Problem statement: Neural networks and fuzzy inference systems are becoming wellrecognized tools of designing an identifier/controller capable of perceiving the operating environment and imitating a human operator with high performance. Also, by combining these two features, more versatile and robust models, called "neuro-fuzzy" architectures have been developed. The mo Approach: Motivation behind the use of neuro-fuzzy approaches was based on the complexity of real life systems, ambiguities on sensory information or time-varying nature of the system under investigation. In this way, the present contribution concerns the application of neuro-fuzzy approach in order to perform the responses of the speed regulation, ensure more robustness of the overall system and to reduce the chattering phenomenon introduced by sliding mode control which is very harmful to the actuators in our case and may excite the unmodeled dynamics of the system. Results: In fact, the aim of such a research consists first in simplifying the control of the motor by decoupling between two principles variables which provoque the torque in the motor by using the feedback linearization method. Then, using sliding mode controllers to give our process more robustness towards the variation of different parameters of the motor. However, the latter technique of control called sliding mode control caused an indesirable phenomenon which harmful and could leads to the deterioration of the inverter's components called "chattering". So, here the authors propose to use neuro-fuzzy systems to reduce this phenomenon and perform the performances of the adopted control process. The type of the neuro-fuzzy system used here is called": Adaptive Neuro Fuzzy Inference System (ANFIS)". This neuro-fuzzy is destined to replace the speed fuzzy sliding mode controller after its training process. Conclusion: Therefore, from a control design consideration, the adopted neuro-fuzzy system has opened up a new direction that allows for the design of robust controllers for uncertain non-linear dynamical systems without resorting to system model simplifications and linearization and without imposing structural conditions on system uncertainties. On the other hand, it is important to say that this approach permits to improve the performance of the controlled system only by choosing the appropriate form of the membership functions (shape, triangular…) and a good partionnement of the universe of discourse of the diverse variables. Finally the obtained simulation results prove that the objectives of the authors where attempt by a significant reduction of the chattering and a good robustness of the process towards parameter variation and external perturbation (load torque).

The principal objective of this work consists in reducing the chattering phenomenon and ensuring more improvement of performances the proposed control scheme especially ensuring the decoupling between the control of the two principle variables of the motor "speed and flux" and also the system's robustness towards parameter's variations and external perturbations. In fact and as a first step a control scheme based on the combination of feedback linearization control is first proposed. Then, and in order to ensure more robustness of the process towards parameter's variations and external perturbation...ect, sliding mode controllers are used. These controllers are based on a piecewise function for smoothing in order to reduce the effect of the chattering phenomenon which is harmful of the actuator. However, this smooth function is relied closely to the upper bound of uncertainties, which include parameter variations and external disturbances. So, fuzzy sliding mode controllers are investigated to solve this difficulty based on the method introduced by Ben-Galia et al. By replacing the gain and the signum of the attractivity function of sliding mode controllers by fuzzy maps. However, it seems that this technique couldn't reduced the chattering considerably which let the authors to propose another type of controllers based on Gaussian Radial Basis Function Neural Network (GRBFNN) which revealed some very interesting features and prove that such controllers are efficients to attempt the required objectives.

No abstract

No abstract