摘要
针对单个分类器方法在滚动轴承故障诊断中精度较低、故障样本标记稀缺、特征空间维度高等问题,提出一种将协同训练与集成学习相结合的Co-Forest轴承故障诊断算法。Co-Forest是半监督学习中的协同训练算法,包含多个基分类器,通过投票实现协同训练中的置信度估算。从滚动轴承的振动信号中提取时域、频域特征指标。利用少量带标签和大量未标记样本重复地训练基分类器。集成基分类器,实现对滚动轴承故障的诊断。实验结果表明,与同类型的协同训练算法(Co-Training、Tri-Training)相比,Co-Forest算法在轴承故障诊断中具有更高的正确率,与当前针对特征向量高维、标记样本稀缺问题的ISS-LPP算法,SS-LLTSA算法相比,Co-Forest算法在保持很高诊断正确率的情况下,不需要降维、参数设置简单,具有一定的实际应用价值。
Aiming at the problems of single classifier method in rolling bearing fault diagnosis,such as low accuracy,scarce fault sample mark and high feature space dimension,a co-forest bearing fault diagnosis algorithm combining collaborative training and ensemble learning is proposed.Co-forest is a collaborative training algorithm in semi-supervised learning,which contains multiple base classifiers and realizes confidence estimation in collaborative training through voting.The time domain and frequency domain characteristic indexes are extracted from the vibration signal of rolling bearing.A small number of labeled and large number of unlabeled samples are used to train the base classifier repeatedly.The fault diagnosis of rolling bearing is realized by integrating base classifier.Experimental results show that the co-forest algorithm has a higher accuracy in bearing fault diagnosis than the same type of cooperative training algorithm(co-training,tri-training).Compared with the ISS-LPP algorithm and SS-LLTSA algorithm for the problem of high dimension of feature vectors and scarcity of labeled samples,the co-forest algorithm has certain practical application value under the condition of maintaining a high diagnostic accuracy rate without dimensional reduction and simple parameter setting.
作者
王得雪
林意
陈俊杰
WANG Dexue;LIN Yi;CHEN Junjie(School of Digital Media,Jiangnan University,Wuxi,Jiangsu 214122,China;Jiangsu Key Laboratory of Media Design and Software Technology,Wuxi,Jiangsu 214122,China;SIEMENS,China Institute,Beijing 100102,China)
出处
《计算机工程与应用》
CSCD
北大核心
2020年第12期273-278,共6页
Computer Engineering and Applications
基金
中央高校基本科研业务费专项(No.JUSRP51902B)。
关键词
协同训练
集成学习
故障诊断
置信度
cooperative training
ensemble learning
fault diagnosis
confidence
