An Artificial Neural Network Based Method for Seismic Fragility Analysis of Highway Bridges

Abstract

Fragility functions have become widely adopted in the seismic risk assessment of highway bridges, or even a transportation network. The computational effort required for a fragility analysis of highway bridges using incremental dynamic analysis (IDA) can become excessive, far beyond the capability of modern computing systems, especially when dealing with the structural parameter uncertainty in generating the fragility functions. In this paper, an artificial neural network (ANN) based prediction scheme for the generation of analytical fragility curves for highway bridges is presented. And the extremely time-consuming process in traditional analytical fragility methodologies is replaced by properly trained ANNs. The implementation of ANNs is focused on the simulation of median value and standard deviation of IDA curves at a given intensity level. The uniform design method (UDM) is proposed for selecting the training datasets for establishing a well-trained ANN model. It is observed that the proposed procedure can provides accurate estimates of fragility curves with relative short time compared to conventional procedures. The sensitivity study also reveals importance of material or geometric uncertainty in developing fragility curves of highway bridges.