基于电流特征的CRDM驱动机构故障诊断方法  

Fault Diagnosis Method of CRDM Drive Mechanism Based on Current Features

作　　者：云浩 黄四和 王岩 陈子溪 Yun Hao;Huang Sihe;Wang Yan;Chen Zixi(China Nuclear Power Operation Technology Corporation,Wuhan 430073,Hubei,China;Fuqing Nuclear Power Co.,Ltd.,Fuqing 350318,Fujian,China;Nuclear and Radiation Safety Center of the Ministry of Ecology and Environment,Beijing 102400,China)

机构地区：[1]中核武汉核电运行技术股份有限公司,湖北武汉430073 [2]福清核电有限公司,福建福清350318 [3]生态环境部核与辐射安全中心,北京102400

出　　处：《电站辅机》2024年第4期11-18,共8页Power Station Auxiliary Equipment

摘　　要：针对核电控制棒驱动机构(CRDM)故障诊断问题,提出一种基于集合经验模态分解(EEMD)特征的长短记忆神经网络(LSTM)算法的故障诊断方法。首先,根据CRDM系统电源驱动电流,通过电流数值大小和电流上升时间的对比,分析不同故障状态下电流特征的变化情况,并对CRDM的电路故障和机械故障进行划分。其次,通过构建EEMD故障特征,分析不同故障下本征模态分量(IMF)的情况,论证EEMD故障特征向量对CRDM故障的可区分性,并引入能量熵构建故障特征向量。再次,采用LSTM算法建立控制棒驱动机构故障诊断模型,识别故障特征向量中包含的故障信息,实现CRDM故障的诊断。最后,设计不同诊断模型进行对比试验。结果表明,所提方法能有效实现控制棒驱动系统故障的诊断识别,且平均辨识准确度可达98%。To resolve fault diagnosis of control rod drive mechanism(CRDM)in nuclear power,a fault diagnosis method with a Long Short-Term Memory(LSTM)neural network algorithm based on Ensemble Empirical Mode Decomposition(EEMD)feature is proposed in this study.First,according to the driving current of the CRDM system,by comparing the current value and the current rise time,the change of current characteristics under different fault conditions was analyzed,Then,the circuit fault and mechanical fault conditions of the CRDM were divided.After that,by constructing EEMD fault features,the situation of intrinsic mode function(IMF)under different faults was analyzed;the distinguishability of the EEMD fault feature vector to CRDM fault was demonstrated;and the energy entropy was introduced to construct a fault feature vector.Finally,the LSTM algorithm was employed to establish the fault diagnosis model of the control rod drive mechanism,identify the fault information contained in the fault feature vector,and realize the diagnosis of CRDM fault status.Through the comparative test of different diagnosis models,the results show that the proposed method can effectively diagnose and identify control rod drive system faults,and the average identification accuracy can reach 98%.

关 键 词：控制棒驱动机构 电路故障 机械故障 电流特征 集合经验模态分解 长短记忆神经网络 

分 类 号：TH86[机械工程—仪器科学与技术] TL362[机械工程—精密仪器及机械]