多源信号融合的感应电机转子断条故障诊断方法研究  

Research on induction motor rotor broken bar fault diagnosis method with multi-source signal fusion

作　　者：孙浩 石颉 崔宪 吴宏杰[1,2] SUN Hao;SHI Jie;CUI Xian;WU Hongjie(School of Electronics and Information Engineering,Suzhou University of Science and Technology,Suzhou 215009,China;Jiangsu Provincial Key Laboratory of Building Intelligent Energy Conservation,Suzhou 215009,China)

机构地区：[1]苏州科技大学电子与信息工程学院,江苏苏州215009 [2]江苏省建筑智慧节能重点实验室,江苏苏州215009

出　　处：《微电子学与计算机》2025年第4期88-97,共10页Microelectronics & Computer

基　　金：国家自然科学基金(62073231)。

摘　　要：鉴于感应电机系统的复杂性和运行状态的多变性,以及故障特征之间的相互重叠和干扰,提出了一种基于改进斑马优化注意力机制的卷积神经网络(Improved Convolutional Neural Networks for Optimizing Attention mechanisms in Zebra,IZOA-Attention-CNN)多源信号融合方法,用于感应电机转子断条故障诊断。首先,采用卷积神经网络(Convolutional Neural Network,CNN)自动提取原始信号中的有效故障特征。然后,通过引入注意力机制,对卷积神经网络提取的信号进行加权处理,以提高故障特征信号的表征能力。此外,为更好地综合多源信号的信息,分别使用电流信号和振动信号作为输入进行特征融合;使用改进的斑马优化算法对网络超参数进行寻优,从而训练转子断条故障诊断模型。实验结果表明,该模型在转子断条故障诊断方面表现出了高效性,其诊断精确率达到了98.65%。In view of the complexity of the induction motor system and the variability of the operating state,as well as the mutual overlap and interference between fault features.In this paper,a multi-source signal fusion method based on Improved Convolutional Neural Networks for Optimizing Attention mechanisms in Zebra(IZOA-Attention-CNN)is proposed to be for induction motor rotor broken bar fault diagnosis.Firstly,Convolutional Neural Network(CNN)is used to automatically extract the effective fault features from the original signals.Then,the signals extracted by the CNN are weighted by introducing the attention mechanisms in order to improve the characterization ability of the fault feature signals.In addition,in order to better synthesize the information of multi-source signals,current signals and vibration signals are used as inputs for feature fusion,respectively;and the improved zebra optimization algorithm is used to find the optimal network hyperparameters,so as to train the rotor broken bar fault diagnosis model.The experimental results show that the model exhibits high efficiency in rotor broken bar fault diagnosis,and its diagnostic accuracy reaches 98.65%.

关 键 词：多源信号 特征融合 感应电机 IZOA-Attention-CNN 斑马优化算法 故障诊断 

分 类 号：TM343[电气工程—电机]