This work introduces a new methodology to design transportation fuels offering improved efficiency and reduced emissions, aimed to complement both traditional and emerging engine technologies. Many of these emerging technologies are centered around exploiting low temperature combustion (LTC) strategies that offer improved efficiency and reduced emissions. However, a standardized fuel does not exist to effectively operate in a LTC mode. Through engine simulations, supervised machine learning, and multivariate optimization, this work provides a tool to create fuels tailored to a specific LTC engine application. Engine simulations in this work characterize the LTC performance for hundreds of fuel samples, quantified by a fuel performance metric called the LTC index. Supervised machine learning provides correlations between measured fuel infrared absorbance spectra to various fuel performance metrics: the LTC index, Research Octane Number (RON) and Motor Octane Number (MON). The predictive models of these metrics circumvent the need for costly and time consuming engine experiments to explore the performance of uncharacterized fuels.
The success of this work is heavily reliant on the ability of the models to make accurate predictions of the fuel performance metrics, viz. the LTC index, which is assessed through two efforts. The first effort is generating a robust validation data set to benchmark model predictive performance. Here, multivariate optimization is also used to create surrogates for the FACE gasolines. These surrogates are paired with engine simulations to determine the true LTC index and compare to predicted values. LTC indices of most FACE gasolines are found to be accurately predicted within 3 units, and at worst within 6. The second effort toward predictive model validation is a novel experimental campaign, measuring first and second-stage fuel-spray ignition delays in a constant volume combustion chamber (CVCC). The novelty comes by using ultra-short injections to enhance mixing prior to ignition phenomena, quantified by a state-of-the-art optical diagnostic---developed in this work---that images spatiotemporal fuel-spray concentrations. The mixture data are compared to a spray-ignition model with simplified fluid mechanics, found to capture spray mixing exceedingly well. The validated spray physical model is paired with a comprehensive chemical mechanism to predict and compare to experimentally-obtained spray-ignition onsets, thereby establishing a novel way to assess the accuracy of chemical mechanisms and surrogate fuels.
The validated LTC index predictive model is finally paired with multivariate optimization techniques to design a fuel tailored to a dual mode engine. In theory, this engine would operate in LTC mode where possible for efficiency improvements, and switch to traditional spark-ignition where needed. The tools developed in this work simultaneously point fuel and vehicle technology in a unique direction for designing high efficiency, next-generation combustion systems.