Luke E. K. Achenie
Department of Chemical Engineering, Unit 3222
University of Connecticut, Storrs, CT 06269, USA
email: achenie@engr.uconn.edu
Computational mathematics has been an integral part of Chemical Engineering since the early sixties. Unfortunately, researchers within the Applied and Computational Mathematics community are only minimally aware of this fact. In this presentation, the author will discuss a sample of challenging, unsolved or partially solved problems in Chemical Engineering which require computational mathematics tools. An example challenge problem comes from plant-wide process simulation that involves hundreds and sometimes thousands of process variables and/or highly coupled nonlinear process constraints. One of the issues is propagating uncertainties in the process model parameters throughout the flowsheet. Specifically we would like to address the question: given a set of uncertain parameters with known bounds, can all the process constraints be satisfied for all realizations of the uncertain parameters? How large can the domain of an uncertain parameter be without violating any process constraint? What is the limiting process constraint? These considerations lead to what is termed "flexibility analysis" within the Chemical Engineering literature. Flexibility analysis sub-problems have been modeled as "maxminmax" nonlinear programs. These problems are inherently non-linear, non-convex and non-smooth. Can interval methods be used to solve flexibility analysis sub-problems? It is the speaker's view that this may very well be the case.