Algoritmos genéticos aplicados na escolha da taxa de amostragem em identificação de sistemas

Abstract

The present work has as the main goal to introduce a new method to select the sample time of input and output signals used in the identification process using NARMAX representation. To achieve this goal is proposed a genetic algorithm wich uses a supersampled signal, i.e., a signal sampled in the most high frequency available, and later decimation rates are used to create different individuals from the high frequency sample signal. The individuals evaluation uses a system identification with NARMAX representation. The evaluation of the proposed method used a genetic algorithm developed in the software Matlab®. The proposed method was applied in the process identification of a polimeric membrane fuel cell temperature model and the results are presented.