Reduced-order, multivariable models with pure time delays



Marshall's method of model reduction (1966) neglects the modes of the high order system which contribute little to the overall system response. This approach is extended to handle complex and/or repeated eigenvalues and is then modified so that pure time delays can be included in the control variables of the reduced order model. Open- and closed-loop examples show that the inclusion of the time delays tends to compensate for the eliminated modes and leads to better agreement between the high and low order models.