Vulnerability Analysis of Large Concrete Dams using the Continuum Strong Discontinuity Approach and Neural Networks

Abstract

Probabilistic analysis is an emerging field of structural engineering which is very significant in structures of great importance like dams, nuclear reactors etc. In this work a Neural Networks (NN) based Monte Carlo Simulation (MCS) procedure is proposed for the vulnerability analysis of large concrete dams, in conjunction with a non-linear finite element analysis for the prediction of the bearing capacity of the Dam using the Continuum Strong Discontinuity Approach. The use of NN was motivated by the approximate concepts inherent in vulnerability analysis and the time consuming repeated analyses required for MCS. The Rprop algorithm is implemented for training the NN utilizing available information generated from selected non-linear analyses. The trained NN is then used in the context of a MCS procedure to compute the peak load of the structure due to different sets of basic random variables leading to close prediction of the probability of failure. This way it is made possible to obtain rigorous estimates of the probability of failure and the fragility curves for the Scalere (Italy) dam for various predefined damage levels and various flood scenarios. The uncertain properties (modeled as random variables) considered, for both test examples, are the Young’s modulus, the Poisson’s ratio, the tensile strength and the specific fracture energy of the concrete.