Preprint
Aug 2026
Physics-Informed Neural Networks for Biot's Model via Fixed-Stress Splitting and Energy Natural Gradient Descent
A robust PINN-based solver is proposed, termed FS-ENGD-PINN, which synergistically integrates physics-based decoupling with geometry-aware optimization, and which consistently outperforms standard PINN formulations in terms of accuracy and robustness.
Kexin Sun, Qiang Liu, Ming-Cheng Feng et al.
· 0 citations