A Novel hybrid Multi-Objective optimization framework based on ELM-NSGA-II-TOPSIS: Case study of rare earth electrolytic cell

Abstract

Chemical equipment involves complex reaction mechanisms and multi-physical field coupling, posing monitoring challenges and limiting data-driven optimization. Integrating CFD-generated data with machine learning offers a promising approach for multi-objective optimization. However, practical implementation still faces significant challenges. In this study, an optimization strategy for chemical equipment was proposed, integrating machine learning, multi-objective optimization, and multi-criteria decision-making. Rare earth electrolytic cell was selected as a representative case. This system exemplifies a complex multi-physics system involving high temperatures, electric fields, and chemical reactions. Firstly, an electrothermal coupling numerical model was established and validated based on the physical property parameters collected on-site. The distributions of the electric field and temperature field were obtained. Moreover, the response surface method (RSM) model and the extreme learning machine (ELM) model were employed to model the electrolytic cell structure. After comparisons with the ELM and backpropagation (BP) neural network models, The R² of RSM, ELM and other BP neural network models are RSM (97.55%), BP (81.56%), BP-GA (93.20%), and BP-PSO (96.30%), ELM (98.32%), among which the ELM model has the highest accuracy. On the basis of adopting ELM model, the NSGA-II algorithm is combined with TOPSIS method to solve the multi-objective optimization problem of rare earth electrolytic cell. Different from the traditional NSGA-II method, this method uses objective weights to select the optimal solution, and effectively balances the conflicting optimization objectives. For the electrolytic cell studied, the optimized structure reduces the voltage drop by 90%, the current density inhomogeneity by 93%, and the error is less than 7.21%.

This study innovatively combined ELM model, NSGA-II algorithm and TOPSIS method to provide a new method for the optimization of complex chemical equipment with multi-physics coupling and data-model collaborative optimization under extreme working conditions taking rare earth electrolytic cell as an example.