Interface reduction of linear mechanical systems with a modular setup

The modular design of technical products offers many advantages. For example, development costs can be reduced if single problem-solving components can be reused in various settings. Taking the same approach in a simulative environment leads to substructured models, such as in elastic multibody systems or substructured finite-element models. To reduce simulation times, model order reduction is applied to each component of the system. The goal is to store reduced models that may be reused in various settings in a common database. The more precise the information about acting forces on the models is, the more modern model order reduction schemes, like moment matching with Krylov subspace methods, can make use of their inherent advantages. If the separate components are connected at many points, then the straightforward application of block Krylov methods leads to large reduced systems. Therefore, the “exact” information about the interfaces, that is, all admissible directions of force application, must be reduced to the most important ones. The novel contribution in this paper is a unifying framework for interface reduction techniques applied to substructured systems. The presented reduction technique is able to provide models for a database and can be automatized. Besides the theoretical derivation of the method, a numerical benchmark model is evaluated, and benefits and drawbacks are discussed.