Fault Feature Enhancement and Diagnosis Method for UAV Rotor Bearings Based on Amplitude-Modulated Graph Fourier Spectrum

Guangshuai Yang; Meng Li; Daigeng Jiao; Ling Shi; Kun Zhang; Yonggang Xu

doi:10.37965/jdmd.2026.1382

Authors

Guangshuai Yang School of Mechanical and Energy Engineering, Beijing University of Technology, Beijing 100124, China
Meng Li School of Mechanical and Energy Engineering, Beijing University of Technology, Beijing 100124, China https://orcid.org/0009-0003-9269-3034
Daigeng Jiao School of Mechanical and Energy Engineering, Beijing University of Technology, Beijing 100124, China
Ling Shi School of Mechanical and Energy Engineering, Beijing University of Technology, Beijing 100124, China https://orcid.org/0000-0002-3859-1112
Kun Zhang School of Mechanical and Energy Engineering, Beijing University of Technology, Beijing 100124, China https://orcid.org/0000-0002-3513-8988
Yonggang Xu School of Mechanical and Energy Engineering, Beijing University of Technology, Beijing 100124, China

DOI:

https://doi.org/10.37965/jdmd.2026.1382

Keywords:

fault diagnosis; graph signal; rotating bearing; spectral modulation spectrum

Abstract

With the increasing integration and diversification of functions in rotating machinery systems, their reliability faces continuous challenges, leading to a corresponding rise in failure rates. Under long-term operating conditions, motor bearings, as key components of the power system, have become one of the most prone to failure due to the large and continuous loads they bear. To address the a forementioned issues, this paper proposes a bearing fault diagnosis method based on an amplitude modulation spectrum model. First, an amplitude modulation framework using the short-time Fourier transform (STFT) is constructed, and the linear kurtosis of the envelope spectrum (LSES) indicator is introduced to quantitatively evaluate fault feature information under different modulation structures. Subsequently, the signal corresponding to the maximum LSES value is selected to construct a graph signal model. Finally, the last K-order spectral components in the graph frequency domain are retained for feature extraction and enhancement. Validation using both simulated signals and experimental signals of bearing inner and outer race faults demonstrates that the proposed method has good feasibility and effectiveness.