Performance analysis of a novel cyclone separator using RBFNN and MOPSO algorithms

Abstract

The current work is aimed at optimizing the performance of a cyclone with variable-pitch-length spiral baffles when the cone of the cyclone is roughened. The numerical simulations are performed utilizing the Reynolds stress method (RSM) and Eulerian-Lagrangian scheme to predict the collection grade efficiency, the Euler number (Eu), and the cut-off size diameter for various combinations of the performance objectives of the cyclone. The numerical results are utilized as inputs of the Radial Basis Function Neural Network (RBFNN) to create a black box and Multi-Objective Particle Swarm Optimization (MOPSO) algorithm for the optimization. Four optimal designs, including C53, D7, E28, and F1 are obtained using the Pareto solution and compared with the Hoekstra cyclone. It is demonstrated that the application of spiral baffles and the roughened cone has a major role in improving the collection efficiency of solid particles, especially the ones with a diameter smaller than 2 μm.