Estimating pitting descriptors of 316 L stainless steel by machine learning and statistical analysis

This research paper presents a hybrid rule-based/machine learning (ML) approach for estimating pitting corrosion descriptors from high-throughput data obtained using Scanning Electrochemical Cell Microscopy (SECCM) on 316L stainless steel. Linear regression and artificial neural networks (ANNs) were employed to determine pitting corrosion descriptors. Non-parametric density estimation assessed the central tendencies of the distributions. Conditional medians provided more accurate results than conditional means due to their robustness to outliers. The study observed a passive range shortening with increased testing aggressiveness, attributed to delayed passive film stabilization rather than early passivity breakdown.

Leonardo Bertolucci Coelho, Daniel Torres, Vincent Vangrunderbeek, Miguel Bernal, Gian Marco Paldino, Gianluca Bontempi, Jon Ustarroz