Machine Learning Approaches to Model Galvanic Corrosion of Coated Al Alloy Systems

Abstract

Previous studies have shown how galvanic coupling susceptibility between stainless steel 316 or titanium alloy fasteners and coated aluminum alloy 7075-T6 depends on the chosen coating system and environmental factors such as relative humidity (RH) and chloride concentration. In this study, several machine learning models were developed to predict, analyze, and quantify galvanic corrosion arising between relatively noble fasteners and coated aluminum alloy panels. Different independent factors including pretreatment, primer coating, topcoat, RH, chloride concentration, fastener material, fastener quantity, existence of a defect, type of environment, and time of wetness were evaluated for their effect on galvanic coupling lost volume. Artificial neural networks (ANN), random forest regression (RFR), and multiple linear regression (MLR) were used to develop damage functions for galvanic corrosion. ANN, RFR, and MLR models all showed a reasonable fit for lost volume as a function of different inputs.