Affiliation:
1. School of Mechanical and Electronical Engineering, Lanzhou University of Technology, Lanzhou 730050, China
Abstract
As one of the important parts of modern mechanical equipment, the accurate real-time diagnosis of rolling bearing is particularly important. Traditional fault diagnosis methods have some disadvantages, such as low diagnostic accuracy and difficult fault feature extraction. In this paper, a method combining Wavelet transform (WT) and Deformable Convolutional Neural Network (D-CNN) is proposed to realize accurate real-time fault diagnosis of end-to-end rolling bearing. The vibration signal of rolling bearing is taken as the monitoring target. Firstly, the Orthogonal Matching Pursuit (OMP) algorithm is used to remove the harmonic signal and retain the impact signal and noise. Secondly, the time-frequency map of the signal is obtained by time-frequency transform using Wavelet analysis. Finally, the D-CNN is used for feature extraction and classification. The experimental results show that the accuracy of the method can reach 99.9% under various fault modes, and it can accurately identify the fault of rolling bearing.
Funder
National Natural Science Foundation of China
Subject
Mechanical Engineering,Mechanics of Materials,Geotechnical Engineering and Engineering Geology,Condensed Matter Physics,Civil and Structural Engineering
Reference27 articles.
1. Fault diagnosis method for rolling bearings based on short-time Fourier transform and convolution neural network;H. Li;Journal of Vibration and Shock,2018
2. Rolling bearing fault diagnosis using recursive autocorrelation and autoregressive analyses;G. Reza
3. Gear fault diagnosis method based on ensemble empirical mode decomposition energy entropy and support vector machine;C. Zhang;Journal of Central South University (Natural Science Edition),2012
4. Bearing fault detection based on hybrid ensemble detector and empirical mode decomposition
5. Observer-biased bearing condition monitoring: From fault detection to multi-fault classification
Cited by
29 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献