Numerical modelling and multi-objective optimization of the novel hydrocyclone for ultra-fine particles classification

Abstract

A novel hydrocyclone with arc inlet and convex cone is proposed to obtain the higher particles classification sharpness than the conventional design. The classification of the micron scale manganese dioxide particle is taken as a study case, and various methods are used to perform the modelling and optimization of the novel geometrical structure. Two performance indexes have been taken into consideration in the multi-objective optimization which are the classification sharpness, Ss and total pressure drop, ΔP. The numerical experiments designed via response surface methodology are carried out using computation fluid dynamics simulations of Eulerian–Eulerian strategy. The obtained data sets are utilized for modelling the performance indexes by support vector machine (SVM) and radial basis function neural network (RBFNN) approaches. The optimal structure with the classification sharpness of 0.956 is searched using the genetic algorithm (GA) in contrast to the conventional design of 0.849. Flow field analysis illustrates that the extended radial space, strengthened centrifugal field and inlet pre-classification effect of the optimal design improve the classification sharpness. However, the narrower inlet cross area leads to a higher pressure drop. The Pareto front of the multi-objective optimization is obtained using the NSGA-II algorithm to provide alternatives for the optimal performance of the novel hydrocyclone.