Bearing Fault Diagnosis with a Feature Fusion Method Based on an Ensemble Convolutional Neural Network and Deep Neural Network

Hong-mei, LI; Jin-ying, Huang; Ji, Shuwei

doi:10.3390/s19092034

Cited by 88 publications

(51 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The stride parameter should also be selected carefully, because an overly large stride parameter will inevitably produce undesired shift-variant features. To improve accuracy, several studies attempted to fuse features produced by multiple 1D CNN [127] and GRU [128]. In the end-to-end learning regime, although rarely reported, AE-based models for vibration data have been investigated; see [129]- [131].…”

Section: ) Vibration Datamentioning

confidence: 99%

A Review on Deep Learning Applications in Prognostics and Health Management

et al. 2019

View full text Add to dashboard Cite

Deep learning has attracted intense interest in Prognostics and Health Management (PHM), because of its enormous representing power, automated feature learning capability and best-in-class performance in solving complex problems. This paper surveys recent advancements in PHM methodologies using deep learning with the aim of identifying research gaps and suggesting further improvements. After a brief introduction to several deep learning models, we review and analyze applications of fault detection, diagnosis and prognosis using deep learning. The survey validates the universal applicability of deep learning to various types of input in PHM, including vibration, imagery, time-series and structured data. It also reveals that deep learning provides a one-fits-all framework for the primary PHM subfields: fault detection uses either reconstruction error or stacks a binary classifier on top of the network to detect anomalies; fault diagnosis typically adds a soft-max layer to perform multi-class classification; prognosis adds a continuous regression layer to predict remaining useful life. The general framework suggests the possibility of transfer learning across PHM applications. The survey reveals some common properties and identifies the research gaps in each PHM subfield. It concludes by summarizing some major challenges and potential opportunities in the domain.

show abstract

Section: ) Vibration Datamentioning

confidence: 99%

A Review on Deep Learning Applications in Prognostics and Health Management

et al. 2019

View full text Add to dashboard Cite

show abstract

“…In the experiments, the proposed method is compare with the other five representative methods, including BPNN and SVM with 10 time-domain features and 9 frequency-domain features manually extracted from the raw signals [8], and 1D-CNN [9], CNNEPDNN [31], M2D-CNN [19] with the raw signals.…”

Section: ) Results and Analysismentioning

confidence: 99%

“…Chen et al [23] developed a fault diagnosis method based on CNN and discrete wavelet transform. Li et al [31] developed a network based on an ensemble CNN and deep neural network (CNNEPDNN). Gong et al [19] designed a modified 2D-CNN for fault diagnosis.…”

Section: Introductionmentioning

confidence: 99%

Fault Diagnosis of Rotating Machinery Based on Combination of Deep Belief Network and One-dimensional Convolutional Neural Network

Zou

Jiang

et al. 2019

IEEE Access

View full text Add to dashboard Cite

The traditional intelligent diagnosis methods of rotating machinery generally require feature extraction of the raw signals in advance. However, it is a very time-consuming and laborious process for extracting the sensitive feature information to improve classification performance. Deep learning method, as a novel machine learning approach, can simultaneously achieve feature extraction and pattern classification. With the characteristics of Deep Belief Network (DBN) and one-dimensional Convolutional Neural Network (1D-CNN) (e.g. learning complex nonlinear, sparse connection and weight sharing), a precise diagnosis method based on the combination of DBN and 1D-CNN is proposed. Firstly, the DBN composed of three pre-trained restricted Boltzmann machines (RBMs) is constructed to achieve feature extraction and dimensionality reduction of the high-dimensional raw data. Secondly, the low-dimensional features extracted by DBN are fed into 1D-CNN for further extracting the abstract features. Finally, Soft-max classifier is employed to identify different operating conditions of rotating machinery. The superiority of the proposed method is validated by comparison with several state-of-the art fault diagnosis methods on two experimental cases. Meanwhile, the proposed method is tested in different background noises and on the imbalanced datasets. The results show that it has higher efficiency and accuracy than the state-of-the art fault diagnosis methods. INDEX TERMS Deep belief network (DBN), one-dimensional convolutional neural network (1D-CNN), rotating machinery, feature extraction, intelligent fault diagnosis.

show abstract

“…Feature Extraction Classification Accuracy (%) [30] HOSA + PCA "one-against all" SVM 96.98 [31] Time-frequency domain ANN 93.00 [32] Time-and frequency-domains SVM 98.70 [33] IMFs decomposed by EEMD SVM with parameter optimized by ICD 97.91 [34] EEMD-MPE SSDAE 99.60 [35] CNNEPDNN CNNEPDNN 98.10 [36] FF_FC_MIC SVM 99.17 [37] HHT-WMSC SVM…”

Section: Referencementioning

confidence: 99%

A Rolling Bearing Fault Diagnosis Method Based on EEMD-WSST Signal Reconstruction and Multi-Scale Entropy

Niu

et al. 2020

Entropy

View full text Add to dashboard Cite

Feature extraction is one of the challenging problems in fault diagnosis, and it has a direct bearing on the accuracy of fault diagnosis. Therefore, in this paper, a new method based on ensemble empirical mode decomposition (EEMD), wavelet semi-soft threshold (WSST) signal reconstruction, and multi-scale entropy (MSE) is proposed. First, the EEMD method is applied to decompose the vibration signal into intrinsic mode functions (IMFs), and then, the high-frequency IMFs, which contain more noise information, are screened by the Pearson correlation coefficient. Then, the WSST method is applied for denoising the high-frequency part of the signal to reconstruct the signal. Secondly, the MSE method is applied for calculating the MSE values of the reconstructed signal, to construct an eigenvector with the complexity measure. Finally, the eigenvector is input to a support vector machine (SVM) to find the fault diagnosis results. The experimental results prove that the proposed method, with a better classification performance, can better solve the problem of the effective signal and noise mixed in high-frequency signals. Based on the proposed method, the fault types can be accurately identified with an average classification accuracy of 100%.Entropy 2020, 22, 290 2 of 28 functions (IMFs) by EMD [10]. Aiming at the shortcomings of the EMD, such as modal mixture and end effect, which affect the accuracy of signal decomposition [11], Wu et al. [12] improved the EMD method with auxiliary noise and proposed the ensemble empirical mode decomposition (EEMD) method.The EEMD method results in signal decomposition with anti-noise characteristics, reduces reconstruction errors, and improves the quality of IMFs by adding Gaussian white noise to the original signal. Therefore, the EEMD method is widely used in signal processing and fault diagnosis [13]. However, when the EEMD method is applied independently for decomposing signals and reducing noise, the information in the high-frequency components is also lost as some IMFs are discarded [14]. The WT method has a good performance on the suppression of random noise by having the properties of multi-scale, low entropy, and decorrelation [15]. Jumah et al. [16] proposed a method using wavelet transform and various thresholding techniques, which has a good effect on removing one-dimensional Gaussian white noise. However, there is no uniform standard for the operation of the wavelet threshold method, which has a greater impact on the final result. The wavelet hard threshold denoising method generates discontinuous points and loses some vital information [17]. The wavelet soft threshold denoising method causes distortion phenomena, such as edge blur effect [18].Although the wavelet threshold denoising method effectively removes noise from high-frequency signals, the effect is not ideal, and the useful signals still mix with noise signals. Nevertheless, the wavelet semi-soft threshold with the advantages of the hard and soft threshold can not only preserve the integrity of the signal but al...

show abstract

Bearing Fault Diagnosis with a Feature Fusion Method Based on an Ensemble Convolutional Neural Network and Deep Neural Network

Cited by 88 publications

References 29 publications

A Review on Deep Learning Applications in Prognostics and Health Management

A Review on Deep Learning Applications in Prognostics and Health Management

Fault Diagnosis of Rotating Machinery Based on Combination of Deep Belief Network and One-dimensional Convolutional Neural Network

A Rolling Bearing Fault Diagnosis Method Based on EEMD-WSST Signal Reconstruction and Multi-Scale Entropy

Contact Info

Product

Resources

About