An enhanced optimum design of a Takagi-Sugeno-Kang fuzzy inference system for seismic response prediction of bridges

This paper introduces an efficient training method aimed at simulating bridge behavior and predicting seismic responses. Addressing the critical need for accurate and adaptive models in seismic response simulation, this study develops a method that integrates an adaptive fuzzy inference system with the Takagi-Sugeno-Kang (TSK) model, optimized through a Nelder-Mead-assisted chaos game optimization algorithm. By leveraging analytical data from previous time steps, the method optimizes TSK model parameters to simulate seismic behavior more accurately. Three training cases, utilizing 2, 4, and 8 previous time steps, were tested to assess prediction accuracy and adaptive capability. Data for training were gathered from dynamic time-history analysis results from 60 accelerometers with diverse seismic characteristics, while 6 additional accelerometers evaluated the system's performance. Performance assessment was conducted on the column and deck of an integrated concrete curved bridge. The results show that the optimized TSK model improves prediction accuracy, reducing the mean square error by 67.41 % and demonstrating its effectiveness as a computational tool for predicting complex bridge responses under seismic conditions.