Machine learning-based prediction of corrosion resistance in pulse electrodeposited coatings

Predicting corrosion properties of electrodeposition enabled alloy coatings remains challenging owing to incomplete and heterogeneous data generated under complex, dynamic corrosive conditions and multiple interacting process parameters. We introduce a data-driven machine learning (ML) approach to predict corrosion potential ( E corr ) and corrosion current density ( i corr ) of multi-element alloy coating based on alloy composition and process-system parameters of pulse electrodeposition (PED). A curated dataset of 96 experiments integrating pulse parameters, alloy descriptors, and electrolyte features collected from the literature was used to train and evaluate multiple algorithms. Gaussian process models achieved overall high accuracy for predicting E corr (R² = 0.935) and i corr (R² = 0.67) over other models. Feature attribution via permutation importance and SHAP analysis revealed distinct mechanistic hierarchies: E corr was predominantly governed by alloy thermodynamic descriptors (electron affinity, mixing entropy, melting temperature difference, melting temperature), while i corr was primarily influenced by electrolyte composition (chloride, sulfate ion concentration), thermodynamic descriptor (melting temperature difference) and pulse parameter (duty cycle). Partial dependence plots (PDP) and individual conditional expectation (ICE) plots further revealed optimal mix design parameters for achieving the desired corrosion properties ( E corr and i corr ).