GLOBAL OPTIMIZATION OF AN AGGREGATED MODEL FOR

LORENA BERGAMINI,; AGUIRRE, PIO; NICOLAS SCENNA,

Rio das Pedras, Brasil

Congreso; 2nd. Mercosur Congress on Chemical Engineering. 4th. Mercosur Congress on Process System Engineering; 2005

An aggregated model for multicomponent distillation columns is first presented. The model allows to predict the internal temperatures and flows for conventional columns operating at minimum reflux. The column is considered as a set of four mass exchanger modules, two for the stripping section, and two for the rectifying section. Each section is divided in two modules, in order to impose the pinch placement. Constraints in the model involve total and individual mass balances, equilibrium equations at the feed point, pinch equations in each section, feasibility for the mass exchange at the inlet and outlet of each module, and product recovery constraints. The model thus generated is a nonconvex NLP. The model was validated with several simulations using HYSYS for mixtures involving up to five components. A new version of our Global Optimization ? Logic Based ? Outer Approximation algorithm was used to overcome the difficulties of non-convexities on finding the global optimal solution. A lower bounding master problem is set up, where the nonconvexities of the original model are replaced by piecewise estimators. The algorithm solves iteratively the bounding problem, which provides a lower bound for the objective function, and a reduced NLP that provides an upper bound for the problem. Convergence is achieved when the bounding problem is infeasible, showing that there are no better solutions. Several examples where solved to illustrate the performance of the algorithm. The aggregated model will then be embedded in a superstructure that corresponds to a sequence of distillation columns performing the separation of a multicomponent feed into pure components.