Prediction of electrochemical corrosion behavior of magnesium alloy using machine learning methods

The corrosion of magnesium (Mg) alloys is a major challenge in many industrial applications because of their extreme reactivity, this greatly limits their applicability. The traditional methods in the prediction of corrosion rate depend on experimental or empirical models that can be time-consuming and expensive. Besides how to integrate experiments with faster machine-learning approaches is still a challenge of research. In recent years machine learning (ML) algorithms have proved and promised better magnesium corrosion predictability. In this study, different training algorithms such as linear regression (LR), decision tree (DT), extra tree (ET), random forest (RF), K-nearest neighbor (KNN), extreme gradient boosting (XGBoost), and Artificial Neural Network (ANN) were used to develop and predict corrosion rate of magnesium alloys based on magnesium chemical alloy composition as independent input variables. The data used for model preparation and validation were obtained from literature and experimental studies respectively. The feature selection methods were used to identify the most important and influential input parameters for magnesium alloy thus Shapley Additive Explanations (SHAP) interpretation was employed in this manuscript to facilitate feature visualization of the model. The results of the study show that the random forest (RF) algorithm has the greatest prediction impact than other machine learning algorithms. The magnesium alloy corrosion prediction model provides a better ability for the prediction of magnesium corrosion alloys. The study demonstrates the potential of ML models in the prediction of the corrosion behavior of Mg alloys and these models can be used to optimize an alloy composition for designing improved corrosion resistance of new Mg alloys.