Synchronously enhancing growth efficiency and corrosion resistance of plasma electrolytic oxidation coatings on magnesium alloys via interpretable machine learning

This article aims to synchronously enhance the performance of plasma electrolytic oxidation (PEO) coatings under a wide range of engineering services whilst minimizing process energy consumption upon magnesium alloys as model system through guidance of interpretable machine learning. Experimental data were compiled from literature and our own work. Key PEO descriptors, including electrolyte formulation and electrical parameters, were extracted and quantified as input features. Predictive models were developed for film formation rate, surface porosity, and corrosion current density. Among the three selected regression algorithms, extreme gradient boosting (XGBoost) achieved the highest prediction accuracy. SHAP analysis was then applied to interpret the dominant factors governing coating performance. Film formation rate is mainly controlled by fluoride concentration, whereas surface porosity strongly depends on silicate concentration. Excessive silicate increases coating porosity, while an appropriate oxidation time favors improved corrosion resistance. Based on these insights, a Pareto optimization framework was employed to design a baseline electrolyte and electrical parameter set to validate their scientific soundness and generalization of the model. The coating produced under such a scheme exhibits a combination of high growth efficiency and superior corrosion resistance, with a film formation rate of 1.8 µm/min, a surface porosity of 5.4%, and a corrosion current density of about 1.3 × 10 −8 A/cm². The proposed framework, which integrates interpretable machine learning with multi-objective optimization, offers an effective route for developing PEO process parameters, enabling the fabrication of low energy, high performance PEO coatings on magnesium alloys and providing a general data driven strategy for related engineering applications.