ISSN: 1692-7257 Volumen 2 Número 6 año - 2005 Universidad de Pamplona I. I. D. T. A. 34 Revista Colombiana de Tecnologías de Avanzada AN EVOLUTIONARY ALGORITHM FOR LINEAR SYSTEMS IDENTIFICATION UN ALGORITMO EVOLUTIVO PARA IDENTIFICACIÓN DE SISTEMAS LINEALES Ph.D. Edmary Altamiranda, Ing. Rodrigo Calderón Ph.D. Eliezer Colina Universidad de Pamplona Instituto de Investigación y Desarrollode Tecnologías Aplicadas IIDTA {edmarya, rodcalderon,ecolina}@unipamplona.edu.co Pamplona, Norte de Santander - Colombia Abstract: Abstract: This paper presents a systems identification method, for discrete time linear systems, based on an evolutionary approach, which allows achieving the selection of a suitable structure and the parameters estimation, using non conventional objective functions. This algorithm incorporates parametric crossover and parametric mutation along a weighted gradient direction (Tang and Wang, 1997). The performance of the proposed method is illustrated with computer simulations using ARX model structures, where parameters, model dynamical order and input-output delay values are estimated Resumen: En este trabajo se presenta un algoritmo para identificación lineal de sistemas en tiempo discreto, basado en un enfoque evolutivo, el cual permite llevar a cabo la selección de una estructura apropiada basada en modelos tipo ARX y de la estimación de los parámetros, orden dinámico y retardo entrada/salida del sistema, utilizando funciones objetivos no convencionales. Este algoritmo incorpora operadores de cruce paramétrico y mutación paramétrica utilizando ascenso por gradiente (Tang and Wang, 1997). La eficiencia del algoritmo evolutivo propuesto es ilustrada a través de simulaciones computacionales.. Keywords: Systems identification, Linear systems, Evolutionary algorithms, Parameter Estimation. 1. INTRODUCTION The identification of a dynamic process from a set of possibly noisy input-output data has been a classical problem in control engineering. System parametric identification is usually achieved in two steps: the first step is selecting a model family from which a candidate model is produced by minimizing some error criterion. The second step is validating the identified candidate model in terms of verifying some performance indicators, such as noise independence, error auto-correlation, input/error correlation and real response following among others. If results do not satisfy some of the