Introducing a Hybrid Physics-Informed Neural Network and Finite Element Model for Predicting Structural Deformation Under Dynamic Load

This study introduces a novel hybrid framework that integrates Physics-Informed Neural Networks (PINNs) with the Finite Element Method (FEM) to accurately predict structural deformation under dynamic loading conditions. While FEM remains a powerful tool in structural mechanics, its computational cost rises significantly with complex geometries and time-dependent simulations. To address this, the proposed hybrid model leverages the domain knowledge embedded in partial differential equations through PINNs, which are trained on both synthetic FEM data and governing physics laws. The model enables faster and more generalizable predictions of displacement fields by learning from limited simulation data while enforcing physical consistency. Numerical experiments on beam and plate structures subjected to varying dynamic loads demonstrate that the hybrid approach achieves high accuracy with substantially reduced computational effort compared to traditional FEM-only simulations. This work highlights the potential of combining data-driven and physics-based modeling to support real-time structural health monitoring and decision-making in engineering systems.