发电机绕组故障时振动的关联维数分析及诊断  

Application of Vibration Signal Correlation Dimension to Generator Winding Fault

作　　者：万书亭[1] 李和明[2] 李永刚[2] 

机构地区：[1]华北电力大学机械工程学院,保定071003 [2]华北电力大学电气工程学院,保定071003

出　　处：《振动．测试与诊断》2005年第3期210-213,共4页Journal of Vibration,Measurement & Diagnosis

基　　金：原国家电力公司电力青年科技促进费(编号:SPQKJ02-10);河北省教育厅科学指导计划项目(X2004455)

摘　　要：运用分形理论的相空间重构算法,分析了发电机在正常运行、转子励磁绕组短路和定子绕组故障3种状态时定转子振动信号的关联维数。分析结果表明,3种状态具有不同的关联维数,揭示了绕组短路的电气故障不仅引起发电机电气参数的变化,也将引起发电机振动特征的变化,且不同绕组故障对气隙磁场畸变效果不同,引起发电机定转子不同的动力学特性。最后,以定子振动的关联维数作为特征向量输入到BP网络中,将实测M JF-30-6型故障模拟发电机的定子振动的关联维数作为样本,通过学习样本的学习和对测试样本的测试分析的结果表明,该诊断方法能正确识别发电机的3种状态。Based on fractal theory and phase space reconstruction arithmetic,the generator vibration correlation dimension is analyzed when the generator is in normal condition,rotor excitation winding short circuit and stator winding fault condition.The results show that the three conditions have different correlation dimensions.On the other hand it is reflected that the generator winding faults not only can excite electric signal changs,but also can excite mechanical vibration changs,and different winding faults have different distorted air-gap fields and different generator mechanism dynamics characteristics.Finally,a new method of generator vibration fault diagnosis is presented based on the correlation dimension and artificial neural network(ANN),which uses practically acquired MJF-30-6 generator data as learning samples.The results show that the method can identify the three conditions effectively.

关 键 词：人工神经网络 关联维数 故障诊断 发电机振动 

分 类 号：TM31[电气工程—电机] TH133[机械工程—机械制造及自动化]