摘要
针对目前许多基于深度学习的滚动轴承故障诊断方法在检测含有噪声的信号以及载荷变化时,其诊断性能会有所下降的问题。提出一种基于卷积胶囊网络的故障诊断方法;该模型使用两个卷积层的卷积网络直接对原始的一维时域信号进行特征提取,并将其送入胶囊网络,输出每种故障类型的诊断结果;为了验证该模型的诊断性能,选用凯斯西储大学轴承数据库来进行验证,并与常见的卷积神经网络和深度神经网络进行对比。试验结果表明,相比于其它深度学习方法,该方法在载荷变化以及信号受到严重噪声污染时,依然拥有良好的诊断性能。
Aiming at the phenomenon that the diagnostic performance of many rolling bearing fault diagnosis methods based on deep learning will be degraded when the noise-containing signal was detected and load variation,a fault diagnosis method based on convolution capsule network was proposed.The model uses a convolutional network with two layers of large convolution kernels to extract features from the original one-dimensional time domain signal and send it to the capsule network.The output of the model was the diagnostic results for each type of fault.In order to verify the diagnostic performance of the model,the Case Western Reserve University bearing database was selected for experiment,and compared the model with common convolutional neural network and deep neural network.The experimental results show that this method still has superior diagnostic performance in the test signals with serious noise and load variation compared with other deep learning methods.
作者
杨平
苏燕辰
张振
YANG Ping;SU Yanchen;ZHANG Zhen(School of Mechanical Engineering,Southwest Jiaotong University,Chengdu 610031,China)
出处
《振动与冲击》
EI
CSCD
北大核心
2020年第4期55-62,68,共9页
Journal of Vibration and Shock
关键词
卷积网络
胶囊网络
故障诊断
滚动轴承
convolution network
capsule network
fault diagnosis
rolling bearing
