摘要
基于符号有向图(SDG)深层知识模型能够表达复杂的因果关系,具有包容大规模潜在信息的能力,在流程系统领域是完备性较好的一种故障诊断方法。但固有的缺陷限制了它的进一步应用。该文通过采用主元统计-有向图方法,可以有效避免传统SDG在确定节点状态和阈值时的单变量统计的缺点;并对某些定性特征一样而定量值不同的故障模式,为了能有效区分,根据节点相对变化增益,在SDG上加入定量信息,并去构造隶属函数,然后由模糊数学最大隶属度原则去进一步确定故障。案例研究表明基于改进SDG方法可以进行有效的诊断。
The Signed Directed Graph （SDG） deep knowledge model can be used to express the complex cause and effect relations, and has very large capacity of containing process potential information, it is a self-contained method to effectively diagnosis system failures, but many limitations restrict it applies in fault diagnosis. The shortcoming of single variable analysis in modifying node state and threshold value can be avoided by combine the Principal Component Analysis （PCA） with SDG; and the patterns that can not be distinguished are diagnosed by using the relative gain of corresponding nodes to form a qualitative and quantitative model, and then the fault can be distinguished by utilizing the maximum membership grade principle of fuzzy mathematics. The case studies show the improved SDG has better resolution.
出处
《中国电机工程学报》
EI
CSCD
北大核心
2005年第23期124-128,共5页
Proceedings of the CSEE
基金
华北电力大学博士学位教师科研基金(20041209)资助
