Time Domain Method for Parameter System Identification

Abstract

In this paper a method of parameter identification for a multi-degree-of-freedom structural system in a noisy environment is presented. The method involves an iterative procedure in which initial parameter estimates are obtained by relying on a least squares kind of approximation. This estimate is used in an adaptive Kalman filter to obtain an improved estimate of the system state. The improved estimate is then utilized in the least squares approximation to produce refined estimates of the system parameters. The iteration is repeated until it converges within an acceptable margin. The parameter errors are compensated during filtering by adding pseudonoise to the system equation; the noise itensity is updated in each iteration. Results of a simulation study conducted for a two-degree-of-freedom system indicate that the method can yield, for a relatively low computational cost, reliable estimates of system parameters, even when the data record is short.