基于EWT与排列熵的SVM汽轮机转子故障诊断  

Fault Diagnosis of Steam Turbine Rotor Based on EWT and Permutation Entropy and SVM

作　　者：石志标[1] 陈斐 曹丽华[2] 

机构地区：[1]东北电力大学机械工程学院,吉林132012 [2]东北电力大学能源与动力工程学院,吉林132012

出　　处：《汽轮机技术》2017年第6期439-442,共4页Turbine Technology

基　　金：国家自然科学基金(51576036);吉林省科技发展计划项目(20100506)

摘　　要：针对汽轮机转子振动信号在强噪声下难以提取瞬态冲击的故障特征的问题,提出一种基于经验小波变换(EWT)与排列熵相结合的特征提取方法。将转子故障振动信号进行经验小波变换,得到一系列本征模态函数(AM-FM),根据相关度原则选取故障信号敏感的本征模态函数计算其排列熵值构建故障特征向量。通过ZT-3转子模拟实验台获得振动故障信号,分别用EWT与排列熵和EWT与样本熵获得故障特征值,使用支持向量机(SVM)识别验证,结果表明:EWT与排列熵构建的特征向量识别率比EWT与样本熵高6.11个百分点,达到了较理想的识别准确率。Aiming at the problem of turbine rotor vibration signals difficult to extract fault characteristics of transient shock of fault characteristics under strong noise, put forward a method of feature extraction based on empirical wavelet transform （EWT） and permutation entropy. Rotor vibration signals was decomposed by empirical wavelet transformobtained a series of intrinsic mode function （ AM - FM）, according to the correlation principle to select intrinsic mode function sensitived to fault signal calculated the permutation entropy to constructfault feature vector. The vibration signals obtained from the ZT -3 rotor, EWT and permutation entropy and EWT and sample entropy are used to obtain the fault characteristic value, and the support vector machine（SVM） is used to identify the fault. The results show that the recognition rate of EWT and permutation entropy is 6.11 percentage points higher than that of EWT and sample entropy, andit reached the ideal identification accuracy.

关 键 词：经验小波变换 排列熵 支持向量机 转子 故障诊断 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]