基于证据理论的高压防爆电机转子故障在线诊断方法研究  

Study on on-line diagnostic method for rotor faults of high voltage explosion-proof motor based on evidence theory

作　　者：张建文[1] 于江[2] 

机构地区：[1]中国矿业大学信息与电气工程学院,江苏徐州221008 [2]国家防灾技术高等专科学校自动化技术系,河北三河065201

出　　处：《华北电力大学学报（自然科学版）》2008年第2期35-41,共7页Journal of North China Electric Power University：Natural Science Edition

基　　金：国家自然科学基金资助项目(50504015);中国矿业大学科研基金资助项目(0C4499)

摘　　要：提出了适用于矿井爆炸性环境的高压防爆电机转子故障在线诊断方案。运用扩展Park矢量方法将转子断条故障特征频率与基频有效分离,与气隙偏心故障特征频率所处频段进行统一量化,形成BP网络的输入模式矢量,将网络输出作为证据理论的基本概率赋值;基于大量实测数据的分析结果,提取出基于启动电流信号小波变换的断条故障最小频带能量比判据和基于单相电流频谱的偏心故障特征分量相对比值判据,并运用自行构造的模糊数学隶属函数分别予以描述。通过将隶属函数扩充为与基本概率赋值函数等维数的向量,获得了诊断转子故障的3种证据。诊断实例表明,运用证据理论可显著提高故障诊断的准确性和置信度;同时指出,应用证据理论诊断复合故障时,不能只依靠所有证据的最终组合结果,还应当重视证据组合的中间结果,否则可能产生漏诊断。A project for on-line diagnosing rotor faults of high voltage explosion-proof was designed to adapt to the volatile environment of mine. The characteristic frequency of rotor broken-bar faults was effectively separated from its fundamental frequency by the extended Park vector approach; the characteristic frequency-band of broken-bar and gaseccentricity faults was normalized together to act as the input vector of a back propagation （BP） neural network, and the output of the network was defined as the basic probability value （BPV） of evidence theory （ET）. Based on lots of on-site experimental data, the minimum frequency-band energy ratio criterion for diagnosing broken-bar was picked up according to wavelet transform of startup current, the similar relative ratio criterion for diagnosing gas-eccentricity was presented according to single current spectrum, and the two criteria was described by different fuzzy membership functions （MF）. The MFs were extended to the same dimensions to the BPV functions, and three evidences for diagnos- ing rotor faults were gained. Practical diagnostic examples show that the fault diagnostic accuracy and confidence are markedly promoted by using the ET. It is also pointed out that, while the ET is used to diagnose compound faults, the diagnostic conclusion isn＇t only based on the final fusion results of all evidences, but also the middle fusion results of evidences fusion, otherwise some faults may be missed.

关 键 词：防爆电机 在线检测 故障诊断 转子故障 证据理论 

分 类 号：TM343[电气工程—电机]