摘要
针对利用卷积神经网络(convolutional neural networks,CNN)对滚动轴承进行故障诊断时可采用的振动信号处理方法较多的情况,设计了基于CNN的振动信号处理方法对比实验,采用不同的振动信号处理方法对滚动轴承在不同工况下的采样数据进行处理,再将动信号输入CNN故障诊断模型进行训练及测试,根据测试精度比较处理方法对故障诊断精度的影响。采用CNN中的AlexNet作为实验模型,选择模型中的最后3个全连接层,以达到快速训练的目的。对比不同信号处理方法对应的检测准确率可知,基于小波变换的滚动轴承故障诊断模型的检测准确率最高。
In view of that many vibration signal processing methods can be used for fault diagnosis of rolling bearings with convolutional neural networks(CNN),a comparative experiment of vibration signal processing methods based on convolutional neural networks is designed.The sampled data under different working conditions were processed by different vibration signal processing methods and input into convolutional neural network fault diagnosis model for training and testing.The impact of these processing methods on the accuracy of fault diagnosis was compared according to test accuracy.In the experiment,the convolutional neural network AlexNet was used as the experimental model.The last three fully connected layers in the model were selected to be trained to achieve fast training.Comparing the detection accuracy of different signal processing methods,it can be seen that the detection accuracy of rolling bearing fault diagnosis model based on wavelet transform is the highest.
作者
周冠禄
江永全
陈锦雄
梅桂明
ZHOU Guanlu;JIANG Yongquan;CHEN Jinxiong;MEI Guiming(State Key Laboratory of Traction Power, Southwest Jiaotong University, Chengdu 610031, China)
出处
《中国科技论文》
CAS
北大核心
2020年第7期729-734,共6页
China Sciencepaper
基金
四川省科技厅重点研究项目(2018GZ0365)。
关键词
故障诊断
卷积神经网络
滚动轴承
振动信号处理
fault diagnosis
convolutional neural networks(CNN)
rolling bearing
vibration signal processing
作者简介
第一作者:周冠禄(1996—),硕士研究生,主要研究方向为深度学习、故障诊断;通讯作者:江永全,助理研究员,主要研究方向为人工智能、计算机仿真,Email:yqjiang@swjtu.edu.cn。
