Etude de l'effet des méthodes de débruitage sur le diagnostic des défauts mécaniques

Abstract

For the detection of a bearing defect, it is useful to use temporal and spectral analysis. These last allow the detection of the defect but do not determine the nature and the gravity of this defect. This requires a priori knowledge of the characteristic’s faults and the kinematic parameters of the machine. To remedy this problem, an unsupervised classification method OPTICS has been proposed for the diagnosis of real and simulated faults. Nevertheless, dynamic classification methods and even signal processing methods are influenced by several parameters such as the preprocessing methods used to reduce or cancel the noise. As a result, our work is based on a comparative study between two denoising methods, the wavelet method and filtering by the Fourier transform FFT.