基于聚焦式模糊聚类算法的数据挖掘故障诊断方法  

Fault Diagnosis by Data Mining Based on Focusing Fuzzy Clustering Algorithm

作　　者：杨苹[1] 

机构地区：[1]华南理工大学电力学院,广州510640

出　　处：《动力工程》2006年第4期511-515,共5页Power Engineering

基　　金：国家重大基础研究计划资助项目(G1998020308)

摘　　要：针对现有的火电厂大型设备故障诊断精度较低的问题，提出一种基于聚焦式模糊聚类算法的数据挖掘故障诊断方法。它采用分段相关分析的方法在火电厂SCADA系统历史数据库查找故障征兆变量，然后利用聚焦式量化算法对故障征兆变量进行离散化，最后应用双向模糊聚类算法找出对应故障类型的关键数据。该方法避免了为诊断故障而附加的专门测试或试验，在降低费用的同时，减少了试验对设备造成的潜在威胁。故障诊断实例表明：其诊断精度在不同的月份介于91％～95％之间，可以满足现场应用的要求。Aiming at the precision problem of current fault diagnosis of large capacity equipment in fossil fired power plants, a new approach of diagnosis is being proposed, which makes use of data mining based on focusing fuzzy clustering algorithm. Divisional correlation analysis is first used to detect fault indicating variables in the SCADA system＇s historical data bank of power plants, which are further discretized by the focusing nonlinear method, and at last the key data associated with the existing kind of fault is found with the help of crossed fuzzy algorithm. This method makes it possible to avoid the necessity of carrying out special additional measurements or tests; wherewith simultaneously with reduction of expenditures, risks, that the equipment may incur during tests, are avoided. Actual examples of fault diagnosing show that the diagnosing accuracy lies between 91% to 95% for different months in the year, which satisfies on-site requirements. Refs 3.

关 键 词：自动控制技术 故障诊断 数据挖掘 模糊聚类 聚焦 

分 类 号：TP206.3[自动化与计算机技术—检测技术与自动化装置]