A seismic response prediction method based on a self-optimized Bayesian Bi-LSTM mixed network for high-speed railway track-bridge system

The construction of China’s high-speed railway (HSR) network has reached earthquake-prone regions, necessitating a timely and accurate post-disaster quick prediction approach to ensure the safety of the HSR systems’ transportation lifeline. This study proposes a fast prediction method utilizing a Bayesian self-optimized bi-directional long short-term memory (Bi-LSTM) network to develop a fast prediction framework for the seismic response of the HSR track-bridge system. It describes a hierarchical clustering algorithm based on discrete wavelet decomposition. The results indicated that the proposed framework effectively predicts the nonlinear seismic response of HSR bridge structures. The model also showed the performance of the work migrate ability and robustness. In addition, the impact of different prediction locations on the HSR track-bridge system is minimal. The hierarchical clustering method based on wavelet decomposition can effectively reduce the number of inputs to the seismic training dataset while ensuring prediction accuracy.