h1

% IMPORTANT: The following is UTF-8 encoded.  This means that in the presence
% of non-ASCII characters, it will not work with BibTeX 0.99 or older.
% Instead, you should use an up-to-date BibTeX implementation like “bibtex8” or
% “biber”.

@PHDTHESIS{Jpel:981366,
      author       = {Jäpel, Ronald},
      othercontributors = {Buyel, Johannes Felix and Blank, Lars M.},
      title        = {{O}vercoming hurdles in the development of chromatographic
                      models},
      school       = {RWTH Aachen University},
      type         = {Dissertation},
      address      = {Aachen},
      publisher    = {RWTH Aachen University},
      reportid     = {RWTH-2024-02968},
      pages        = {1 Online-Ressource : Illustrationen},
      year         = {2023},
      note         = {Veröffentlicht auf dem Publikationsserver der RWTH Aachen
                      University 2024; Dissertation, RWTH Aachen University, 2023},
      abstract     = {Chromatography is widely used for the purification of
                      biopharmaceutical proteins but can be a major cost driver
                      and time factor during production and process development.
                      Such costs and time requirements can be limited by the
                      model-driven optimization of chromatographic separation,
                      which reduces experimental screening to the most relevant
                      operational conditions. However, the establishment of a
                      useful chromatographic model poses multiple challenges, such
                      as high computational demands during model calibration, high
                      variability within the experimental datasets, and
                      discrepancies between model assumptions and experimental
                      realities. This thesis aims to ease the model establishing
                      process by investigating these challenges and offering
                      options for improvement. The long computation time during
                      model calibration was reduced with the implementation of a
                      novel calibration algorithm based on Bayesian optimization,
                      called BayesOpt. The fitting of a complete model parameter
                      set based on a typical experimental dataset was ~10-fold
                      faster with the new BayesOpt than the industry standard
                      gradient descent algorithm using a single CPU core (46 min
                      vs 478 min). This drastically improved the model calibration
                      workflow, during which the model calibration algorithm has
                      to be run frequently, as multiple datasets have to be
                      calibrated multiple times to compare different model
                      assumptions and setup-options. Using this calibration
                      algorithm, chromatographic models were created based on
                      several experimental ion exchange chromatography datasets.
                      Potential sources for model errors were discovered and
                      mitigation strategies were suggested. Lastly, a key
                      assumption within a model for hydrophobic interaction
                      chromatography was challenged, as it led to unrealistic
                      model predictions. Two alternative models were proposed,
                      implemented, and evaluated on in silico and experimental
                      datasets. The new models were found to have eliminated the
                      unrealistic predictions while also outperforming the
                      original model on two out of three experimental datasets.
                      Together, these advancements can offer substantial increases
                      for the speed and probability of success of setting up
                      chromatographic models.},
      cin          = {162910 / 160000},
      ddc          = {570},
      cid          = {$I:(DE-82)162910_20140620$ / $I:(DE-82)160000_20140620$},
      typ          = {PUB:(DE-HGF)11},
      doi          = {10.18154/RWTH-2024-02968},
      url          = {https://publications.rwth-aachen.de/record/981366},
}