[P] P24~P35 Computational Science

[P27] Analysis of Alloy-Dependent Features Reflected in Fracture Surfaces

○Yuma Takenaka¹, Tokuteru Uesugi¹, Naoyuki Hamada², Shunsuke Kita², Genshiro Sawa², Tomotake Hirata² (1. Osaka Metro. Univ., 2. Osaka Research Institute of Industrial Science and Technology)

Keywords:

Fracture surface analysis,Convolutional neural network,Class activation map,Explainable artificial intelligence,Alloy type classification

Alloy-dependent features in fracture surface images were analyzed using fracture morphology and alloy classification models. Correlation analysis of Grad-CAM maps revealed that alloy-specific features contribute to fracture morphology classification.