Hybrid Simulation-Optimization Algorithms for Distillation Design

Abstract This work addresses the rigorous design of distillation columns using a mixed approach that combines mathematical programming with explicit equations, and the rigorous models that are available in commercial Chemical Process Simulators. A superstructure that has embedded all the potential configurations is proposed. Based on this superstructure representation the problem is formulated as an optimization problem using generalized disjunctive programming (GDP) to minimize the total cost of the process, subject to design specifications. The method determines the optimal number of equilibrium stages and operation conditions to obtain the specified product separation. The model is solved at different levels. Mass balances, purity specifications and other constrains are included as explicit equations in the model. Properties, like specific enthalpies, and flash equilibrium are calculated by implicit models at the level of process simulator (Simulis Thermodynamics™). In this way we increase the robustness of pure equation based models because we take advantage of tailored numerical methods maintaining the flexibility of mathematical programming based approaches for distillation design. Two simple examples (easy to reproduce) are used to show the performance of the model.