We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
A new fault diagnosis method based on convolutional neural network and compressive sensing.
- Authors
Ma, Yunfei; Jia, Xisheng; Bai, Huajun; Liu, Guozeng; Wang, Guanglong; Guo, Chiming; Wang, Shuangchuan
- Abstract
Compressive sensing is an efficient machinery monitoring framework, which just needs to sample and store a small amount of observed signal. However, traditional reconstruction and fault detection methods cost great time and the accuracy is not satisfied. For this problem, a 1D convolutional neural network (CNN) is adopted here for fault diagnosis using the compressed signal. CNN replaces the reconstruction and fault detection processes and greatly improves the performance. Since the main information has been reserved in the compressed signal, the CNN is able to extract features from it automatically. The experiments on compressed gearbox signal demonstrated that CNN not only achieves better accuracy but also costs less time. The influencing factors of CNN have been discussed, and we compared the CNN with other classifiers. Moreover, the CNN model was also tested on bearing dataset from Case Western Reserve University. The proposed model achieves more than 90 % accuracy even for 50 % compressed signal.
- Subjects
ARTIFICIAL neural networks; CASE Western Reserve University; FAULT diagnosis; DIAGNOSIS methods; MONITORING of machinery
- Publication
Journal of Mechanical Science & Technology, 2019, Vol 33, Issue 11, p5177
- ISSN
1738-494X
- Publication type
Article
- DOI
10.1007/s12206-019-1007-5