An expert control system using neural networks for the electrolytic process in zinc hydrometallurgy

The final step in zinc hydrometallurgy is the electrolytic process, which involves passing an electrical current through insoluble electrodes to cause the decomposition of an aqueous zinc sulfate electrolyte and the deposition of metallic zinc at the cathode. For the electrolytic process studied, the most important process parameters for control are the concentrations of zinc and sulfuric acid in the electrolyte. This paper describes an expert control system for determining and tracking the optimal concentrations of zinc and sulfuric acid, which uses neural networks, rule models and a single-loop control scheme. The system is now being used to control the electrolytic process in a hydrometallurgical zinc plant. In this paper, the system architecture, which features an expert controller and three single-loop controllers, is first explained. Next, neural networks and rule models are constructed based on the chemical reactions involved, empirical knowledge and statistical data on the process. Then, the expert controller for determining the optimal concentrations is designed using the neural networks and rule models. The three single-loop controllers use the PI algorithm to track the optimal concentrations. Finally, the results of actual runs using the system are presented. They show that the system provides ot only high-purity metallic zinc, but also significant economic benefits. (C) 2002 Elsevier Science Ltd. All rights reserved.