• CN: 11-2187/TH
  • ISSN: 0577-6686

Journal of Mechanical Engineering ›› 2026, Vol. 62 ›› Issue (16): 406-418.doi: 10.3901/JME.260727

Previous Articles    

Bearing Fault Diagnosis Method Based on Multi-source Weighted Spectrum Reconstruction Model

HU Zhongshuo1,2,3, YANG Jianwei1,3, YAO Dechen1,3, WANG Jinhai1,3, HUO Jiyuan1,3, SUN Runtao1,2,3, REN Yuteng1,3   

  1. 1. School of Mechanical-electronic and Vehicle Engineering, Beijing University of Civil Engineering and Architecture, Beijing 100044;
    2. School of Mechanical, Electronic and Control Engineering, Beijing Jiaotong University, Beijing 100044;
    3. Beijing Key Laboratory of Performance Guarantee on Urban Rail Transit Vehicles, Beijing University of Civil Engineering and Architecture, Beijing 100044
  • Received:2025-08-14 Revised:2026-03-03 Published:2026-10-14

Abstract: Rolling bearings are critical components of urban rail trains, and their health monitoring holds significant engineering importance for operational safety. However, signals from a single measurement point are susceptible to noise interference such as wheel-rail impacts and vibrations from system components. Therefore, this paper proposes a multi-source weighted spectrum reconstruction(MWSR) model. First, the sub-band energy distribution of the weighted entropy spectrum from multi-point measurement information is calculated and analyzed to dynamically adjust the sub-band boundaries, making them approximate the energy distribution of a preset target. Second, a frequency-band coupling relationship between multiple measurement points is constructed based on an ARMA model, iteratively refining the sub-band boundaries for each measurement point’s information. Finally, a filter bank based on the optimized sub-band boundaries is designed to fuse and reconstruct signals from multiple measurement points, thereby enhancing bearing fault features. Fault diagnosis is achieved through envelope spectrum analysis. Compared to traditional signal decomposition methods, the proposed method maintains the highest fault characteristic energy ratio(FCER) across varying signal-to-noise ratios(SNRs), resulting in more identifiable fault characteristics that provide highly distinguishable features for fault diagnosis.

Key words: multi-source information fusion, fault diagnosis, autoregressive moving average model, weighted spectral entropy

CLC Number: