Predicting and Optimizing Two-Stroke Engine Performance Using Multidimensional CFD

One-dimensional unsteady gas dynamics dominate the prediction and optimization of two-stroke engine performance. Its application in engines with complicated geometry is, however, limited because the flow through the engine is three dimensional in nature. Multidimensional CFD has the capacity to capture the effect of complicated flow fields. However, most existing CFD studies include either only one cylinder with a partial exhaust system or just a separate exhaust manifold, and boundary conditions need to be fed from experimental data. It is found in this study that such simplifications may yield misleading results. In a previous study, the authors extended a multidimensional CFD code, KIVA to simulate a multi-cylinder engine together with a full exhaust manifold. The need for exhaust pressure boundary conditions was thus eliminated. In continuation of this study, a crankcase model was first developed to dynamically predict the crankcase pressure. A reed valve model proposed by Blair, et al. was used to predict the flow into the crankcase through the reed valve. Thus the code is capable of predicting engine performance without any input of dynamic boundary conditions. The developed code was first rigorously tested against a single cylinder engine at a wide range of engine speeds. The computed exhaust pressure trace agrees well with the measurement. It shows that the predicted trapped oxygen mass is a good indicator of engine power. The test was then extended to a twin-cylinder loop scavenged engine with a folded exhaust manifold. The predicted exhaust pressure matches experiments, and the predicted trapped oxygen mass agrees well with the estimate based on the measured fuel consumption and emission data. The code was finally used to optimize a production two-stroke engine, and the improvement predicted by the simulation was realized in dynamometer tests.