基于证据推理的电力推进系统轴承多特征融合故障诊断研究  

Research on Fault Diagnosis of Electric Propulsion Bearing Multi-feature Fusion Based on Evidence Reasoning

作　　者：张雪琴 盛晨兴[1,2,3] 欧阳武 ZHANG Xue-qin;SHENG Chen-xing;OUYANG Wu(School of Energy and Power Engineering,Wuhan University of Technology,Wuhan 430063,China;National Engineering Research Center for Water Transport Safety,Wuhan University o£Technology,Wuhan 430063,China;Key Laboratory of Marine Power Engineering and Technology of Ministry of Transport,Wuhan University of Technology,Wuhan 430063,China)

机构地区：[1]武汉理工大学能源与动力工程学院,武汉430063 [2]武汉理工大学国家水运安全工程技术研究中心,武汉430063 [3]武汉理工大学船舶动力工程技术交通行业重点实验室,武汉430063

出　　处：《武汉理工大学学报》2021年第4期27-34,共8页Journal of Wuhan University of Technology

基　　金：工业和信息化部高技术船舶项目(MC-201917-C09);NSFC-浙江两化融合联合重点基金(U1709215)。

摘　　要：针对船舶电力推进系统滚动轴承发生故障时不易诊断的问题,建立了基于数据驱动的证据推理(ER)诊断模型。从滚动轴承试验样本数据集中提取出时域、频域和小波包节点能量等特征证据,通过对特征数据的统计分析,从数据中获得证据的置信分布、可靠性因子和重要性因子值。最后对多条诊断证据进行ER规则融合,诊断滚动轴承典型故障类型。整个过程满足了严格的概率推理,具有很强的可解释性,在实际工程中为技术人员监测检修提供更多参考依据。通过与其他模型预测结果对比,该模型实现了多特征的有效融合,诊断准确率达98.57%。Aiming at the problem that is difficult to diagnose when rolling bearing of marine electric propulsion system failure occurs,a data-driven Evidence Reasoning(ER)diagnosis model was established.First,characteristic evidences such as time domain,frequency domain and wavelet packet node energy from the rolling bearing test sample data set were extracted.Through statistical analysis of the characteristic data,we obtained the confidence distribution of the evidence,the initial value of the reliability factor and the importance factor from the data.Finally,multiple pieces of diagnostic evidence were fused with ER rules to diagnose typical failure types of rolling bearings.The whole process satisfies strict probabilistic reasoning,has strong interpretability,and provides more reference basis for technicians to monitor and repair in actual engineering.Compared with the prediction results of other models,it is shown that the model realizes the effective fusion of multiple features,and the diagnostic accuracy rate reaches 98.57%.

关 键 词：滚动轴承 故障诊断 证据推理 多特征融合 

分 类 号：TH16[机械工程—机械制造及自动化]