Improving Pitting Corrosion Prediction Accuracy in 316L Stainless Steel through the Synergy of Predictive Machine Learning and Reflective Microscopy

Accurately forecasting the locations of future pit formation sites on stainless steels (SS) holds significant practical value in both fundamental science and the corrosion industry, yet it poses a significant experimental challenge due to the need to localize imperfections in the passive film at a sub-nanometer scale. In this study, we tackle this issue by utilizing the combination of in-situ Reflective Microscopy (RM) instrumentation, optical modeling, and predictive machine learning (ML) methods, focusing on the prediction of pits location in SS316L alloy in chloride-rich media. Our findings indicate that the appearance of future pits is most often linked to slight ( 0.5%) decrease in the intensity of light reflected from the SS surface, yet this relationship is not always reciprocal. Optical modeling reinforced with ex-situ XPS suggest a chromium oxide deficiency in the surface films over these zones. This study provides proof of concept that RM enforced with ML approaches can be used as an accessible, highly precise tool to define zones of preferential pitting corrosion and highlights future promising directions for developing predictive corrosion monitoring systems.