LAUSR

Abstract This paper presents a procedure for three-dimensional optimization of a transonic centrifugal compressor impeller with splitter blades by integrating 3D blade parameterization method, a genetic algorithm (GA), an artificial neural network, and a CFD solver. Because computational fluid dynamics (CFD) is a time-consuming method, an artificial neural network is coupled with GA to evaluate the objective function. SRV2-O, a typical high-pressure ratio centrifugal impeller, is selected as the test case. A good understanding of flow characteristics in the passage of SRV2-O is obtained using 3D Reynolds Averaged Navier-Stokes solver. Twenty-eight design variables defining the impeller blade angle distribution are used to parametrize the blade geometry. Isentropic efficiency of the impeller is selected as the objective function while the total pressure ratio and mass flow rate are defined as constraints. The optimization results indicate that the performance of the optimum geometry is improved in comparison with the original impeller at both design and off-design conditions. The isentropic efficiency is increased by 0.97% at the design point, and total pressure ratio and mass flow rate are increased by 0.74%, 0.65%, respectively.