Thesis (M.A.Sc.)--University of Toronto (Canada), 2003
This thesis presents a new approach to the design of input signals for the identification of 2 x 2 ill-conditioned multivariable systems. The proposed approach builds on the correlated input design developed by Kuong and MacGregor (1991). The input designs were implemented on a simulated distillation column. It is shown that the correlated design is more reliable at producing stable models as compared to the conventional designs (uncorrelated designs) for systems comprised of similar dynamic elements. The input designs proposed in the thesis allows for better identification results than the original input designs for systems consisting of a mixture of slow and fast dynamic elements. In order to validate the input designs, model-based controllers were developed and tested for their performance. The proposed input designs produce models for controllers that provide superior performance over the original input designs for systems comprised of similar dynamics and for systems consisting of both slow and fast dynamics
