% 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{vomLehn:847057,
author = {vom Lehn, Florian Alexander},
othercontributors = {Pitsch, Heinz and Mani Sarathy, S.},
title = {{C}ombustion kinetic modeling and fuel design based on
uncertainty quantification and machine learning methods},
school = {Rheinisch-Westfälische Technische Hochschule Aachen},
type = {Dissertation},
address = {Aachen},
publisher = {RWTH Aachen University},
reportid = {RWTH-2022-05057},
pages = {1 Online-Ressource : Illustrationen},
year = {2022},
note = {Veröffentlicht auf dem Publikationsserver der RWTH Aachen
University; Dissertation, Rheinisch-Westfälische Technische
Hochschule Aachen, 2022},
abstract = {The required reduction of green-house gas emissions in the
transport sector motivates the use of alternative,
potentially CO2-neutral fuels. However, the efficient
identification of suitable candidates from the multitude of
possible fuel molecules constitutes a challenge.
Quantitative structure-property relationship (QSPR) models
can form the basis for a streamlined selection of
high-priority molecules. In turn, accurate chemical kinetic
models are required for the most promising candidates to
assess their combustion behaviors in detail. In the first
part of the present thesis, a framework of methodologies for
the uncertainty-quantification-based development and
analysis of detailed kinetic models is established.
Particular emphasis lies on the consideration of
uncertainties in the thermochemical properties of species
and their underlying groups, as well as in the
quantification of their impacts on model predictions.
Consequently, sensitivity analysis, uncertainty
quantification, and optimization methods commonly applied to
the kinetic rate parameters alone are extended toward a
comprehensive analysis of the impacts of thermochemical
property values and rate parameters as well as their joint
optimization based on sets of experimental target data.
These are applied to exemplary fuel molecules, establishing
a broad overview of the types of species and groups as well
as elementary reactions and reaction classes for which high
sensitivities can be expected and for which highly accurate
values are thus of highest importance. The framework is
complemented by a model-based experimental design method,
which identifies those conditions for which experimental
data are most effective in terms of model uncertainty
minimization. In the second part of the thesis,
artificial-neural-network-based QSPR models for the
prediction of research and motor octane numbers, octane
sensitivity, and laminar burning velocity as global
indicators of fuel combustion performance are developed,
accompanied by a systematic analysis of the dependence of
these properties on molecule structure and the establishment
of corresponding fuel design guidelines. The QSPR models are
then used to predict the property values for a large number
of fuel molecules in a database of spark-ignition engine
fuels. This facilitates the straightforward selection of
promising fuel candidates for specific applications based on
defined constraints, which is finally demonstrated by
ranking potential blending agents with gasoline for high
efficiencies of future engines.},
cin = {411410},
ddc = {620},
cid = {$I:(DE-82)411410_20140620$},
typ = {PUB:(DE-HGF)11},
doi = {10.18154/RWTH-2022-05057},
url = {https://publications.rwth-aachen.de/record/847057},
}
guest ::