|dc.description.abstract||Pentane isomers are among the most abundant constituents of worldwide gasoline. Their autoignition has been extensively studied in fundamental reactors (shock tube, rapid compression machine (RCM), jet-stirred reactor, etc.) but rarely in practical engines.
This thesis therefore investigates the autoignition of the three pentane isomers, n-, neo- and iso-pentane, in a Cooperative Fuel Research (CFR) engine. The research octane number (RON) and motor octane number (MON) are first measured. Both RON and MON show that the reactivity of the pentanes are in the order of n-pentane > neo-pentane > iso-pentane. Modeling of engine autoignition are then conducted using a two-zone model  and two detailed, optimized chemical models. Reasonable agreement is observed between the modeling and experiment in simulating the knock onset timing of pentane in the CFR engine.
Given the inherent uncertainty in kinetic models, uncertainty of the engine autoignition modeling is then quantified to determine whether a chemical model that is validated with fundamental experiments can reproduce the engine autoignition to be within the engine measurement uncertainty. A deterministic approach, Bound-to-Bound Data Collaboration , is used for this purpose to investigate the pentane model from the National University of Ireland Galway. Sensitivity analysis is first conducted to identify 29 most important reactions in pentane oxidation, for which the uncertainties of the rate coefficient are estimated. Extensive Computer experiments are then performed to build the modeling surface using the 29 reactions, from which the engine autoignition modeling uncertainty is determined. Results show that the modeling uncertainty, propagated from the uncertainties in the rate coefficients of 29 sensitive reactions, is 14 times greater than the engine measurement uncertainty. The modeling uncertainty can be reduced significantly from constraining the modeling with 15 fundamental experiments (6 ignition delays, 7 species concentrations and 2 flame speeds), which, however, is still 5 times of the measurement uncertainty. The impacts of three important in-cylinder conditions on modeling uncertainty are also investigated, which are found less significant than the 29 reactions. Multiple strategies are then proposed to reduce the modeling uncertainty, where including CFR engine experiment is found to be the most effective constraint.
A set of new CFR engine experiments is then conducted. It is found that these data are able to further reduce the modeling uncertainty to be smaller than the individual engine measurement uncertainty. These engine experiments are further combined with 17 fundamental experiments and the combined dataset significantly reduces the modeling uncertainty, with the smallest uncertainty being 0.4 CA, much smaller than the engine measurement uncertainty. To understand the significance of such modeling uncertainty in term of octane number, further CFR engine experiments are conducted. It is found that the smallest modeling uncertainty (0.4 CA) is equivalent to a change of 1.35 in octane number, which is over the tolerance allowed for standard octane number measurements. This indicates that the most constrained model for pentanes is not accurate enough for application to engine combustion modeling. This situation is likely to be ameliorated by including more experimental data, particularly from the CFR engine, as constraints. This work demonstrates that well calibrated CFR engine experiments complements the fundamental experiments and should be used together for developing more accurate chemical models.||en_US