Accurate prediction of pitting corrosion in aluminum alloys via integrated multi-model methods

Abstract

As one of the major corrosion forms of aluminum alloys in atmospheric environments, pitting corrosion is characterized by stochastic development and non-uniform progression, which challenges the accurate prediction of pitting corrosion behavior. This study employs an integrated approach combining laboratory-accelerated corrosion testing with finite element modeling (FEM) to elucidate critical environmental factors governing localized corrosion behavior. The incorporation of micro-galvanic current density parameters derived from FEM analysis demonstrates significant prediction capacity enhancement, achieving 17.9 ​% and 35.5 ​% (MAE values) improvements in pit area and depth prediction accuracy respectively compared to conventional experimental data-driven machine learning approaches. Furthermore, the developed machine learning framework enables probabilistic prediction of pit dimension distributions, establishing a holistic methodology for comprehensive early-stage pitting assessment in aluminum alloys.