Convenient representations of structured systems for model order reduction

In control theory there exist two convenient representations of a model, which are transfer functions and state-space matrices. However, structure in the transfer functions is not clearly seen in the state-space form, and on the other hand structure in the state-space form is not clearly seen in the transfer functions. The main reason for this is that Laplace transformation destroys most types of structures. The goal of this paper was introducing such representations that clearly reflect structure in both frequency and time domain. Such representations are obtained by introducing auxiliary signals which define the interactions within the structure. The auxiliary signals lower level of abstraction of input-output mappings, thus providing an insight into physical properties of a system.