Skip to Main content Skip to Navigation

Du diagnostic au pronostic de pannes des entraînements électriques

Abstract : Faults diagnosis and prognosis of electrical drives play a key role in the reliability and safety of production tools especially in key sectors (military, aviation, aerospace and nuclear, etc.). The research presented in this thesis aims to introduce new methods for faults diagnosis and prognosis of an induction motors and roller bearings. These methods use measured data collected from sensors placed on the system (induction motor, roller) in order to construct a feature vector which indicates the state of the system. Supervised and unsupervised classification methods are developed to classify measurements (observations) described by the feature vector compared to known or unknown operating modes, with or without failures. Defects were created in the rotor and the bearing of the induction motor, fed by a voltage inverter. The unsupervised classification technique, based on artificial ant-clustering, allows analyzing the unknown and unexplored observations to highlight classes with similar observations. This allows improving the classification and the detection of new operating modes. The supervised classification, based on hidden Markov models, allows associating a degree of similarity when we affect an observation to one or more classes. This defines a reliability index which allows the detection of new operating modes. These methods are not limited to diagnose faults; they can also contribute to the prognosis of faults. Indeed, the prognosis can be defined as an extension of the problem of diagnosis. The prognosis of faults is carried out by three methods based on hidden Markov models for the detection of an impending failure and by two methods based on the neuro-fuzzy system (ANFIS for Adaptive Neuro fuzzy Inference System and the neo-fuzzy neuron) to estimate the remaining time before its appearance. A set of historical data collected on roller bearings is used to validate the proposed methods. The obtained results show the effectiveness of the proposed methods for faults diagnosis and prognosis of electrical drives
Document type :
Complete list of metadata

Cited literature [74 references]  Display  Hide  Download
Contributor : Abes Star :  Contact Connect in order to contact the contributor
Submitted on : Monday, January 18, 2016 - 5:27:05 PM
Last modification on : Monday, September 13, 2021 - 2:44:03 PM
Long-term archiving on: : Tuesday, April 19, 2016 - 12:20:38 PM


Version validated by the jury (STAR)


  • HAL Id : tel-01258196, version 1


Abdenour Soualhi. Du diagnostic au pronostic de pannes des entraînements électriques. Energie électrique. Université Claude Bernard - Lyon I, 2013. Français. ⟨NNT : 2013LYO10146⟩. ⟨tel-01258196⟩



Record views


Files downloads