Seismic safety of gravity dams:: From snake table experiments to numerical analyses

The purpose of this paper is to present shake table experiments conducted on four 3.3-m-high plain concrete gravity dam models to study their dynamic cracking and sliding responses. The experimental results are then compared with a smeared cracked finite-element simulation using a nonlinear concrete constitutive model based on fracture mechanics. For the sliding mechanism, the numerical simulations use rigid body dynamics with frictional strength derived from the Mohr-Coulomb criterion. From the cracking test, it is shown that a single triangular acceleration pulse could initiate and propagate a crack. The numerical correlation with the observed response is good. However, viscous damping varies experimentally from 1% in an uncracked situation to over 20% in a partially cracked case. For the sliding mechanism when the critical acceleration is exceeded, it is shown experimentally that sliding could occur due to a single triangular acceleration pulse. For actual seismic records, the cumulative sliding displacement for a given peak ground acceleration due to low frequency western North American records can be 3-4 times larger than the corresponding sliding displacements due to high frequency eastern records. For simplified pseudostatic or pseudodynamic sliding analyses, the concept of an effective acceleration is developed.