Efficient algorithm for optimal model reduction

In this paper we present a new algorithm for obtaining the optimal approximation of high-order linear continuous time systems. The reduced-order model is determined by a recursive optimization procedure which minimizes a positive definite function of the integral impulse response error between the full and reduced-order model. The proposed optimization algorithm involving the performance index of the complicated nonlinear type is formulated in terms of linear equations which can be easily solved in a recursive manner. By introducing a Routh table and its reverse elimination procedure for solving the linear equations, the computational efficiency is greatly improved.