Uncertainty Quantification of Compressor Performance Affected by Data-Driven Blade Geometric Deviations

Abstract

Accurate uncertainty quantification of compressor performance arising from blade geometric deviations is conducive to blade optimization design, blade error verification, etc. In order to investigate the impact of actual blade geometric deviations on compressor performance, this study initially conducted measurements of geometric deviations on three sections of 100 blades. Then, the probability density distributions of various geometric deviations were obtained through kernel density estimation. Subsequently, combining data-driven nonintrusive polynomial chaos with Halton sequence, the distribution of sampling points and the construction approach of the response model were determined. Based on the parameterization of a subsonic rotor geometric model, blade samples with different geometric features were generated. Utilizing numerical simulation results of the aerodynamic performance of each sample, the impact of blade geometric deviations on compressor performance was quantified, and sensitivity analysis was conducted using Sobol' index. It was observed that the total pressure ratio was most sensitive to the stagger angle deviation at 50% blade height, while the sensitivity of the isentropic efficiency to each geometric deviation varied with operating conditions. Then, the flow field was divided into five regions based on different flow loss mechanisms, and a viscous loss coefficient was introduced to quantify the flow losses in each region. It was found that various geometric deviations at 50% span section, as well as leading edge radius deviation and stagger angle deviation at 95% span section, have a significant impact on the flow field losses.