Model reduction in the physical domain

Model reduction for dynamic systems in the physical domain has recently raised an increasing interest. To this day, the identification of the components/subsystems relevant to a given dynamic behaviour remains a problem. This paper shows that, the dynamic elements' energy exchange pattern can be used as a measure of their relevance. Accordingly, a model reduction procedure is presented which allows one to identify the elements that make major contribution to various behaviours of a system. The approach is applicable to linear and nonlinear systems. A few examples are presented to illustrate the concepts.