基于主成分分析的多源特征融合故障诊断方法  

Feature-level Fusion Fault Diagnosis Based on PCA

作　　者：巫茜[1] 蔡海尼[2] 黄丽丰[1] 

机构地区：[1]重庆理工大学计算机科学与工程学院,重庆400050 [2]重庆大学软件学院,重庆400044

出　　处：《计算机科学》2011年第1期268-270,共3页Computer Science

基　　金：重庆自然科学基金项目(2008BB2065);重庆理工大学青年基金项目(2010ZQ21)资助

摘　　要：目前故障诊断的实际应用中,因噪声的干扰,基于单传感器的故障诊断稳定性较差,很难达到满意的诊断精确度。提出了一种多传感器多特征数据融合的故障诊断方法。该方法利用多传感器从不同部位获取同一部件的运行状况,并通过构建多源特征融合模型,提高特征信息的抗干扰性,最后通过融合特征信息来完成部件的故障诊断。在将新方法应用于滚动轴承故障诊断的试验中,可以看到新方法能够获得较好的性能,比基于单传感器故障诊断的精确度更高。In most of current application,fault diagnosis based on single sensor show bad performance and is difficult to achieve satisfactory accuracy owning to noises.In this paper,a multi-sensor multi-feature fusion fault diagnosis method was proposed.The method firstly collects part operating conditions using multiple sensors installed on it.Then a fusion model based on principal component analysis was constructed to fuse all extracted features from the different sensors.Finally,fault diagnosis was carried out according to the fused results.Experiments and results show that new method can achieve perfect performance which is also better than that achieved by single-sensor.

关 键 词：多源信息融合 故障诊断 主成分分析 特征级 分类 

分 类 号：TH212[机械工程—机械制造及自动化] TH213.3