东北大学学报（自然科学版）

东北大学学报:自然科学版 ›› 2019, Vol. 40 ›› Issue (7): 974-979.DOI: 10.12068/j.issn.1005-3026.2019.07.012

• 机械工程 • 上一篇    下一篇

PSO优化深度神经网络诊断齿轮早期点蚀故障

李嘉琳1， 何巍华1,2， 曲永志3   

  1. (1. 东北大学 机械工程与自动化学院， 辽宁 沈阳110819; 2. 伊利诺伊大学芝加哥分校 机械及工业工程系， 伊利诺伊州 芝加哥60607; 3. 武汉理工大学 机电工程学院， 湖北 武汉430070)
  • 收稿日期:2017-04-29 修回日期:2017-04-29 出版日期:2019-07-15 发布日期:2019-07-16
  • 通讯作者: 李嘉琳
  • 作者简介:李嘉琳(1993-)，男，辽宁沈阳人，东北大学博士研究生； 何巍华(1961-)，男，美国芝加哥人，东北大学教授，博士生导师.
  • 基金资助:
    国家自然科学基金资助项目(51505353，51675089).

Diagnosis of Gear Early Pitting Faults Using PSO Optimized Deep Neural Network

LI Jia-lin1， HE David1,2， QU Yong-zhi3   

  1. 1. School of Mechanical Engineering & Automation， Northeastern University， Shenyang 110819， China; 2. Department of Mechanical & Industrial Engineering， University of Illinois at Chicago， Chicago 60607， USA; 3. School of Mechanical and Electronic Engineering， Wuhan University of Technology， Wuhan 430070， China.
  • Received:2017-04-29 Revised:2017-04-29 Online:2019-07-15 Published:2019-07-16
  • Contact: HE David
  • About author:-
  • Supported by:
    -

RichHTML

0

PDF (PC)

644

摘要： 基于数据驱动方法诊断齿轮故障时一般会用傅里叶变换等进行特征提取，特征提取方法的选取对诊断结果影响很大.提出应用深度神经网络来诊断齿轮早期点蚀故障，直接以采集的振动信号作为网络输入，可以避免特征提取环节产生误差.此外，应用粒子群算法优化深度神经网络，使训练过程更稳定、诊断率更高.在分析结果时应用主成分分析法对网络输出进行降维.用实验采集的数据训练并测试网络，诊断正确率能达到90%之上，证明所提出的方法是合理、可用的.

关键词: 齿轮, 早期点蚀, 粒子群算法, 深度神经网络, 主成分分析

Abstract: When gear faults were diagnosed based on the data-driven method， feature extraction was generally performed by Fourier transform， etc. The feature extraction method used has a great influence on the diagnosis results. Therefore， deep neural network(DNN) was proposed to diagnose early gear pitting faults and the vibration signals are directly used as the network inputs to avoid errors caused by feature extraction. In addition， the particle swarm optimization(PSO) algorithm was applied to optimize the DNN for obtaining a more stable training process and better diagnosis results. Principal component analysis(PCA) algorithm was used to reduce the dimensions of the DNN outputs. The data collected from the experiment was used to train and test the DNN. The fault diagnostic accuracy can reach over 90%， which proves that the proposed method is reasonably effective.

Key words: gear, early pitting, PSO algorithm, deep neural network(DNN), principal component analysis(PCA)

中图分类号: