Background: The Hall-Herault aluminum electrolysis process is the only industrial method for extracting and producing pure aluminum from alumina. This process produces molten aluminum by dissolving alumina in a cryolite bath and applying an electric current. Despite the widespread use of this method, there are significant challenges in the productivity and production rate of this method that affect the cost and efficiency of the process. Therefore, optimizing the operating parameters (electric current, voltage, temperature, cryolite bath composition, etc.) actively and accurately can have a direct impact on the cost, efficiency, and operational stability of this process.
Objective: The main objective of this research is to find the control values of the aluminum production cell so that the highest production rate is achieved with the lowest amount of energy.
Methodology: The method of this research is based on a hybrid approach in which numerical simulation, machine learning, and metaheuristic algorithms are used in an integrated manner. In this research, data is generated by a simulator based on ODE dynamic equations and is used to train a machine learning model. Finally, the NSGA-II algorithm finds the best control parameters with the help of the artificial intelligence model.
Findings: Based on the proposed model, the final Pareto Front set was constructed and the optimal operating point was selected based on the management policies.
Conclusion: The results of this research showed that the combination of machine learning to create fast predictive models and multi-objective optimization algorithms simultaneously allows the production of a comprehensive and diverse set of optimal solutions. With the geometric analysis of the Pareto line along with quantitative decision criteria, it is finally possible to choose an operating point for the aluminum electrolysis cell in an informed, reasoned and systematic manner. This approach not only optimizes efficiency and cost