基于高维多尺度核函数的模糊SVM剩余寿命预测  

Fuzzy SVM Residual Life Prediction Based on High-dimensional Multi-scale Kernel Function

作　　者：石慧[1] 李芷萱 彭壮壮 SHI Hui;LI Zhi-xuan;PENG Zhuang-zhuang(School of Electronic and Information Engineering,Taiyuan University of Science and Technology,Taiyuan 030024,China)

机构地区：[1]太原科技大学电子信息工程学院,山西太原030024

出　　处：《系统工程》2022年第6期148-155,共8页Systems Engineering

基　　金：国家自然科学基金青年科学基金资助项目(61703297,72071183,71701140);山西省基础研究计划(自由探索类)面上项目(20210302123206);山西省回国留学人员科研项目(2021-135,2021-134);山西省留学回国人员科技活动择优项目(20220029);山西省高等学校科技创新项目(2021L322)。

摘　　要：为提高齿轮剩余寿命预测精度,本文提出基于高维多尺度核函数的模糊SVM齿轮剩余寿命预测方法。首先结合多个退化特征量作为SVM的高维输入来反映设备的退化过程,而不同的退化特征量因其自身的特性不同,在剩余寿命的建模中贡献度也不相同,通过熵权法确定不同变量权重,实现高维变量加权系数的自动求取。其次考虑到高维数据具有异构特性,针对每一维变量分别输入多尺度核函数进行映射,同时基于样本的分布特征利用梯形模糊隶属度函数弱化离群点,对每个训练点赋予不同的权值,以此来构建高维多尺度核函数的模糊SVM模型,实现小样本数据潜在信息的最大挖掘。最后通过齿轮箱的试验对模型进行验证,表明本文提出方法与SVM、多尺度核SVM、模糊SVM相比,可提高预测准确度。In order to improve the prediction accuracy of the remaining useful life of gears, a fuzzy SVM bearing remaining life prediction method based on high-dimensional multi-scale kernel function is proposed. First, multiple degradation feature quantities are used as high-dimensional input of SVM to reflect the remaining service life of the equipment, and different degradation feature quantities have different contributions in the modeling process of the remaining life due to their different characteristics. The weight method determines the weights of different variables, and realizes the automatic calculation of the weighting coefficients of high-dimensional variables. Secondly, considering the heterogeneous characteristics of high-dimensional data, a multi-scale kernel function is input for each dimensional variable to map, and a trapezoidal fuzzy membership function is used to weaken outliers based on the distribution characteristics of the samples, and each training point is assigned a different the fuzzy SVM model of high-dimensional multi-scale kernel function is constructed to realize the maximum mining of potential information of small sample data. Finally, the prediction model is verified by the experiment of the gearbox. The results show that the proposed method has higher prediction effect and prediction accuracy than fuzzy SVM, multi-scale kernel SVM, and SVM.

关 键 词：剩余寿命预测 高维多尺度核函数 高维输入 模糊SVM 熵权法 

分 类 号：TH17[机械工程—机械制造及自动化]