Inclusion of model uncertainty in a computational framework for dynamic operability assessment

This paper deals with the treatment of model uncertainty in a Q-parametrization framework for dynamic operability assessment. Structured nonlinear and/or time-varying model uncertainty is considered using the l1 robust control theory of Khammash and Pearson (1991). In the case of unstructured model uncertainty, dynamic operability assessment is posed as a convex quadratic programming problem and solved efficiently using sparse matrix techniques. If the uncertainty is structured, the resultant problem is nonconvex and is solved at present using a hybrid approach, with more sophisticated global optimization methods being investigated. The approach is applied to a multivariable distillation column containing all four performance limiting factors and the results are discussed.