This study aims to address the urgent need for accurate prediction of the long-term corrosion performance of low-alloy steel in tropical marine atmospheric environments and to quantify the effects of environmental factors and steel composition on corrosion.
This study compared the predictive accuracy of four machine learning algorithms: Support Vector Regression, Multi-Layer Perceptron, Random Forest and Extreme Gradient Boosting (XGBoost). Subsequently, the best-performing XGBoost model was interpreted using SHapley Additive exPlanations (SHAP) and Accumulated Local Effects (ALE) to quantitatively assess the influence of features on corrosion.
The XGBoost model demonstrated the best predictive accuracy on the independent dataset and showed good generalization ability (R² = 0.841, MAE = 5.37 µm/a). SHAP and ALE quantified the influence of features and revealed the nonlinear threshold effects of features on corrosion.
This study provides insights into the long-term prediction of the corrosion performance of low-alloy steel and the formulation of targeted, condition-responsive corrosion prevention strategies.
