Faleh, AbdallahBoukadoum, Ahcene2024-03-212024-03-212023-03-20http://dspace.univ-skikda.dz:4000/handle/123456789/502three-phase induction motors are the mainly used rotating machine in diverse sectors (industries, military, aerospace, aviation, human housing), due to their advantages such as robustness, and lower maintenance costs. However, in different environments of working, these machines expose to various internal stress such as electrical, mechanical, external high temperature, humidity or both. Prognostic and health management plays a crucial role in the safety, reliability and continuity of production of these machines. The research presented in this thesis aims to introduce new data-driven methods for fault diagnostic of three-phase induction motors fed by inverter and prognostic of roller element bearing degradation. These methods are based on analysing the electrical measurement from the sensors to define the health state of the system. Where, a new health indicator propose based on combined temporal features extracted from electrical signals (current and voltage), which use as input to the K-Nearest Neighbour to diagnose and classify different health states of three-phase induction motors including bearing wear and different broken rotor bars. In case of prognostic of remaining useful life of bearing degradation, first of all, a new proposed health monitoring extract from the few first historic vibration signal, that determine the point between the health state and degraded state in order to start the prediction phase. Secondly, after detecting this point, begin the phase of feature extraction and selection of the best monotone features depending on the monotonicity criteria, and reduce the selected features by Principal Component Analysis. Finally, the fused feature is used as input to Support Vector Regression to predict the remaining useful life. Where, the obtained results are attractive compared to the real remaining useful life.enThesis