A wavelet packet transform-based deep feature transfer learning method for bearing fault diagnosis under different working conditions

Yu, Xiao; Liang, Zhongting; Wang, Youjie; Yin, Hongshen; Liu, Xiaowen; Yu, Wanli; Huang, Yanqiu

doi:10.1016/j.measurement.2022.111597

Cited by 55 publications

(16 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…MK-MMD is an UDA algorithm designed to improve the performance of cross-domain learning by minimizing the differences between the source and target domains [39][40][41]. As shown in figure 3, this algorithm utilizes a distance metric called MMD to quantify the distance between two distributions, allowing the mapping of the source and target domains into a feature space and computing their distances in that space.…”

Section: Construction Of Mk-mmdmentioning

confidence: 99%

Innovative integration of multi-scale residual networks and MK-MMD for enhanced feature representation in fault diagnosis

Li,

Yuan,

et al. 2024

Meas. Sci. Technol.

View full text Add to dashboard Cite

This article proposes a method based on multi-scale expansion of residual neural networks to address challenges in the operation of rotating components, such as bearings and gears, under complex conditions where they are often affected by environmental noise. This interference leads to weaker fault characteristics, making feature selection difficult and increasing the presence of extraneous information features. To tackle this issue, the proposed method first employs a multi-scale feature residual neural network to extract features from vibration signals of rotating machinery. The method decomposes the signal into multiple sub-signals of different scales, extracting local features at each scale. It then uses residual connections to combine these local features to obtain a global feature representation. Furthermore, the article introduces a construction of the maximum mean discrepancy and minimization of entropy boundaries to adapt to the differences between two domains. The method utilizes multiple kernel functions to calculate distances between data at different scales and combines these distances to obtain a comprehensive measure. By employing the maximum mean discrepancy and minimization of entropy boundary approach, the method can more accurately determine whether signals at different scales belong to the same category, thereby improving diagnostic accuracy and robustness. Experimental results demonstrate the effectiveness of the proposed method in unsupervised cross-domain fault diagnosis tasks. Future work will focus on further optimizing the architecture of residual neural networks, enhancing feature extraction capabilities, and exploring advanced data augmentation methods to further improve the model's generalization performance.

show abstract

Section: Construction Of Mk-mmdmentioning

confidence: 99%

Innovative integration of multi-scale residual networks and MK-MMD for enhanced feature representation in fault diagnosis

Li,

Yuan,

et al. 2024

Meas. Sci. Technol.

View full text Add to dashboard Cite

show abstract

“…, u N ] ∈ R N×N in order to make it as ideal as possible. The subspace representation model of SSC is equation (12),…”

Section: Sparse Subspace Clustering (Ssc)mentioning

confidence: 99%

“…Contrastingly, the WPT offers a more precise decomposition of high-frequency components. For example, [12] devised a distinctive WPT time-frequency feature map construction method using WPT for time-frequency analysis on nonlinear and non-stationary vibration signals. He and Ye [13] utilized WPT to extract the spectrum of signals, which is then fed into a neural network for parameter adjustment.…”

Section: Introductionmentioning

confidence: 99%

Fault diagnosis method for bogie bearings of high-speed trains based on weighted-sparse subspace clustering

Yang,

Liu,

Chen

et al. 2023

Meas. Sci. Technol.

View full text Add to dashboard Cite

The bogie bearings are crucial components of the high-speed train transmission system, and any failure of them can severely impact the normal functioning of the vehicle. Nowadays, sparse subspace clustering (SSC) is an effective technology for diagnosing mechanical system faults. However, SSC is easily affected by noise. To address this issue, we introduce the weighted-SSC algorithm, which incorporates weighting coefficients to enhance the connections between similar sample points. Our approach involves extracting the fault characteristic parameters of the vibration signal through Wavelet Packet Transform (WPT) and Singular Value Decomposition (SVD). Subsequently, these parameters are employed to construct corresponding weighted-sparse subspaces. We also examined the selection of a hyperparameter in the weighted coefficients. The resultant sparse representation vectors are then harnessed for the purpose of transductive simi-supervised learning clustering and diagnosing the specific type of fault. To validate the effectiveness of our proposed method, we designed and built a bearing experimental platform. We compare our method with existing clustering algorithms, including K-means, SSC, Dimension Reduction Graph (DRG)-based SSC, and Ant Colony Optimization (ACO)-K-means clustering algorithms. The results demonstrate that our proposed weighted-SSC method achieves higher accuracy in fault diagnosis than the other clustering algorithms.

show abstract

“…For this problem, Yu X developed a processing method of an ensemble empirical mode decomposition with adaptive noise (CEEMDAN) joint wavelet packet threshold to process ultrasonic, non-destructive testing defect signals, and concluded that the method can reduce the reconstruction error and iteration time and ensures the integrity of the signal. However, it has a relatively complex calculation process and has strict requirements on the installation position of the sensor [ 19 ].…”

Section: Introductionmentioning

confidence: 99%

Research on Signal Feature Extraction of Natural Gas Pipeline Ball Valve Based on the NWTD-WP Algorithm

Liu

Wang

et al. 2023

Sensors

View full text Add to dashboard Cite

The measured signals of internal leakage detection of the large-diameter pipeline ball valve in natural gas pipeline systems usually contain background noise, which will affect the accuracy of internal leakage detection and sound localization of internal leakage points due to the interference of noise. Aiming at this problem, this paper proposes an NWTD-WP feature extraction algorithm by combining the wavelet packet (WP) algorithm and the improved two-parameter threshold quantization function. The results show that the WP algorithm has a good feature extraction effect on the valve leakage signal, and the improved threshold quantization function can avoid the defects of the traditional soft threshold function and hard threshold function, such as discontinuity and the pseudo-Gibbs phenomenon, when reconstructing the signal. The NWTD-WP algorithm is effective in extracting the features of the measured signals with low signal/noise ratio. The denoise effect is much better than that of the traditional soft and hard threshold quantization functions. It proved that the NWTD-WP algorithm can be used for studying the existing safety valve leakage vibration signals in the laboratory and the internal leakage signals of the scaled-down model of the large-diameter pipeline’s ball valve.

show abstract

A wavelet packet transform-based deep feature transfer learning method for bearing fault diagnosis under different working conditions

Cited by 55 publications

References 42 publications

Innovative integration of multi-scale residual networks and MK-MMD for enhanced feature representation in fault diagnosis

Innovative integration of multi-scale residual networks and MK-MMD for enhanced feature representation in fault diagnosis

Fault diagnosis method for bogie bearings of high-speed trains based on weighted-sparse subspace clustering

Research on Signal Feature Extraction of Natural Gas Pipeline Ball Valve Based on the NWTD-WP Algorithm

Contact Info

Product

Resources

About