TY - JOUR
T1 - 基于自适应带宽核密度估计的压气机叶片加工误差统计分析方法
AU - Ren, Yubin
AU - Tan, Miaolong
AU - Wu, Baohai
AU - Zhang, Ying
AU - Gao, Limin
N1 - Publisher Copyright:
© 2024 China Aerospace Science and Industry Corp. All rights reserved.
PY - 2024/6/1
Y1 - 2024/6/1
N2 - Aiming at the problem of statistical analysis and modeling of machining error of compressor blades, this paper proposes a nonparametric kernel density estimation method with adaptive bandwidth, and realizes the probability density modeling of machining error of compressor blades. Firstly, based on the thumb rule, the fixed optimal bandwidth in the fixed bandwidth kernel density estimation is solved as the initial bandwidth. Then, based on the fixed optimal bandwidth, the sensitivity factor is introduced to construct the adaptive bandwidth function, and the kernel density estimation is regulated by the adaptive bandwidth function. On this basis, the accuracy MSE and sensitivity S of kernel density estimation are defined as objective functions, and the elite strategy genetic algorithm is used to optimize the sensitivity factor to obtain the optimal sensitivity factor. Then the optimal bandwidth corresponding to different data sample points is calculated, so that the bandwidth can be adaptively adjusted according to the density of data samples. Finally, six kinds of blade profile errors of 134 groups of machined blades are statistically modeled, and the generalization performance of adaptive bandwidth kernel density estimation is verified by cross-test method. The experimental results show that the method has good applicability and high precision, and avoids the local adaptability problem of traditional fixed bandwidth kernel density estimation. The statistical analysis method proposed in this paper can accurately obtain the distribution characteristics of compressor blade machining error under the existing process capability. It provides an effective means for the optimization and improvement of blade aerodynamic design.
AB - Aiming at the problem of statistical analysis and modeling of machining error of compressor blades, this paper proposes a nonparametric kernel density estimation method with adaptive bandwidth, and realizes the probability density modeling of machining error of compressor blades. Firstly, based on the thumb rule, the fixed optimal bandwidth in the fixed bandwidth kernel density estimation is solved as the initial bandwidth. Then, based on the fixed optimal bandwidth, the sensitivity factor is introduced to construct the adaptive bandwidth function, and the kernel density estimation is regulated by the adaptive bandwidth function. On this basis, the accuracy MSE and sensitivity S of kernel density estimation are defined as objective functions, and the elite strategy genetic algorithm is used to optimize the sensitivity factor to obtain the optimal sensitivity factor. Then the optimal bandwidth corresponding to different data sample points is calculated, so that the bandwidth can be adaptively adjusted according to the density of data samples. Finally, six kinds of blade profile errors of 134 groups of machined blades are statistically modeled, and the generalization performance of adaptive bandwidth kernel density estimation is verified by cross-test method. The experimental results show that the method has good applicability and high precision, and avoids the local adaptability problem of traditional fixed bandwidth kernel density estimation. The statistical analysis method proposed in this paper can accurately obtain the distribution characteristics of compressor blade machining error under the existing process capability. It provides an effective means for the optimization and improvement of blade aerodynamic design.
KW - Adaptive bandwidth kernel density estimation
KW - Compressor blade
KW - Genetic algorithm
KW - Machining error
KW - Sensitive factor
UR - http://www.scopus.com/inward/record.url?scp=105005397053&partnerID=8YFLogxK
U2 - 10.13675/j.cnki.tjjs.2304013
DO - 10.13675/j.cnki.tjjs.2304013
M3 - 文章
AN - SCOPUS:105005397053
SN - 1001-4055
VL - 45
JO - Tuijin Jishu/Journal of Propulsion Technology
JF - Tuijin Jishu/Journal of Propulsion Technology
IS - 6
M1 - 2304013
ER -