Effective optimization on Bump inlet using meta-model multi-objective particle swarm assisted by expected hyper-volume improvement

This paper presents an efficient multi-objective optimization method, focusing on aerodynamic optimization of a diverterless supersonic inlet (DSI) in transonic and supersonic flight conditions. The DSI inlet, through scrutinizing the Bump shape has potential to attain greater aerodynamic performance on exit plane of inlet. However, the high cost of computational fluid dynamic (CFD) simulations raises a significant challenge in the DSI optimization process. In order to obtain solution set in few numbers of objective function calls, a meta-model multi-objective particle swarm optimization (MOPSO) method is proposed based on a self-adaptive Kriging surrogate model, and applied to solve this kind of costly black-box optimization problem. The Kriging model is updated by using a dynamic expected hyper-volume improvement (EHVI) sample metric, which is developed by analyzing disadvantages of the original sample criterion. With the help of the dynamic sample metric, simulation results show that the surrogate-based MOPSO algorithm can obtain plenty enough non-dominated solutions and achieve high precision in the approximation of the Pareto front. In terms of DSI inlet optimization, the bump shape is parameterized by free form deformation (FFD) method, and the total pressure distortions of inlet exit plane are treated as two minimization objectives under transonic and supersonic flight conditions. A well distributed non-dominated solution set is generated by the proposed algorithm within the context of a small call number of cost evaluations, and optimized inlet configurated by the selected solution has better aerodynamic characteristics compared with the initial inlet.