Dynamic optimization based reactor synthesis and design under uncertainty for liquid multiphase processes

Abstract

New process structures for the substitution of petrochemical feedstocks by renewables are required to cope with the upcoming scarcity of hydrocarbon resources and to decrease the anthropogenic impact on climate change. These challenges are triggered and increased by rising demands of a steadily growing world population and their increasing living standards. The development of therefore required innovative processes is a complex, interdisciplinary, multi-scale challenge including various fields of natural science and engineering. One of the main stimuli of and contributions to this endeavor are provided by computer-aided tools on every time and length scale of the process development procedure. With the Collaborative Research Center TR 63 “InPROMPT” a trans-regional project was founded (i) to investigate the utilization of innovative solvent systems for the functionalization of long chain substrates from renewable feedstocks and (ii) to develop the required tools for fast and efficient design of the corresponding process systems. As part of this project, the presented work had the goal to design an optimal reactor unit for the hydroformylation of 1-dodecene in a thermomorphic multicomponent solvent system. This process is an example of the very important class of homogeneously catalyzed liquid multiphase processes, which is highly promising for the task of substitution of petrochemical feedstocks. In order to accomplish this task, methodical approaches for the synthesis and design of chemical reactors have been developed and a detailed reactor design study as well as technical realization have been carried out. An approach for the qualitative synthesis of reactor-networks was created within the process design framework of the methodology of elementary process functions. The key notion of the underlying methodology is the dynamic optimization of mass and energy control fluxes manipulating a Lagrangian fluid element on its way through the process in order to make it follow the optimal route in the thermodynamic state space. By analyzing the optimal control fluxes resulting from this dynamic optimization, the developed flux profile analysis enables the derivation of reactor-network candidates of different synthesis levels. This includes functions of reaction, separation, and recycling, and allows for a rational selection of promising reactor-network candidates based on reaction engineering fundamentals. The applicability of this approach was proven by comparison to well-known literature examples, which used the attainable region approach and superstructure optimization. Moreover, its application on the hydroformylation process revealed highly performing process candidates and gained interesting insights into the reaction characteristics. The second methodical development touches the design of reactors under uncertainty within the aforementioned process design framework. Since past process design studies within this framework neglected the impact of arising uncertainties in the process model and disturbances of the process operation, the aspired development of a probabilistic design approach was necessary and reasonable. In order to include uncertainties in the dynamic optimization of the fluid element, the unscented transformation was used. For the reactor design under uncertainty the arising types and sources of uncertainty have been identified and classified with respect to their static or dynamic nature of appearance, and the mathematical formulation of the probabilistic dynamic optimization problem has been derived. The approach was applied on the hydroformylation process example considering model parameter uncertainties on the one hand, and imperfect realization of optimal control profiles on the other hand. Furthermore, the first study was extended in order to identify the most sensitive model parameters by a global sensitivity analysis. In the final step, both novel approaches were applied to design an optimal reactor for an existing miniplant of the hydroformylation process. This retrofit included the synthesis of an optimal reactor-recycle-network and its detailed technical design and realization under the constraints of the miniplant. The resulting in silico design indicated an increase of the yield with respect to the desired product of 17 % and 23 % for the operation without and with closed side product recycle, respectively, whereby the first prediction was experimentally validated and confirmed. Within this work it is demonstrated how computer-aided synthesis and design tools significantly improve and accelerate the development of chemical processes, even for a complex liquid multiphase process. The presented dynamic optimization based approaches for reactor synthesis and design under uncertainty have proven to be expedient extensions for the methodology of elementary process functions, and they still provide potentials for further improvements and application on other processes.