On the relative and multiplicative model reductions

A new relative error model reduction method is proposed using frequency weighted balanced realization and inner-outer factorizations. Explicit L/sub /spl infin// norm error bounds are also derived for the relative error and multiplicative error. A computer example is also given in the paper showing that the improved error bounds by Wang and Safonov (1990, 1992) are in fact wrong. We consider the problem of reducing the order of a linear multivariable dynamical system.<
>