摘要
近年来随着传感器的技术进步,使得滚动轴承的故障数据获取成几何倍数增长。然而,传统的深度学习方法在处理海量数据时,常常会出现效率低、计算量与占用内存过大的问题。为解决这些问题,文中提出了一个双层宽核卷积神经网络(Two Wide Kernel Convolutional Neural Network,TWCNN)模型用于滚动轴承故障诊断。该模型以一维振动信号作为输入(1D-TWCNN),通过在前两个卷积层中采用宽卷积核提取特征,实现了以较少的参数来获取更大的感受野,因此大幅地减少了网络模型的连接参数,使得模型的计算量大幅减少,效率提升。与传统的优秀轻量化模型MobileNetV3(Small)的变体和ShuffleNetV2相比,文中所提出的1D-TWCNN模型不仅总参数量远小于这两个模型。而且在滚动轴承的故障诊断中的诊断精度更高。
With the technological progress of sensors and the reduction of production costs,the acquisition of fault data of rolling bearings can be geometrically multiplied.However,when processing massive data,traditional deep learning methods often have problems of low efficiency,large amount of computation and memory consumption.In order to solve these problems,a two-wide Kernel Convolutional Neural Network(TWCNN)model is proposed for rolling bearing fault diagnosis.The model takes one-dimensional vibration signals as input(1D-TWCNN),and uses wide convolution kernels in the first two convolution layers to extract features,so it can obtain larger receptive fields with fewer parameters.In this way,the connection parameters of the network model are greatly reduced,the calculation amount of the model is greatly reduced,and the efficiency is improved.Compared with the excellent traditional lightweight model MobileNetV3(Small)variant and ShuffleNetV2,the 1D-TWCNN model proposed in this paper not only has a much smaller total number of parameters than these two models.And the diagnosis accuracy is higher in the fault diagnosis of rolling bearing.
作者
陈洪明
孟威
谭力
王建景
林群煦
CHEN Hongming;MENG Wei;TAN Li;WANG Jianjing;LIN Qunxu(School of Rail Transportation,Wuyi University,Jiangmen 529020,China)
出处
《机械工程师》
2022年第11期59-62,共4页
Mechanical Engineer
基金
2019年度普通高校认定类科研项目(2019GKQNCX104)。
关键词
滚动轴承故障诊断
轻量化模型
宽卷积核
卷积神经网络
rolling bearing fault diagnosis
lightweight model
wide convolution kernel
convolutional neural network
