Integrating optimal design of experiments and superstructure optimization under uncertainties for the design of membrane processes

Membrane processes are attractive options for fluid separations as they enable highly selective separations under mild process conditions. The design and evaluation of membrane processes is however a complex task, due to the large variety of different separation principles and corresponding types of membrane processes, a variety of different membrane materials, and alternative process structures, as well as the limited availability of data and models that is especially a hurdle in the early-stage design of chemical processes. A model- and optimization-based approach can help to design membrane separation processes systematically but requires the determination of a sufficiently accurate process model that usually requires time-consuming experimental work. In the early stage of process design, there are still many options regarding the process structure and operating parameters, and it is desirable to focus the experimental and modelling efforts only on promising process variants and operating conditions. For this purpose, the current work presents an integrated methodology that enables the application of optimization-based methods in early phase process design applied to a reaction-separation process. The application focusses on the selection and design of the membrane separation as an important aspect of the overall flowsheet. By performing a superstructure optimization in which uncertainties are explicitly considered, a set of promising flowsheet variants can be selected. Instead of a separated design of experiments for the model identification, an integral approach is pursued by selecting the most informative experiments on the basis of the impact of the model uncertainties on the design decisions, in order to quickly determine the most relevant parameters. In the specific case study, the methodology is used for the conceptual design of a membrane process where the membrane is used to separate a co-product from a complex reaction mixture while retaining the expensive homogeneous catalyst. Using the integrated approach only one set of experiments in addition to the initial screening experiments is necessary to identify the optimal flowsheet for the case study. • Superstructure optimization and optimal design of experiments are integrated for early-phase process design • Model parameters that are relevant for the design decisions are identified • Lab experiments are performed to determine model parameters more precisely • The methodology is applied to a membrane process to retain the homogeneous catalyst