基于改进SVDD算法的升降机轴承故障检测研究  

Bearing Fault Detection Method of Elevator Based on Improved SVDD Algorithm

作　　者：刘俊辰 唐文秀[1] 金剑桥 吴俊英 LIU Junchen;TANG Wenxiu;JIN Jianqiao;WU Junying(School of Mechanical and Electrical Engineering,Northeast Forestry University,Harbin 150040,China)

机构地区：[1]东北林业大学机电工程学院

出　　处：《重庆理工大学学报（自然科学）》2019年第7期66-73,共8页Journal of Chongqing University of Technology：Natural Science

基　　金：国家自然科学基金资助项目(61473001);黑龙江省自然科学基金资助项目(E201403)

摘　　要：为解决升降机轴承故障数据不易收集导致的数据不均衡问题,提出一种基于改进支持向量描述算法即带负类样本的支持向量数据描述算法(SVDD NE)的升降机轴承故障检测分类方法。该方法在支持向量描述算法(SVDD)的基础上加入新的惩罚因子,考虑所有的训练样本集,使得超球面能很好地将所有的正类样本包围起来,同时拒绝负类样本,让球面达到最小状态。将SVDD NE算法与SVDD算法用于升降机轴承数据故障检测实验中,结果表明:SVDD NE算法适合处理不均衡小样本数据分类问题,且该算法的检测分类准确率较SVDD算法有显著提高。To solve the problem of data imbalance caused by the difficulty in collecting fault data of elevator bearings,an improved support vector data description algorithm (SVDD-NE) based on improved support vector description algorithm (SVDD-NE) with negative class samples was proposed for fault detection and classification of elevator bearings.This method adds a new penalty factor based on support vector description algorithm (SVDD) and considers all training sample sets,so that hypersphere can surround all positive class samples well,while rejecting negative class samples,so that the sphere reaches the minimum state.The SVDD-NE algorithm and SVDD algorithm are applied to the fault detection experiment of elevator bearing data.The results show that the SVDD-NE algorithm is suitable for dealing with the problem of unbalanced small sample data classification,and the detection and classification accuracy of the algorithm is significantly improved than that of the SVDD.

关 键 词：升降机 轴承 故障检测 支持向量数据描述 负类样本 

分 类 号：TH133.3[机械工程—机械制造及自动化]