Uncertainty quantification based optimization of centrifugal compressor impeller for aerodynamic robustness under stochastic operational conditions

Abstract

Centrifugal compressor impeller operates at stochastic boundary conditions. The operational uncertainties cause performance deviation from design value and consequently affect the reliability of the compressor. Considering the stochastic operational uncertainties in the early design stage, this paper presents an uncertainty quantification based optimization of centrifugal compressor impeller with splitter blades for aerodynamic robustness. The nonlinear response relation between the design variables, the aerodynamic boundary uncertainties and the impeller performance is modelled by a combination of the Latin Hypercube Sampling, the three dimensional CFD, the Kriging surrogate model and the Non-intrusive Probability Collocation method. A sensitivity analysis with single and multiple random geometry variations is carried out to identify the most sensitive parameters. The effects of rotor speed uncertainty on pressure ratio and efficiency are quantified. A case study is conducted to search for optimal impellers with higher performance and lower sensitivity to boundary uncertainty using NSGA-II. The optimization is verified by the Monte Carlo method. The results demonstrate that the aerodynamic robustness of the compress impeller with splitter blades is enhanced by the proposed approach, which provides references for compressors, turbines and other turbo machinery.