Model construction of corrosion resistance of alloying elements for low alloy steel in marine atmospheric corrosive environment based on machine learning

The study focused on constructing a machine learning model, considering the interaction of alloying elements on corrosion resistance of low alloy steels in the marine atmospheric environment. Spearman's analysis was applied, and the relationship between alloying element and corrosion rate was evaluated based on random forest (RF) importance and Shapley additive explanation (SHAP) analysis. The prediction performance of the six models (RF, multilayer perceptron (MLP), ridge regression (RR), K-nearest neighbor regression (KNN), logistic regression (LR), and support vector machine (SVM) was compared by using the preferred dominant elements as input variables. Afterwards, a high-precision corrosion rate prediction model based on RF was constructed. Finally, the generalizability of the model was demonstrated using 10 lines of steel corrosion data from several new marine atmospheric environments.