THE SUPPRESSION OF DRIFT AND TRANSIENT EFFECTS FOR FREQUENCY-DOMAIN IDENTIFICATION

The identification of linear system dynamics from measurements corrupt ed by low-frequency aperiodic noise is discussed for situations where a periodic excitation can be applied to the system. It is shown that, by considering the low-frequency harmonic content of the disturbing no ise, accurate elimination of the disturbance can be achieved for typic al linear/nonlinear trend characteristics. The removal of transient ef fects from the initial response of the system to the excitation is als o discussed. This is applicable when the experimental time is limited, as can be the case for systems with long time constants. The removal of the transient effects can be coupled with estimation of the drift p arameters, thus allowing the effects of both the transients and the dr ift to be removed simultaneously.