Machine Learning Assisted Prediction of Corrosion Characteristics in Zinc-Nickel Electroplating Processes

Abstract

Corrosion resistance of electroplated surfaces is a crucial functionality desired across multiple manufacturing industries. Zinc-Nickel (Zn-Ni) coatings are widely used for their excellent corrosion resistance and low costs. Corrosion resistance of the electroplated surface is defined by corrosion potential, corrosion current density, and percentages of coated metals. These characteristics are heavily influenced by the operating parameters of the electroplating process, such as bath pH, electrolyte temperature, current density, molar mass of additives, and metal ion concentrations. Determining the coating characteristics experimentally is time-consuming and resource-intensive. In this work, data published through multiple Zn-Ni electroplating studies are leveraged to compile a literature-derived dataset. A data-driven methodology is developed to accurately predict the coating characteristics directly from the process parameters. Following data cleaning and standardization, principal component analysis (PCA) is performed for dimensionality reduction. Support vector regression (SVR) and multilayer perceptron (MLP) regression models are trained for predicting coating characteristics. The regression models’ performance is evaluated on an unseen test dataset. The SVR and MLP regression models achieve R² = 0.84 to 0.87 across all four characteristics, with mean absolute errors (MAE) of 44 mV for E_corr, 0.53 mA/cm² for I_corr, and 1.5 wt. % for zinc and nickel weight fractions. The best models capture more than 84 % of the variance in the compiled dataset. Based on these performance metrics, the model predictions can be further used as effective screening tools to eliminate operating conditions while experimentally developing new electrolyte solutions for the electroplating process.