Otimização de controladores fuzzy por algoritmos genéticos multiobjetivos no domínio Wavelet

Abstract

Due to the increasing competitiveness in the industry, it is imperative to use more efficient tuning techniques that can in fact find controllers with the desired performance. With this proposal, optimization techniques can be used to obtain the controller parameters according to an evaluation criterion, which should encode how good a particular controller is, properly expressing the desired specifications, so that the utilized algorithm can find the desired controller. The methods traditionally used in tuning present a difficulty in expressing the desired specifications. The difficulty is due to the fact that the traditionally used criteria, in general, only use the total error information, through indices such as the Integral Absolute Error (IAE) or the Integral Square Error (ISE), which do not describe aspects of system behavior, such as if the response is very aggressive and oscillatory, steady state error, rise time and stabilization time, as a human designer would do. Some of these impressions are not well defined for references other than the step, lacking generality. Thus the optimization algorithm responsible for obtaining the controller parameters according to an evaluation function, which must actually be able to encode how good a given controller is, adequately expressing the desired specifications, so that the optimization algorithm employed can find the controller that best satisfies such a function. In view of this, a generic methodology for using wavelet analysis will be presented along with multiobjective optimization techniques to express more closely and closely related to the human behavior of the controlled system, allowing a more accurate optimization. In the proposed methodology, wavelet analysis, very present in the literature, focused on other applications, especially in the analysis of signals, sounds and images, is used to obtain descriptors that describe aspects of system behavior, such as its steady state behavior, behavior In the transient, no amplification of noise and rejection of disturbances, these descriptors become objectives that will be optimized by multiobjective techniques. The study carried out used Multiobjective Genetic Algorithm (MOGAs) techniques for optimization, due to their being widely used in the literature and known for their simplicity and efficiency.