Physics informed neural networks for phase field fracture modeling enhanced by length-scale decoupling degradation functions
Haojie Lian, Peiyun Zhao, Mengxi Zhang, Peng Wang, Yongsong Li
Abstract
The paper proposed a novel framework for efficient simulation of crack propagation in brittle materials. In the present work, the phase field represents the sharp crack surface with a diffuse fracture zone and captures the crack path implicitly. The partial differential equations of the phase field models are solved with physics informed neural networks (PINN) by minimizing the variational energy. We introduce to the PINN-based phase field model the degradation function that decouples the phase-field and physical length scales, whereby reducing the mesh density for resolving diffuse fracture zones. The numerical results demonstrate the accuracy and efficiency of the proposed algorithm.