Fracture Analysis of Functionally Graded Plates under Thermo-Mechanical Loading Using Mesh Enriched Physics-Informed Neural Networks (M-PINNs)

Recently, physics-informed neural networks (PINNs) have emerged as a viable approach for solving various computational mechanics problems. In the present work, a method based on an enriched meshed physics-informed neural network framework is developed to model crack problems in functionally graded material (FGM) plates under combined thermal and mechanical loading. In this method, to capture singular behavior near the crack tip, the standard PINN formulation is enriched by including a crack tip function. In functionally graded materials, the material properties change along the width according to a gradation model, and transverse loads are applied alongside the thermal load. These learnable parameters of enriched PINNs are optimized to satisfy the singular field near the crack tip. The present formulation was implemented in PYTHON and validated against existing published benchmark Stress Intensity Factor (SIF) data. The developed method demonstrates excellent validation and computational efficiency under thermal as well as combined thermo-mechanical conditions.