Stochastic System Identification: Uncertainty of the Estimated Modal Parameters

In this paper, two stochastic system identification methods are compared. Stochastic means that the method has to cope with output-only data (stochastic, unknown input). The first method is a stochastic subspace algorithm that finds the system matrices of a stochastic state space model. It is a linear method but said to be suboptimal: it is not based on the minimization of a criterion. The second method is a prediction error method that finds the parameters of a multivariable ARMA-model in a nonlinear, iterative way. Both methods are discussed and applied to experimental data from a dynamic test on a concrete beam. The uncertainties of the estimated modal parameters are compared. The determination of these uncertainties is very relevant for structural monitoring based on dynamic measurements. In this way, changes of the dynamic characteristics due to structural modifications (e.g. damage) can be separated from random changes which fall within the uncertainties of the estimated modal parameters.