Truncation error propagation in model order reduction techniques based on substructuring.

Abstract

Several model order reduction techniques split a system in components after which these are reduced individually, where the dynamic response of individual components is typically approximated with a modal truncation of component modes. By an appropriate selection (which usually means selecting enough modes) the truncation error is expected to decrease, but generally no guarantee for the associated error found after reassembling the reduced component models into a single reduced model can be given. In this contribution we investigate how the truncation error arising from the applied reduction techniques for a separate component, propagates to the assembled models. This gives insight on how accurate the model description of separate component needs to be to obey a global overall accuracy of the assembled reduced model and can lead to a different selection criterium for the reduced model. This work is based on an error estimator for modal truncation and the work by Voormeeren [1] on error propagation techniques.