L.R. Castro, J.L. Figueroa, and O.E. Agamennoni (Argentina)
Nonlinear systems, identification, piecewise linear techniques.
In this paper we propose a nonlinear output error (NOE) model structure for black-box identification of nonlinear dynamic systems. This model structure allows the im plementation of an identification algorithm in which the degrees of freedom of the model can be easily increased or decreased during the identification process. This is done using High Level Canonical Piecewise Linear (HL CPWL) functions with an increasing (decreasing) grid di vision. Therefore, the algorithm can start using a linear estimation of the model. The parameters of the HL CPWL functions are updated using a simple algorithm based on an improved steepest descent method with an adaptive learn ing rate. Furthermore, an algorithm to evaluate straight forwardly the parameters of the HL CPWL approximation when going from a coarse division of the simplicial parti tion of the domain to a finer one, is given. A simple con dition to update the parameters preserving BIBO stability is also derived. In order to show the behavior of the pro posed algorithm, the identification of a well known dynam ical system example from the literature is presented.
Important Links:
Go Back
