Multivariable model identification from frequency response data

A strategy for synthesising high-fidelity, MIMO models from transfer function data has been developed. It takes advantage of the strengths of three separate model ID methods. The eigensystem realisation algorithm is used to generate a model, and then a curve fit based on the complex log of the transfer functions is used to reduce the model. Finally, the model is fine-tuned using a curve fit to the transfer function data directly. This method has been shown to work well on the SERC Interferometric Testbed, a system with many lightly-damped, closely spaced modes. One facet of identifying a system model from the interferometer data was the absence of noise in the measured transfer functions. Future work will examine performance of this method on noisier data and suggest modifications.<<ETX>>