Abstract: To investigate the impact of introducing changes in the dimensionality and correlation of twist deviation on the quantitative results of compressor performance, based on the measurement results of twist deviation of a certain compressor rotor blade, a data-driven non-intrusive polynomial chaos method was used to quantitatively study the uncertainty effect of blade twist deviation on the aerodynamic performance of a single-stage axial subsonic compressor, and sensitivity analy-sis was conducted. The research results indicate that the standard deviation of compressor performance at various oper-ating conditions decreases with the increase of the dimension of twist deviation introduced. Through quantitative compari-son, it was observed that considering twist deviations of the blade tip and root sections is sufficient to ensure the conver-gence of compressor performance standard deviation and performance probability distribution form. However, in sensitiv-ity analysis, it was revealed that the more dimensions of twist deviation introduced, the more accurate the identification of the most twist deviation region in compressor performance can be achieved. As the correlation between twist deviations on various sections gradually strengthens, the fluctuation degree of compressor performance at different operating condi-tions also increases. The difference in the first-order sensitivity indices of compressor performance to twist deviations also gradually decreases, but the ranking of compressor performance sensitivity to twist deviations on various sections remains unchanged.

Key words: data driven, twist deviation, uncertainty quantification, sensitivity analysis, compressor