Construction and Optimization of Corrosion Map in a Broad Region of Acidic Soil via Machine Learning

Abstract

Machine learning has been widely applied to exploring the key affecting factors for metal corrosion in some local regions. However, there is a lack of systemic research and a practicable prediction model for metal corrosion in a broad region. In this paper, the corrosion map of Q235 steel in a broad region of acidic soils of Hunan province of Central China was constructed and optimized via field experiment and machine learning. Both the experimental and optimized corrosion maps confirmed that the corrosion rate of the steel decreased from the western to the eastern part of the province. The concentrations of pH, F−, Cl−, NO3−, HCO3−, K+, and Mg2+ were the key affecting factors in the broad region of acidic soils of the province. Among them, the contribution rate of the HCO3− concentration was higher than that of other factors. The optimization model based on the ordinary least squares could be used for the optimization of the corrosion map of steels in a broad region of acidic soils. The optimized corrosion map was a good alternative to the estimation methods for the corrosion rate of steel in soil.