Transformations of linear systems under external equivalence (Q1118558)

We consider systems of linear differential and algebraic equations in which some of the variables are distinguished as ``external variables''. Two systems are called equivalent if the set of solutions for the external variables is the same for both systems. We give an operational form for this definition of equivalence, i.e., we describe a set of system transformations having the property that two systems are equivalent if and only if they can be taken into each other by transformations from that set. Next, an algorithm is described to transform a given system in general form to a system in minimal state- space form. This algorithm differs from existing methods in that it first takes the equations to first-order form, so that each subsequent step can be formulated and interpreted in state-space terms. We also compute the ``structure indices'' in terms of a state-space description in nonminimal form and use this to prove the minimality of the end result of the algorithm. Finally, an application is shown to the problem of ill- posedness of feedback connections.