Experimental Model Building of Timevarying Systems by Linear Optimal Filtering

Abstract

The experimental model-building of time-varying systems is discussed. The estimation of parameters of time discrete models is closely related to optimal filtering theory. But for the identification of industrial processes it is necessary to reduce the amount of a-priori information required by theory. Therefore additional algorithms are developed. The properties of the modified algorithms are investigated by application to simulated discrete time-varying systems. Systems with stochastically varying gain, time-lags and damping ratio are studied. Error functions are developed to describe the relations between the rate of change of parameters, the level of output noise and the available a-priori information quantitatively.