To detect a bearing fault, it is useful to use temporal and spectral analysis. These enable fault detection, but do not determine the nature and severity of the fault. This requires a priori knowledge of the fault characteristics and kinematic parameters of the machine. To overcome this problem, an unsupervised OPTICS classification method has been proposed for diagnosing real and simulated faults. However, dynamic classification methods and even signal processing are influenced by several parameters, such as pre-processing methods to reduce or cancel noise. For this reason, our work is based on a comparative study of two denoising methods: wavelet filtering and Fourier transform filtering (FFT).