Optimization of the performance of a GDI engine under lean conditions: an experimental and numerical study

Split injection events are being considered also in gasoline direct injection (GDI) engines. Under some regimes, they are already employed to increase the temperature and maximize the efficiency of the catalytic converter or to reduce the engine tendency to knock-ing. Alternatively, under moderate or low loads and speeds, split injection may be adopted for properly stratifying the air-to-fuel ratio around the spark plug. Due to the complexity of the in-cylinder phenomena, the start of injection (SOI) and the dwell time between the injection events must be correctly chosen in conjunction with the time of spark advance. This last is well known as strongly affecting the rate of heat release when all the other relevant variables are kept constant. Therefore, the solution of an optimization problem can be considered as a viable way to achieve the task of engine efficiency maximization. The effect of the synchronization of multiple injection events and spark advance within the engine working cycle is studied into detail in the present work through computational fluid dynamics optimization. An experimental and numerical characterization of the energy conversion process taking place in a GDI engine equipped with a multi-hole high-pressure injector is first performed. The engine under study is a single cylinder research DISI (Direct Injection Spark Ignition) engine mounting the same injection system of a commercial turbocharged engine. Tests are carried out using commercial gasoline. The injection duration is fixed in order to obtain lean combustion conditions. A 3D engine model is developed within the AVL FireTM environment. An in-house developed sub-model allows simulating the spray dynamics under various injection strategies. Pollut-ants formation is also considered and assessed by comparison with the measurements of ma-jor harmful species at the exhaust. After validation, the 3D engine model is coupled with a multi-objective optimization software, that employs a genetic algorithm to select the value of the input variables in the design of ex-periment (DOE space), and point out the solution better fulfilling prescribed objectives of power output and pollutants amount at the exhaust.