STAR-CD fueling injector primary breakup
Jarrod Sinclair and Chris Seeling, Victorian Partnership for Advanced Computing, Centre for Computational Prototyping, Australia. Peter Murdoch, GM Holden, Australia.

Automotive engines of the future must meet increasingly stringent emission and fuel consumption demands. Efficiency improvements to the fuel injection system is one key activity to achieve this. However, a current lack of fundamental knowledge of the injection process is limiting the potential for future injector designs. In particular, a clearer understanding of the physical processes involved as liquid fuel exits a fuel injector and fragments into droplets is desirable. With this knowledge one could more effectively design the ideal combustible air/fuel cloud within the combustion system, and therefore significantly improve engine performance using currently available technologies.

Breakup or atomization are the terms used to describe the formation of droplets from a continuous stream of liquid or from larger droplets. Primary breakup is the initial phase of liquid fragmentation to form these larger droplets, ligaments or network structures. A liquid stream or sheet undergoes breakup by various modes that are classified by the Reynolds, Weber and Ohnesorge dimensionless ratios. The breakup of a high speed jet is more likely to fall within the turbulent atomization regime, where the exact physical processes leading to fragmentation are currently unclear. This regime was the subject of this work.

Experimental spray test rigs and measuring probes can be used to investigate primary breakup at the nozzle tip, however this is quite demanding. A resolution of the order of one micro-meter, and a time-scale of the order of one nano-second are required to capture the physics. In addition, it can be relatively costly and difficult to vary the system parameters in order to adequately map the injector response. Therefore, STAR-CD was used in this work to numerically investigate the primary breakup process.

A series of 3-dimensional multiphase analyses were conducted using STAR-CD on an angular segment of an annulus injector nozzle. The flow was defined by a Reynolds number of 3000, a Weber number of 3000, and an Ohnesorge number of 0.02. The physics of cavitation, important in many injector flows, was not accounted for as it does not play a major role for the type of injector under investigation.

The volume-of-fluid (VOF) two-phase approach was used to model the free surface between air and fuel. In STAR-CD, the CICSAM advection scheme was used to minimize artificial diffusion of the interface between the two immiscible fluids. To ensure solution stability and accuracy, a target Courant number of 0.3 was enforced throughout the entire simulation runtime. This corresponded to a time-step size of approximately 1 nano-second. Surface tension between air and fuel was accounted for using the continuum surface force (CSF) approach.

As breakup considered in this work is classified in the turbulent atomization regime, capturing turbulence accurately is critical. Therefore, a fine resolution Large-Eddy Simulation (LES) approach was adopted. This ensured that turbulent structures larger than a few cells were directly simulated by STAR-CD, while sub-grid structures were modeled using the one-equation k-l sub-grid model. Particular care was taken to ensure that the y+ values were less than one next to the internal channel walls. It is considered that this turbulence treatment is more appropriate than Reynolds averaged (RANS) models in predicting the separation of the boundary layer flow from the injector nozzle outlet and its transition into the shear layer formed further downstream.

The computational domain consisted of the nozzle outlet zone with an opening height of 20 micro-meters. The domain was extended about 1mm upstream on the inside of the nozzle channel, and 4mm downstream of the exit plane. From experimental and analytical results, this domain was considered to be large enough to contain the primary breakup region. A submicron mesh size was created at the nozzle outlet and core primary breakup region. Several layers of mesh refinement were used to reduce the cell count away from the region of interest without enforcing unrealistic boundary conditions. A mesh containing approximately 10 million cells was used for each simulation, highlighting the need for high-performance cluster computing.

Simulations were run on a 16-CPU Intel Itanium II cluster running HP-UX, taking on average five days to complete for each. To obtain time-averaged values, and to wash out the initial conditions, the total simulation time was chosen to be five times the fuel residence time in the domain. Waves form on the surface of the fuel after it has traveled some distance from the nozzle outlet plane. These waves grow in amplitude while maintaining a constant wavelength. A transition occurs as the waves form more of a localized 3- dimensional structure, most likely dominated by surface tension effects. Large-scale breakup of the sheet takes place quite rapidly at this transition point. It can be seen that a finer network structure comprising of transverse ligaments begins to form further downstream. This increase in fuel surface area causes the formation of pre-atomization droplets which would eventually compose a finely atomized spray cloud beyond the domain of the current simulations. The root cause of the initial instability, and a characterization of the turbulent structures on the amount of breakup are currently under investigation.

Conclusion

From this work, STAR-CD has shown to be a valuable simulation tool in capturing the critical phases of fuel injector primary breakup, and has given key insights into the governing physical processes. The solver provided a near-linear parallel speed-up, giving accelerated throughput on such a large model. The techniques developed using STAR-CD in this work are showing promise as a design tool for future injectors.

Breaking up isn’t hard to do…
Marco Buonfiiglioli, CD-adapco UK

Always in tune with our customers, CD-adapco has independently been devoting significant effort to the study of primary break up and atomization, using a similar VOF and LES based approach.

The CD-adapco study concentrated on analyzing the effect of inlet turbulence on spray formation, a phenomenon that is very difficult to measure empirically.

Since the flow at the inlet is not fully developed, a “synthetic turbulence” condition was applied (developed in collaboration with the University of Manchester). By varying the inflow velocities in both time and space, the vortex-like method generates synthetic but realistic turbulent structures at inflow. This approach allowed the use of a relatively small nozzle length (just three diameters), removing the need to use a computationally expensive long inlet duct.

The intensity of the synthetic turbulence at inflow was systematically reduced from an original highly turbulent level to almost zero.

Comparisons with the limited experimental data available indicated a good agreement in the prediction of both break-up length and spray angle. Figure 3 shows instantaneous iso-surfaces for VOF concentrations of both 1% and 99% heavy fluid, illustrating both the core and the external spray boundaries.

Figures 4, 5 and 6 show a section through the jet for high, medium and low inlet turbulence levels and clearly show the extent to which the break-up length and cone angle are influenced by the upstream condition.

An impressive aspect of this project is the surprisingly short run times required in order to produce impressive results, as both LES and VOF are often regarded to be highly computationally expensive. Each of the simulations was run until the spray reached a statistical steady state (from a quiescent initial condition). Using 4 2.8GHz processors of a Linux cluster was possible to compute 4ms of simulation in a single day.