A NOE Identification Algorithm with BIBO Stability Property

LILIANA RAQUEL CASTRO; JOSE LUIS FIGUEROA; OSVALDO ENRIQUE AGAMENNONI

Rio Cuarto

Workshop; Reunión de trabajo en Procesamiento de la Información y Control; 2005

In this paper we present a nonlinear output error (NOE)model structure that allows the implementation of an identification algorithm in which the degrees of freedom of the model can be easily increased or decreased during the identification process. This attractive feature is achieved by using High Level Canonical Piecewise Linear (HLCPWL) functions with an increasing grid division. The algorithm then starts using a linear estimation of the model and the parameters of the HLCPWL functions are updated using a simple algorithm based on an improved steepest descent method with an adaptive learning rate. Furthermore, an algorithm to evaluate straightforwardly the parameters of the HLCPWL approximation when going from a coarse division of the simplicial partition of the domain to a finer one, is given. Besides, a simple condition to update the parameters preserving BIBO stability is also derived and included in the algorithm. In order to show the behavior of the proposed algorithm, two examples are presented.