Gui-Li Yang

Stress Field Prediction and DIC-Informed Spatial Regularization for AlSi10Mg Alloy Tensile Specimens Based on Graph Neural Networks

Highlights FiLM-GNN predicts von Mises stress with R2 = 0.9984 and RMSE = 4.60 MPa. The independent force branch reconstructs dense curves with R2 = 0.9957. DIC-informed spatial regularization improves DIC spatial shape loss by 3.21% with 0.232 MPa mean correction. The framework enables rapid full-field prediction with...