A new model reduction scheme for linear time-invariant systems with polynomial.

Abstract

This thesis presents two new model reduction techniques for the approximation of high order linear time-invariant dynamic systems, one from the state space point of view and the other from the input/output point of view. The two techniques give reduced models having good steady-state and transient behavior when the inputs are polynomials. In the state space method, the original system is transformed into an identical new system which when reduced gives an improved steady-state and transient response. In the second model reduction technique, we use a mixture of Markov parameters and moments to obtain an approximate realization having a better steady-state response than the approximate realization resulting from using only Markov parameters. The two proposed techniques are developed for both continuous-time and discrete-time dynamic systems. Several examples demonstrate the accuracy of the approximations and confirm the above claims.