Analysis and optimization of multi-inlet gas cyclones using large eddy simulation and artificial neural network

The present study is aimed at optimizing the performance of multi-inlet gas cyclones. The current contribution is threefold. First, a design of experiments (DoE) has been conducted for three variables viz. the flow rate through the secondary inlet, the (square) cross-sectional area of the secondary inlet and the location of the top of the main inlet from cyclone roof. Second, the numerical simulations are performed using large eddy simulation (LES) to predict the Euler number, cut-off size and the collection efficiency for different combinations of the independent variables. The CFD simulation results are used to train an artificial neural network for three responses, namely the Euler number, the cut-off diameter and the overall collection efficiency. Moreover, the simulation results explain how the variations of the design variables affect the flow pattern and performance. Furthermore, the fitted surrogate model demonstrates that the most significant factors are the ratio of flow rates and the area ratio. Third, single-objective and multi-objective optimization studies are carried out using artificial neural network. The optimum design results in better performance than the conventional cyclones. (C) 2017 Elsevier B.V. All rights reserved.