Research on dynamic response prediction in multi-degree-of-freedom systems using convolutional neural networks

Predicting dynamic responses is a classic problem in structural dynamics, categorized under forward problems. Traditional methods have been extensively developed but rely heavily on dynamic model parameters. In contrast, recent popular methods based on deep learning for dynamic response prediction do not depend on precise dynamic model parameters and hold promising application prospects. This paper proposes a method for estimating the dynamic response of linear elastic systems using convolutional neural networks. The performance of the prediction model is studied in the time domain for multi-degree-of-freedom systems. Additionally, the noise resistance of the proposed method is investigated in noisy environments. The reliability of the proposed method is demonstrated through numerical simulations and experimental validation.