A Piecewise Continuous Approach For On-line Identification Of Continuous-time Plants

Abstract

This paper proposes an original method for recursive on-line plant identification.
The method identifies directly a linear continuous time state space model of the plant
from full state measurement through a black box approach. The basic principle consists
in using a reference model, called clone, which runs in parallel so as to reproduce the
behaviour of the plant. To do so, the state error between the plant and its clone is minimised
by means of an adaptive algorithm that tunes the clone's varying parameters iteratively.
These parameters represent the estimators of those of the plant. The originality lies in
the particular structure of the clone. The latter is in fact a Piecewise Continuous System
(PCS) characterised by exogenous switchings of its state. The method is appropriate for real
time applications, where real plants are controlled by digital calculators. Results from a
computer simulation and a real time application are given.