Stochastic vibration prediction of long-span bridges under traffic load based on deep neural network of multi-mode information fusion

The accuracy prediction of the bridge response is crucial in the structural health monitoring and safety assessment of long-span bridges under vehicular loads. Hence, a novel method based on multi-mode information fusion for stochastic vibration prediction of long-span bridges is proposed in this study. Specifically, the dynamic equations for the random vehicle-bridge interaction system in the state-space form are deduced to provide a theoretical basis and physical guidance for the proposed method. On this basis, a surrogate model consisting of a variational mode decomposition and long short-term memory networks is developed for vibration prediction of bridges. In particular, the variational mode decomposition can be employed to extract the vibration mode information of the bridge in multi-frequency bands, which significantly improves the time-series analysis capability of the long short-term memory networks. The accuracy and robustness of the proposed method are verified by a numerical case and real bridge field measurements. The results show that the maximum prediction error of the proposed method is <4 %.