A Novel and Efficient Hybrid Method to Develop the Fragility Curves of Horizontally Curved Bridges

This study presents a new hybrid method to develop seismic fragility curves for horizontally-curved steel I-girder bridges using Artificial neural network and logistic regression methods. The approach for developing fragility curves based on the assumption that engineering demand parameters follow the lognormal distribution for calculating the probability of damage occurrence. A sufficient number of input data including a set of earthquake ground motion records and macro-structural parameters together with the output data resulting from nonlinear structural analyses was assigned to neural network structure to achieve satisfactory approximations of responses. Logistic regression statistical method was used to determine the probability of occurrence or non-occurrence of limit states for earthquake ground motion parameters and structural characteristics. In this study, based on the estimation of engineering demand parameters, the proposed method is compared with the neural network method, simplified mathematical model and analytical method. The nonlinear time history analysis of three dimensional horizontally curve bridges were performed using the OpenSEES software. The statistical results indicate the accuracy and efficiency of the predicted limit state occurrence of the proposed method at a low computational cost. Comparison of fragility curves using the mentioned methods represent a proper estimation for slight, moderate, extensive and collapse limit states at different levels of seismic intensity.