Large Eddy Simulation as an Effective Tool for GDI Nozzle Development

The new homologation cycles with expanded test conditions in terms of temperature and altitude, the inclusion of RDE as well as longer and more stringent in-service conformity requirements are driving an increase in the performance requirements for GDI injectors. The further reduction of engine-out Particle Numbers (PN) emissions is of primary importance even when a gasoline particulate filter is used in order to optimize filter size and account for filter efficiency evolution over vehicle lifetime.

Within the engine development process, it is now standard practice to individually tailor the spray plumes to the combustion system for optimized fuel consumption and PN emissions. Both the spray targeting (jet directions) and the mass distribution are optimized for mixture homogeneity and surface wetting in the combustion chamber. For downsized engines, it is also critical to optimize the nozzle geometry for reduced spray penetration to minimize wall film formation and oil dilution. In addition, the prevention of tip deposit formation allows to avoid PN drift [1].

Spray formation, injector tip coking and PN drift involve highly complex physical and chemical processes. Nozzle design optimization is a key element to achieve the engineering targets. While much has been achieved experimentally to parametrically optimize the product, greater progress in injector design is possible with a better understanding of the fundamental physics concerning the microscopic behavior of nozzle flow dynamics and the spray formation process using new simulation technology. The available measurement techniques often cannot resolve the gas-liquid interface scales and the transience in the near-nozzle spray under high pressure injection conditions. Moreover, today no measurement technique is available for characterizing the field turbulence and the complex vortex structure in the nozzle.

High-fidelity, high-resolution Large Eddy Simulation (LES), enhanced with the Coupled Level Set Volume of Fluid (CLSVOF) technique for interface tracking, can provide a detailed insight into the internal nozzle flow and the near-nozzle spray morphology and a clearer understanding on the underlying phenomena. This paper will present an example of how Computational Fluid Dynamics (CFD) can be used to build physical understanding in support of the development of GDI nozzle hole geometries for the purpose of spray control and injector coking reduction.

The performance of a typical nozzle under accelerated tip coking conditions and the injector's subsequent response to increasing pressure is presented in Figure 1. The injector is first tested in its clean or new state at a range of pressures (red bars). These results indicate a clear trend of improving PN with higher pressures, mainly due to improved atomization and air-fuel mixing. The injector is then aged for 10 h at 100 bar causing the formation of a fine layer of deposits on the injector tip. The emissions performance is then tested again (blue bars) and shows a significant increase in PN at the respective pressure levels. Endoscopic visualization of the combustion event (lower row of photographs), reveals that the aged injector now has a bright diffusion flame in the vicinity of the nozzle tip after the main combustion event. The temporal integral of this flame size correlates with the PN increase [1] and indicates the cause to be liquid fuel remaining on the injector tip. Increasing the fuel pressure is an effective measure to clean the deposits or stop their formation, eliminating the diffusion flame and stabilizing the PN of the aged injector to very low levels. Considering the substantial improvements in particle formation and drift performance associated with higher pressures, the Multec 14 injector, FIGURE 2, was introduced as the first to market 400-bar-capable GDI system. Currently, development is progressing on the next-generation 600-bar system with the target to remain compatible with the current production process. From an injector hardware point of view, the increased system pressure requires the adaptation of the injector pressure vessel components as well as a revision of the weld interface geometries and process parameters. The increased system pressure brings the advantages of faster injection giving longer time for fuel-gas mixing and evaporation, increased dynamic range using pressure modulation and a further significant reduction of the PN.

Despite the effectiveness of high fuel pressure for injector coking and PN reduction being well known, the question as to which fluid dynamics processes are causing this effect and how to take maximum advantage of them remains. How to optimize the nozzle tip for both deposit formation and spray penetration then requires an understanding of phenomena which are not directly available by measurement but to which CFD can bring new light. LES analysis of the influence of injection pressure and counter bore dimension on spray and coking cleaning force is presented below. A further investigation of the flow during injector needle closing with focus on the complex process of sac fuel evacuation, fuel vaporization, vapor condensation and gas suction can be found in [2].

The simulation methodology development has been focused on the following aspects: complex physics modelling to describe the real world phenomena, high-fidelity simulation and high efficiency techniques such as adaptive mesh refinement and node-interpolation-based moving mesh approach to support development tasks. On the basis of the previous CLSVOF LES methodology developed for diesel injection primary breakup analysis [3-5], a number of new functionalities have been added. A fully compressible model has been introduced to account for the compressibility of all phases. A cavitation model based on the barotropic Equation of State (EoS) based on [6] has been implemented in Ansys Fluent v19 using User Defined Functions (UDF) and applied to the present study. This model produces the correct bubble collapse speed in the cavitation vapor condensation process and has been applied in a successful cavitation erosion simulation study [7]. The Wall Adapted Local Eddy viscosity (WALE) subgrid model [8] is adopted owing to its correct behavior near the wall. For the computational domain inside the injection nozzle, high resolution hexahedral LES mesh is designed based on the estimation of the Taylor microscales [9]. Wall refinement is applied to reach an average wall resolution of y⁺=1. For the benefit of computational efficiency, an adaptive grid refinement strategy is adopted in the spray region. A base mesh of 64 μm is used with successive levels of refinement until a minimum size of 4 μm. The mesh adaption is made on the basis of the normalized liquid volume fraction gradients (refined for a normalized value above 0.1 and coarsened for a value below 0.03) and the subgrid viscosity ratio (refined for values above 5). This ensures a consistent resolution for the multi-scale turbulent and multi-phase processes. As an example, a plot of the instant subgrid viscosity ratio and the corresponding mesh is presented in Figure 3.

An implicit VOF method is used to track the fuel-air interface using a compressive scheme which is based on a second-order reconstruction method with a slope limiter of 2 [10]. A second- order bounded central differencing scheme is employed for the momentum equation, a body-force-weighted scheme is employed for pressure interpolation and a first-order accurate upwind scheme for density advection together with a first order implicit-in-time scheme to ensure numerical stability. A coupled pressure-velocity solver is applied and a time-step corresponding to Courant number ~1 in the spray hole is chosen to have a good compromise of computational efficiency and sharpness of gas- liquid interfaces.

A comparison for the near-nozzle spray and the corresponding contours of the shear force on the counter bore surface, which can be used for the cleaning of deposits at various fuel pressures is presented in Figure 4. The time averaged surface shear intensity and the impacted surface area, are also provided. At 100 bar fuel pressure, the jet is narrow and only marginally impinges on the counter bore side surface, leading to low wall shear with low impacted area. With increasing fuel pressure, the root of the spray becomes wider in the counter bore and the cleaning force increases significantly. As expected, the spray momentum also increases proportionally to the available fuel pressure increase, which is expected to produce a better air entrainment, air-fuel mixing, and bring more turbulence into the combustion chamber. All these features are advantageous for reducing injector tip coking and PN formation.

The distribution of liquid and vapour over a symmetrical plane of the spray hole is illustrated in Figure 5. It is observed that cavitation in the spray hole is almost suppressed for all fuel pressure levels due to the high convergence taper and a long length-to-diameter ratio, l/d. In contrast, cavitation occurs in the spray core region immediately downstream of the spray hole outlet. This cavitation helps to open and break the continuous liquid jet inside the counter bore, resulting in a spray morphology without a liquid core. This observation is consistent with previously measured and published x-ray spray visualizations [11], but is not visible to the available optical measurement techniques. Increasing fuel pressure increases cavitation, leading to an extended vapour distribution and a wider spray root in the counter bore, which enhances the counter bore wall shear force for deposit cleaning. Considering the high flow efficiency and good atomization performance, this hole configuration will not only show good coking resistance at all pressures, but will also improve engine combustion and PN emissions at high injected quantities.

VOF LES was also applied to investigate the counter bore diameter impact on the spray angle, the near-nozzle spray morphology and the wall shear force for deposit cleaning. The results are demonstrated in Figure 6. In this case, the spray hole geometry and the counter bore length are fixed. A narrow counter bore is observed to limit the root spray dispersion angle while producing a high surface shear force for deposit cleaning. On the other hand, a very wide counter bore can lead to an excessively weak surface cleaning force while bringing little advantage to the near-nozzle spray dispersion angle. A suitable balancing of the counter bore dimension is therefore required to meet the right compromise between the cleaning force and the spray dispersion angle for penetration reduction.

As illustrated in Figure 5, cavitation is predicted in the core region of the initial jet for a long l/d high taper spray hole nozzle, which helps to break up the liquid jet. This is a new observation revealed by the simulation and is consistent with the spray images from the x-ray phase-contrast imaging visualization showing no liquid core and massive portions of gaseous phase in the near-nozzle spray. This result provides an explanation for the good atomization performance experimentally observed for some nozzles using this type of spray hole design, and is evidence of cavitation induced atomization.

In Figure 7, a vortex induced atomization mechanism is presented, which is similar to our previous observation in high-pressure diesel injection [4]. For that purpose, the flow vortex structures in the nozzle and in the near-nozzle region together with the corresponding instantaneous spray morphology are demonstrated. It is observed that the vortex shedding events inside the spray hole are transferred downstream outside of the nozzle, producing pulsating ligament formation events and wavy structures in the near-nozzle spray. A similarity can be observed between the vortex structures outside of the nozzle and the near nozzle spray, both presenting "periodic" structures. These events are schematically highlighted by the black brackets. Therefore, vortices are another significant driving mechanism for primary breakup. The standard spray optimization is based on nozzle geometry adaptation where vortex-controlled atomization mechanisms play an important role. The insight that this vortex formation can purposefully be influenced by the sack volume and spray hole geometry opens up new possibilities for injection nozzle optimization to specifically improve the performance with regard to wall wetting, mixture preparation and PN drift behavior.

High-resolution, high-fidelity LES enhanced with complex physics modelling of multiphase flow and cavitation and with the CLSVOF approach for the interface tracking has been applied to understand the nozzle flow and near-nozzle spray behavior. The results provide a new understanding on how pressure and the nozzle-tip geometry influence the physical processes controlling spray atomization and robustness to coking. The applied methods are helping to drive the development of proved nozzle geometries which significantly reduce the PN and PN drift in current engine applications. The work also showed the benefit of further increasing system pressure, which is currently being pursued at Delphi Technologies, in particular to improve PN emissions and drift, as well as CO and HC emissions.