耦合双层参数更新的改进 PCE 模型在燃油离心泵不确定性分析中的应用

Blade manufacturing and extreme operating conditions in aircraft engines introduce uncertainty factors that significantly impact the performance and flow field variability of fuel centrifugal pumps. This paper proposes a synchronized analysis method based on CFD mechanistic models and surrogate models. Firstly,based on the Karhuben-Loève(KL)transform theory,uncertainty modeling of three-dimensional blade profile errors in the centrifugal pump was conducted. Secondly,nesting the Least Angle Regression(LAR)algorithm,we applied a double-layer parameter update to the Polynomial Chaos Expansion(PCE)model,constructing a high-precision surrogate model. Finally,employing a specific type of fuel centrifugal pump as the research object,we accomplished uncertainty analysis of the centrifugal pump based on CFD simulations,experimental verification,and the PCE surrogate model. The research demonstrates that the KL transform efficiently describes uncertainty in the three-dimensional blade profile error using only 9 input parameters. The improved PCE model with coupled double-layer parameter updating exhibits an average increase of 27.6% in accuracy metrics across multiple centrifugal pump conditions compared to the unimproved model. Controlling the blade profile error within −0.3 to 0.3 mm at the hub significantly reduces uncertainty in centrifugal pump performance. The blade profile errors at the hub have a greater impact on flow field uncertainty than those at the shroud and midsection,while speed is a crucial operational parameter affecting uncertainty in the fuel pump flow field.