A combined state-space identification algorithm applied to data from a modal analysis experiment on a separation system

This paper discusses identification of state-space models from impulse response or initial value experiments. Kung´s geometrical realization algorithm is combined with classical nonlinear parametric optimization to improve the quality of the estimated state-space model. These ideas are applied on real data originating from a modal analysis experiment on a separation system. The results indicate that the parametric optimization step increases the model quality significantly compared with the initial model the realization algorithm provides