In this study, the piston bowl geometry and injector design of a light-duty GCI engine were co-optimized using computational fluid dynamics (CFD) and advanced machine learning (ML) techniques to maximize the capability of a GCI technology. This study was performed at low- (6 bar), mid- (11 bar) and high-load (22 bar) indicated mean effective pressure (IMEP) conditions. The 3-D CFD setup was first validated against experimental data. Then, the injector and the piston bowl were simultaneously co-optimized. In total, 13 (ten piston- and three injector-related) design parameters were considered. At each load condition, 128 DoE cases were generated, and the features and performance of the top three designs were analyzed. The best DoE solution was selected by defining weighted merit values to provide one best design across all load conditions. The simulated dataset was used for further optimization using advanced ML techniques. It was found that a flatter design with low center height, wide and shallow bowl, and the low lip was preferable in the low- and high-loads. At the low-load partially premixed compression ignition (PPCI) mode, spray targeting the upper lip and divided into the bowl and squish zones for enhanced mixing is preferred. At the high-load mixing-controlled diffusion combustion mode, where injection occurs near the top dead center (TDC) and diffusion burn is the dominant combustion mode, targeting the lower lip is favorable. At the mid-load with a high premixed ratio, the combustion was close to homogeneous charge compression ignition (HCCI), and the piston bowl design had limited effect. Regarding the optimum injector parameters, larger number nozzles with smaller diameters were favorable at low-load to control partially premixed charge. At high-load, larger and fewer number nozzles are recommended. Lastly, the optimum ML design provided similar performance at mid-load but a 3.8–4.5 % reduction in fuel consumption compared to the baseline cases at low- and high-load conditions.
Read full abstract