摘要
针对近年来对永磁同步电机故障诊断的需求,提高故障诊断的精度。提出了一种基于多尺度特征融合与空洞卷积金字塔模型的永磁同步电机诊断方法,可以通过电机运行时的定子电流数据直接对电机进行故障诊断。利用多尺度特征融合模块提取图像不同尺度、不同分辨率的特征,提高单一图像的信息利用率;通过在特征融合模块中添加注意力机制使网络中不同通道的特征权重保持高度一致,进一步确保了网络提取图像特征的能力;通过在空间池化金字塔中引入空洞卷积核来构建空洞卷积金字塔,在解决了网络对同一特征反复提取、节约计算成本的同时,增强了模型的感受野,提高模型对不同故障的诊断精度。实验结果表明,所提方法对不同类型的电机故障均具有较高的诊断精度。对比传统的智能算法,其算法精度与损失函数都得到了明显改进。
To meet the demand of permanent magnet synchronous motor fault diagnosis in recent years and improve the accuracy of fault diagnosis,a diagnosis method of permanent magnet synchronous motor is proposed based on multi-scale feature fusion and atrous convolution pyramid model,which can directly diagnose the motor fault through the stator current data.Firstly,the multi-scale feature fusion module is used to extract the features of different scales and resolutions of images to improve the information utilization rate of a single image.At the same time,by adding the attention mechanism to the feature fusion module,the feature weights of different channels in the network are highly consistent,which further ensures the ability of the network to extract image features.Finally,the atrous convolution kernel is introduced into the spatial pooling pyramid to construct the atrous spatial pyramid pooling,which not only solves the problem of repeated extraction of the same feature by the network,but also enhances the receptive field of the model and improves the diagnostic accuracy of the model for different faults.The experimental results show that this method has high diagnostic accuracy for different types of motor faults.Compared with those of the traditional intelligent algorithm,the accuracy and loss function of the proposed algorithm are improved obviously.
作者
郭又铭
吴钦木
GUO Youming;WU Qinmu(School of Electrical Engineering,Guizhou University,Guiyang 550025,China)
出处
《无线电工程》
2024年第5期1294-1307,共14页
Radio Engineering
关键词
永磁同步电机
改进ResNet
多尺度特征融合
空洞卷积金字塔
故障诊断
permanent magnet synchronous motor
improved ResNet
multi-scale feature fusion
atrous convolution pyramid
fault diagnosis
