From Lab to Plant: Data-driven Modelling of an Electroplating Process

Electroplated zinc-nickel (Zn-Ni) coatings are commonly applied as corrosion protection for steel components in demanding industrial environments. However, due to high process complexity, coupled with a limited process transparency and irregular chemical monitoring, achieving resource efficiency and predictive process control still has its challenges. This paper presents a data-driven approach for monitoring and modelling of an alkaline Zn-Ni barrel plating process at process and product level using machine learning (ML) techniques. A methodology was developed within a cyber-physical production system framework for training the models based on a controlled laboratory environment (20 L bath volume) and knowledge transfer on an industrial electroplating plant (13,000 L bath volume). Different supervised algorithms were trained with a limited set of input features to predict metal concentrations of the bath, coating thickness, nickel ratio in the coating and aggregated coating quality indicator. Despite data scarcity and process variability, the selected best models demonstrate accurate prediction rates (> 86%). The results show that experimentally generated data can improve product-level modelling in the industrial environment, while process-level transfer remains limited when relevant industrial interventions, such as zinc dosing, are not recorded.