Advanced system simulation, parameter estimation and process design in preparative chromatography

  • Erweiterte Systemsimulation, Parameterschätzung und Prozessdesign in der präparativen Chromatographie

He, Qiaole; Wiechert, Wolfgang (Thesis advisor); Mitsos, Alexander (Thesis advisor)

Aachen (2018)
Dissertation / PhD Thesis

Dissertation, RWTH Aachen University, 2018


Preparative chromatography is one of the prevailing processes in academia and industry to extract or purify one or several target components from a mixture solution. Mathematical modeling can support rational design and analysis of experiments that are usually time-consuming and expensive. In this thesis, model based contributions are made to advanced simulation and efficient optimization of preparative chromatography. Bayesian inference is applied to parameter estimation of chromatographic models in batch mode. To this end, a Markov Chain Monte Carlo (MCMC) algorithm is enhanced for improved convergence and efficiency. Experimental data for a case study, using lysozyme as model protein, was provided by a collaboration partner. Parameter estimation is performed in several stages, each resulting in probability distributions for the sought parameters and using previous results as prior information. Uncertainty of the estimated parameters is quantified using credible intervals. Throughput of chromatographic processes can be increased by using continuous instead of batch operation, e.g., simulated moving bed (SMB), which is a network of single columns operated in counter-current mode. A modeling strategy for SMB processes based on weak coupling of individual column models is developed, which has flexibilities in not only supporting different network configurations but also choosing between various modeling options for the columns and dead volumes. The one-column analog has been published, while operating-splitting appears to be novel at least in the field of chromatography modeling. Four numerical methods are developed in one software such that a fair comparison between them is presented. The introduced methods for solving inverse problems and efficiently simulating SMB chromatography are applied to optimally design ion-exchange SMB (IEX-SMB) units for protein separations. The major difficulty in determining operating conditions of IEX-SMB processes is the nonlinearity that results from its binding kinetics. Consequently, the commonly used triangle theory is not applicable, and inverse methods are applied in this study. A binary and a ternary mixture are used as case studies; the process design of ternary separation using closed-loop IEX-SMB is novel. Suitable network configurations have been successfully determined to separate the proteins. Lastly, the author has implemented the MCMC algorithm (CADET-MCMC) and different solution approaches for SMB models (CADET-SMB) in software packages based on the CADET project. They are published as open-source software and freely distributed under the GPL license.