A methodology for CFD predictions of spark-ignition direct-injection engine conical sprays combining improved physical submodels and system optimization

The motivation of this study is to improve the modeling of conical sprays used in Spark Ignition Direct-Injection (SIDI) engines via submodel development and an integrated system optimization approach. Direct-injection combustion strategies have been identified as a possible solution to meet future stringent fuel economy regulations for passenger vehicles. A common fuel injector of choice has been the pressure-swirl atomizer having either an inwardly or outwardly opening nozzle. Several numerical investigations of these injectors have been undertaken over the past decade; however, adequately simulating these sprays over the entire SIDI operating range has posed many challenges. One reason for the discrepancy between CFD simulations and experimental results is the deficiencies in the models describing droplet behavior due to injector design, primary atomization, and secondary atomization. Both the initial spray that emerges from the injector, as well as the primary atomization process, are difficult to model due to a high level of uncertainty in quantifying the controlling physics in a region of the spray where accurate experimental measurements are difficult to obtain. Secondary atomization is complicated by the uncertainty in modeling high-speed droplet breakup due to the combined effects of droplet deformation and momentum coupling with the ambient medium. A unique methodology for improving predictions of high-speed conical sheet sprays is outlined in this document. Initially, the limitations of a select group of submodels will be assessed on a zero-dimensional basis outside of the multidimensional code. From a submodel development standpoint, the physics of the atomization submodels will be enhanced to better represent the physical processes. Next, from a numerical perspective, a physically-based configuration and integration of the entire family of spray submodels within a multidimensional simulation will be evaluated. Finally, an optimization algorithm will be used to improve the accuracy of numerical predictions when compared to various complementary experimental measurements obtained over a wide range of operating conditions.