Computers and Chemical Engineering 28 (2004) 1223–1236 Processes synthesis and design of distillation sequences using modular simulators: a genetic algorithm framework Jose Leboreiro, Joaquin Acevedo ∗ Department of Chemical Engineering, Instituto Tecnologico y de Estudios Superiores de Monterrey, Av. E. Garza Sada 2501, 64849 Monterrey, NL,Mexico Received 31 October 2002; received in revised form 24 June 2003; accepted 20 October 2003 Abstract An optimization framework is proposed in this work for the synthesis and design of complex distillation sequences, based on a modified genetic algorithm (GA) coupled with a sequential process simulator. The use of a simulator facilitates the formulation of rigorous models for different process alternatives, while the genetic algorithm allows the solutions of the complex non-convex mathematical problem, involving discrete and continuous decisions. To reduce the computational requirements of the optimization procedure, several strategies are proposed, including a novel stopping criterion, which provides an efficient way to end the calculations when the optimal solution has been found. The implementation of these strategies resulted in reductions up to 60% in CPU time for the synthesis of complex distillation systems, succeeding in problems where deterministic mathematical algorithms had failed. © 2003 Elsevier Ltd. All rights reserved. Keywords: Genetic algorithms; Process simulation; Process optimization; Distillation sequences 1. Introduction Distillation is still one of the most important operations in process engineering, but, from a thermodynamic point of view, the efficiency of direct distillation systems is rarely greater than 10% (Kister, 1992). The application of math- ematical programming tools to improve this efficiency has been vastly studied in the past years, and different structural and operational alternatives have been proposed. The first systematic methodology for synthesis and design of distillation sequences based on the optimiza- tion of a superstructure was proposed by Sargent and Gaminibandara (1976). Andrecovich and Westerberg (1985) presented a mixed-integer linear programming (MILP) simplification to solve the synthesis of heat-integrated dis- tillation sequences. Floudas and Paules (1988) extended this model including the non-linearities of the condensers and reboilers and, later, Aggarwal and Floudas (1992) pro- posed the generation of a superstructure containing all the combinations of a sequence of heat-integrated columns for sharp and non-sharp separations, modeled through a ∗ Corresponding author. E-mail address: jacevedo@itesm.mx (J. Acevedo). mixed-integer non-linear programming (MINLP) formula- tion. Viswanathan and Grossmann (1993) included a rig- orous stage-by-stage MINLP model to obtain the optimal feed stage and number of stages, considering the existence of each stage as a discrete decision. Smith and Pantelides (1995) and Dunnebier and Pantelides (1999) proposed a superstructure based on a state task network, modeling each column through a rigorous mathematical formulation without predefined separations. The generation of complex superstructures through heat and mass transfer modules was studied by Ismail, Pistikopoulos, and Papalexandri (1999), Ismail, Proios, and Pistikopoulos (2001) for the separation of azeotropic systems and for reaction/separation systems. Recently, Shah and Kokossis (2002) developed a MILP model for the synthesis of complex distillation systems, which can examine and reduce complex superstructures. The formulation of such models, however, is difficult and time consuming. In addition to the time and expertise needed to formulate these models, the synthesis and design of dis- tillation sequences pose other difficulties for their solution. The use of rigorous design and thermodynamic models leads to very large non-convex models, very difficult to converge. Moreover, taking into account structural and design deci- sions, such as the existence of stages, columns, condensers 0098-1354/$ – see front matter © 2003 Elsevier Ltd. All rights reserved. doi:10.1016/j.compchemeng.2003.06.003