CFD-based optimization of dynamic cyclones with variable vortex length using GMDH artificial neural network

Dynamic cyclone separators with adjustable vortex length are widely used in industrial applications such as particle collection and air pollution control. However, optimizing their performance remains a challenge due to complex fluid–particle interactions. This research introduces a three-step multi-objective optimization framework for dynamic cyclones with adjustable vortex length. Initially, computational fluid dynamics (CFD) simulations are utilized to examine airflow behavior in different cyclone designs. The Reynolds-averaged Navier-Stokes (RANS) equations, combined with the Reynolds stress turbulence model, are employed to model turbulence. The Eulerian-Lagrangian approach is used to track particle motion, while the Discrete Random Walk technique simulates velocity variations. In the second phase, data obtained from the numerical simulations is used to construct objective function models, focusing on minimizing pressure drop and maximizing collection efficiency. These models are developed using artificial neural networks based on the Group Method of Data Handling (GMDH). The final step involves optimizing the cyclone designs through the non-dominated sorting genetic algorithm (NSGA). The Pareto front is generated and analyzed, offering valuable insights into cyclone design improvements. The findings highlight that an optimized design for dynamic cyclones with variable vortex length can only be achieved through a systematic multi-objective optimization approach.