Effect of soil–structure interaction and spatial variability of ground motion on seismic risk assessment of bridges - Seismic Risk Analysis and Management of Civil Infrastructure Systems - page 597

independently from the characteristics of the incoming seismic motion. On
the other hand, the dynamic characteristics of a bridge subjected to strong
ground motion are also evolutionary in time due to damage accumulation
in terms of frequencies of vibration and contribution of individual modes.
The collapse of the Hanshin Expressway during the Kobe (1995) earth-
quake due to signifi cant detrimental SSI effects (up to 100% increase) is
a notable example of the critical interplay between the soil, bridge and
earthquake characteristics (Mylonakis et al. , 2006b). Other studies (Sextos
et al. , 2003b) have also highlighted the importance of local soil and site
conditions when considering coupled soil-bridge interaction issues. The
impact of SSI effects on seismic bridge displacements may be underesti-
mated by a maximum factor of 1.5-2.5, if the local 1D/2D/3D basin site
effects have not been properly incorporated in SSI analysis. Numerous
works (Zhang and Makris, 2002a,b) have also shown that mechanical prop-
erties and geometry of the abutments signifi cantly affect the dynamic
response of short to moderate length overpasses; thus it may be highly
unconservative to introduce the abutment-embankment system fl exibility
without due consideration of the incoming seismic energy to the bridge
boundaries.

22.2.3 Consideration of soil nonlinearity

As the earthquake intensity increases, the behavior of the supporting soil
deposits quickly becomes nonlinear, thus introducing additional fl exibility
and damping at the soil-foundation interface. A large number of sophisti-
cated constitutive laws for soil materials have been developed and incor-
porated into modern FE codes (Sextos, 2011). Nevertheless, the numerical
simulation of concurrent inelastic mechanisms developing at the soil, foun-
dation and structure simultaneously is still computationally demanding and
a major challenge due to material and epistemic uncertainties. This can be
primarily attributed to the following:

• Despite the signifi cant advances in modern computer software, it is dif-

fi cult to achieve a balanced modeling refi nement for all materials as
most computer codes do not simultaneously support built-in features
for yielding mechanisms specialized for soil, concrete, steel and rubber.
Furthermore, coupling of different material laws often leads to signifi -
cant analysis convergence diffi culties.

• The data required to prescribe various yield parameters (i.e. von Mises,

Willam-Warnke, Mohr-Coulomb, Drucker-Prager) cannot always be
reliably assessed, thus, implicitly undermining the foreseen analysis
refi nenment. This can be to some extent tackled through a probabilistic
analysis framework.