Development of Seismic Fragility of Concrete Bridge with Column Ductility Measure and Neural Network Approach

Document Type : Research Paper


Department of Structural Engineering, Science and Research Branch, Islamic Azad University, Tehran, Iran


The objective of the present research approach based on soft computing (neural network) in the evaluation of seismic fragility of the highway bridge. In addition to the empirical methods and expert’s judgmental, seismic fragility curves are often determined by using analytical method in Structures, recently. The derivation of seismic fragility curves of the horizontal curved bridge based on the neural network with a focus on concrete column ductility measure by using 129 seismic records is performed. Earthquake records have been chosen from the PEER strong motion database and scaled on 0.1g to 1.3g. By using 1677 nonlinear dynamic analysis, incremental dynamic analysis (IDA) curves was drawn. Characteristics of earthquake ground motion as input and extraction of nonlinear dynamic analysis of columns ductility as output, are variables in building the neural network. Feature Extraction different records in different seismic intensity represents a seismic record neural network. Obviously, these characteristics are properly associated with structural damage. By transforming collection features different seismic record (matrix n×m) to the data with the same characteristics of seismic input (matrix p×m, p<m) through factor analysis, Neural network prediction in order to determine the response of structures, computational effort is much reduced with acceptable accuracy.