Détection des défauts de roulement par une méthode de reconnaissance de forme

Abstract

Bearings are among the most stressed components of rotating machinery and are a frequent source of failure. The faults encountered can be of several origins: chipping, seizing, corrosion (which leads to chipping), false Brinell effect, etc. This thesis aims to combine the vibration analysis with a pattern recognition method in order on the one hand to improve the detection of bearing defects and on the other part of better monitoring the evolution of the degradation of bearing defects over time. To do this, we have proposed an unsupervised density-based pattern recognition method, OPTICS (Ordering Points To Identify Clustering Structure), for the dynamic detection of the appearance of a defect on the inner and / or outer race of the bearing.