Les travaux de recherche présentés dans cette thèse concernent les petites machines à aimants permanents, à flux axial et à auto-commutation électronique ayant la topologie d’un rotor intérieur discoïdal avec des aimants Nd-Fe-B montés en surface et de deux stators extérieurs identiques, chacun avec enroulement triphasé distribué dans des encoches. Après l’examen des topologies candidates pour applications d’entraînement direct basse-vitesse, une modélisation électromagnétique analytique de pré-dimensionnement d’un prototype de telles machines est réalisée. Ensuite, une approche numérique originale est développée et couplée à l’optimisation géométrique des aimants rotoriques en vue de réduire les harmoniques d’espace de l’induction magnétique dans l’entrefer et aussi les ondulations du couple électromagnétique. Des nombreux tests expérimentaux sont effectués sur le prototype de machine pour vérifier son dimensionnement ainsi que pour valider la stratégie d’auto-commutation électronique et de contrôle de base / The research work presented in this thesis is concerned with small electronically-commutated axial-flux permanent-magnet (AFPM) machines having the double-sided topology of an inner rotor with surface-mounted Nd-Fe-B magnets, sandwiched between two outer slotted stators with distributed three-phase windings. After reviewing the small double-sided AFPM machine candidate topologies for low-speed direct-drive applications, the thesis hinges on the size equations and the analytical electromagnetic design of the inner-rotor AFPM (AFIPM) machine topology under study. Original methods of modelling and design optimization of a small prototype AFIPM machine are then proposed with the view to reducing the airgap flux density space-harmonics and the torque ripple by rotor-PM shape modification. Extensive experimental tests are carried out on the small three-phase AFIPM machine prototype in order to validate its proper design and to check its electronic commutation and basic control technique

The dissertation thesis deal with the design and the optimization of the permanent magnet synchronous machine (SMPM) based on the artificial intelligence. The main target is to apply potential optimization methods on the design procedure of the machine and evaluate the effectiveness of optimization and the optimization usefulness. In general, the optimization of the material properties (NdFeB or SmCo), the efficiency maximization with given nominal input parameters, the cogging torque elimination are proposed. Moreover, the magnet shape optimization, shape of the air gap and the shape of slots were also performed. The well known Genetic algorithm and Self-Organizing migrating algorithm produced in Czech were presented and applied on the particular optimization issues. The basic principles (iterations) and definitions (penalty function and cost function) of proposed algorithms are demonstrated on the examples. The results of the vibration generator optimization (VG) with given power 7mW (0.1g acceleration) and the results of the SMPM 1,1kW (6 krpm) optimization are practically evaluated in the collaboration with industry. Proposed methods are useful for the optimization of PM machines and they are further theoretically applied on the low speed machine (10 krpm) optimization and high speed machine (120 krpm) optimization.