基于逆残差卷积的轻量级电机轴承故障诊断方法  

A Fault Diagnosis Method of Lightweight Motor Bearing Based on Inverse Residual Convolution

作　　者：徐丽红[1] XU Lihong(School of Mechanical and Energy Engineering,Shanghai Technical Institute of Electronics&Information,Shanghai 201411,China)

机构地区：[1]上海电子信息职业技术学院,机械与能源工程学院,上海201411

出　　处：《微型电脑应用》2023年第5期155-158,共4页Microcomputer Applications

摘　　要：针对深度神经网络故障诊断方法存在参数量大、难以应用在移动设备上的问题,提出一种基于逆残差卷积的轻量级电机轴承故障诊断方法。提出用电机轴承原始振动信号数据的小波包节点系数重构二维小波包图像,具体来说,以轻量级残差网络(ResNet)为基础,将其中卷积改为更轻量级的深度可分离卷积,从而达到了缩短模型训练时间之目标。实验结果验证了所提方法用于轴承故障诊断的有效性。For fault diagnosis method used deep neural network has the problems of large number of parameters and difficulty in use of mobile devices.In this paper,a lightweight motor bearing fault diagnosis method based on inverse residual convolution is proposed.The method reconstructs a two-dimensional wavelet packet image with the wavelet packet node coefficients of the original vibration signal data of the motor bearing.Specifically,the lightweight residual network(ResNet)is used as the basis,in which the convolution is changed to a more lightweight depth-separable convolution,thus the goal of reducing the model training time is achieved.The experimental results verify the effectiveness of the proposed method for bearing fault diagnosis.

关 键 词：电机轴承故障诊断 小波包变换 深度神经网络 ResNet 深度可分离卷积 

分 类 号：TM307[电气工程—电机] TP277[自动化与计算机技术—检测技术与自动化装置]