Rolling Bearings Fault Diagnosis via 1D Convolution Networks

This paper proposes a convolution neural network based fault diagnosis for rolling bearings. Analysis about why convolution networks fit for vibration signals is first introduced. Then, an end to end network that reflects the raw vibration signals to fault types is built. Hyper parameters are tuned guided by the similarities and differences between images and vibration signals. Some conclusions about hyper parameters are drawn from hyper parameters tuning process, which helps us get a deeper understanding of convolution network’s application to vibration signals. The experiment is taken using Case Western Reserve University bearings dataset and 101 categories including different motor speeds are divided from the raw data. The result shows that our model achieves about 99% accuracy. What’s more, even the frequency differences among motor speeds are very small (about 20 RPM, less than 0.5 Hz), our model can also classify them with high accuracy.

[1]  G. Lewicki,et al.  Approximation by Superpositions of a Sigmoidal Function , 2003 .

[2]  H. W. Ngan,et al.  Detection of Motor Bearing Outer Raceway Defect by Wavelet Packet Transformed Motor Current Signature Analysis , 2010, IEEE Transactions on Instrumentation and Measurement.

[3]  Noureddine Zerhouni,et al.  Bearing Health Monitoring Based on Hilbert–Huang Transform, Support Vector Machine, and Regression , 2015, IEEE Transactions on Instrumentation and Measurement.

[4]  Raghunathan Rengaswamy,et al.  A review of process fault detection and diagnosis: Part I: Quantitative model-based methods , 2003, Comput. Chem. Eng..

[5]  Arthur K. Kordon,et al.  Fault diagnosis based on Fisher discriminant analysis and support vector machines , 2004, Comput. Chem. Eng..

[6]  Geoffrey E. Hinton,et al.  ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[7]  Yoshua Bengio,et al.  Gradient-based learning applied to document recognition , 1998, Proc. IEEE.

[8]  Ronald M. Summers,et al.  Deep Convolutional Neural Networks for Computer-Aided Detection: CNN Architectures, Dataset Characteristics and Transfer Learning , 2016, IEEE Transactions on Medical Imaging.

[9]  Geoffrey E. Hinton,et al.  Rectified Linear Units Improve Restricted Boltzmann Machines , 2010, ICML.

[10]  Ping Zhang,et al.  A comparison study of basic data-driven fault diagnosis and process monitoring methods on the benchmark Tennessee Eastman process , 2012 .

[11]  Daming Lin,et al.  A review on machinery diagnostics and prognostics implementing condition-based maintenance , 2006 .

[12]  Nitish Srivastava,et al.  Dropout: a simple way to prevent neural networks from overfitting , 2014, J. Mach. Learn. Res..

[13]  Hamid Reza Karimi,et al.  Data-driven design of robust fault detection system for wind turbines , 2014 .

[14]  Yaguo Lei,et al.  Condition monitoring and fault diagnosis of planetary gearboxes: A review , 2014 .

[15]  Ali Farhadi,et al.  You Only Look Once: Unified, Real-Time Object Detection , 2015, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[16]  Yu Yang,et al.  A fault diagnosis approach for roller bearing based on IMF envelope spectrum and SVM , 2007 .

[17]  S. Joe Qin,et al.  Multivariate process monitoring and fault diagnosis by multi-scale PCA , 2002 .

[18]  Kurt Hornik,et al.  Multilayer feedforward networks are universal approximators , 1989, Neural Networks.

[19]  Yoshua Bengio,et al.  How transferable are features in deep neural networks? , 2014, NIPS.

[20]  Sergey Ioffe,et al.  Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift , 2015, ICML.