Machine learning-assisted risk assessment of pitting corrosion susceptibility of AA1050 in ethanol-containing fuels

Abstract

The ability to assess the risk of corrosion of metallic structures in particular environments holds considerable significance in the field of automotive industry. In recent years, machine learning has evolved into a crucial tool to evaluate the complex and multidimensional corrosion phenomena. In this paper, the special case of non-aqueous alcoholate pitting corrosion of AA1050 in ethanol-blended fuels with water and chloride contamination is examined via supervised machine learning techniques in order to distinguish between safe and unsafe conditions. The data space was created by conducting dedicated experiments with varying ethanol–fuel–water ratios, temperatures, and surface preparations. The classifier's performance rating of 0.87 (balanced accuracy) indicates an outstanding predictive ability and highlights the model's usefulness as decision support for subsequent experiments.