| Abstract |
Our paper deals with the synthesis and the comparison of two robust controllers since embedded applications require fixed parameters. In order to quantify robustness in relation to parameter variations and actuator and measure disturbances, we only consider the most restrictive robustness indicators: the module and delay margins. In a similar way, noise rejection is quantified by sensitivity functions, which represent the influence of noise on the input or the output. The first method is dedicated to H infinity optimization of controllers. The main part of our work deals with the designing of the weighting functions using sensitivity functions (Fig. 1), which ensures the convergence of the optimization process and the realization of the different margins. The controller is simplified by the balanced model reduction and sampled by the Tustin transformation, which keeps the H infinity norm. The second method deals with polynomial RST corrector, which is synthesized by robust pole assignment using a stochastic genetic algorithm and Fuzzy logic rules (Fig. 2). It takes into account constraints of robustness while satisfying the requirements of performances. Our new approach deals with the synthesis of a robust corrector, which satisfies the same set of conditions on output disturbance rejection and stability robustness on errors of modeling. The two methods are compared with the same schedule, which define the noise rejection and robustness objectives.
They are applied in order to synthesize the current correctors on a vector field oriented control of an induction machine. |
