The aim of this article is to show the interest the wavelet transforms in improving the sensitivity of the scalar indicators (kurtosis and crest factor) in the context of conditional maintenance by vibratory analysis of rotating machinery. When the bearing is damaged, the appearance of a crack on the gear tooth disturbs signal. This change is due to the presence of periodic pulses. Nevertheless, the presence of noise induced by the random excitation can have an inuence on the values of these time indicators. Denoising of these signals by wavelet transform allows improve the sensitivity of these indicators and to increase the reliability of diagnosis. To simulate the signal to be analyzed, we voluntarily introduced a default. We selected wavelet Daubchies type that is well suited to this type of problem. The purpose is to try to master the various parameters related to the wavelet analysis for reliable and inexpensive detection, namely, the order of the wavelet and the level of decomposition. The approach is to observe the kurtosis and the crest factor for several wavelet orders depending on the gravity of the fault.