Mixed-integer optimization of distillation sequences with Aspen Plus: A practical approach

A modified Generalized Benders Decomposition (mGBD) algorithm for the mixed-integer nonlinear problem (MINLP) optimization of distillation processes is presented and applied to the optimization of an extractive distillation process in Aspen Plus. The mGBD algorithm is implemented in Visual Basic for Applications and communicates via a COM interface with Aspen Plus. The mGBD algorithm accounts for nonconvexities of the MINLP problem by employing the so-called discard-fix-and-decompose strategy. Optimization runs starting from four different integer sets demonstrate that the algorithm can find high-quality solutions independent of the initial integer values. The advantage of the mGBD algorithm is that the Aspen Plus RadFrac model can be used with minor customizations.