Identificação por predição de erro e síntese de controladores robustos

Abstract

This work investigates the estimation of models and uncertainty sets for control objectives. Firstly, one presents relative concepts to the prediction error identification, closed loop identification and uncertainty sets $Di$. Later, it is carried through a study on factors that influence the quality of the region $Di$ with the objective of robust control design. It is verified that, the power spectrum of the signal used during the stage of record data, influences the size of the region $Di$, could be used as project parameter for identification and robust control. Thus, the main contribution of this work is to proporse an iterative procedure for identification and robust control. This procedure obtained a region $Di$, tuned for robust control design with certain probability level to stabilize and to achieve the prescribed closed loop performance with the real life system. Two key points of the purpose procedure are: (1) integration of robust analysis tools, in the synthesis of controllers using Genetic Algorithms and (2) the use of worst case performance information to redesign the identification experiment, case the estimate region $Di$ is classified as it invalidates. The procedure was applied in two benchmarks generally used for the community of identification for control. The results suggest that, experimental conditions can be chosen in function of the performance specifications when the objective is to estimate regions $Di$ tuned for robust control.