基于自适应特征选择k子凸包的滚动轴承故障诊断  

ROLLING BEARING FAULT DIAGNOSIS BASED ON ADAPTIVE FEATURE SELECTION k⁃SUB CONVEX HULL

作　　者：胡爱孺 吴占涛[1,2] 杨宇[1,2] 程军圣[1,2] HU AiRu;WU ZhanTao;YANG Yu;CHENG JunSheng(College of Mechanical and Vehicle Engineering,Hunan University,Changsha 410082,China;Hunan Provincial Key Laboratory of Equipment Service Quality Assurance,Hunan University,Changsha 410082,China)

机构地区：[1]湖南大学机械与运载工程学院,长沙410082 [2]湖南大学装备服役质量保障湖南省重点实验室,长沙410082

出　　处：《机械强度》2024年第2期255-263,共9页Journal of Mechanical Strength

基　　金：国家重点研发计划项目(2020YFB2009602);国家自然科学基金项目(51875183,51975193);湖南省教育厅科学研究项目(21A0017)资助。

摘　　要：滚动轴承故障诊断中往往将特征选择和分类器的设计分别进行研究,从而难以获得满意的分类精度。将特征选择和分类器寻优结合起来,提出了一种自适应特征选择k子凸包(Adaptive Feature Selection K-sub Convex Hull, AFSKCH)的分类模型,从而实现了故障特征自适应选择和分类的一体化。首先,利用凸包距离函数保持数据流形上的局部邻域结构,通过交替构造k子凸包得到特征权值矩阵;其次,采用线性规划接近度方法求解k子凸包距离,利用乘子交替方向法得到自适应特征空间;最后,根据测试点到k子凸包的最小重构距离进行分类。滚动轴承故障振动信号分析结果表明,该方法特征选择性能优于其他特征选择方法,且具有较高的分类精度。Feature selection and classifier design are often studied separately in rolling bearing fault diagnosis,so it is difficult to obtain satisfactory classification accuracy.An adaptive feature selection k⁃sub convex hull(AFSKCH)classification model was proposed by combining feature selection and classifier optimization,which realized the integration of adaptive feature selection and classification.Firstly,the convex hull distance function was used to maintain the local neighborhood structure on the data manifold,and the feature weight matrix was obtained by alternately constructing k⁃sub convex hulls.Secondly,the distance was solved by the method of linear programming proximity,and the adaptive feature space was obtained by using the multiplier alternating direction method.Finally,the classification was carried out according to the minimum reconstruction distance from the test point to the k⁃sub convex hull.The analysis results of rolling bearing fault vibration signals show that the feature selection performance of this method is better than other feature selection methods,and the classification accuracy is higher.

关 键 词：自适应特征选择 邻域嵌入 k子凸包 滚动轴承 故障诊断 

分 类 号：TH165.3[机械工程—机械制造及自动化]