An Iterative Procedure for Model Updating Based on Selective Sensitivity

Model updating of a structural system may require a large number of parameter to be identified simultaneously. Due to the ill-conditionedness, large errors in identified parameter values will occur when errors are present in the measurements. One solution for this problem is using the concept of selective sencitivity [

1

]. The method allows to reduce ill-conditioning by providing specific excitations causing model responses sensitive to a small number of model parameters. Thus, only a few parameters are estimated at a time. However, defining such excitations generally involves the knowledge of all parameters to be identified. Therefore, an iterative experiment procedure is suggested (e.g the method of multi-hypothesis testing [

2

]) which is normally a time-consuming process.

This paper presents the theory for an alternative iterative procedure for dynamical excitation of undamped, linear structures. The approach is developed using the concept of predictive control and then is incorporated into a Bayesian updating methodology to reduce the uncertainty in the system parameters. Simulation examples of a multi-storey frame structure and a continuously supported beam under hamonic excitation demonstrate the potential of the proposed method.