Knock onset prediction of propane, gasoline and their mixtures in spark ignition engines

Gaseous fuels, such as Liquefied Petroleum Gas (LPG) and Natural Gas (NG), thanks to their excellent mixing capabilities and high knocking resistance, allow complete and cleaner combustion than gasoline in Spark Ignition (SI) engines, resulting in lower pollutant emissions, above all if particulate matter is considered. In previous works [1,2] the authors proved how the simultaneous combustion of gasoline and gaseous fuel (NG or LPG) may strongly reduce both fuel consumption and pollutant emissions with respect to pure gasoline operation without a significant power loss. These very encouraging results were obtained thanks to the strong knock resistance increase obtained adding gaseous fuel to gasoline, which allowed the use of stoichiometric mixtures and better spark advances, even at full load. The introduction of such a kind of combustion in series production engines would however require the use of properly calibrated simulation models, capable to adequately predict the performance and efficiency of engines fuelled by gaseous fuel-gasoline mixtures; in particular, specific combustion models are needed, together with reliable knock onset prediction sub-model. The total absence of such sub-models in the scientific literature induced the authors to investigate the knocking resistance of gasoline-propane mixtures and calibrate a proper knock onset prediction sub-model to be implemented in the zero dimensional thermodynamic models usually employed for engine performance optimization. To this purpose several light knocking in-cylinder pressure cycles have been recorded on a CFR engine, fuelled by gasoline, propane and their mixtures, varying the most important knock-related parameters: compression ratio, spark advance, inlet mixture temperature and fuel mixture composition. The collected data have been used to calibrate two different models, compared in terms of knock onset prediction accuracy: the Knock Integral model (KI) and the Ignition Delay model (ID). Both models revealed a good reliability in predicting the onset of knocking phenomena, with maximum errors around 4 crank angle degrees. The Knock Integral model showed a slightly higher accuracy, which, together with its lower computational effort, makes it preferable for the implementation in the commonly employed thermodynamic engine models.