Multi-objective optimization of cyclone performance via inlet velocity and particle properties using CFD and response surface methodology

Cyclone-based solid–solid classification is widely used in mineral processing and polymer recycling. However, operators still lack a clear and scientific method for selecting the inlet velocity. This motivated the development of a framework that relates operating conditions to the inertial difference of the particles. In this study, a modified Stokes number was adopted as the governing parameter, since it captures how differences in particle size and density affect their trajectories. A Large Eddy Simulation (LES) model was combined with a two-way-coupled Discrete Phase Model (DPM). The model was validated against benchmark data with less than 3 % tangential-velocity error and less than 5 % pressure-drop error. LES was selected instead of RANS because preliminary trials showed that unsteady vortex structures were excessively smoothed under RANS, reducing prediction reliability. The validated CFD method was integrated with a design of experiments approach. The logit-transformed response-surface model predicts that separation efficiencies above 85 % can be achieved for large inertial contrasts (Δd = 30 µm and Δρ = 6259 kg/m³) at the model-identified optimal inlet velocity. In the practical CFD runs, the highest achieved efficiency reached 64.3 %, yet the operational maps clearly indicate potential for further improvement when inlet velocity is tuned according to Stk*. Operating within the Stk*-guided range reduced the pressure drop by about 20 %. This reduction was most evident in simulations where the inlet velocity was slightly lowered from typical industrial settings. Overall, the Stk*-based model and the derived operational guideline for selecting inlet velocity in solid–solid cyclone systems. The findings also offer an alternative to the usual inconsistent methods used for setting inlet velocity.