Propagation of variances of FRFs through FRF-based coupling calculations

In many engineering cases, uncertainties will exist on the frequency response functions (FRFs) of component models. Possible sources of uncertainty in the mechanical structure are uncertain material properties, loading conditions or geometry. When bringing together the component models together and interconnecting them into an assembly, it is of high interest to be able to predict the uncertainty of the response of the assembled system. This paper outlines a technique allowing such an uncertainty characterization. The underlying idea of the approach is based on variance propagation through FRF-based coupling techniques. An analytical approximation using a mean value first order second moment approach (MV) is proposed, which significantly reduces the computational cost involved in a probabilistic uncertainty quantification through the common Monte Carlo (MC) simulation technique. A two-component carbody-subframe assembly model indicates that the overall efficiency of the analytical approach is superior to the classical Monte Carlo simulation approach, while the accuracy obtained with both approaches is very similar.