| This dissertation presents an experimental and analytical study of autoignition and knock in a spark ignition engine. A 2.2 liter dyno engine was instrumented with in-cylinder pressure and block-acceleration transducers, and run at steady-state, full load conditions at engine speeds ranging from 1200 to 4000 rpm. A flexible engine control system was built, which enabled individual stewing of the spark angle, to induce various grades of knock in each cylinder. A suite of routines was written for the automatic processing of the data, including heat release analysis for autoignition combustion characterization, and fourier and wavelet transformation for knock recognition and measurement.;Autoignition was characterized using heat release analysis. An algorithm to determine the autoignition angle at different engine speeds, based on the first and second derivatives of the energy release history, was developed. The progression of the flame-propagated combustion if autoignition did not occur was estimated by tuning a weibe function, an empirical function commonly-used to describe the burning history of the fuel, using feedback combustion data up to the autoignition point. The mass of fuel consumed explosively was then estimated by subtraction from the actual heat release history. A method for modeling the heat release of autoigniting cycles in spark ignition engines, based on a weighted double weibe function, was also developed.;The severity and onset of autoignition were correlated with the commonly-used knock measurement methods based on the cylinder pressure. A good correlation was found between the onsets of autoignition and in-cylinder pressure oscillations. The delay of the latter, relative to the former, was in the order of a few degrees, and was a function of engine speed. However, the magnitude of knock generally did not correlate well with the mass of fuel consumed by autoignition and other related quantities. The block acceleration signal of the ensuing structure-borne vibration, was analyzed using the computationally-efficient discrete wavelet transform. Knock severity metrics obtained from wavelet analysis correlated well with direct, pressure-based measurement methods, even at the higher engine speeds. Moreover, information about the onset of knock was also obtained using wavelet analysis of the block acceleration signal. |
