Machine learning approach for process optimization of black nickel electroplating

Electroplating enhances the mechanical, thermal, and tribological properties of components in industries like aerospace, computing, pharmaceuticals, and telecommunications. To achieve high-quality electroplated coatings, control of process parameters is essential. This study focuses on the challenges associated with the black nickel electroplating process, particularly its nonlinearity which makes traditional linear methods inadequate. We employed machine learning techniques to develop models capable of predicting defects, coating colour, and coating mass specifically for black nickel electroplating process adhering to the MIL-P-18317 specification, a boric acid-free method that is more environmentally friendly but more sensitive to bath conditions, hence making process optimization more challenging. Our research addresses significant gaps in the literature, focusing on boric acid-free black nickel plating, which requires unique modelling approaches different from other electroplating processes. Unlike previous studies that mainly focused on predicting coating mass, our models also predict defects and colour, ensuring comprehensive quality control. The best-performing models achieved an F1 score of 0.9875 for defect prediction, an F1 score of 0.95 for colour prediction, and R \(^{2}\) $_{}^2$ = 0.9561, MAE = 0.0039, and MSE = 0.00001 for mass prediction. Additionally, we developed a novel optimization algorithm based on these models to fine-tune process parameters, ensuring that the resulting coatings meet strict standards for quality, appearance, and productivity. The validity of our approach was confirmed through experimental results. This research demonstrates how machine learning can help surface finishers, by providing effective strategies for optimizing processes in nonlinear scenarios, thereby improving product quality and productivity.