基于IMF-MFDE和GRU的水电机组故障诊断  

Fault Diagnosis of Hydroelectric Units Based on IMF-MFDE and GRU

作　　者：朱文鑫 王淑青[1] ZHU Wen-xin;WANG Shu-qing(School of Electrical and Electronic Engineering,Hubei University of Technology,Wuhan 430068,China)

机构地区：[1]湖北工业大学电气与电子工程学院,湖北武汉430068

出　　处：《水电能源科学》2024年第4期173-177,共5页Water Resources and Power

摘　　要：针对水电机组振动信号非平稳、非线性及强噪声的特点,提出了一种IMF多尺度波动散布熵(MFDE)结合门控循环单元(GRU)的故障诊断方法。首先,采用跳蛛优化算法(JSOA)寻找变分模态分解(VMD)最优参数,达到振动信号最佳分解降噪效果;其次,对分解得到的本征模态函数(IMF)进行重构,计算有效IMF的多尺度波动散布熵(MFDE)作为故障特征向量;最后,将特征向量输入GRU构建水电机组故障识别器。所提方法对实际水电站机组故障样本数据的故障识别率达97.83%,验证了该方法的有效性。In response to the characteristics of non-stationary,nonlinearity,and strong noise in the vibration signals of hydroelectric units,a fault diagnosis method combining IMF multi-scale fluctuation dispersion entropy(MFDE)and gated cyclic unit(GRU)is proposed.Firstly,the jumping spider optimization algorithm(JSOA)is used to optimize the parameters of variational mode decomposition(VMD)for achieving the optimal decomposition and noise reduction effect of vibration signals.Secondly,the eigenmode function(IMF)obtained from the decomposition and noise reduction is reconstructed,and the multi-scale fluctuation dispersion entropy(MFDE)of the effective IMF is calculated as the fault feature vector.Finally,a fault identifier for hydroelectric units is established by choosing feature vectors as the input of GRU.Taking the actual fault sample data of hydroelectric power plant units as an example,the fault recognition rate reached 97.83%,verifying the effectiveness of the proposed method.

关 键 词：水电机组振动信号 故障诊断 跳蛛优化算法 变分模态分解 多尺度波动散布熵 

分 类 号：TV734.1[水利工程—水利水电工程] TK05[动力工程及工程热物理]