Computación y Sistemas
versão impressa ISSN 1405-5546
FLORES-ANDO, Fortunato; AGUILAR-IBANEZ, Carlos; SUAREZ-CASTANON, Miguel S e CUEVAS DE LA ROSA, Francisco. An Identification Genetic Algorithm for a Family of Duffing's System. Comp. y Sist. [online]. 2003, vol.7, n.2, pp. 102-112. ISSN 1405-5546.
This paper shows a simple way to recover the whole unknown parameters set of the Duffing's oscillator by using a genetic algorithm. The fact that the system is observable and constructible with respect to a suitable output helps in obtaining an integral parameterization of the output. Subsequently an integral parameterization of the output which depends upon the unknown parameters, and, a random estimation of the output is proposed, assuming that the set of unknown parameters are contained into a bounded set. This random estimation is chosen provided that the error between the actual output and the estimated output minimizes the errors of a quadratic function. The minimization problem and the random estimations of the output are formulated directly in terms of a genetic algorithm. A population of chromosomes is codified with the parameters of the Duffing's oscillator system. A fitness function is established to evaluate the chromosomes, in such a way that it minimizes the errors of a quadratic function. The chromosomes' population evolves till a fitness average threshold is obtained. This method is numerically possible and easy to implement in a digital computer.
Palavras-chave : Mechanical Oscillator; Chaos; Genetic Algorithms; Reconstruction.