Modelisation et optimisation d'une machine a reluctance variable par algorithmes intelligents

Abstract

The theme dealt with in this thesis focuses on the modeling of a variable reluctance motor (VRM) adapted to the process of optimizing its performance we have developed and tested. Knowing that the performance of VRM depends as well on the geometric structure, the nonlinear characteristics of the materials used and the converter control parameters, we therefore paid particular attention to the associated mathematical model, based on an analytical approach taking into account the peculiarity of the VRM run in saturated state. Despite the accuracy of numerical approaches such as finite element method, researchers are moving increasingly towards hybrid methods combining their analytical models as better suited to an optimization process; this is the approach we have adopted here. Thus we have oriented our work towards finding hybrid models (analytical- numerical) sufficiently accurate while being rapid execution. The validation of these models was made by comparing the results obtained and the numerical simulations by finite elements in Flux- 2D. Validation and exploitation of results of different simulations allowed the continuation of work to optimize the geometric parameters for stator and rotor teeth of the studied prototype. After state of the art different algorithms of optimizations present in the literature, our choice fell on Genetic Algorithms and Swarm particles. The various simulations in MATLAB environment allowed get optimized dental angles and significantly improve the average couple.