A local transient feature extraction method via periodic low rank dynamic mode decomposition for bearing incipient fault diagnosis

Zhang, Qixiang; Lv, Yong; Yuan, Rui; Li, Zhaolun; Li, Hewenxuan

doi:10.1016/j.measurement.2022.111973

Cited by 14 publications

(5 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Then, a robust tensor low rank component extraction framework is constructed in MDMD for simultaneous extraction of bearing fault features from the multivariate signal, which enhance the noise robustness of the proposed method. In our previous work [33], we have verified that the fault features are low rank distributed in the dynamical system matrix. As the multivariate extension of traditional DMD algorithm, multivariate fault features are also low rank distributed in the dynamical system tensor (DST).…”

mentioning

confidence: 56%

Multivariate Dynamic Mode Decomposition and Its Application to Bearing Fault Diagnosis

Zhang

Yuan

et al. 2023

IEEE Sensors J.

View full text Add to dashboard Cite

In practical engineering applications, the multivariate signal contains more fault feature information than the single channel signal. How to realize synchronous extraction of fault features from the multivariate signal is of great significance in fault diagnosis of rotary machinery. Dynamic mode decomposition (DMD) has attracted much attention due to its excellent dynamic feature extraction ability. However, DMD lacks mode aliasing property when dealing with the multivariate signal, which may lead to the loss of critical fault feature information. Cater to this problem, this paper proposed a multivariate dynamic mode decomposition (MDMD) algorithm which is the multivariate extension of DMD. First, snapshot tensors are defined to convert multivariate signals to tensor format. Then, MDMD algorithm is proposed by introducing tensor operations into the original DMD algorithm, where a tensor low tubal rank component extraction framework is constructed to enable simultaneous extraction of bearing fault features from the multivariate signal, to enhance the robustness and effectiveness of the algorithm. Finally, both numerical simulations and experiments verify the proposed MDMD has higher fault diagnosis accuracy than other multivariate signal processing methods.

show abstract

mentioning

confidence: 56%

Multivariate Dynamic Mode Decomposition and Its Application to Bearing Fault Diagnosis

Zhang

Yuan

et al. 2023

IEEE Sensors J.

View full text Add to dashboard Cite

show abstract

“…To validate the effectiveness of the presented algorithm, rubbing fault simulation signals y (t) were first constructed to simulate rubbing faults. The transient characteristics of the rubbing fault signals can be considered as periodic pulses submerged in noise and interference [30]. In this paper, y 1 (t) represents the feature frequency of rubbing failure and its integer multiple frequency components; y 2 (t) represents the j/2 (j = 1,2,4,5,7,8) fractional multiple frequency components of rubbing fault characteristic frequency; y 3 (t) represents the j/3 (j = 1,2,4,5,7,8) fractional multiple frequency components of rubbing fault characteristic frequency; and y 4 (t) represents the rotational frequency.…”

Section: Simulation Signals Of Rubbing Faultmentioning

confidence: 99%

A new method of noise reduction grounded on the Hankel matrix and its application in rubbing fault diagnosis

Zhang,

Yu,

Feng

et al. 2024

Meas. Sci. Technol.

View full text Add to dashboard Cite

In processing signals with singular value decomposition (SVD), one of the keys lies in building an appropriate Hankel matrix from signals. To address the difficulty in extracting the feature information of rubbing faults between rotor and stator, by taking advantage of the nature of rubbing fault information closely related to the rotation period of equipment, a new method of SVD is presented according to the Hankel matrix built from the periodicity of rotation machine. Firstly, with the periodicity of rub-impact fault as the basis, the interval step size between Hankel vectors was determined to self-adaptively build a Hankel matrix of signal. Secondly, the newly-built Hankel matrix was denoised through singular value differential spectrum (SVDS). Thirdly, to reduce the loss of data as much as possible, a strategy was proposed to rebuild signals according to the first and last rows of denoised signals. Fourthly, features of rubbing faults were extracted according to the frequency spectrum of reconstructed signals and faults were identified. To verify the applicability and effectiveness of presented algorithm, various types of simulation signals and tester signals from different states were brought in; meanwhile, the presented algorithm was compared with a variety of classical methods. The result proves that the proposed method not only can effectively restrain noise interference, but also highlight fault feature information and correctly identify rub-impact faults.

show abstract

“…In fact, low-rank representation has been widely used in PHM [32][33][34][35]; see Li [36] for a recent comprehensive review. However, it is found that LRSR-based transfer subspace learning has not been investigated in the field of fault diagnosis.…”

Section: Introductionmentioning

confidence: 99%

Robust transfer subspace learning based on low-rank and sparse representation for bearing fault diagnosis

Yu,

Xiu,

et al. 2024

Meas. Sci. Technol.

View full text Add to dashboard Cite

With the development of industrial intelligence, data-driven fault diagnosis plays an important role in Prognostics and Health Management (PHM). However, there is usually a large amount of unlabeled data from different working conditions, making cross-domain fault diagnosis unstable and inflexible. To deal with this issue, we propose two novel transfer subspace learning methods based on the Low-Rank Sparse Representation (LRSR), called LRSR-G and LRSR-R. Specifically, LRSR-G integrates an additional matrix with LRSR to characterize the Gaussian noise for robustness, as well as capture global and local structures. Furthermore, LRSR-R adaptively learns the label matrix from samples instead of using the binary labeling matrix in LRSR-G, thus providing the possibility to improve the flexibility. In addition, we develop two efficient algorithms using the Alternating Direction Method of Multipliers (ADMM) to solve the proposed LRSR-G and LRSR-R. Extensive experiments are conducted on the Case Western Reserve University (CWRU) dataset and Jiangnan University (JNU) dataset. The results show that the proposed LRSR-G and LRSR-R perform better than the existing methods, while LRSR-R has more potential in cross-domain fault diagnosis tasks.

show abstract

A local transient feature extraction method via periodic low rank dynamic mode decomposition for bearing incipient fault diagnosis

Cited by 14 publications

References 35 publications

Multivariate Dynamic Mode Decomposition and Its Application to Bearing Fault Diagnosis

Multivariate Dynamic Mode Decomposition and Its Application to Bearing Fault Diagnosis

A new method of noise reduction grounded on the Hankel matrix and its application in rubbing fault diagnosis

Robust transfer subspace learning based on low-rank and sparse representation for bearing fault diagnosis

Contact Info

Product

Resources

About