一种基于改进卷积神经网络的齿轮故障诊断方法
Method of gear fault diagnosis based on improved convolutional neural network
田彪 1张周锁 1李想1
作者信息
- 1. 西安交通大学机械工程学院,陕西西安 710049
- 折叠
摘要
由于工作环境和复杂工况的影响,齿轮容易损坏,以致造成巨大经济损失和人员伤亡,因此,齿轮故障的早期诊断越来越重要.为了解决齿轮故障早期诊断问题,文中提出了一种基于改进卷积神经网络的齿轮故障诊断方法.该方法基于经典的卷积神经网络(Convolution Neural Networks,CNN),引入了全局平均池化层替代全连接层用于提高神经网络的诊断效率,且加入并行模块结构用于提高故障诊断的准确率.使用齿轮故障数据集进行了试验验证,结果表明:提出的神经网络相比CNN能够有效提高齿轮的故障诊断效率和准确率,具有重要的工程应用意义.
Abstract
Due to the working environment and complex working conditions,the gear is easy to damage,thus resulting in huge economic loss and casualty.Therefore,gear fault diagnosis in the initial stage becomes increasingly important.In order to solve this problem,in this article a method of gear fault diagnosis based on the improved convolution neural network is proposed.This method,based on the classical CNN(Convolution Neural Network),introduces the global average pooling layer,instead of the full connection layer,to improve the neural network's diagnosis efficiency;in addition,the parallel module is added to im-prove the accuracy in fault diagnosis.The gear-fault data set is used for experimental verification.The results show that compared with CNN,this improved convolutional neural network greatly enhances efficiency and accuracy of gear fault diagnosis,which can be applied in engineering projects.
关键词
卷积神经网络/齿轮/故障诊断/全局平均池化/并行模块Key words
CNN/gear/fault diagnosis/global average pooling/parallel module引用本文复制引用
出版年
2024
NETL
NSTL
万方数据