Multi-objective optimization of radially stirred tank based on CFD and machine learning

Structural optimization is essential to improve the performance of mixing equipment. An efficient optimization strategy based on computational fluid dynamics, machine learning, and the multi-objective genetic algorithm was proposed to predict and optimize the performance of the stirred tank. Single-factor analysis was performed to study the effects of structural parameters on power consumption and mixing time, which were reduced by 16.0% and 1.4%, respectively, in the optimized stirred vessel. To further optimize the stirred tank geometries and maximize the integrated performance, XGB coupled NSGA-II were utilized to minimize the power consumption and mixing time. The optimal design parameters from the Pareto front were identified by two well-known decision-making methods (LINMAP and TOPSIS), which decreased power consumption and mixing time by 12.3% and 13.4% compared to the stirred tank with the baseline structure. This research further confirmed the accuracy and reliability of the machine learning-based optimization method.