Etude des signaux vibratoires d’une boite de vitesses d’éolienne

Abstract

The wind energy sector has grown considerably over the last few decades. Although this development is very encouraging, it is still marked by failures in wind turbine components. Bearings are critical and widely used components in various rotating machines such as wind turbines, and their failure is one of the main causes of breakdowns in rotating machines. For this reason, the main objective of monitoring is to minimize as far as possible the high costs incurred by unplanned downtime and equipment failure, which can lead to considerable losses. Intelligent diagnostic methods for bearings have become a key focus of research. These methods require the construction of a shape vector (SV), usually composed of indicators calculated from time-sampled vibration or velocity signals. However, under non-stationary conditions, these signals require the application of complex methods that take a very long time to calculate. For this reason, the use of angular resampling techniques is recommended, as they make it possible to avoid velocity fluctuations and use simple processing methods. In the operational phase of intelligent diagnosis. We used angularly resampled acceleration signals for intelligent diagnosis of bearings in non-stationary conditions. This involves constructing several VFs. Next, selection algorithms are used to select the most relevant indicators and minimize computation time. Finally, classification is performed using a Cubic Support Vector Machine (Cubic SVM) for the detection and identification stages of the various bearing failure conditions. The proposed method proves its effectiveness for diagnosing bearing faults under variable speed conditions. In addition, this approach will be tested on more complex systems operating under non-stationary conditions, such as wind turbine gearboxes, and the method could also be used to build an on-line diagnostic system