Transactions of FAMENA, Vol. 50 No. 3, 2026.
Izvorni znanstveni članak
https://doi.org/10.21278/TOF.503082125
A Cross-Domain Fault Feature Extraction Method Based on Multi-Statistic Multi-Scale Permutation Entropy
Longqiu Shao
; Faculty of Data Science, City University of Macau, Macau 999078, China; School of Automation, Guangdong University of Petrochemical Technology, Maoming 525000, China
*
Shengyi Jiang
; School of Information Science and Technology, Guangdong University of Foreign Studies, Guangzhou 510000, China
Jianbin Xiong
; School of Automation, Guangdong Polytechnic Normal University, Guangzhou 510000, China
Aisong Qin
; School of Automation, Guangdong University of Petrochemical Technology, Maoming 525000, China
Qin Hu
; School of Automation, Guangdong University of Petrochemical Technology, Maoming 525000, China
* Dopisni autor.
Sažetak
Addressing the insufficient adaptability of existing fault features in the cross-domain fault diagnosis of key transmission components of rotating machinery, this paper proposes an improved method based on permutation entropy theory. By introducing diverse statistics and different time scales, the original vibration signal is subjected to coarse-graining processing to construct a multi-statistic and multi-scale permutation entropy method, which is used to quantify the dynamic characteristics of vibration signals under variable operating conditions. This method overcomes the limitations of traditional permutation entropy, realizes feature extraction with both class discrimination and domain invariant characteristics, thus improving the efficiency of the cross-domain fault diagnosis. The verification of the measured dataset shows that compared with multi-scale permutation entropy, the proposed features can effectively weaken the interference of working condition fluctuations, accurately identify fault patterns under various operating conditions, have stronger migration and adaptation capabilities, break through the bottleneck of high-quality fault feature extraction in cross-domain diagnosis, and provide a new method for cross-domain fault diagnosis feature learning.
Ključne riječi
multi-statistic multi-scale permutation entropy; cross-domain fault diagnosis; fault feature extraction; transfer learning; rotating machinery
Hrčak ID:
351043
URI
Datum izdavanja:
16.6.2026.
Posjeta: 0 *