Pronostic de défaillances basé l'nalyse vibratoire pour la maintenance des systémes mécaniques tournants

Abstract

Failure prognosis is considered like a key process in modern predictive maintenance strategies. However, in practice, prognostic tools still require developments in terms of stability and robustness and generalizability. The objective of this end-of-study project is to propose a prognostic tool guided by vibration data to predict the remaining useful life of bearings. The proposed approach is based on vibration analysis and artificial neural networks. The proposed approach is experimentally validated on the NASA database: Bearing data set.