基于证据理论聚类算法的故障诊断  

Fault Diagnosis Based on Dempster Evidential Clustering Algorithm

作　　者：洪贝[1] 胡昌华[1] 姜学鹏[2] 

机构地区：[1]第二炮兵工程学院302教研室,西安710025 [2]海军航空工程学院七系,烟台264001

出　　处：《南京航空航天大学学报》2011年第B07期134-137,共4页Journal of Nanjing University of Aeronautics & Astronautics

摘　　要：为了正确高效地诊断故障,以并发故障为例,建立了基于证据理论聚类算法的诊断模型,提出了基于Dempster理论的聚类算法,该聚类算法既分析每个单类,还考虑单类的各种组合,并利用质心特性,将单类聚类原型推广为适合组合类特性的聚类原型。通过故障实例说明,本文模型既适合普通故障诊断,更适合并发故障诊断,且计算的时间复杂度较小。To diagnose fault correctly and efficiently, a clustering algorithm based on evidence theory diagnostic model is established by taking the simultaneous fault as an example. The algorithm is proposed based on the Dempster theory, which analyzes each single class, and also considers various combinations of a single class. Using centroid feature, a single-class clustering prototype combination of class properties is promoted to fit the cluster prototype. Through the examples of the fault diagnosis of examples shows that the model is suitable for general diagnosis, especially for concurrent fault diagnosis, with less complexity of computation time.

关 键 词：故障诊断 证据理论 聚类 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]