Corrosion Potential Prediction of Marine Engineering Steel Based on Machine Learning

Corrosion is a major cause of failure in marine engineering steels, resulting in large economic losses worldwide. This study combines marine corrosion knowledge with machine learning techniques to predict corrosion potential. Using experimental data collected from many published studies, five machine learning models were built in Python: K-Nearest Neighbors (KNN), Extreme Gradient Boosting (XGBoost), Gradient Boosting Regressor (GBR), Stacked Generalization (Stacking), and a Weighted Average Ensemble. Model prediction accuracy was improved through feature engineering and data augmentation. The XGBoost model performed best and achieved a coefficient of determination (R²) of 0.80 on the training set and 0.62 on the test set. Its mean absolute error (MAE) was 0.07 V and root mean square error (RMSE) was 0.09 V. The generalization gap was 0.179. Feature importance analysis revealed that Mn, Cr, and the Cr×Mo interaction are key factors influencing corrosion potential. This approach provides a accurate and interpretable technical solution to predict corrosion potential for marine engineering steels. This study offers valuable insights for optimizing steel composition and enhancing corrosion-resistant design.