Construction of a batch-normalized autoencoder network and its application in mechanical intelligent fault diagnosis
Open accessThe DOI matches its Crossref record; 33 of the 33 references it cites were found in a registry or an index.
Abstract
Among various fault diagnosis methods, deep learning has shown state-of-the-art performance in processing mechanical big data. This paper investigates a reliable deep learning method known as autoencoder, which is most suitable for automatic feature extraction of fault signals. However, traditional autoencoders have two deficiencies: (1) the multi-layer structure of autoencoder has an internal covariate shift problem, which will cause great difficulty for the network training. (2) The application of autoencoder in the case of rotating speed fluctuation is not mature. To overcome the aforementioned deficiencies, batch normalization strategy is employed in every layer of the autoencoder network to obtain a steady distribution of activation values during training. It can regularize the network without parameter adjustment, and deal with the speed fluctuation problem perfectly. So, a new network named batch-normalized autoencoder is first proposed for intelligent fault diagnosis. The raw vibration signals are directly fed into the network and the extracted features are employed to train a softmax classifier for health state identification. A bearing and a gearbox data set are finally used to confirm the effectiveness of the proposed method. The results manifest that the proposed method can extract salient features from the raw signals and handle the fault diagnosis problem under the speed fluctuation problem.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Citations
Cites
(2008) · Journal of Machine Learning Research · Cited by 33,051
Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
Ioffe, S., & Szegedy, C. (2015) · arXiv · doi:10.48550/arxiv.1502.03167 · Cited by 27,818
(2010) · International Conference on Machine Learning · Cited by 13,995
Deep learning and its applications to machine health monitoring
Zhao, R., Yan, R., Chen, Z., Mao, K., Wang, P., & Gao, R. X. (2019) · Mechanical Systems and Signal Processing · doi:10.1016/j.ymssp.2018.05.050 · Cited by 2,894 · cited 1×
“In modern industries, internet of things, smart manufacturing and data-driven systems have been revolutionizing manufacturing through enabling computers to collect the massive amount of data from monitored machines and turn the mechanical big data into operational information [1].”
BackgroundJia, F., Lei, Y., Lin, J., Zhou, X., & Lu, N. (2016) · Mechanical Systems and Signal Processing · doi:10.1016/j.ymssp.2015.10.025 · Cited by 1,710 · cited 1×
“Jia et al [12] proposed stacked autoencoder (SAE)- based DNNs for roller bearing and planetary gearbox fault diagnosis with input as frequency spectra after Fourier transform.”
BackgroundSupport vector machine in machine condition monitoring and fault diagnosis
Widodo, A., & Yang, B.-S. (2007) · Mechanical Systems and Signal Processing · doi:10.1016/j.ymssp.2006.12.007 · Cited by 1,650 · cited 1×
“Conventional data-driven methods are usually to design a right set of features, and then put the features into some shallow machine learning models such as Naive Bayes [7], support vector machines (SVMs) [8] and logistic regression [9].”
MethodFeng, Z., Liang, M., & Chu, F. (2013) · Mechanical Systems and Signal Processing · doi:10.1016/j.ymssp.2013.01.017 · Cited by 998
A sparse auto-encoder-based deep neural network approach for induction motor faults classification
Sun, W., Shao, S., Zhao, R., Yan, R., Zhang, X., & Chen, X. (2016) · Measurement · doi:10.1016/j.measurement.2016.04.007 · Cited by 698 · cited 2×
“Sun et al [16] adopted a regularization technique called ‘dropout’ to mask portions of output neurons randomly in the sparse autoencoder, so as to reduce overfitting in the training process.”
MethodBearing fault diagnosis based on wavelet transform and fuzzy inference
Lou, X., & Loparo, K. A. (2004) · Mechanical Systems and Signal Processing · doi:10.1016/s0888-3270(03)00077-3 · Cited by 689 · cited 1×
“The bearing data set provided by Case Western Reserve University [26] is analyzed in this section.”
MethodJia, F., Lei, Y., Lu, N., & Xing, S. (2018) · Mechanical Systems and Signal Processing · doi:10.1016/j.ymssp.2018.03.025 · Cited by 587 · cited 1×
“To illustrate the strong feature extraction ability of the BNAE network, 12 weight vectors of W trained by BNAE are displayed in figure 9(a), which is used to show the patterns of the network weights learning [31].”
Method(2007) · The MIT Press eBooks · doi:10.7551/mitpress/7496.001.0001 · Cited by 536
Jia, F., Lei, Y., Guo, L., Lin, J., & Xing, S. (2018) · Neurocomputing · doi:10.1016/j.neucom.2017.07.032 · Cited by 517 · cited 3×
“Jia et al [12] proposed stacked autoencoder (SAE)- based DNNs for roller bearing and planetary gearbox fault diagnosis with input as frequency spectra after Fourier transform.”
BackgroundScaling Learning Algorithms toward AI
Bengio, Y., & LeCun, Y. (2007) · Large-Scale Kernel Machines · doi:10.7551/mitpress/7496.003.0016 · Cited by 455
Ngiam, J., Chen, Z., Bhaskar, S. A., Koh, P. W., & Ng, A. Y. (2011) · neural information processing systems · Cited by 297
Text classification based on deep belief network and softmax regression
Jiang, M., Liang, Y., Feng, X., Fan, X., Pei, Z., Xue, Y., & Guan, R. (2018) · Neural Computing and Applications · doi:10.1007/s00521-016-2401-x · Cited by 268 · cited 1×
“Then, we combine the learned feature set with the health label set ¶ f , li ©N i=1 to train softmax regression [25].”
MethodBatch-normalized deep neural networks for achieving fast intelligent fault diagnosis of machines
Wang, J., Li, S., An, Z., Jiang, X., Qian, W., & Ji, S. (2019) · Neurocomputing · doi:10.1016/j.neucom.2018.10.049 · Cited by 242
Self-citationMuralidharan, V., & Sugumaran, V. (2012) · Applied Soft Computing · doi:10.1016/j.asoc.2012.03.021 · Cited by 241 · cited 1×
“Conventional data-driven methods are usually to design a right set of features, and then put the features into some shallow machine learning models such as Naive Bayes [7], support vector machines (SVMs) [8] and logistic regression [9].”
MethodStacked Sparse Autoencoder-Based Deep Network for Fault Diagnosis of Rotating Machinery
Qi, Y., Shen, C., Wang, D., Shi, J., Jiang, X., & Zhu, Z. (2017) · IEEE Access · doi:10.1109/access.2017.2728010 · Cited by 236 · cited 1×
“On the other hand, with fast-developing computer systems and sensors, data-driven fault diagnosis systems have drawn increasing public attention [4, 5].”
BackgroundJiang, X., Wang, J., Shi, J., Shen, C., Huang, W., & Zhu, Z. (2019) · Mechanical Systems and Signal Processing · doi:10.1016/j.ymssp.2018.07.014 · Cited by 234 · cited 1×
“In contrast with top-down modeling proposed by physics-based fault diagnosis models, data-driven systems provide a bottom-up model for detecting the occurrence of machinery faults [2, 3].”
Self-citationRolling Bearing Fault Diagnosis Based on STFT-Deep Learning and Sound Signals
Liu, H., Li, L., & Ma, J. (2016) · Shock and Vibration · doi:10.1155/2016/6127479 · Cited by 213
Intelligent fault diagnosis of rolling bearings using an improved deep recurrent neural network
Jiang, H., Li, X., Shao, H., & Zhao, K. (2018) · Measurement Science and Technology · doi:10.1088/1361-6501/aab945 · Cited by 178
Degradation Assessment and Fault Modes Classification Using Logistic Regression
Yan, J., & Lee, J. (2005) · Journal of Manufacturing Science and Engineering · doi:10.1115/1.1962019 · Cited by 166 · cited 1×
“Conventional data-driven methods are usually to design a right set of features, and then put the features into some shallow machine learning models such as Naive Bayes [7], support vector machines (SVMs) [8] and logistic regression [9].”
MethodWavelet leaders multifractal features based fault diagnosis of rotating mechanism
Du, W., Tao, J., Li, Y., & Liu, C. (2013) · Mechanical Systems and Signal Processing · doi:10.1016/j.ymssp.2013.09.003 · Cited by 145 · cited 1×
“In [28], wavelet leaders multifractal features and an SVM model were used to represent ten health conditions of bearings under four loads, and finally obtained 88.”
MethodLin, J., & Chen, Q. (2013) · Mechanical Systems and Signal Processing · doi:10.1016/j.ymssp.2012.12.014 · Cited by 136 · cited 1×
“Lin et al [29] proposed a bearing diagnosis method using multifractal detrended fluctuation and achieved 96.67% accuracy.”
MethodInfluentialImageNet pre-trained models with batch normalization
Simon, M., Rodner, E., & Denzler, J. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1612.01452 · Cited by 130 · cited 2×
“In addition, internal covariate shift is also a tricky problem during the training process [19].”
BackgroundMultifeatures Fusion and Nonlinear Dimension Reduction for Intelligent Bearing Condition Monitoring
Guo, L., Gao, H., Huang, H., He, X., & Li, S. (2016) · Shock and Vibration · doi:10.1155/2016/4632562 · Cited by 117
Self-citationLi, W., Zhang, S., & He, G. (2013) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2013.2245180 · Cited by 109 · cited 1×
“[27] proposed a method that combined 19 time-domain and frequency-domain features with a self-organizing map for the bearing diagnosis under four loads, and achieved 95.8% testing accuracy.”
MethodInfluentialFault diagnosis method study in roller bearing based on wavelet transform and stacked auto-encoder
Junbo, T., Weining, L., Juneng, A., & Xueqian, W. (2015) · The 27th Chinese Control and Decision Conference (2015 CCDC) · doi:10.1109/ccdc.2015.7162738 · Cited by 78
Qian, W., Li, S., Wang, J., An, Z., & Jiang, X. (2018) · Measurement Science and Technology · doi:10.1088/1361-6501/aad101 · Cited by 66 · cited 1×
“In contrast with top-down modeling proposed by physics-based fault diagnosis models, data-driven systems provide a bottom-up model for detecting the occurrence of machinery faults [2, 3].”
Self-citationFault diagnosis of hydraulic pump based on stacked autoencoders
Huijie, Z., Ting, R., Xinqing, W., You, Z., & Husheng, F. (2015) · 2015 12th IEEE International Conference on Electronic Measurement & Instruments (ICEMI) · doi:10.1109/icemi.2015.7494195 · Cited by 50 · cited 1×
“In the work of [17], rectified linear unit (ReLU) activation function and dropout were adopted in the SAE-based model.”
MethodXu, Z., Li, Y., Wang, Z., & Xuan, J. (2015) · Neurocomputing · doi:10.1016/j.neucom.2015.12.015 · Cited by 36 · cited 1×
“Xu et al [30] applied multiple domain features and ensemble fuzzy ARTMAP neural networks to distinguish the health conditions and 99.”
MethodZero-bias autoencoders and the benefits of co-adapting features
Konda, K., Memisevic, R., & Krueger, D. (2014) · arXiv (Cornell University) · doi:10.48550/arxiv.1402.3337 · Cited by 33 · cited 1×
“Note that the bias b1 is not used, since its effect will be canceled in the following mean subtraction process [22].”
BackgroundThe Loss Surface of Residual Networks: Ensembles and the Role of Batch NormalizationPreprint
Littwin, E., & Wolf, L. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1611.02525 · Cited by 15 · cited 1×
“Through several recent studies [19 – 21], batch normalization has shown its effectiveness in decreasing the internal covariate shift problem.”
Cited by
Applications of machine learning to machine fault diagnosis: A review and roadmap
Lei, Y., Yang, B., Jiang, X., Jia, F., Li, N., & Nandi, A. K. (2020) · Mechanical Systems and Signal Processing · doi:10.1016/j.ymssp.2019.106587 · Cited by 2,997
Zhao, B., Zhang, X., Li, H., & Yang, Z. (2020) · Knowledge-Based Systems · doi:10.1016/j.knosys.2020.105971 · Cited by 386
Cao, H., Shao, H., Zhong, X., Deng, Q., Yang, X., & Xuan, J. (2022) · Journal of Manufacturing Systems · doi:10.1016/j.jmsy.2021.11.016 · Cited by 290
Liu, D., Cui, L., & Cheng, W. (2024) · Measurement Science and Technology · doi:10.1088/1361-6501/acf390 · Cited by 256
“Wang et al [66] developed a batch-normalized AE (BNAE) model, in which batch normalization was introduced into each layer of the AE model to extract more robust features, and then a softmax layer was applied for fault classification.”
A Review on Deep Learning Applications in Prognostics and Health Management
Zhang, L., Lin, J., Liu, B., Zhang, Z., Yan, X., & Wei, M. (2019) · IEEE Access · doi:10.1109/access.2019.2950985 · Cited by 232
“In the end-to-end learning regime, although rarely reported, AE-based models for vibration data have been investigated; see [129]–[131].”
Background
Metadata
- Authors
- Wang, J., Li, S., Han, B., An, Z., Xin, Y., Qian, W., & Wu, Q.
- Affiliations
- Nanjing University of Aeronautics and Astronautics (CN)
- Shandong University of Science and Technology (CN)
- Wuhan Ship Development & Design Institute (CN)
- Corresponding author
- Wang, Jinrui
- An, Zenghui
- Venue
- Measurement Science and Technology
- Publisher
- IOP Publishing
- Published
- 2018-12-14
- Volume
- 30
- Issue
- 1
- Pages
- 015106
- ISSN
- 0957-0233 (ISSN-L)
- 1361-6501
- OpenAlex
- W2901137651
- Cited by
- 61
- References
- 33
- Citations by year
- 2026: 1 · 2025: 2 · 2024: 1 · 2023: 4 · 2022: 15 · 2021: 14 · 2020: 16 · 2019: 8
- FWCI
- 5.88
- Type
- journal article
- Language
- en
- Open access
- bronze
- doi.org
- Licence
- publishingsupport.iopscience.iop.org/iop-standard/v1 (vor)
- iopscience.iop.org/info/page/text-and-data-mining (tdm)
- Locations
- Measurement Science and Technology (publishedVersion)
- Funders
- National Key Research and Development Plan of China (2016YFD0700800)
- Advance research field fund project of China (6140210020102)
- National Natural Science Foundation of China (51675262)
- Topics
- Physical Sciences › Engineering › Control and Systems Engineering › Machine Fault Diagnosis Techniques
- Physical Sciences › Engineering › Control and Systems Engineering › Fault Detection and Control Systems
- Physical Sciences › Engineering › Mechanical Engineering › Gear and Bearing Dynamics Analysis
- Keywords
- mechanical fault diagnosis, internal covariate shift, speed fluctuation, vibration signal feature extraction, bearing defects, gearbox failure, softmax classifier, deep learning
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.