Professional Documents
Culture Documents
1502
SECOND INTERNATIONAL CONFERENCE ON PERFORMANCE-BASED
DESIGN IN EARTHQUAKE GEOTECHNICAL ENGINEERING
May 28-30, 2012 - TAORMINA (ITALY)
Civil Engineer, MSc, PhD Candidate, Department of Civil Engineering, Aristotle University of
Thessaloniki, Greece, e-mail: gtsinidi@civil.auth.gr
2
Professor, Department of Civil Engineering, Aristotle University of Thessaloniki, Greece, e-mail:
kpitilak@civil.auth.gr
Detail
0. 3
Invertheaveup
to0.5m
.4
Backfillfoam
concrete
Previous
shotcrete
shell
Figure 1. Collapse of Bolu tunnel during the 1999 Kocaeli earthquake (after Kontoe et al., 2008)
Generally, moderate to heavy damages were observed for PGAs larger than 0.5g, whereas for PGAs
smaller than 0.2g, none to slight damages were reported. The lining type and the soil-lining interface
conditions are of prior importance for the seismic behavior of a tunnel. Unlined tunnels or tunnels
constructed by masonry found to be more vulnerable.
Tunnels have geometrical and conceptual features that make their seismic behavior very distinct from
aboveground structures (i.e. Owen and Scholl, 1981 etc.). The ground deformations, introduced by the
surrounding soils, are prevailing, while the inertial forces are of secondary importance. During an
earthquake, tunnels can be affected by: (i) ground shaking and/or (ii) permanent displacements by ground
failure (i.e. liquefaction, slope instabilities, fault displacements). During ground shaking the tunnel can be
deformed in various modes, both in the longitudinal and transverse direction, i.e. longitudinal axial
deformation, longitudinal bending, cross sectional compression and cross sectional ovaling (Owen and
Scholl, 1981). The latter is of prior importance, as it can cause large stresses on the tunnels lining.
Seismic analysis methods
Several methods have been proposed in the literature for the seismic design, ranging from closed form
solutions (i.e. to compute the lining internal forces) and simplified uncoupled methods, to the more
accurate full dynamic time history analysis of the soil-tunnel system (i.e. Wang, 1993, Penzien, 2000,
Hashash et al., 2001, ISO 23469, 2005, FWHA, 2009 etc). A comprehensive review is made by Pitilakis
and Tsinidis (2012). The results of these methods may significantly vary, demonstrating the relative lack
of knowledge regarding the seismic behavior and design of tunnels.
The proper estimation of several crucial parameters like of the design input motion, the distribution and
the magnitude of the seismic shear stresses around the tunnel and the impedance functions, adequate for
circular tunnels, are among the open issues that need further research. To this end, dynamic centrifuge
tests were carried out on circular tunnel-models embedded in dry sand. The tests were performed in 2007
at the geotechnical centrifuge facility of the University of Cambridge (Schofield Center), by researchers of
University of Napoli Federico II, within the framework of the ReLUIS Project (2005-2009) (i.e. Bilotta et
al., 2009, Lanzano et al., 2009, Lanzano et al., 2010 etc).
(b)
(c)
(a)
Figure 2. (a) Turner beam centrifuge, (b) Tunnel-model with strain gauges, (b) Tunnel-model
placement in the laminar box (after Bilotta et al., 2011)
500mm
124mm
124mm
Accelerometersensing
directiontowardsleft
110mm
LVDT045
LVDT059
130mm
Acc9
Acc6
Acc14
Acc12
SW
NE
Acc3
Acc4
Acc1
Acc10
ScaleFactorN=80
SE
Accelerometersensing
directionvertical
Referenceaccelerometer
recordinginputsignal
Acc16
Acc7
125mm
NW
Tunneld=75mm
57.5mm
45mm
57.5mm
290mm
Acc8
144mm
Acc15
Acc11
Acc5 DryFraction
Esand
Dr=75.9%
Straingauge
LVDT
Acc13
Shakedirection
G/Go
23.2m
1g
0.001
GD
0.01
0.1
50
0.8
40
0.6
30
Fitting
TSTests
RCtests
0.4
20
0.2
Displacement
constrains
a(t)
(a)
DT(%)
110mm
10
0
0.0001
0
0.001
(b)
0.01
(%)
0.1
40.0m
Figure 4. (a) Numerical model in ABAQUS, (b) G--D curves adopted in the analyses (black solid
line) vs. triaxial shear and resonant column tests results (experimental results after Visone, 2009)
Table 1 Mechanical properties in case of visco-elastic dynamic analyses of the coupled system
Test
case
Vsm
(m/sec)
(t/m3)
Gm
(MPa)
EQ1
EQ2
EQ3
EQ4
136
129
126
116
1.55
28.8
25.8
24.5
20.9
All
116
1.55
20.9
Rayleigh damping
parameters
a
b
Equivalent linear analyses
3.0
0.1371
0.0039
5.0
0.2218
0.0047
0.333
7.0
0.4423
0.0062
12.0
0.8815
0.0103
Elasto-plastic analyses
0.333
10.0
0.734
0.0085
v
Damping
(%)
()
()
c (MPa)
38
0.001
Gmax(MPa)
20
40
10
60
Depth(m)
0.06
0.12
0.17
0.23
0.29
Depth(m)
0.00
15
Shearmodulus(kPa)
20
25
30
0.00
0.06
0.12
0.17
0.23
0.29
TestresultsEQ1
TestresultsEQ3
EERAEQ1
EERAEQ3
TestresultsEQ2
TestresultsEQ4
EERAEQ2
EERAEQ4
(a)
(b)
Figure 5 (a) Small strain shear modulus distribution to depth at the final iteration of the
procedure, (b) Comparisons of the mobilized shear moduli computed from the 1D equivalent linear
analyses in EERA with the derived from experimental data shear moduli (after Lanzano et al.,
2010)
NUMERICAL PREDICTIONS VS EXPERIMENTAL RESULTS
In this section, indicative numerical predictions are presented, compared with experimental records, and
discussed, in terms of accelerations and dynamic increments of bending moments of the tunnel lining. All
the comparisons are made in model scale.
Accelerations
As mentioned the acceleration time histories were recorded at several points in the soil deposit and on the
laminar box, by miniature accelerometers forming three arrays, namely the reference array, the free field
array and the tunnel array. In Figures 6-10 indicative computed acceleration time histories are presented
and compared with the recorded data. It is noted that both time histories were filtered. Considering the
general character of the assumptions we made for the numerical simulation, as we did not have all the
experimental data, the numerical predictions are compared to the experimental results reasonably well.
t(s)
0
0.5
1.5
Acc4
0.5
8
4
0
4
8
A(g)
t(s)
Acc7
0.5
0.55
0.6
0.65
0.7
t(s)
0.5
0.55
0.6
0.65
Acc7
Numericalprediction
t(s)
0
1.5
A(g)
8
4
0
4
8
t(s)
Acc8
8
4
0
4
8
A(g)
8
4
0
4
8
A(g)
A(g)
Acc4
0.5
1.5
8
4
0
4
8
Acc8
A(g)
8
4
0
4
8
0.7
t(s)
0.5
0.55
0.6
0.65
0.7
Experimentalresults
Figure 6. Indicative acceleration time histories for EQ1; Experimental records vs. equivalent linear
analysis results
Acc7
t(s)
0
0.5
1.5
Acc4
0.5
A(g)
t(s)
0.5
0.55
0.6
0.65
8
4
0
4
8
0.7
t(s)
0.55
0.6
0.65
Acc7
Numericalprediction
t(s)
0
Acc7
0.5
Acc8
1.5
A(g)
8
4
0
4
8
t(s)
8
4
0
4
8
A(g)
A(g)
A(g)
8
4
0
4
8
0.5
1.5
Acc8
8
4
0
4
8
A(g)
Acc4
8
4
0
4
8
0.7
t(s)
0.5
0.55
0.6
0.65
0.7
Experimentalresults
Figure 7. Indicative acceleration time histories for EQ1; Experimental records vs. Mohr-Coulomb
analysis results
0.2
0.4
0.6
0.8
1
Acc3
0.2
0.4
0.6
0.8
1
Acc6
8
4
0
4
8
0.4
0.45
t(s) 0.5
Acc9
t(s)
0
0.2
0.4
0.6
0.8
1
Acc9
8
4
0
4
8
A(g)
A(g)
A(g)
8
4
0
4
8
t(s)
8
4
0
4
8
A(g)
A(g)
t(s)
Acc6
8
4
0
4
8
A(g)
Acc3
8
4
0
4
8
0.4
Acc7
Numericalprediction
0.45
t(s) 0.5
0.4
0.45
t(s) 0.5
Experimentalresults
Figure 8. Indicative acceleration time histories for EQ2; Experimental records vs. Mohr-Coulomb
analysis results
14
7
0
7
14
Acc14
t(s)
0.45
Acc5
0.45
t(s) 0.9
Acc14
14
7
0
7
14
0.25
t(s)
0.3
0.35
0.45
t(s) 0.9
Acc15
14
7
0
7
14
0.25
A(g)
A(g)
14
7
0
7
14
0.25
0.9
A(g)
Acc15
14
7
0
7
14
A(g)
A(g)
A(g)
14
7
0
7
14
0.3
Acc7
Numericalprediction
t(s) 0.35
0.3
t(s) 0.35
Experimentalresults
Figure 9. Indicative acceleration time histories for EQ3; Experimental records vs. equivalent linear
analysis results
Acc5
20
10
0
10
20
Acc6
0.45
Acc5
0.3
t(s) 0.35
Acc8
0.45
t(s) 0.9
Acc6
20
10
0
10
20
0.25
0.45
20
10
0
10
20
0.25
t(s) 0.9
Acc8
A(g)
A(g)
20
10
0
10
20
0.25
t(s) 0.9
A(g)
20
10
0
10
20
A(g)
A(g)
A(g)
20
10
0
10
20
0.3
Acc7
Numericalprediction
t(s) 0.35
0.3
t(s) 0.35
Experimentalresults
Figure 10. Indicative acceleration time histories for EQ4; Experimental records vs. Mohr Coulomb
analysis results
Experimentaldata
0.6
0.3
0
0.9
0.6
0.3
0
100
200
f(Hz)
300
MCanalyses
1.2
Amplitude
0.9
EQLanalyses
1.2
Amplitude
Amplitude
1.2
0.9
0.6
0.3
0
100
200
f(Hz)
300
100
200
f(Hz)
Figure 11. Fourier Spectra of Acc8 as computed from numerical and experimental results for EQ3
The relatively minor discrepancies are mainly attributed to the difference of the assumed soil stiffness and
damping with the actual one. Generally, the predictions of the Mohr-Coulomb analysis were closer to the
experimental results. This is again attributed to the different soil stiffness assumed in the two analyses
cases. Actually, when applying the Mohr-Coulomb criterion, the small strain shear modulus was assumed
to be smaller than the equivalent static analyses for the test cases EQ1, EQ2 and EQ3.
300
M(Nmm/mm)
Experimentaldata
NW
SW
6
4
NE
SE
2
0
2 0
3
t(s)
EQ1
0.05
0.04
0.03
0.02
0.01
0
60
0.1
0.14
0.12
0.1
0.08
0.08
0.06
0.04
0.02
0
60
0.3
EQ3
EQ4
0.25
|Mpk|(Nmm/mm)
|Mpk|(Nmm/mm)
EQ2
0.12
|Mpk|(Nmm/mm)
|Mpk|(Nmm/mm)
0.06
0.2
0.15
0.06
0.04
0.02
0
0.1
0.05
60
0
0
2|Mpk|
60
EQLanalyses
Experimentaldata
MCanalyses
Wang(1993)
Figure 13. Bending moment increments; Numerical predictions vs. experimental data
CONCLUSIONS
A first attempt to model numerically a dynamic centrifuge test of a circular tunnel embedded in dry sand is
presented. The simulation was done in the framework of a blind prediction test within the Round Robin
Tunnel Test (RRTT) organization. One test (T3), of a series of centrifuge tests that were performed on
circular tunnels within the ReLUIS Project, at the Schofield Centre of University of Cambridge, is used as
the benchmark case. The test was performed on an aluminum circular tunnel model embedded in dry sand
and excited with sine wavelets of increasing amplitude and frequency, under centrifugal acceleration of
80g. Full dynamic time history analyses of the coupled soil-model system were performed using
ABAQUS, under plane strain conditions. We were forced to make several assumptions regarding crucial
parameters of the problem as the experimental data were not known at that time. In a first series of
analyses the soil non-linear behavior was modeled as a linear visco-elastic material according to the
equivalent linear approach, while in the final analysis an elastoplastic Mohr Coulomb material was
adopted. The soil-tunnel interface was modelled using a Coulomb friction model.
Figure 14. Plastic deformations around the tunnel at the end of the dynamic analysis
The numerical predictions in terms of soil acceleration time histories and dynamic increments of bending
moments at several locations of the tunnel lining are in good, if not excellent (i.e. accelerations)
agreement, considering the absence of detailed experimental data. Having all the experimental data,
further analysis will be performed, trying to better model the test and better understand the seismic
behavior of this type of structures.
ACKNOWLEDGEMENTS
The support we had by Prof. Emilio Bilotta and Prof. Francesco Silvestri, during the running of the Round
Robin numerical test program, is kindly acknowledged. The work is partially supported by the SERIES
research project.
REFERENCES
ABAQUS (2009). Analysis Users Manual Volumes I - IV v6.9 [Computer Program], Dassault
Systmes, SIMULIA Inc, USA.
Bardet, J. P., Ichii ,K. and Lin, C. H. (2000). EERA a Computer Program for Equivalent-linear Earthquake
site Response Analyses of Layered Soil Deposits. Univ. of Southern California, Dep. of Civil Eng.
Bilotta, E., Lanzano, G., Madabhushi, S.P.G., Russo, G., Santucci de Magistris, F. and Silvestri, F. (2011)
RRTT, Round Robin numerical Test on Tunnels under seismic loading A joint venture between
TC104, TC203 and TC204.
Bilotta, E., Lanzano, G., Russo, G., Silvestri, F. and Madabhushi, S.P.G. (2009). Seismic analyses of
shallow tunnels by dynamic centrifuge tests and finite elements. Proc. 17th Int. Conf on Soil
Mechanics and Geotechnical Engineering, Alexandra, Egypt.
Brennan, A. J., Thusyanthan, N. I. and Madabhushi, S.P.G. (2005). Evaluation of shear modulus and
damping in dynamic centrifuge tests. Journal of Geotechnical and Geoenvironmental Engineering,
ASCE, Vol. 131, No. 12, pp. 1488 1497.