ID 249

Multi-objective optimization of high-pressure gasoline injector nozzles using Genetic Algorithms coupled with Computational Fluid Dynamics (CFD): exploiting the manufactural design space

 

Abstract:

Challenging emission regulations demand novel optimization strategies to tap the full potential of homogeneous Gasoline Direct Injection (GDI) engines. High development effort is put on the nozzle layout of multi-hole high-pressure gasoline injectors to reach optimal combustion with minimum emissions and highest possible efficiency. Flexible design tools are required to identify sensitivities of the key valve seat parameters towards nozzle flow dynamics as well as spray pattern and to meet engine specific requirements. Previous investigations have revealed the importance of a holistic optimization approach concerning nozzle flow and spray characteristics. In the present paper the importance of coupling the nozzle flow with the spray simulation in order to correctly predict the key objectives while taking into account all relevant constraints is highlighted. Shadowgraphy measurements show that the spray characteristics, such as penetration length, are correctly reproduced by the simulation. Then, a constrained 9-dimensional Design-of-Experiment (DoE) has been calculated with the optimization software OPAL++ (OPtimization Algorithm Library++) using high performance computing to understand the interdependencies between design variables and objectives. Second, approximation models have been built to identify sensitivities regarding nozzle flow and spray formation. Finally, a strategy is introduced to extract an initialization population from the large DoE for a pure CFD-based optimization, delivering valid designs for a specific problem definition. Applying NSGA2 (a multi-objective genetic optimization algorithms) the Pareto front containing optimal designs regarding spray penetration and spray plume cone angle has been identified. Meaningful non-dominated designs are analyzed and conclusions are derived regarding optimal geometry parameter.