A new synthesis framework screens and examines complex distillation sequences. Con®entional superstructure de®elopments are replaced in fa®or of a no®el representation that assumes the form of a supertask model. The supertask is based on simple tasks that accommodate for basic sequences. Hybrid tasks account for complex columns and sloppy splits. Discrete instances of simple tasks are combined with hybrid transformations to optimize operating conditions. The optimization problem is formulated as a simple MILP problem that is possible to sol®e to global optimality. The proposed representation can de®elop different noncon®entional and no®el designs featuring fully integrated columns, parallel sequences, and multiple-effect columns. The approach is illustrated with se®eral literature and industrial problems. In all cases solutions are reported in the form of noncon®entional designs that perform as optimal or near-optimal schemes.
IntroductionEffective screening of separation systems constitutes a critical stage, as engineers need to review and understand trade-offs ahead of detailed modeling and simulation. In the separation, it is often desired to explore the use of complex rather than simple columns, because the complex units reduce mixing losses, use available vapor and liquid more effec-Ž tively, and improve the separation efficiency Triantafyllou . and Smith, 1992 . Despite its recognized potential in energy savings, complex distillation applications are limited due to their difficult and demanding design and synthesis assignment. Synthesis challenges and operability issues that arise from a more complex dynamic behavior have discouraged wider industry acceptance. A prohibitive number of configurations emerge from different allocations of side-rectifiers, side-strippers, prefractionators, and side-draw columns. Such options are difficult to enumerate and assess. The design alternatives increase rapidly and the trade-offs are impossible Ž . to assess with an exhaustive implicit or explicit enumeration of the options. Although this article does not address dynamCorrespondence concerning this article should be addressed to A. C. Kokossis at this current address: Dept. of Chemical and Process Engineering, University of Surrey, Guildford GU2 5XH, U.K.Current address of P. B. Shah: AEA Technology-Hyprotech, 707-8th Avenue SW, Suite 800, Calgary, Alberta, Canada T2P 3V3.ics, it presents a simultaneous, systematic, and rigorous approach for the automated development of optimal designs. The approach is computationally inexpensive, reliable, and very efficient to implement. It provides the engineer with a selected set of optimal designs on which further attention to dynamics and operability would yield the final design.Previous efforts have focused on the development of shortcut methods that were designed to evaluate fixed configurations and initialize simulation models. Stupin and Lockhart Ž . 1972 developed the equivalent arrangements of simple columns to represent complex configurations. Several other Ž researchers Fi...