GATE RECURRENT NEURAL NETWORK PREDICTION MODEL FOR DYNAMIC MECHANICAL RESPONSE OF PLATES UNDER EXPLOSIVE SHOCK LOADING

The deformation and damage modes exhibited by the target plate under explosive shock demonstrate significant nonlinearity, presenting a highly coupled relationship between temporal and spatial variables throughout the dynamic response process. Precise mathematical models to comprehend such physical phenomena are lacking, often necessitating reliance on data-driven approaches. However, obtaining comprehensive real-time experimental results remains challenging. To address these limitations, this study introduces a spatiotemporal decoupling network based on deep learning algorithms. By incorporating the time variable, the network examines the dynamic deformation process of the target plate. The proposed network employs fully connected units and gate recurrent units to construct a spatiotemporal decoupling framework for the dynamic response process variables. Consequently, the model facilitates the prediction of target plate deformation under arbitrary initial conditions such as charge and burst distance, enabling instant solutions for target plate deformation within blast fields.