Surrogate modelling of corrosion inhibition finite element simulations using machine learning

Abstract A modelling approach that combines a previously developed 2D continuum finite element model with machine learning to support the design and evaluation of corrosion-inhibiting coatings. The FEM simulates the leaching of corrosion inhibition pigments from an organic coating and the resulting protection of the metal surface. This is conducted for a system of aluminium alloy 2024-T3 with an active protective coating loaded with lithium carbonate particles. A generated dataset from FEM results was used to train ML models to predict inhibitor concentration and corrosion current density based on geometric and material input parameters. A feature importance analysis was conducted to identify the most influential input variables, providing insight into the factors controlling the achievement of corrosion inhibition. Furthermore, a blind test was performed using five unseen cases that were not involved in the training phase. Finally, the trained models were applied to explore their use in coating design.