An Optimized Inference System for Predicting Seismic Responses of Bridges Considering Structural Behavior

Artificial intelligence has emerged as a powerful tool for simulating nonlinear behavior and predicting the seismic response of structural systems. This study presents an adaptive approach based on an optimization algorithm to train the Takagi-Sugeno-Kang (TSK) fuzzy inference model for estimating seismic responses of bridge structures using analytical data. The optimization algorithm is employed to fine-tune the TSK model parameters by minimizing prediction errors on the training dataset. The adaptive training process leverages feedback from ground acceleration and structural accelerations measured with sensors. Data for training is derived from time history analyses using 30 distinct ground motion records with varying seismic characteristics, and an additional 6 records are used to test the model’s performance. The charged system search (CSS) algorithm is used in this research to adjust the parameters of the TSK model and its performance is compared with Adaptive Neuro-Fuzzy Inference System (ANFIS) and multi-layer perceptron (MLP) models. The efficacy of the proposed inference system is evaluated on a real-scale reinforced concrete bridge and with structural behavior considered. The charged system search (CSS) algorithm demonstrated superior performance in optimizing the TSK model parameters, with prediction accuracy notably improving the prediction. Results indicate that the TSK model, enhanced with an optimization algorithm, provides an efficient computational method for predicting seismic responses.