基于非平稳高斯过程的叶栅加工误差不确定性量化

Uncertainty Quantification of Cascade Manufacturing Error Based Non-Stationary Gaussian Process

关键词:  非平稳高斯过程  多项式混沌  稀疏网格  加工误差  不确定性量化

Key words:  Non-stationary Gaussian random process  Polynomials chaos  Sparse grid  Manufacturing error  Uncertainty quantification

摘要:

基于非平稳高斯过程描述叶片加工误差，结合Karhunen-Loeve展开方法，建立了由于加工误差导致的叶片型线几何不确定性表征模型。耦合非嵌入式多项式混沌展开、稀疏网格技术与Reynolds-Averaged Navier-Stokes (RANS)方程求解技术，提出了叶栅加工误差不确定性量化方法，研究量化了加工误差所导致的叶型几何不确定性对典型高负荷Pak-B叶栅气动性能的影响。结果表明，在加工误差影响下，叶片负荷相对于设计值变化[±1%]以上的概率为0.56，总压恢复系数相对于设计值降低1%以上的概率为0.12。详细气动分析表明，斜切部分和尾缘的加工制造精度对Pak-B叶栅气动性能影响显著，相应位置的加工误差应严格控制。

Abstract:

A parametric model of manufacturing error on geometric variability is derived using a non-stationary Gaussian random process and Karhunen-Loeve expansion. Combined with the non-intrusive polynomials chaos expansion，sparse grid technique and Reynolds-Averaged Navier-Stokes (RANS) solver technique，an uncertainty quantification method for cascade manufacturing error was proposed. Uncertainty quantification of effects of manufacturing error on the aerodynamic performance of a representative high load blade named Pak-B was carried out. Statistics of stochastic output quantities of interest indicate that the probability of variation of the blade loading more than [±1%] relative to the design value is about 0.56 and the probability of reduction of the total pressure recovery coefficient more than 1% relative to the design value is about 0.12. Detailed aerodynamic analysis shows that the manufacturing precision of blade oblique section and trailing edge has a significant effect on Pak-B blade aerodynamic performance and it is possible to minimize the manufacturing error of that to obtain a better aerodynamic performance.