Multi-Objective Multi-Variable Large-Size Fan Aerodynamic Optimization by Using Multi-Model Ensemble Optimization Algorithm

The constrained multi-objective multi-variable optimization of fans usually needs a great deal of computational fluid dynamics (CFD) calculations and is time-consuming. In this study, a new multi-model ensemble optimization algorithm is proposed to tackle such an expensive optimization problem. The multi-variable and multi-objective optimization are conducted with a new flexible multi-objective infill criterion. In addition, the search direction is determined by the multi-model ensemble assisted evolutionary algorithm and the feature extraction by the principal component analysis is used to reduce the dimension of optimization variables. First, the proposed algorithm and other two optimization algorithms which prevail in fan optimizations were compared by using test functions. With the same number of objective function evaluations, the proposed algorithm shows a fast convergency rate on finding the optimal objective function values. Then, this algorithm was used to optimize the rotor and stator blades of a large axial fan, with the efficiencies as the objectives at three flow rates, the high, the design and the low flow rate. Forty-two variables were included in the optimization process. The results show that compared with the prototype fan, the total pressure efficiencies of the optimized fan at the high, the design and the low flow rate were increased by 3.35%, 3.07% and 2.89%, respectively, after CFD simulations for 500 fan candidates with the constraint for the design pressure. The optimization results validate the effectiveness and feasibility of the proposed algorithm.