Abstract: To solve the uncertainty quantification (UQ) problem of scarce measurement data and complex distribution in practical engineering, a novel data-driven UQ method is developed by using preconditioned matrix. The effectiveness and ac-curacy of the developed method are validated by robustness analysis, reconstruction of orthogonal basis functions and nonlinear test functions. Based on the measurement data of blade leading-edge radius and incoming incidence, the influence of uncertain parameters on the aerodynamic performance of a high subsonic compressor cascade is quantitatively evaluated by using the self-developed UQ method. Under the design and high incidence conditions, results show that the probability that the actual total pressure loss coefficient is higher than the nominal value is 83.6 % and 69.9 %, respectively. The dispersion of total pressure loss at high incidence is about 2.4 times of that at design condition. The uncertainties of leading-edge radius and incidence cause large fluctuations in the leading-edge flow process, which is the main factor leading to the overall performance degradation and performance dispersion of the cascade.

Key words: Data-driven, Uncertainty quantification, Compressor, Leading-edge radius, Incidence, Aerodynamic performance