Please wait a minute...
Tsinghua Science and Technology  2019, Vol. 24 Issue (06): 750-762    doi: 10.26599/TST.2018.9010144
REGULAR ARTICLES     
A Deep Adaptive Learning Method for Rolling Bearing Fault Diagnosis Using Immunity
Yuling Tian*, Xiangyu Liu
∙ Yuling Tian and Xiangyu Liu are with the College of Information and Computer, Taiyuan University of Technology, Taiyuan 030000, China. E-mail: 18713340041@163.com.
Download: PDF (1331 KB)      HTML
Export: BibTeX | EndNote (RIS)      

Abstract  

The extraction of rolling bearing fault features using traditional diagnostic methods is not sufficiently comprehensive and the features are often chosen subjectively and depend on human experience. In this paper, an improved deep convolutional process is used to extract a set of features adaptively. The hidden multi-layer feature of deep convolutional neural networks is also exploited to improve the extraction features. A deterministic detection of low-confidence samples is performed to ensure the reliability of the recognition results and to decrease the rate of false positives by evaluating the diagnosis of the deep convolutional neural network. To improve the efficiency of the continuous learning elements of the rolling bearing fault diagnosis, a clone learning strategy based on cloning and mutation operations is proposed. The experimental results show that the proposed deep convolutional neural network model can extract multiple rolling bearing fault features, improve classification and detection accuracy by reducing the false positive rate when diagnosing rolling bearing faults, and accelerate learning efficiency when using low-confidence rolling bearing fault samples.



Key wordsdeep learning      fault diagnosis      feature extraction      clone selection strategy     
Received: 27 November 2018      Published: 20 June 2019
Corresponding Authors: Yuling Tian   
About author:

Xiangyu Liu is a PhD candidate of Information and Computer Department, Taiyuan University of Technology, China. His research interests include deep learning and fault diagnosis.

Cite this article:

Yuling Tian, Xiangyu Liu. A Deep Adaptive Learning Method for Rolling Bearing Fault Diagnosis Using Immunity. Tsinghua Science and Technology, 2019, 24(06): 750-762.

URL:

http://tst.tsinghuajournals.com/10.26599/TST.2018.9010144     OR     http://tst.tsinghuajournals.com/Y2019/V24/I06/750

Fig. 1 Self-adaptive DCNN fault diagnosis model.
Training batchesConvolution layer numberNumber of feature mapsConvolution kernel sizeTraining set accuracy (%)Test set accuracy (%)
Number selection of feature maps50430×50×80×1003×399.8798.86
50460×60×60×603×399.7798.14
504100×80×50×303×399.8399.19
Selection of convolution kernel size504100×80×50×303×399.8399.19
503100×80×504×499.9399.13
503100×80×505×599.9698.62
Selection of training batch size403100×80×504×499.9498.76
603100×80×504×499.9699.16
803100×80×504×499.9698.75
Table 1 Main parameters selection of network model.
Image size (pixel)Network layer numbersConvolution kernel sizePooling sizeTrain set accuracy (%)Test set accuracy (%)
15×1574×42×298.8396.10
20×2084×42×299.6998.19
25×2584×42×299.8797.95
28×2894×42×299.8898.00
32×3294×42×299.8398.19
Table 2 Time domain signal size selection.
Image size (pixel)Network layer numbersConvolution kernel sizePooling sizeTrain set accuracy (%)Test set accuracy (%)
15×1574×42×299.8097.90
20×2084×42×299.9399.10
25×2584×42×299.9199.38
28×2894×42×299.9399.14
32×3294×42×299.9699.16
Table 3 Frequency domain signal size selection.
Fig. 2 DCNN time domain model.
Layer numberNumber of feature mapsFeature map sizeLayer numberNumber of feature mapsFeature map size
1132×326303×3
210029×297301×1
310015×158501
45012×12971
5506×6
Table 4 Time domain signal model structure.
Layer numberNumber of feature mapsFeature map sizeLayer numberNumber of feature mapsFeature map size
1125×255504×4
210022×226301×1
310011×117501
4508×8871
Table 5 Frequency domain signal model structure.
Fig. 3 Normal state time signal.
Fig. 4 Normal state time and frequency domain signal sample.
Fig. 5 Slight rolling element fault time domain signal.
Fig. 6 Slight rolling element fault time and frequency domain signal sample.
Fig. 7 Slight inner ring fault time domain signal.
Fig. 8 Slight inner ring fault time and frequency domain signal sample.
Fig. 9 Slight outer ring fault time domain signal.
Fig. 10 Slight outer ring fault time and frequency domain signal sample.
Technical methodTrain set accuracy (%)Test set accuracy (%)
DCNN-TD-Softmax99.8398.19
DCNN-FD-Softmax99.9199.38
EMD-SVM92.2592.00
EMD-BP86.0381.64
EEMD-SVM91.5690.67
EEMD-BP85.1980.86
FA-SVM81.8381.08
FA-BP80.5580.23
Table 6 Comparison of DCNN and feature extraction techniques.
Algorithm modelInput dataNetwork layer numberTrain set accuracy (%)Test set accuracy (%)
DCNN-TD32×32999.8398.19
DCNN-FD25×25899.9199.38
DBN[21]1024499.40none
DBN-TD20×20425.0025.00
25×25425.0025.00
30×30425.0025.00
DBN-FD20×20425.0025.00
25×25425.0025.00
30×30425.0025.00
Table 7 Comparison of DCNN and DBN.
Algorithm modelInner ring fault (%)Outer ring fault (%)Rolling element fault (%)Severe inner ring fault (%)Severe outer ring fault (%)Severe rolling element fault (%)
DCNN-TD99.9299.8310099.7599.6799.58
DCNN-FD99.8399.9299.9399.9299.9099.86
Table 8 Train set accuracy for various fault types.
Algorithm modelInner ring fault (%)Outer ring fault (%)Rolling element fault (%)Severe inner ring fault (%)Severe outer ring fault (%)Severe rolling element fault (%)
DCNN-TD98.3397.3399.1197.0097.3399.27
DCNN-FD99.2499.4399.5699.2596.0099.67
Table 9 Test set accuracy of various fault types.
Fault typeTrain set (%)Test set (%)Test false positive rate (%)Test missed rate (%)
Sel-DCNNKnown fault99.9299.420.280.57
Unknown faultnone98.322.161.85
DCNN-TDKnown fault99.8398.192.671.89
DCNN-FDKnown fault99.9199.380.340.72
Table 10 Sel-DCNN model diagnosis results.
Fig. 11 Comparison of accuracy of identification.
[1]   Long H., Li C., and Liu H., Feature extraction method of rolling bearing fault signal based on EEMD and cloud model characteristic entropy, Entropy, vol. 17, no. 12, pp. 6683-6697, 2015.
[2]   Ince T., Kiranyaz S., Eren L., Askar M., and Gabbouj M., Real-time motor fault detection by 1-D convolutional neural networks, IEEE Transactions on Industrial Electronics, vol. 63, no. 11, pp. 7067-7075, 2016.
[3]   Cabrera D., Sancho F., Li C., Cerrada M., Sanchez R. V., Pacheco F., and de Oliveira J. V., Automatic feature extraction of time-series applied to fault severity assessment of helical gearbox in stationary and non-stationary speed operation, Applied Soft Computing, vol. 58, pp. 53-64, 2017.
[4]   Dey D., Chatterjee B., Dalai S., Munshi S., and Chakravorti S., A deep learning framework using convolution neural network for classification of impulse fault patterns in transformers with increased accuracy, IEEE Transactions on Dielectrics and Electrical Insulation, vol. 24, no. 6, pp. 3894-3897, 2017.
[5]   Park D., Kim S., An Y., and Jung J. Y., LiReD: A light-weight real-time fault detection system for edge computing using LSTM recurrent neural networks, Sensors, vol. 18, no. 7, pp. 1-15, 2018.
[6]   Tran V. T., AlThobiani F., Tinga T., and Ball A., Single and combined fault diagnosis of reciprocating compressor valves using a hybrid deep belief network, Journal of Mechanical Engineering Science, vol. 232, no. 20, pp. 3767-3780, 2018.
[7]   Xu F., Tse W. T., and Tse Y. L., Roller bearing fault diagnosis using stacked denoising autoencoder in deep learning and Gath-Geva clustering algorithm without principal component analysis and data label, Applied Soft Computing, vol. 73, pp. 898-913, 2018.
[8]   Meng Z., Zhan X. Y., Li J., and Pan Z. Z., An enhancement denoising autoencoder for rolling bearing fault diagnosis, Measurement, vol. 130, pp. 448-454, 2018.
[9]   Ahmed H. O. A., Wong M. L. D., and Nandi A. K., Intelligent condition monitoring method for bearing faults from highly compressed measurements using sparse over-complete features, Mechanical Systems and Signal Processing, vol. 99, pp. 459-477, 2018.
[10]   Tao J., Liu Y., and Yang D., Bearing fault diagnosis based on deep belief network and multisensor information fusion, Shock and Vibration, vol. 2016, no. 7, pp. 1-9, 2016.
[11]   Tran V. T., Althobiani F., and Ball A., An approach to fault diagnosis of reciprocating compressor valves using Teager-Kaiser energy operator and deep belief networks, Expert Systems with Applications, vol. 41, no. 9, pp. 4113-4122, 2014.
[12]   Appana D. K., Ahmad W., and Kim J. M., Speed invariant bearing fault characterization using convolutional neural networks, in Proc. 11th International Workshop on Multi-disciplinary Trends in Artificial Intelligence, Gadong, Brunei, 2017, pp. 189-198.
[13]   David V., Ferrada A. S., Droguett E. L., and Meruane V., Deep learning enabled fault diagnosis using time-frequency image analysis of rolling element bearings, Shock and Vibration, vol. 2017, pp. 1-17, 2017.
[14]   Hasan M. J. and Kim J. M., Bearing fault diagnosis under variable rotational speeds using stockwell transform-based vibration imaging and transfer learning, Applied Sciences-Basel, vol. 8, no. 12, pp. 1-15, 2018.
[15]   Muhammad S., Cheol-Hong K., and Jong-Myon K., A hybrid feature model and deep-learning-based bearing fault diagnosis, Sensors, vol. 17, no. 12, pp. 1-16, 2017.
[16]   Hart E. and Davoudani D., An engineering-informed modelling approach to AIS, in Proc. 10th International Conference on Artificial Immune Systems, Cambridge, UK, 2017, pp. 240-253.
[17]   Montechiesi L., Cocconcelli M., and Rubini R., Artificial immune system via euclidean distance minimization for anomaly detection in bearings, Mechanical Systems and Signal Processing, vols. 76&77, pp. 380-393, 2016.
[18]   Vatefipour O., A novel electric load consumption prediction and feature selection model based on modified clonal selection algorithm, Journal of Intelligent and Fuzzy Systems, vol. 34, no. 4, pp. 2261-2272, 2018.
[19]   Bayar N., Darmoul S., Gabouj S. H., and Pierreval H., Fault detection, diagnosis and recovery using artificial immune systems: A review, Engineering Applications of Artificial Intelligence, vol. 46, pp. 43-57, 2015.
[20]   Smith W. A. and Randall R. B., Rolling element bearing diagnostics using the Case Western Reserve University data: A benchmark study, Mechanical Systems and Signal Processing, vol. 64, no. 12, pp. 100-131, 2015.
[21]   Li W. H., Shan W. P., and Zeng X. Q., Bearing fault identification based on deep belief network, Journal of Vibration Engineering, vol. 29, no. 2, pp. 340-347, 2016.
[22]   Li Q., Liu Y., and Liang H. J., A new fault diagnosis method based on HHT-CNNs and its application in rolling bearing fault diagnosis, in Proceedings of the 36th Chinese Control Conference, Dalian, China, 2017, pp. 7021-7026.
[23]   Zhang W., Peng G., Li C., Chen Y. H., and Zhang Z. J., A new deep learning model for fault diagnosis with good anti-noise and domain adaptation ability on raw vibration signals, Sensors, vol. 17, no. 2, pp. 425-446, 2017.
[1] Weiwei Jiang, Lin Zhang. Geospatial Data to Images: A Deep-Learning Framework for Traffic Forecasting[J]. Tsinghua Science and Technology, 2019, 24(1): 52-64.
[2] Qi Dang, Jianqin Yin, Bin Wang, Wenqing Zheng. Deep Learning Based 2D Human Pose Estimation: A Survey[J]. Tsinghua Science and Technology, 2019, 24(06): 663-676.
[3] Kaiming Nan, Sicong Liu, Junzhao Du, Hui Liu. Deep Model Compression for Mobile Platforms: A Survey[J]. Tsinghua Science and Technology, 2019, 24(06): 677-693.
[4] Xiaocheng Feng,Lifu Huang,Bing Qin,Ying Lin,Heng Ji,Ting Liu. Multi-Level Cross-Lingual Attentive Neural Architecture for Low Resource Name Tagging[J]. Tsinghua Science and Technology, 2017, 22(6): 633-645.
[5] Xijia Liu,Xiaoming Tao,Ning Ge. Fast Remote-Sensing Image Registration Using Priori Information and Robust Feature Extraction[J]. Tsinghua Science and Technology, 2016, 21(5): 552-560.
[6] Zhenlong Yuan,Yongqiang Lu,Yibo Xue. DroidDetector: Android Malware Characterization and Detection Using Deep Learning[J]. Tsinghua Science and Technology, 2016, 21(1): 114-123.
[7] . On the Application of PCA Technique to Fault Diagnosis[J]. Tsinghua Science and Technology, 2010, 15(2): 138-144.
[8] . SDG-Based HAZOP and Fault Diagnosis Analysis to the Inversion of Synthetic Ammonia[J]. Tsinghua Science and Technology, 2007, 12(1): 30-37.
[9] Zunjing Wu, Zhigang Cao. Improved MFCC-Based Feature for Robust Speaker Identification[J]. Tsinghua Science and Technology, 2005, 10(2): 158-161.