Interpretable AI Links Surface Roughness and Film Thickness to Pitting Resistance (IMAGE)
Caption
Decoupling of physics-informed descriptors revealing a model-supported transition near FT ≈ 2,200 nm in the higher-morphology regime.
An interpretable machine-learning framework relates four physical descriptors—film thickness (FT), surface roughness (SQ), crystallite size (CS), and substrate dislocation density index (FS)—to the pitting potential of steam-coated aluminum alloy. Random Forest modeling, SHAP, and Accumulated Local Effects analyses were used to examine descriptor importance and interactions. Within the higher-roughness range studied, the modeled SQ–FT interaction changed sign near FT ≈ 2,200 nm. The proposed shift from protective coverage to defect-related transport pathways is a model-supported interpretation requiring further experimental validation.
Credit
Professor Takahiro Ishizaki from SIT, Japan
Usage Restrictions
Credit must be given to the creator.
License
Original content