CFD-guided optimization of the injector and injection parameters for light-duty GCI engine

Abstract

In this study, computational fluid dynamics (CFD) and machine learning (ML) were used to investigate the effects of and optimize the injector design parameters for light-duty gasoline compression ignition (GCI) engine. This study was performed at part- (6 bar) and high-load (22 bar) indicated mean effective pressure (IMEP) conditions. The effects of number of nozzles (nNoz), spray angle (SA) and plume angle (PA) while maintaining total nozzle area (TNA) and start of injection (SOI) were first investigated. The increased nNoz and PA enhanced fuel/air mixing, especially in the bowl region, by even spray distribution at part-load condition, but the effects are negligible at high-load mixing-driven combustion due to high in-cylinder temperature and pressure at the time of main injection. On the other hand, SA had significant effect on the air utilization and hence engine performance and emissions at both part- and high-load conditions. The best design from this manual parametric study produced a balanced air utilization between piston bowl and squish region resulting from the fuel spray targeting the upper lip of the bowl. The second phase of this study focused on the optimization of the injector and injection parameters including nNoz, SA, nozzle hole diameters (dNoz) and SOI using design of experiment (DoE) approach. 32 DoE cases were generated and best design was selected at each load point. It was found that the SA and SOI are the most influential injection parameters. Specifically, two optimum SOI regimes at part-load conditions have been identified. These are a) early injection resulting in retarded combustion and low pressure rise rate and b) late injection yielding high combustion efficiency and low hydrocarbon emission. At high-load condition, SOI right before top dead center (TDC) is most preferable and early injection should be avoided to minimize pressure rise rate. The narrow SA (90 degrees (deg) in this study) as well as wide SA were found to produce optimum performance. Subsequently, ML algorithm was used to further optimize the injector and injection parameters. As a result of this optimization study, 10.7% reduction of fuel consumption at part-load and 7.5% reduction at high-load were achieved.