Predictive maintenance of machinery is becoming more and more important every day, and within maintenance, reliability analysis of equipment operation. During the performance of vibration analysis, noise, present in most of the industrial situations, in which predictive maintenance plays a fundamental role, is one of the main problems that have to be faced while analyzing the vibration levels of machinery. The main purpose of this study is to analyze and identify the different regions in which the main failures that can originate in the bearings, which are located in all industrial machinery. The objective of this work is to identify and isolate from noise, the harmonics coming from defects in bearings. These results enable more accurate vibration analysis to be carried out, optimizing predictive maintenance processes. To carry out this study, a vibration analysis test bench is used in which bearings with different failures are tested and with this it is possible to associate the problem of each bearing not with specific vibration levels but with the probability bands in which are the peaks obtained in their characteristic vibration spectra.