dimensional models developed with ANSYS finite element software.

of the four elements is equal to the stiffness of the packer
(when it is assumed to have linear behaviour). The dowels
are not modelled.
The first ring is supported in the axial direction. In the
radial direction the surrounding ground is modelled as
linear springs, following the Duddeck theory. The radial
forces are also according to Duddeck theory (350 kN/m2),
while in the axial direction the TB M driving forces (14j acks,
max. force at the bottom of 3500 kN, min. force at the top of

1000 kN) are modelled. When considering the construction

stage, the last ring within the TBM shield was directly
loaded by the jack forces without any load or support in the
radial direction. In the serviceability stage, all rings are
loaded and supported in the same way.
As mentioned above, the ground loads on rings 1 and 2
(Fig. 7) are modelled using the Duddeck theory, while it is
assumed that the grout behind the TBM changes from
liquid to solid. In this way, a situation is created that is

Contact elements

Figure 8. Segment assembly.

Volume 14, Number 2, 1999

TUNNELLING ANDUNDERGROUND SPACETECHNOLOGY 221

l
Figure 9. Side view of TBM and lining.
similar to that of the Second Heinenoord Tunnel, when the
TBM stood still for more than eight hours. When the driving
process was restarted, some damage occurred at the tunnel
segments, which was also one of the reasons for performing
a three-dimensional analysis.
In reality, of course, the ring within the TBM is not just
"suddenly there", but is assembled segment by segment. To
be able to analyse the stress development per segment in
time, new kinds of contact elements are applied, which can
be "switched off' together with the volume elements, representing the concrete segments. By switching them on again,
a stepwise ring assembly can be simulated. The problem in
this case is, where is the next segment going to be placed
when the existing structure is deformed?

3.2 Construction Stage, Including Grouting

To be able to analyse the influence of the grout directly
behind the TBM, together with the overall asymmetrically
axial loads, it was necessary to model a larger part of the
tunnel (12 rings; see Fig. 9). The axial force from the TBM
asymmetrically loads the lining with close to the resultant
hydrostatic pressure at the excavation front, but working in
the opposite direction. Furthermore, floating of the lining
behind the TBM can occur in the liquid grout (depending on
grout properties and distribution around the tunnel). This
causes an upward force on the back of the TBM, which must
be compensated through the jack forces.
The question is, how does the lining react at the moment
that it is being subjected to the asymmetric, axial force in
combination with the upward force caused by the liquid
grout? Another question is, how does
the strength of the grout develop in
time, and which situation occurs when
the grout is completely h a r d e n e d
(Duddeck-like situation or something
else)?
Of the 12 tunnel rings modelled,
one ring is not loaded and not supported (located in the TBM); five consequent rings behind it are loaded and
supported by grout; and the other six
rings are loaded and supported by the
ground. Thus, it is assumed that aP~er
placement of rings, the grout is completely hardened and has the same
properties as the surrounding ground.
From the first to the sixth grout ring,
the bedding stiffness increases from 0

to the stiffness of the surrounding ground

(linearly; see Fig.10). It is also assumed
that when the grout has become solid,
the Duddeck theory can be applied. It is
clear that in the liquid grout situation,
Duddeck's force distribution is not realistic, because hi theory does not take
into account this kind of circumstance.
The loading history has been established as follows. On the first five rings
behind the TBM, a "floating force" was
applied. The basic principle is that the
radial grout pressure should be equal to
the overburden around the lining. This
force gradually changes (linearly) to
Duddeck forces. On the other five rings,
the full Duddeck theory was applied.
Under the assumption that the radial grout pressure on the lining is equal
to the vertical ground pressure or the
liquid grout pressure, the total upward
force cannot be taken by the grout itself.
To reach equilibrium, this force has to be
spread out to the sides (Fig. 11), i.e. to the TBM and to the
grout that is already solid. Of course, there is not a sharp
line between liquid and solid grout, and the transition area
also depends on the TBM driving speed. Between the "fully
liquid" and the "fully solid" stage of the grout, there is an
area where the grout stiffness takes an average value.
The jack force distribution blocks, which act as an intermediary between the TBM jacks and the tunnel segments,
are also modelled. In this model, they are capable only of
moving in the axial direction and of rotating around the
tangential axis.

4.0 Results, Conclusions and Recommendations

Some results from the calculation models described above
are briefly given here. Figure 12 shows the tangential
stresses along the lining. It is obvious that the value of the
stresses does not change much, because of the hardening
grout, but the distribution around the lining does. Figure 13
shows the deformed shape of the lining as a result of the
loads previously described. In these calculations, a flexible
packing material was applied (plywood), so the rings can
rotate relatively strongly in relation to each other. When
modelling concrete-to-concrete contact, this rotation will be
much smaller.
The construction behaviour found in the mentioned
application is definitely influenced by the assumed ground
and grout behaviour, as well as by the properties of the

7.5 '

LINING

Figure 10. Distribution of ground stiffness behind the TBM.

222 TUNNELLINGANDUNDERGROUNDSPACETECHNOLOGY

Volume 14, Number 2, 1999

Figure 11. Applied loads.

applied packing materials. It is clear that further research
into these matters is of utmost importance. The other very
important parameter is the fact that the tunnel rings have
been surrounded by solid grout, and once were surrounded
by liquid grout, so a phased simulation seems to be unavoidable for obtaining accurate predictions.
The influence on assembling stresses of the imperfection of dimensions of segments and non-accurate assembling should be investigated. The results of this study
could be applied as requirements for segment production
and assembling procedures (e.g. protocol of assembling).
The problem of assembling stresses could be solved through
reduction of deformation differences by applying the following measures:
S u p p o r t s t r u c t u r e . Usuallycalled"Reformer",this
structure can be used to maintain the circular form of
the rings, which are already placed behind the shield.
L o n g e r s h i e l d tail. More rings could be assembled
within the shield. The distance between the deformed
rings behind the shield and the undeformed ring in
the shield will then be longer.
T h i c k e r Hning. When the stiffness of the lining
increases, then the deformation differences between

Figure 12. Tangential stresses.

Volume 14, Number 2, 1999

the rings will decrease. Because a thicker

lining provides better stress distribution
within the segments, its influence of the
axial stresses on tunnel lining behaviour
will be reduced.

S t e e l - f i b r e c o n c r e t e . The stress
concentrations caused by the assembling process can lead to cracks
in segments. Steel-fibre concrete
can absorb these stresses.

C o n t r o l l e d g r o u t i n g . With a special grout consistency, considering

the time of grout hardening, the
influence of external loads on the
lining deformations behind the
shield could be controlled.
Q u a l i t y c o n t r o l . A very impor
tant matter is dimension tolerances. A
discrepancy of only 1.0 mm per segment
length gives, at the end, 7.0 mm of problematic space.
There may be too much space or not enough space for
placing a key segment.
The future model analyses have to be verified through
appropriate experimental data and measurements in situ
as well as in laboratories. Furthermore, the conventional
models that are usually applied in engineering practice
should fulfill the requirements obtained from these verifications.

5.0 L a b o r a t o r y T e s t i n g
In order to test the described hypotheses and calculation
models concerning assembly stresses under laboratory circumstances, full-scale laboratory tests are being conducted.
For this purpose, three tunnel rings, consisting ofthe Botlek
Railway Tunnel segments, are being built in the Stevin II
laboratory at Delft University. Different conditions that can
occur during the construction of the tunnel lining can be
simulated. In particular, the influence of a "not-fitting key
segment" and varying jack forces are to be researched.
Figure 14 shows an artist impression of the testing facility.
The main goals that should be achieved through this
experiment are to analyse the complex lining behaviour
under conditioned circumstances and to predict realistic
structure behaviour with more certainty. Nevertheless, the
calibration of 3-D finite element models can take a place in
the design process, aiming at an economical, optimum
solution related to functional requirements. The expected
results are related to the general lining behaviour (3D
stress and strain paths in the lining), structure behaviour
during assembly (influences of radial loaded and not loaded
rings, assembling imperfections and tolerances, placing of
the key segment, jack forces, grout process, loose sands,
etc.), and structural behaviour in the serviceability stage
(i.e. freezing loads, dynamic loads, etc.)

Figure 13. Deformations.

TUNNELLING ANDUNDERGROUND SPACETECHNOLOGY 223

7.0 References
Bakker, K. J.; Boer, F. de; Admiraal, J. B. M.; and Jaarsveld, E. P.
van. 1999. Monitoring pilot projects using bored tunnelling:
the Second Heinenoord Tunnel and the Botlek Rail Tunnel.
Tunnelling and Underground Space Technology 14 (2), 121130.
Bakker, K. J.; van Schelt, W.; and Plekkenpol, J. W. 1996.
Predictions and a monitoring scheme with respect to the
boring of the Second Heinenoord tunnel. Proc. TC28

Symposium, City University, London, April 1996.

Blom, C. B. M.; Duurland, H. C. W.; van Oosterhout, G. P.;
Jovanovic, P. S. 1998. Three-dimensional structural analyses
and design of segmented tunnel lining at construction stage.

Proceedings of a Conference on Numerical Methods in

Geotechnical Engineering, 1998, Udine, Italy.

Figure 14. Test set-up.

6.0 Conclusion
I n most cases of common practice, engineers follow the
existing construction traditions: the s t r u c t u r e already exists, so let us find out how it behaves. This is the opposite
approach to asking, "How do we w a n t a s t r u c t u r e to behave?", a n d t h e n designing the s t r u c t u r e to behave t h a t
way. To be able to use the l a t t e r as the basic principle in
designing t u n n e l s , the construction b e h a v i o u r has to be
known and has to be predictable. The investigations described herein can contribute much to r e a c h i n g this goal.

224 TUNNELLINGAND UNDERGROUNDSPACE TECHNOLOGY

van der Horst, E. J. 1998. Invloed van voegmaterialen op de

krachtswerking in gesegmenteerde betonnen boortunnels (in
Dutch). Technical University of Delft.
Duddeck, H. 1980. Empfehlungen zur Berechnung yon Tunneln
im Lockergestein (i/1 German). Bautechnik 10, 349-356.
Blom, C. B. M. and van Oosterhout, G. P. 1997. Tweede orde
evaluatie tunnelconstructie Tweede Heinenoordtunnel (in
Dutch). Gouda.
van de Horst, E. J.; Blom, C. B. M.; van der Veen, C.; Jovanovic,
P. S. 1999. Influence of Packing Materials on Tunnel Lining
Behaviour. Challenges for the 21st Century: Conference

Proceedings, World Tunnel Congress, June 1999, Oslo, Norway,

Vol. 1,363-368. Rotterdam: A. A. Balkema.
Oosterhout, G. P. C.; Blom, C. B. M.; Jovanovic, P. S. '1999.
Analyses of tunnel 1.ining behaviour in soft soil during the
constructionstage, The Second Heinenoord Tunnel. Conference
Proceedings, 1999, RAI-Amsterdam, The Netherlands (invited
paper).
Leendertse, W. L.; Bakker, K. J.; and Teunissen, E. A. H. 1997.
TBM-tunnellingin the Netherlands--an overview of research
and development. Tunnels for People:Proceedingsof the World
Tunnel Congress Vienna '97, Vol. 1I, 593-603. Rotterdam: A.A.
Balkema.

Volume 14, N u m b e r 2, 1999