一类基于类别区分的多重故障诊断模型框架
A class of category discrimination based model frameworks for multiple fault diagnosis

多重故障诊断是故障诊断和容错控制技术中的重难点问题. 本文关注数据驱动故障诊断, 以基于知识的方 法为基本思路, 按照故障表现及征兆的组合形式分析了多重故障和征兆之间的映射关系, 将多重故障诊断转化为一 类由各个组成故障对应于征兆数据集合的类划分问题. 在已有应用于故障诊断的分类和聚类方法的基础上, 分析 了4种基于类别区分的多重故障诊断模型框架. 并讨论了其优缺点和适用的类别区分算法
Multiple fault diagnosis (MFD) is a key issue in fault diagnosis and fault tolerant control technology. This paper is concerned with data-driven based MFD. Knowledge-based approaches are taken as basic thinking ways. An analysis of relationships between multiple faults and symptoms is made according to the fault forms and combinations of fault features. By this way, MFD is resolved into a class of category discrimination problems (the multiple-fault symptom data set corresponding to each constituent fault). Existing classification and clustering based fault diagnosis methods are referred. Then, 4 kinds of category discrimination based MFD model frameworks are analyzed and proposed, and their advantages and disadvantages are separately discussed. Futhermore, a summarization of category discrimination methods that apply to the models is also given.