A Novel Uncertainty Quantification Method for Compressor Performance Based on Functional Response Model

Compressor is one of the core components in the aeroengine. Its performance is inevitably impacted by manufacturing errors, resulting in deviations from the design specifications. The compressor performance uncertainty arising from manufacturing uncertainty should be quantified in the design stage to reduce development risk. In this paper, a novel uncertainty quantification method using functional response model is proposed to quantify the performance uncertainty of the whole characteristic line of compressor. Firstly, the classic Rotor 37 axial compressor is selected as the research object. To facilitate rapid modification of compressor geometry, a parameterized model with 37 independent control parameters is developed to characterize the compressor geometry. Secondly, a widely used simulation method based on computational fluid dynamics (CFD) is employed to obtain characteristic lines of a series of specified compressor geometries which are defined using the sliced Latin hypercube design. Subsequently, considering the functional features of characteristic lines, three functional response models are established based on the Gaussian process regression, emulating the functional relationship between characteristic line and control parameters. Eventually, a surrogate-based Monte Carlo simulation method is used to quantify the performance uncertainty of compressor. The performance uncertainty at any interested operating point on the characteristic line of compressor can be calculated and analyzed. The analysis results yield new insights for compressor designers.