On the use of Frequency-Domain Subspace-based System Identication for Estimating Resonant Systems

In this paper, a parametric identification of high order resonant systems from frequency response data is analyzed. Among many techniques for obtaining black-box models for such systems, subspace-based algorithms stand out. Subspace methods uses geometric projections to obtain specific information necessary to compute the system extended observability matrix and, eventually, to obtain a state-space model representation. Throughout the paper, the method is also shown to use only well conditioned matrices to avoid numerical problems with high order models. The validity of the algorithm is firstly demonstrated on a simulated data of a synthetic 15th order system. Then, in order to state the systems effectiveness, the method is applied to actual data extracted from a resonant beam.