基于声振融合和WR-VMD的电机轴承故障诊断研究  

Research on motor bearing fault diagnosis based on vibro-acoustic signal fusion and WR-VMD

作　　者：金成毅 陈建鹏 程伟 徐正国[2,3] JIN Chengyi;CHEN Jianpeng;CHENG Wei;XU Zhengguo(China Nuclear Power Engineering Co.,Ltd.,Shenzhen 518120,China;Huzhou Institute of Zhejiang University,Huzhou 313000,China;College of Control Science and Engineering,Zhejiang University,Hangzhou 310027,China)

机构地区：[1]中广核工程有限公司,广东深圳518120 [2]浙江大学湖州研究院,浙江湖州313000 [3]浙江大学控制科学与工程学院,浙江杭州310027

出　　处：《热力发电》2024年第11期101-111,共11页Thermal Power Generation

基　　金：中国广核集团科研项目(K-A2023.560)。

摘　　要：在火电和核电机组冷却系统中,电机轴承故障信号因具有微弱性和非线性特征,容易被运行信号和无效信号掩盖,采用单一的振动监测不足以采集到完整的故障信息。对此,提出融和声音和振动(声振融合)信号的方法来监测电机轴承的故障信息。首先,针对采集的电机轴承声振信号,提出一种结合小波脊线(WR)和变分模态分解(VMD)的WR-VMD算法提取声振信号的特征;利用WR分析原信号的成分,随后利用获取的信息确定VMD的参数,这样弥补了原VMD方法需要预先根据经验设定参数的不足;最后,将声振信号融合技术和WR-VMD算法用于实测的电机轴承故障数据,仿真结果显示:对比同类型的方法,WR-VMD方法所提取的特征最明显,干扰信息最少,用于故障诊断的准确率最高;融合声振信号相比单一振动信号或者声音信号用于故障诊断的准确率提升了至少7%。In thermal and nuclear power generating unit cooling systems,the bearing fault signal of the motor is weak and nonlinear,which is easily masked by running signals and invalid signals,and the use of a single vibration monitoring may not be sufficient to collect complete defect information.To address this problem,vibration and sound signals are combined to monitor bearing fault signals,and the collected sound and vibration signal features are fused.In order to process the sound and vibration signals of motor bearings,a WR-VMD algorithm that integrates wavelet ridge(WR)and varational mode decomposition(VMD).The WR is first used to analyze the components of the original signal,and then the acquired information is used to determine the parameters of the VMD,which makes up for the shortcomings of the original VMD method,which requires the parameters to be set empirically in advance.The simulated signal results show that compared with the same type of methods,the features extracted by the WR-VMD method are the most obvious and have the least interference information.Finally,the acoustic and vibration signal fusion technique and the WR-VMD algorithm are applied to the measured motor bearing fault data,and the results show that,compared with other feature extraction algorithms of the same type,the WR-VMD extracts the most obvious features and has the highest accuracy in fault diagnosis;the acoustic and vibration signal fusion has at least a 7%increase in accuracy compared with a single vibration or acoustic signal in fault diagnosis.

关 键 词：轴承故障 声振融合 变分模态分解 小波脊线 特征提取 

分 类 号：TM307[电气工程—电机]