Automatic Characteristic Frequency Association and All-Sideband Demodulation for Detection of a Bearing Fault of a Test Rig
Résumé
This paper proposes advanced signal-processing techniques to improve condition monitoring of operating machines. The proposed methods use the results of a blind spectrum interpretation that includes harmonic and sideband series detection. The rst contribution of this study is an algorithm for automatic association of harmonic and sideband series to characteristic fault frequencies according to a kinematic conguration. The approach proposed has the advantage of taking into account a possible slip of the rolling-element bearings. In the second part, we propose a full-band demodulation process from all sidebands that are relevant to the spectral estimation. To do so, a multi-rate ltering process in an iterative schema provides satisfying precision and stability over the targeted demodulation band, even for unsymmetrical and extremely narrow bands. After synchronous averaging, the ltered signal is demodulated for calculation of the amplitude and frequency modulation functions, and then any features that indicate faults. Finally, the proposed algorithms are validated on vibration signals measured on a test rig that was designed as part of the Eu-ropean Innovation Project KAStrion'. This rig simulates a wind turbine drive * Corresponding author Email address: marcin.firla@gipsa-lab.grenoble-inp.fr (Marcin Firla) Preprint submitted to Mechanical Systems and Signal Processing March 11, 2016 train at a smaller scale. The data show the robustness of the method for localizing and extracting a fault on the main bearing. The evolution of the proposed features is a good indicator of the fault severity.
Origine : Fichiers produits par l'(les) auteur(s)
Loading...