Optimization-based design of energy-integrated distillation processes

Increasing economic competition in the chemical industry and the need to reduce greenhouse gas emissions fosters the desire to design energetically and economically efficient chemical processes. Distillation is still the default option for the separation of liquid mixtures, despite its low thermodynamic efficiency. Fortunately, various concepts for energy integration have been proposed to improve its energy efficiency. However, the selection of the cost-optimal process among the potential alternatives is tedious, as a multitude of different process options need to be compared for each specific separation. Therefore, there is a need for computationally efficient design methods that allow for a comparison of the competing alternatives on the basis of rigorous conceptual models, particularly for the separation of multicomponent non-ideal or azeotropic mixtures. In this thesis, an optimization-based design approach is proposed that enables the efficient, cost-optimal design of the competing process variants based on deterministic superstructure optimization. Such a method can be a significant step towards enabling the systematic consideration of energy integration during the conceptual design of distillation processes.
While the design of distillation processes is commonly based on the equilibrium stage model and HETP values for packed columns to determine the required column height, this design procedure includes significant design risks for certain separations. In such cases, rate-based models that should enable a more reliable column sizing can be applied. However, it is still unknown in which cases rate-based models need to be applied or when the equilibrium stage model suffices. This thesis also investigates the general need for rate-based modeling of distillation processes in packed columns and proposes a novel method for model discrimination that enables a mixture-specific evaluation of potential differences between the equilibrium stage and rate-based model.