基于多层次传感器融合的永磁同步电机故障诊断研究  

Research on Permanent Magnet Synchronous Motor Fault Diagnosis Based on Multi-level Sensor Fusion

作　　者：徐东 唐镜博 张晓飞 XU Dong;TANG Jingbo;ZHANG Xiaofei(92942 Troops of People’s Liberation Army of China,Beijing 100161,China;College of Electrical and Information Engineering,Hunan University,Changsha 410082,China)

机构地区：[1]中国人民解放军92942部队,北京100161 [2]湖南大学电气与信息工程学院,长沙410082

出　　处：《微电机》2023年第6期50-54,共5页Micromotors

摘　　要：针对单一传感器获取的振动信号难以全面表征轴承故障特征的问题,本文提出了一种基于多传感器融合图像、注意力机制和深度残差网络的永磁同步电机故障诊断方法。该方法首先通过对称点模式、矩阵图将三个加速度传感器获取的振动信号分别转换为灰度图像,然后将不同加速度传感器对应的灰度图像作为RGB图像的不同通道进行第一层次的融合,然后基于Resnet、注意力机制设计了一种特征、决策融合的多尺度融合网路,最后将融合后的彩色图像作为所提网络的输入。经过实验验证,对称点模型、矩阵图故障诊断准确率可达96%和91.5%,显著高于采用单一传感器振动信号和单一网络的故障诊断结果。因此与传统单一传感器故障诊断相比,本文所提出的多层次传感器融合方法可以更加全面地表征电机故障的特征,具有更高的电机故障分类准确度。Aiming at the problem that it is difficult to comprehensively characterize the bearing fault characteristics of the vibration signal obtained by a single sensor,a fault diagnosis method for permanent magnet synchronous motor based on multi-sensor fusion image,attention mechanism and deep residual network was proposed.Firstly,the vibration signals obtained by the three accelerometers were converted into grayscale images by symmetrical point mode and matrix diagram,and then the grayscale images corresponding to different accelerometers were used as different channels of RGB images for the first level of fusion,and then a multi-scale fusion network with features and decision fusion was designed based on Resnet and attention mechanism,and finally the fused color image was used as the input of the proposed network.Experimental results show that the fault diagnosis accuracy of symmetry point model and matrix diagram can reach 96%and 91.5%,which is significantly higher than the fault diagnosis results using single sensor vibration signal and single network.Therefore,compared with the traditional single-sensor fault diagnosis,the multi-level sensor fusion method proposed in this paper can more comprehensively characterize the characteristics of motor faults and have higher motor fault classification accuracy.

关 键 词：多层次融合 多传感器融合 深度残差网络 电机故障诊断 

分 类 号：TM351[电气工程—电机] TM341