A Machine Learning Approach for Predicting Passive Current Densities for Cr-passivated Fe-based alloys as a function of Physical, Chemical and Metallurgical Conditions

Stainless steel is widely used across industries due to its superior corrosion resistance, which is continuously optimized to enhance reliability and performance. This study employs machine learning to predict the passivation current density (ipass) of stainless steels in chloride-containing environments under varying pH and temperature conditions, aiming to establish a quantitative relationship between alloy composition and electrochemical passivation behavior. Unlike previous studies that focused on individual grades or limited environmental factors, this work integrates multi-alloy, multi-environment experimental and literature data into a unified dataset, enabling simultaneous analysis of compositional and electrochemical variables. A curated dataset comprising 52 refined literature data points and 202 new electrochemical measurements was used for model training. Ensemble support vector regression (eSVR) and extreme gradient tree boosting (eXGBoost) models demonstrated good generalization performance, with R2 values of 0.7 in the training set and 0.9 in the independent test set, effectively capturing the underlying trends in the data. Post-hoc model explanation identified pH as the most influential input, with Cr content, temperature, NaCl concentration, and Fe content also contributing substantially to the predictions. Increasing Cr content was associated with lower ipass, consistent with its established role as the principal passivating alloying element in Fe-Cr alloys. An interactive web application, CRSSI (Corrosion-Resistant Stainless-Steel Informatics), was developed to facilitate accelerated screening of alloy compositions and environmental parameters, enabling real-time prediction of passivation current and serving as a practical resource for data-driven design of corrosion-resistant stainless steels. Thus, the primary recommendation is to employ the developed ML framework to inform alloy screening and targeted data acquisition.