A Unified Physics-Informed Neural Network for Modeling Coupled Electro- and Elastodynamic Wave Propagation Using Three-Stage Loss Optimization

Researchers developed a Physics-Informed Neural Network (PINN) to model coupled electro- and elastodynamic wave propagation, a complex system governed by partial differential equations. Their model achieved impressively low global relative L2 errors of 2.34% for mechanical displacement and 4.87% for electric potential. This demonstrates PINNs as effective, mesh-free solvers for time-dependent systems, though challenges like error accumulation in stiff systems remain. The work validates AI's potential in scientific computing.