Predicting the Corrosion Rate of Al and Mg Alloys Coated by Plasma Spraying Method with Machine Learning

Abstract

Developing technology has increased the need for materials that are more economical in terms of cost and more reliable in terms of strength, chemical and physical properties in all industrial areas. This has necessitated the development of new materials or the improvement of existing material properties. Surface coating methods are used to improve existing material properties. In this study, Al and Mg alloys, which are considered as an alternative to steel material in terms of being lightweight materials, were coated with Al2O3 and TiO2 at different rates by plasma spraying method, and the corrosion behaviors of the coatings in different environments were predicted using machine learning methods. AA7075 and AZ91 non-metal materials were chosen as the substrate for the study. Different ratios of Al2O3 and TiO2 ceramic materials were coated on the substrates. To determine the corrosion resistance of the coated samples, corrosion experiments were carried out in 3.5% NaCl and 0.3M H2SO4 environments. Using the experimental results, corrosion rate values were estimated using machine learning algorithms such as XGBoost, Random Forest (RF) and artificial neural networks (ANN) methods, depending on the substrate material, corrosive environment and coating rates. At the end of the study, corrosion rate values were estimated with low error rates and the best estimate was obtained with the XGBoost method (0.9968 R2 value).