Towards a new method for identifying systems

The identification problem addressed is one where a set of frequency response functions exists (from measurement) and system matrices of a minimal order are sought which accurately reproduce all of the dynamics of the FRFs. The method outlined comprises three stages : (a) create a model having a very large number of degrees of freedom and a particular structure which matches the measured FRFs to arbitrary accuracy over the range of frequencies measured, (b) keeping the same large number of degrees of freedom, allow the parameters in the system to relax to further reduce residual (c) successively reduce the number of degrees of freedom in the model retaining as much fidelity to the measured response functions as possible and quantifying the loss in accuracy at every step.