Global optimization of multicomponent distillation configurations: 1. Need for a reliable global optimization algorithm

Abstract

Nonazeotropic multicomponent mixtures are often separated into products by distillation configurations containing multiple distillation columns. One method of calculating the minimum vapor duty of a configuration is to sequentially calculate the minimum vapor duty of each mixture as it is split into two streams within a given column starting from the feed column. The other method simultaneously manipulates all the splits to yield the overall minimum vapor duty of the entire configuration. Of these two methods, the sequential minimization is attractive as it can be analytically solved. However, through extensive computations, we find that the sequential minimization method is not a valid substitute for the simultaneous minimization method. As the number of components in the feed increases, the fraction of the basic configurations for which sequential method yields a reasonable estimate decreases rapidly, thereby emphasizing the need for a more robust and reliable global optimization algorithm. © 2012 American Institute of Chemical Engineers AIChE J, 59: 971–981, 2013

Deterministic global optimization in conceptual process design of distillation and melt crystallization

“…Both aspects, however, are beyond the scope of this paper. Inaccuracies induced by the solution method are exemplified by the following case: It is easily possible to discard a good design option when a gradientbased optimization algorithm applied to a nonconvex problem terminates in a poor local optimum, also see [17]. This last problem can be addressed by making decisions based on globally optimal solutions, which is possible by applying deterministic global optimization algorithms as demonstrated in this work.…”

Minimum energy of multicomponent distillation systems using minimum additional heat and mass integration sections

“…The total vapor duty requirement calculation at minimum reflux for any configuration is formulated as a nonlinear programming (NLP) problem based on vapor duty estimation using Underwood's method; and NLP optimization solvers such as GAMS/BARON guarantee that each configuration is solved to global optimality . For illustration purposes, we focus on ideal saturated liquid feed mixture separations.…”

Thermal coupling links to liquid‐only transfer streams: An enumeration method for new FTC dividing wall columns

“…We compared the overall minimum vapor requirement of these two operating‐modes of the FTC DWC for 120 ideal saturated liquid feed mixtures; given by eight unique relative volatility combinations of easy/difficult separation between any two components (Table ), along with 15 representative feed compositions (Table ) that were chosen to span the composition space. For each of these ideal feed mixtures, the two operating‐modes (Figures b and 8) are optimized (with a tolerance of 0.001 for convergence using BARON) using the Global Minimization Algorithm, an approach that determines the globally minimum total vapor requirement of a distillation configuration based on the Underwood's equations, under assumptions of infinite trays and constant relative volatilities between components. The pure products are assumed to be withdrawn as saturated liquids.…”