Model updating using uncertain experimental modal data

Abstract

The propagation of parameter uncertainty in structural dynamics has become a feasible method to determine the probabilistic description of the vibration response of industrial scale �nite element models. Though methods for uncertainty propagation have been developed extensively, the quanti�cation of parameter uncertainty has been neglected in the past. But a correct assumption for the parameter variability is essential for the estimation of the uncertain vibration response. This paper shows how to identify model parameter means and covariance matrix from uncertain experimental modal test data. The common gradient based approach from deterministic computational model updating was extended by an equation that accounts for the stochastic part. In detail an inverse approach for the identi�cation of statistical parametric properties will be presented which will be applied on a numerical model of a replica of the GARTEUR SM-AG19 benchmark structure. The uncertain eigenfrequencies and mode shapes have been determined in an extensive experimental modal test campaign where the aircraft structure was tested repeatedly while it was 130 times dis- and reassembled in between each experimental modal analysis.