Diagnostic Et Classification Des Défauts De Roulement Dans Une Eolienne, En Utilisant L’analyse Vibratoire Et Les Réseaux De Neurones Artificiels.

No Thumbnail Available
Date
2021
Journal Title
Journal ISSN
Volume Title
Publisher
Université Oum El Bouaghi
Abstract
In the industrial environment, production systems are increasingly complex and cannot be free of disturbances and failures, affecting the quality of the product, which can cause the immediate shutdown of a machine and affect the proper functioning of an entire production system. The diagnosis of defects in these machines is essentially based on the monitoring of symptoms related to different degradation conditions. These symptoms can be drawn and extracted from various sources of information, among which, the vibratory analysis occupies a dominating place. This study aims to diagnose, detect, and classify bearing faults from actual signals measured by an accelerometer placed on a bearing of an industrial wind turbine for power generation, where a fault at the inner race of the bearing has been initiated, and the measurements are performed during 50 days of operation, using two methods. The first method is multiresolution wavelet analysis (MWA) to detect and locate bearing defects. The second method used is a classification method called the self-organizing map (SOM). The results show the effectiveness of the MWAR method in detecting and locating defects. In addition, the neural classifier (SOM) provided relevant information about the fault evolution and bearing degradation, as it could automatically cluster the vibration signal into groups corresponding to the life stages of the bearings. Thus, these results can effectively contribute to timely maintenance decisions.
Description
Keywords
cartes auto-organisatrices, Indicateurs scalaires, éolienne
Citation