• My research activities concern fuzzy logic, neural networks, particle swarm and genetic algorithm applied to the identification and control of induction machines.
  The paper referenced [C4] presents a study of the replacement of a fuzzy controller by a neural network one within a vector control scheme of an induction motor. The learnig procedure is also discussed.

  After the developpement of a machine model that takes into account rotor structure [C2], the operating and control aspects of faulty induction motors are under study [J1] [C11].
  
  Lastly, sensorless control is also a research axis that we explore. Our aim is to build a vector control which is simple as regards the expermental implementation aspects and computer time consumption. It also has to be robust and thus relies on the knowledge of minimal parameters for powerful control. The paper [C7] shows this method and the early results. This approach has been used successfully to implement a hybrid sensorless vector control (direct and indirect) for low speed and standstill operation [D3].
  
  Since December 2000, a new approch has been taken, which relies on the injection of a high frequency signal added to the fondamental signal of the feeding voltage (in a vector control scheme). The aim is to extract an image of the speed and position of the rotor. This is the subject of the Phd theis of Imad Al-Rouh.
  
  We also work on parametres identification et optimisation processes using genetic algorithms and PSO (Particle Swarm Optimisation).
  
  

• Co-supervision of the Phd thesis of Imad Al-Rouh, carried out at GREEN - UHP, December 2000 to Jully 2004: "Commande sans capteur de la machine asynchrone".
• Co-supervision of the DEA training session carried out at GREEN - UHP, May 2002 to September 2002 : "Modélisation et Commande de la génératrice asynchrone", DEA PROTEE of Mr A. CHETOUANE.
• Co-supervision of the DEA training session carried out at GREEN - UHP, May 1997 to September 1997 : "Etude et modélisation de machines asynchrones présentant des défauts de structure", DEA PROTEE of Miss A. ABED.
• Research contract GREEN - Jeumont Industrie uppon the developpement of a vector conttol of an induction motor with a onsite assistance (3 days).
• DEA training session at GREEN-UHP, April to Juilly 1995, "Commande vectorielle d'une machine asynchrone à enroulements de mesure de f.e.m.", including theory, simulation and experiments.

There are many possible control strategies with differents consequences (dynamical behaviour, robustness...). Besides, while some give good simulation resultats, they cannot be implanted. With such problems, the machine models must therefore be refined and new research areas must be investigated The studies carried out must be validated only with a test bench. This has led us to develop it.
It is the analysis of the simulation and experimental results that allowed us to tune our control algorithms and bring new ones (based on fuzzy logic and neural networks). These works have been published in international conferences and journals.

Résumé du mémoire de Thèse [J1]
Genetic algorithms, fuzzy logic and neural networks are increasingly used in greatly varied applications. We propose to study them for the identification and the control of the induction motor. More particularly, we use the genetic algorithms in order to identify the parameters of the transient model of the machine. Controllers, based on fuzzy logic and neural networks, are implemented within a rotor-flux oriented control scheme.

The influence of variation of the parameters on the operation of the system is also investigated. Faulty induction machines are considered in the case of rotor broken bars. The diagnosis of these defects and their influence on the behaviour of the control are studied.

Lastly, an improvement of the mechanical-sensorless vector control is presented opening a new way towards "intelligent control". Thanks to the ever increasing power of processors, such techniques can be used to replace the sensor.

A simulation tool elaborated along with the thesis made it possible to develop the studies presented here. Particular care was devoted to the experimental side, the only guarantee of the feasibility and validation of the study.

Keywords : Induction machine, Vector control, Fuzzy logic, Neural networks, Genetic algorithms, Identification, Fault diagnosis, Sensorless control.

Liste des publications Back to homepage