Skip to Main content Skip to Navigation
Theses

Diagnostic par reconnaissance des formes : application à un ensemble convertisseur - machine asynchrone

Abstract : Advances in power electronics, control circuits and automatic have contributed to an increasing use of induction motors in electrical drive systems. The large – scale utilization of induction motors is mainly due to their robustness, their power – weight ratio, and to their manufacturing cost. The appearance of variators making it possible to vary the rotational frequency largely supported its development. Indeed, these variators enter the design of many industrial processes associating static inverters and electric machines. The maintenance and the monitoring of these two systems allow making profitable the installations. Therefore, it is important to develop diagnosis tools in order to detect earlier the faults, which can appear in these machines.
Our approach is based on pattern recognition methods. A vector of features, named pattern vector, is obtained from the measurements made on the machine. These classifications are made according to the different operating conditions, with and without fault.
Faults have been created on both the rotor and the stator sides of the induction machine. This one was fed either from the mains, or from a three – phase voltage inverter.
Fault detection has been made with decision procedure based on the k – nearest neighbors rule, associated with a membership function. This procedure allows detecting the evolution of the operating modes as well as the proven defects. Thereafter, the evolution tracking of these modes is carried out by a Kalman approach: a recursive Kalman estimator is used to determine the parameters of the dynamic model accounting for the evolution of a mode and a Kalman predictor to envisage an evolution towards new zones of space.
The results obtained with these algorithms have proved the efficiency of pattern recognition methods for diagnosis.
Complete list of metadatas

Cited literature [126 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00113102
Contributor : Publications Ampère <>
Submitted on : Friday, November 10, 2006 - 3:48:47 PM
Last modification on : Friday, October 23, 2020 - 4:59:55 PM
Long-term archiving on: : Tuesday, April 6, 2010 - 10:12:14 PM

Identifiers

  • HAL Id : tel-00113102, version 1

Citation

Olivier Ondel. Diagnostic par reconnaissance des formes : application à un ensemble convertisseur - machine asynchrone. Autre. Ecole Centrale de Lyon, 2006. Français. ⟨tel-00113102⟩

Share

Metrics

Record views

1310

Files downloads

7419