基于多元状态评估和相关分析的振动故障诊断  

Vibration fault diagnosis based on multivariate state evaluation and correlation analysis

作　　者：黄杨森 王勇 刘云平 冯小健 HUANG Yangsen;WANG Yong;LIU Yunping;FENG Xiaojian(Dongfang Electric Machinery Co.,Ltd.,Deyang 618000,China)

机构地区：[1]东方电气集团东方电机有限公司,四川德阳618000

出　　处：《振动与冲击》2024年第17期269-277,共9页Journal of Vibration and Shock

基　　金：四川省科技攻关计划(2020YFG0350)。

摘　　要：针对运行状态汽轮发电机轴振、瓦振异常预警及其原因诊断困难问题,提出一种基于多元状态评估(multivariate state evaluation technology,MSET)和相关分析(correlation analysis, CA)的故障诊断方法。首先基于MSET和滑动窗口计算当前评价窗口内参量估计值与其运行值的残差;其次计算当前评价窗口内参数间相关系数与状态矩阵中对应相关系数的残差;再次通过为各参数相对偏差均值或残差均值、各相关系数残差设定阈值,从而提取出异常特征;最后基于异常特征和欧式距离进行振动异常预警和异常故障诊断。利用掌握的汽轮发电机运行数据测试验证,测试结果表明,所提故障诊断方法切实可行,与单一的参数自身变化评价或相关分析相比,可提取更多异常或故障特征,具备多故障诊断分析能力,更有利于振动异常预警和提高故障诊断准确性。Here,aiming at difficulty problems of abnormal warning and cause diagnosis for shaft and bearing vibrations in steam turbine generators under operation condition,a fault diagnosis method based on multivariate state evaluation technology(MSET)and correlation analysis(CA)was proposed.Firstly,based on MSET and sliding window,residuals between estimated parameter values and their operating values within the current evaluation window were calculated.Secondly,residuals between correlation coefficients among parameters within the current evaluation window and corresponding correlation coefficients in state matrix were calculated.By setting thresholds for each parameter’s relative deviation mean or residual mean as well as residual of each correlation coefficient,abnormal features were extracted.Finally,vibration abnormal warning and abnormal fault diagnosis were performed based on abnormal features and Euclidean distance.The obtained operational data of steam turbine generator were used for testing verification.The testing results showed that the proposed fault diagnosis method is practical and feasible;compared with single parameter self-change evaluation or correlation analysis,the proposed method can extract more abnormal or fault features,and have the analysis ability of multi-fault diagnosis;it is more conducive to vibration abnormal warning and improving the correctness of fault diagnosis.

关 键 词：多元状态评估(MSET) 相关分析(CA) 异常预警 故障诊断 

分 类 号：TM621[电气工程—电力系统及自动化]