Multi-objective optimization of a gas cyclone separator using genetic algorithm and computational fluid dynamics

• Published in: Powder technology 2018-02, Vol.325, p.347-360
• Main Authors: Sun, Xun, Yoon, Joon Yong
• Format: Article
• Language: English
• Subjects: Classification; Computational fluid dynamics; Computer applications; Cyclone separator; Cyclone separators; Design; Fitness; Fluid dynamics; Genetic algorithm; Genetic algorithms; Hydrodynamics; Kinetics; Mathematical models; Multi-objective optimization; Multiple objective analysis; Optimization; Pressure; Pressure drop; Reproductive fitness; Response surface methodology; Sorting algorithms; Statistical analysis
• ISSN: 0032-5910; 1873-328X
• DOI: 10.1016/j.powtec.2017.11.012

In the present study, multi-objective optimization of a gas cyclone is performed using a genetic algorithm (GA) and computational fluid dynamics (CFD) techniques to minimize pressure drop and maximize its collection efficiency. The reference model is a well-optimized cyclone from a previous study. First, a screening experiment for seven factors is performed to determine the statistically significant factors. Then, to define the fitness function used in the GA, four of the factors are studied using the central composite design in the response surface methodology. The second-generation non-dominated sorting genetic algorithm is utilized to optimize the four significant factors of the cyclone according to the well-defined fitness functions, and 53 non-dominated optimum cyclone design points are suggested. The reasonable accuracy of the results from the GA is confirmed via CFD validation of five representative optimum points. The obtained Pareto front comprises important design information for the new cyclones. Finally, the performance and flow field of a representative optimal design are compared with those of the classical Stairmand model and the reference model. The optimal design reduces the pressure drop and cut-off size by 7.38% and 9.04%, respectively, compared to the reference model. In addition, compared to the Stairmand model, decreases of 19.23% and 42.09% are achieved for the pressure drop and cut-off size, respectively. [Display omitted] •A multi-objective optimization of a well-optimized cyclone is performed.•Seven geometrical factors and two objective functions are considered.•Genetic algorithm and computational fluid dynamics are used.•The obtained Pareto front includes important design information for new cyclones.