Improved graph-regularized deep belief network with sparse features learning for fault diagnosis

Yang, Jie; Bao, Weimin; Li, Xiaoping; Liu, Yanming

doi:10.1007/s00521-022-06972-5

Cited by 7 publications

(4 citation statements)

References 39 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It can be seen from the previous description that the dynamic weight µ is determined by the fixed factor c and the loss function value E. To increase the decisive role of E, it is appropriate to choose a value greater than 1 for c. To explore the best values of c 1 and c 2 , several sets of experiments were carried out in the transfer case A-D, and the experimental results are shown in figure 13. Among them, the value ranges of c 1 and c 2 were in [0.5, 1.5] and [1,3], respectively. Although higher accuracy can be obtained when c 1 is set to 0.5, considering the comprehensive dynamic weight strategy, c 1 = 1 and c 2 = 2.5 are selected as the optimal parameters.…”

Section: Comparison Of Different Methodsmentioning

confidence: 99%

“…Since the state of rotating mechanical parts has an important impact on the safe operation of machinery, it is necessary to conduct fault diagnosis research on components [1]. In recent years, with further research on deep learning, datadriven intelligent diagnosis algorithms have developed rapidly [2,3]. Their main advantage is that they can perform intelligent feature extraction on data from massive data [4,5].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A transfer-learning fault diagnosis method considering nearest neighbor feature constraints

Zeng¹,

Li²,

Li³

et al. 2022

Meas. Sci. Technol.

View full text Add to dashboard Cite

Aiming at the problem of low diagnostic accuracy of fault diagnosis models due to changes in actual operating conditions, a novel fault diagnosis method based on transfer learning considering nearest neighbor feature constraints (NNFCTL) is proposed. Firstly, nearest neighbor samples are considered to measure data features, as well as a nearest neighbor feature constraint strategy is designed to improve the feature extraction performance of the network. Secondly, a multiple alignment strategy of nearest neighbor samples is proposed to enhance the domain adaptation performance of the network model utilizing multiple alignments. Then, a loss function dynamic weighting strategy is used to improve the convergence of the loss function during model training. Finally, the experimental verification is carried out on the public data set of Western Reserve University and the private data set. The experimental results show that the proposed method exhibits superior transfer performance with reliability and stability compared with the existing methods.

show abstract

Section: Comparison Of Different Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

A transfer-learning fault diagnosis method considering nearest neighbor feature constraints

Zeng¹,

Li²,

Li³

et al. 2022

Meas. Sci. Technol.

View full text Add to dashboard Cite

show abstract

“…More researchers are focusing on deep learning approaches. Yang et al [14] proposed the Gaussian-Bernoulli deep belief network (GDBN). The method applies graph regularization and sparse feature learning to the GDBN, which can generate discriminant features and is highly separable.…”

Section: Fault Diagnosismentioning

confidence: 99%

Two-Stage Edge-Side Fault Diagnosis Method Based on Double Knowledge Distillation

Yang,

Long,

Lin

et al. 2023

Computers, Materials &Amp; Continua

View full text Add to dashboard Cite

With the rapid development of the Internet of Things (IoT), the automation of edge-side equipment has emerged as a significant trend. The existing fault diagnosis methods have the characteristics of heavy computing and storage load, and most of them have computational redundancy, which is not suitable for deployment on edge devices with limited resources and capabilities. This paper proposes a novel two-stage edge-side fault diagnosis method based on double knowledge distillation. First, we offer a clustering-based self-knowledge distillation approach (Cluster KD), which takes the mean value of the sample diagnosis results, clusters them, and takes the clustering results as the terms of the loss function. It utilizes the correlations between faults of the same type to improve the accuracy of the teacher model, especially for fault categories with high similarity. Then, the double knowledge distillation framework uses ordinary knowledge distillation to build a lightweight model for edge-side deployment. We propose a two-stage edge-side fault diagnosis method (TSM) that separates fault detection and fault diagnosis into different stages: in the first stage, a fault detection model based on a denoising auto-encoder (DAE) is adopted to achieve fast fault responses; in the second stage, a diverse convolution model with variance weighting (DCMVW) is used to diagnose faults in detail, extracting features from micro and macro perspectives. Through comparison experiments conducted on two fault datasets, it is proven that the proposed method has high accuracy, low delays, and small computation, which is suitable for intelligent edge-side fault diagnosis. In addition, experiments show that our approach has a smooth training process and good balance.

show abstract

“…Assorted neural network architectures and enhanced techniques, such as Deep Belief Network (DBN) [18], Convolutional Neural Network (CNN) [19][20][21], Recurrent Neural Network (RNN) [22,23], Attention mechanism [24], Transformer [25,26] and their variants [27,28] have also been generally exploited for the fault diagnosis. Jie et al [29] established a novel Gaussian-Bernoulli deep belief network (GDBN) model for intelligent the same architecture. Moreover, the centering and sharpening operations included in the self-distillation with no labels algorithm are integrated into the proposed framework to train the model networks, which enable us effectively avoid the mode collapse during the training procedure.…”

Section: Introductionmentioning

confidence: 99%

A Wavelet Transform and self-supervised learning-based framework for bearing fault diagnosis with limited labeled data

Jin¹,

Hou²,

Du³

et al. 2022

Preprint

View full text Add to dashboard Cite

A novel fault diagnosis framework for bearings with limited labeled data is developed.• Satisfactory diagnosis accuracy is achieved on two bearing fault datasets consisting of only 1% labeled data.• The effect of hyperparameters on fault diagnosis performance and computational complexity of the proposed framework is discussed in detail.• Without any supervised sample, fault-specific features with inter-class separability are observed via visualization technique.

show abstract

Improved graph-regularized deep belief network with sparse features learning for fault diagnosis

Cited by 7 publications

References 39 publications

A transfer-learning fault diagnosis method considering nearest neighbor feature constraints

A transfer-learning fault diagnosis method considering nearest neighbor feature constraints

Two-Stage Edge-Side Fault Diagnosis Method Based on Double Knowledge Distillation

A Wavelet Transform and self-supervised learning-based framework for bearing fault diagnosis with limited labeled data

Contact Info

Product

Resources

About