Uncertainty quantification in experimental frequency based substructuring

Abstract

Numerical conditioning of the subsystems’ interface flexibility matrix is an important issue in experimental frequency based substructuring (FBS) methods. As this matrix needs to be inverted, it is believed that ill numerical conditioning could severely magnify even small random errors in experimentally obtained subsystems, yielding erroneous FRFs of the coupled system. In this paper a method is introduced with which the uncertainty of the coupled system's FRFs can be quantified based on the uncertainties of the subsystem FRFs. This uncertainty propagation method is based on the statistical moment method and uses the measured (time) data of the subsystems as input. The method will be applied to a numerical example, where it is shown that the uncertainty on substructure FRFs can significantly influence the accuracy of the coupled system. Furthermore, the application shows that the method introduced is computationally efficient and highlights some interesting phenomena that can be encountered in experimental dynamic substructuring.