Integrated molecular and process design based on fully predictive methods

 

For sustainable mobility, it is indispensable to reduce CO₂ emissions in the transport sector. An auspicious technology in this regard are liquid fuels whose production is based on renewable energies, biomass and CO₂. The cluster of excellence Fuel Science Center (FSC) researches the most promising fuel candidates from production to propulsion. Within the FSC, this LTT subproject contributes to the development of systematic methods to elaborate sustainable production processes of renewable fuels. However, the design framework developed in this project is not limited to fuel design. The goal is to develop a tool that can design a molecule in-silico for an arbitrary process, respective to an objective function defined by the user.

We integrate Computer-Aided Molecular Design (CAMD) into chemical process design for simultaneous optimisation of molecules and processes. The resulting approach is called Computer-Aided Molecular and Process Design (CAMPD). In CAMPD, thermodynamic property models and process models are used to relate molecular properties to process performance. Thus, the molecular structure is optimised with respect to process performance.

The existing CAMD approach uses an evolutionary algorithm to optimise the molecular structure. The evolutionary algorithm drives the optimization by the results of process models which evaluate how suitable a candidate molecule is for the respective application. The process models require molecular properties as input values. We use quantum chemistry to predict these properties.

 

In several case studies we continuously expand the range of use cases of the CAMPD-Framework. The first case studies focused on the molecular design of extraction solvents for liquid-liquid extraction (1,2).

Following these successful case studies of the CAMPD-Framework, we expanded the scope to reaction solvent design (3, 4). Here we not only considered equilibrium data, but also reaction kinetics for the optimization of molecular design. The solvents do not actively participate in the reaction, however their molecular properties affect the reaction equilibrium and therefore enable the algorithm to steer the optimization.

Recently we were able to extend the applicability of the CAMPD-Framework to automated design of catalysts (6). As catalysts are a key component of many chemical processes and often are responsible for making the process economically or technically feasible in the first place, computer aided catalyst design poses a highly desirable alternative to expensive and time consuming experimental screenings. The application of the CAMPD-Framework for catalyst design was successfully demonstrated in a case study for the carbamate cleavage reaction.

Furthermore, we have also demonstrated that further design criteria and models can be added to the CAMPD-Framework. In previous CAMPD designs for solvents, our group has focused on choosing the most suitable solvent by a thermodynamic evaluation. However, solvents should also be evaluated by their sustainability and environmental impact. Therefore, to achieve a holistic assessment of all candidates we have integrated Life Cycle Assessment (LCA) into the CAMPD-Framework in a cooperation with the Energy Systems Engineering group the LTT (5). The aim is to evaluate the impact of the solvent from Cradle-to-grave. To this end, an artificial neural network is employed to predict the solvent’s environmental impact until usage (cradle-to-gate) and use the data from process models for a gate-to-grave assessment of the candidate. This approach was successfully applied for the design of solvents in a hybrid extraction-distillation process. We were able to show that integrating LCA improves the quality of solvent candidates generated by CAMPD.

In summary, the project yields a framework for automated molecular and process design for chemical processes using consistent thermodynamic properties from quantum chemistry for all candidate solvents. This framework was successfully applied to several case studies. Our results show the mutual dependency of optimal solvents and processes: Only the right choice of solvent, selected with an integrated process design, leads to an optimal process performance.

  Copyright: © LTT

Previous studies used COSMO-RS predominantly as a property model. To enhance our approach, we implement prediction methods for additional properties needed for improved economic process evaluation, e.g. the Perturbed Chain Polar – Statistical Associating Fluid Theory (PCP-SAFT) equation of state. PCP-SAFT is based on five parameters that are usually determined experimentally. However, given the use case that we want to employ CAMPD to circumvent expensive experimental screening, it is necessary to make the PCP-SAFT equation of state fully predictive.

We can predict the equation of state parameters for arbitrary molecules by using a method called Segment-based equation of state parameter prediction (SEPP). Thus, SEPP broadens the applicability of PCP-SAFT for molecular design as it enables us to overcome the limitation of group-contribution methods to determine the necessary parameters. To further improve the scope and accuracy of our property prediction methods, we want to introduce models for additional properties, such as interfacial tension or transport properties. With these additional properties, we aim to expand the design framework’s applicability to a variety of processes and properties.

 

Publications

1. Scheffczyk J., Fleitmann L., Schwarz A., Lampe M., Bardow A., Leonhard K. (2917). COSMO-CAMD: A framework for optimization-based computer-aided molecular design using COSMO-RS
2. Scheffczyk J., Schäfer P., Fleitmann L., Thien J., Redepenning C., Leonhard K., Marquardt W., Bardow A. (2018). COSMO-CAMPD: a framework for integrated design of molecules and processes based on COSMO-RS. Mol. Syst. Des. Eng., 2018, 3, 645-657
3. Gertig, C., Kröger, L., Fleitmann, L., Scheffczyk, J., Bardow, A., Leonhard, K. (2019). Rx-COSMO-CAMD: Computer-Aided Molecular Design of Reaction Solvents Based on Predictive Kinetics from Quantum Chemistry
4. Gertig, C., Fleitmann, L., Schilling, J., Leonhard, K., Bardow, A. (2020). Rx-COSMO-CAMPD: Integrated Computer-Aided Design of Solvents and Reactive Chemical Processes based on Quantum Chemistry
5. Fleitmann, L., Kleinekorte, J., Leonhard, K., Bardow, A. (2021) COSMO-susCAMPD: Sustainable solvents from combining computer-aided molecular and process design with predictive life cycle assessment, Chemical Engineering Science, 245, 116863
6. Gertig, C., Leonhard, K., Hemprich C., Hense J., Bardow, A. (2021). CAT-COSMO-CAMPD: Integrated in silico design of catalysts and processes based on quantum chemistry, Computers & Chemical Engineering, 153, 107438
7. Gertig, C., Leonhard, K., Bardow, A. (2020). Computer-aided molecular and processes design based on quantum chemistry: current status and future prospects
8. Polte L., Raßpe-Lange L., Latz F., Jupke A., Leonhard K.; COSMO‐CAMPED – Solvent Design for an Extraction Distillation Considering Molecular, Process, Equipment, and Economic Optimization (2023) Chemie Ingenieur Technik, 95(3), 416-426.