Aerodynamic analysis and design optimization of a centrifugal compressor impeller considering realistic manufacturing uncertainties

Centrifugal compressor impeller blades inevitably suffer from manufacturing uncertainties. Such manufacturing uncertainties result in geometric deviations of blade profiles, and have been increasingly recognized to be detrimental to compressor performance. However, relevant studies on the uncertainty analysis and design optimization of centrifugal impeller blades are quite rare. The effect of realistic manufacturing uncertainties on impeller aerodynamic performance and flow field still remains unclear. Here we present a comprehensive investigation on deterministic analysis, uncertainty analysis and design optimization of a centrifugal compressor impeller with considerations of three-dimensional realistic manufacturing uncertainties. Through statistical analysis of blade profile measurements of machined centrifugal impellers, both the blade camber and thickness errors are confirmed to obey Gaussian distributions. The deterministic analysis results show that the positive thickness errors near the main blade leading edge are mainly responsible for impeller performance degradations because of aggravated flow separations especially at large flow rate. A bi-fidelity metamodel, namely sparse grid stochastic collocation-support vector regression (SGSC-SVR), is used to balance the efficiency and accuracy of the uncertainty analysis. The mean impeller performance is found to become more deteriorated as the mass flow rate increases while the uncertain band is smaller at the design point than at off design points. Finally, a sequential infilling sampling-based optimization method is employed to solve the design optimization. It is demonstrated that reduced blade angles are able to compensate the adverse effect of positive blade thickness errors and thus enhance the impeller performance robustness against manufacturing uncertainties. The present study lays a theoretical foundation for the further uncertainty quantification and robust design of advanced centrifugal compressors. (C) 2021 Elsevier Masson SAS. All rights reserved.