Two-stroke carbureted engines typically have much higher emissions and lower efficiency than comparable four-stroke engines, due to fuel loss in the scavenging process. Direct injection may be applied to two-strokes in order to correct this, but emissions sources remain. Thus, a research project has been conducted with these goals: to learn what factors are most important in direct injection emissions; to gain an understanding of the physical processes responsible; to apply and evaluate statistical research methods; and to determine directions for additional research.

A 2 factorial experimental design was applied to a direct-injection two-stroke research engine, using injection timing, injector spray type, spark plug reach, and spark plug location as the factors. The experiment was replicated to allow estimation of variance and creation of confidence intervals. Responses recorded include full emissions data, flow information, and cylinder pressure history. Cylinder pressure data was analyzed to determine heat release and thus provide combustion information.

Results indicate that, at the operating condition studied, spray type was the most important factor affecting hydrocarbon emissions from this engine. This was followed, in order of importance, by spark plug reach, spark plug location, and the interaction between spark plug location and injection timing. The dominant mechanism for hydrocarbon emissions was misfiring caused by the fuel cloud not being ignited by the spark. Mixing of fuel and air beyond the lean limit of combustion may be a secondary mechanism. The fuel injector and in-cylinder flows control these mechanisms.

Nitrogen oxide emissions were affected by injection spray, injection timing, and spark plug location. Misfiring also was the dominant mechanism here, possibly with secondary effects of the injection timing on fuel-air mixing and thus on combustion temperature. Carbon monoxide emissions depended significantly only on injection spray, due to its influence on the richness of the air-fuel ratio in the flame. Efficiency depended on spray type, spark plug reach, and spark plug location, with misfire the dominant mechanism.

Statistical methods for analysis of unreplicated factorial experiments were not adequate for identifying significant factors due to the number of active factors, so methods for replicated experiments were required. Further research is recommended into injector spray characteristics and their influence on emissions, and into in-cylinder flow patterns and variability.