Basic HTML Version
42
Geotechnical News • March 2013
www.geotechnicalnews.com
desired average void ratio and relative
density. Of course the procedure of
model making was carefully con-
trolled, since it is well known that the
mechanical behavior of a reconstituted
granular soil is strongly dependent on
the deposition procedure both in the
lab (e.g. [11]) and in the centrifuge
([12]).
Four earthquakes were fired at 80g and
one at 40g, with variable nominal peak
acceleration amplitude and frequency.
The main features of each earthquake
are shown in Table 1, at the model and
prototype (bracketed values) scales;
their time histories, as recorded by
the reference accelerometer (Acc13 in
Fig. 3) are summarised in Fig. 4.
The experimental values of both bend-
ing moment, M, and hoop force, N, in
the lining were derived from the strain
gauges records during each seismic
event. It is apparent (e.g. fig. 5) that,
after the shaking, the residual values
of the internal lining forces are signifi-
cantly different from the initial condi-
tions. This behaviour was observed
almost systematically for any event in
all the models, and seems to indicate
that permanent deformations occurred
around the tunnel during shaking. This
is qualitatively consistent with the
observed densification of the sand dur-
ing the shaking, shown by the surface
settlements.
Round Robin
Tunnel Test
organisation
The Round Robin
Tunnel Test (RRTT)
was jointly pro-
moted by three ISSMGE Technical
Committees, i.e. TC104 (
Physical
modeling in Geotechnics
), TC203
(
Earthquake Geotechnical Engi-
neering
) and TC204 (
Underground
construction in soft ground
). All
participants were enabled to use the
selected test data, i.e. the reference
accelerograms (Fig. 4) and the results
of laboratory tests on LB sand, which
were delivered through website with
a restricted access. The analyses were
initially intended as ‘blind’ predic-
tions of the behavior of the first of the
selected centrifuge model tests (model
T3).
Six of the initial fourteen teams,
belonging to academic departments
of several countries (see Table 2),
completed the analyses in time for a
workshop organised after one year
from the launch, at the 2nd Interna-
tional Conference on PBD in EGE in
Taormina (2012). During the work-
shop the results of the numerical blind
predictions were presented to the
floor, thereafter compared and freely
discussed.
Each group adopted a different numer-
ical code and a different constitutive
model for the soil, as shown in A
significant issue for all the participants
was the calibration of the constitutive
model on the results of the laboratory
tests in order to correctly reproduce
the decay of soil stiffness from small
to large shear strain.
Although the interpretation of tests
to calibrate an advanced constitu-
tive model should be regarded as a
back-analysis of a non-linear boundary
value problem ([13]), it is commonly
accepted that a laboratory test is
Figure 4. Shaking applied to model T3.
Figure 5. Time histories of bending moment and hoop
force during shaking.
Table 2. Main features of numerical analyses.
Group
Adopted constitutive law
Numerical code
AUT (Greece) Visco-elasto-plastic model
ABAQUS (FEM)
UCT (Italy)
Visco-elasto-plastic model
ADINA (FEM)
TUD (Germany Hypoplastic (von Wolffersdorff
model)
TOCHNOG (FEM)
TVG (Italy)
Visco-elasto-plastic model
FLAC (FDM)
UTL (Portugal) Elastoplastic multi-mechanism
(Hujeux model)
GEFDYN (FEM)
NEW (UK)
Generalized plasticity
(PZ-III model)
SWANDYNE II (FEM)