Study of data collected from a linear, passive, time-invariant, closed loop roveals that the linear dependence of the control variable upon the input variable makes data collected from such a system unsuitable for the determination of its dynamics. It is shown that the introduction of a disturbance into the loop can destroy the linear dependence, but special care may still be necessary for the determination of the dynamics from the data. It emerges that : (a) to seek a linear, passive, time-invariant operator, to approximate the active element containing the disturbance, leads to a physically valueless solution of the optimization problem, while (b) the insertion of the disturbance into the loop at a point external to the element studied or (c) the choice of an optimum operator from a linear, active, time-invariant sub-class, makes a physically meaningful solution of the optimization problem possible.
Tijdschrift beschrijving