We present a new knowledge representation and reasoning framework for modeling nonlinear dynamical systems. The goals of this framework are to smoothly incorporate varying levels of domain knowledge and to tailor the reasoning methods --- and hence the search space --- accordingly. Our solution exploits generalized physical networks (GPN), a meta-level representation of idealized two-terminal elements, together with a hierarchy of qualitative and quantitative analysis tools, to produce a dynamic modeling domain whose complexity naturally adapts to the amount of information available about the target system.
Full paper in postscript (186KB) and pdf (194KB).