Bearing fault identification based on convolutional neural network by different input modes
- PDF / 1,528,046 Bytes
- 10 Pages / 595.276 x 790.866 pts Page_size
- 50 Downloads / 158 Views
TECHNICAL PAPER
Bearing fault identification based on convolutional neural network by different input modes Tian Han1 · ZhiXin Tian1 · Zhongjun Yin1 · Andy C. C. Tan2 Received: 4 November 2019 / Accepted: 7 August 2020 / Published online: 15 August 2020 © The Brazilian Society of Mechanical Sciences and Engineering 2020
Abstract Convolutional neural networks (CNNs) have been applied to the field of fault diagnosis as one of the most widely used deep learning architectures. Different input modes of CNN for bearing fault identification were analyzed by researchers to improve recognition accuracy, such as time-domain diagram, grayscale diagram, short-time Fourier transform diagram, and continuous wavelet transform diagram. However, for the data with small sample size and high background noise, the performance of the CNN is constrained. In this paper, one CNN input mode for bearing fault recognition is proposed based on time-domain color feature diagram (TDCF) through adding red color to diagrams. The method significantly enhanced the fault characteristics of the signal, which is beneficial to the CNN extraction of bearing fault features. Convolution visualization illustrates the effectiveness of the proposed method that provides more bearing fault recognition information. Different sample size and color rate were analyzed by bearing vibration data with high noise. The results showed that the bearing fault identification method based on CNN with 0.4 TDCF obtained a highest fault identification accuracy compared with other input mode methods. The feasibility of the proposed method has been validated, which also provides one reference for other faults identification and pattern recognition. Keywords Fault identification · Rolling bearing · Convolutional neural network · Input mode · Image characteristics
1 Introduction Rolling element bearings (REBs) are the essential components in rotary machines. Health conditions of REBs have considerable impacts on the health of machines. According to the literature [1], 45–55% of broken machines are caused by bearing faults. Condition monitoring and fault diagnosis of bearings are essential in the industry to prevent catastrophic failures and eventual shutdown of overall operation. Fault recognition by the traditional methods requires maintenance personnel to have rich knowledge and Technical Editor: Wallace Moreira Bessa, D.Sc. * Tian Han [email protected] 1
School of Mechanical Engineering, University of Science and Technology Beijing, 30 Xueyuan Road, Haidian District, Beijing 100083, China
LKC Faculty of Engineering, University Tunku Abdul Rahman, Sungai Long Campus, Cheras, 43000 Kajang, Selangor, Malaysia
2
experience of the machine and extraction of fault characteristics of the signal through time-domain analysis, frequencydomain analysis, time–frequency analysis, or other methods and calculates the characteristics of the fault bearing by using the empirical formulae to complete the identification of the bearing fault. The most popular way to diagnose bearing
Data Loading...
