Convolutional Neural Networks for Fault Diagnosis Using Rotating Speed Normalized Vibration
Fault diagnosis is vital for the health management of rotating machinery. The non-stationary working conditions is one of the major challenges in this field. The key is to extract working-condition-invariant but fault-discriminative features. Traditional methods use expert knowledge on the machines and signal processing to extract fault features from vibration signals manually. This paper regards this issue as a domain adaption problem and utilizes deep learning technique to learn fault discriminative features automatically. We teach deep Convolutional Neural Networks to pronounce diagnostic results from raw vibration data and propose a Rotating Speed Normalization method to improve the domain adaption ability of the neural network models. A case study of rotor crack diagnosis under non-stationary and ever-changing rotating speeds is presented. Using 95600 signal segments, we compare the diagnostic performance of ours and reported Convolutional Neural Network models. The results show that our model gives solid diagnostic accuracy from non-stationary vibration signals, and the proposed Rotating Speed Normalization method can successfully boost the performance of all investigated CNN models.
KeywordsFault diagnosis Rotating machine Deep learning Domain adaption Convolutional Neural Networks
This research is supported by National Natural Science Foundation of China (51305067), National Key Research and Development Program of China (2017YF- C0108401, 2016YFB1200401), Natural Sciences and Engineering Research Council of Canada (Grant #RGPIN-2015-04897), and Future Energy Systems research under Canada First Research Excellent Fund (CFREF) (FES-T14-P02). The authors would also like to acknowledge the help of Mr. Peng Chen and Mr. Mian Zhang.
- 1.Chen P, Wang K, Feng K (2016) Application of order-tracking holospectrum to cracked rotor fault diagnostics under nonstationary conditions. In: 2016 Prognostics and system health management conference (PHM-Chengdu)Google Scholar
- 2.Chen Y, Zaki MJ (2017) KATE: K-competitive autoencoder for text. In: Proceedings of the 23rd ACM SIGKDD international conference on knowledge discovery and data mining, KDD 2017. ACM, pp 85–94Google Scholar
- 3.Chen Z, Li C, Sanchez RV (2015) Gearbox fault identification and classification with convolutional neural networks. Shock and VibrationGoogle Scholar
- 5.He K, Zhang X, Ren S, Sun J (2016) Deep residual learning for image recognition. In: 2016 IEEE conference on computer vision and pattern recognition (CVPR), pp 770–778Google Scholar
- 17.Simonyan K, Zisserman A (2014) Very deep convolutional networks for large-scale image recognition. CoRR abs/1409.1556Google Scholar
- 19.Xia M, Li T, Xu L, Liu L, de Silva CW (2017) Fault diagnosis for rotating machinery using multiple sensors and convolutional neural networks. IEEE/ASME Trans Mechatron PP(99):101–110Google Scholar
- 20.Xie J, Zhang L, Duan L, Wang J (2016) On cross-domain feature fusion in gearbox fault diagnosis under various operating conditions based on Transfer Component Analysis. In: 2016 IEEE international conference on prognostics and health management (ICPHM)Google Scholar