Aero-mechanical multidisciplinary optimization of a high speed centrifugal impeller
Chenxi Li,Jing Wang,Zhendong Guo,Liming Song,Jun Li
DOI: https://doi.org/10.1016/j.ast.2019.105452
IF: 5.457
2019-01-01
Aerospace Science and Technology
Abstract:An aero-mechanical multidisciplinary optimization was carried out for a high-speed centrifugal impeller, SRV2-O, by integrating a Self-adaptive Multi-Objective Differential Evolution (SMODE) algorithm, RANS solver technique, Finite Element Method (FEM), and data mining technique of analysis of variance (ANOVA). Specifically, the optimization of the impeller was conducted for the maximization of isentropic efficiency and the minimization of maximum stress. During optimization, constraints were imposed on the total pressure ratio at the optimization point and the mass flow rate at the choked point as well. Where, the former constraint intends to guarantee the working capability of the impeller while the latter tries to fix the working range. After optimization, six optimal Pareto solutions are finally obtained. The isentropic efficiency of the optimal solutions is increased by 2.07% at most while the maximum stress is decreased by 6.36% at most among the Pareto solutions. The better performance of optimal designs was demonstrated through detailed aerodynamic and mechanical analysis. Then, the ANOVA is used to explore the effects of variables on performance function in design space, it is found that, the design variables located at the meridian channel and the leading edge of the full blade have significant effect on aerodynamic performance. These variables are crucial to reduce the loss caused by shock wave at impeller inlet and the leakage flow at blade tip. Meanwhile, the design variables located near the leading edge of full blade root section have great effect on strength performance, as they are effective to decrease the bend of blades and thus reduce the maximum stress. Thereby, the better aeromechanical performance can be achieved by dedicated adjustment on the curves of both the shroud of meridional channel and leading edge of full blade. The results of ANOVA are consistent with the aero-mechanical analysis. Therefore, the effectiveness of aero-mechanical optimization and data mining framework is demonstrated.