Simulation of High-Pressure Primary Atomization using a Complex Diesel Injector

Principle Investigator: Dr. Luis Bravo
HPC System: Garnet
Sponsor: ARL, Vehicle Technology Directorate


Gray Eagle Drone Conceptual Animation

Case Four

Combination View: Cases 1-4

Background Details:

The Army is interested in heavy fuel engines for platforms that exclusively rely on direct injection fuel delivery systems. To meet and exceed the range and operational requirements, the engines need to provide significant advancements in fuel flexibility and fuel conversion efficiencies. In propulsion systems, energy conversion by combustion exclusively relies on the use of jet fuel in the compressed liquid form. The initial step in the energy conversion process is the atomization, or disintegration of the coherent liquid core, which has a significant influence on the droplet-size distribution and fuel conversion efficiency. In combusting scenarios, the liquid fuel has to be fully atomized, evaporated and mixed with the carrier gas-phase environment. Hence, the interplay of liquid atomization and spray vaporization is of paramount importance, for it governs how much energy and fuel is supplied to the flame. Despite the relevance of the spray atomization process, its modelling is still among the weakest parts of practical engineering simulation models. This is due in part to the inaccessibility of the optically dense region and the steep resolution demands of the interface scales. With the steady progress in computing power and advances in numerical algorithms, first-principle based simulations of atomization processes are emerging today as a viable tool to study and predict the fuel spray and combustion behavior.

Simulation Details:

The simulations presented here were performed using a direct numerical simulation (DNS) approach to investigate the primary breakup process of diesel injection using large distributed memory parallel computers [1]. The simulations calculate the internal and external multiphase flows across a diesel injector orifice using an interface-tracking approach (for the liquid fuel surface) to capture the primary atomization behavior in the spray dense region. As diesel sprays are mostly optically thick, the measurements in the dense region are limited, thus the present computational study provides a novel science-based diagnostic method [1-2].

The high-pressure common-rail diesel injector geometry studied accounts for the complex internal features including the mini-sac region (0.2mm3), needle valve position, and converging nozzle with 90 μm exit orifice. The injector was scanned at Argonne National Laboratory using x-ray phase contrast imaging techniques with a 5 μm resolution. The fuel pressure was specified at 150 bar imposing a Reynolds and Weber numbers Re = 25,573 and We = 125,806. The physical properties were based on a fuel temperature at 298K as an approximation to the water-cooled injector jacket temperature in the laboratory. The liquid fuel properties employed are density ρ = 686 kg/m3, viscosity μ = 0.475 mPa.s, and surface tension σ = 18.6 mN/m. To specify diesel type conditions, the chamber gas density is set to ρ = 22.8 kg/m3, by using 100% filled gaseous nitrogen at 303K and a backpressure at 20 bar. The simulation initializes with a liquid filled injector and prescribes a rate-of-injection profile with bulk inflow velocities based from nozzle flow measurements. In addition a stochastic turbulent inflow generation condition is employed to capture the transition to internal flow turbulence dynamics. The simulations provide detailed diagnostics in the optically dense region, 0 < x/d < 30 jet diameters, seamlessly calculating droplet statistics, and fuel/air mixture formation processes. The results provide insights and new understanding of the breakup phenomena and droplet formation process for pulsed-injection diesel sprays.

The research was completed on the Cray-XE6 Garnet DoD Supercomputing Resource Center located at the U.S. Army Engineer Research and Development Center (ERDC). Each simulation used 80 million cell nodes, distributed over 5.000 processors and run for 340 wall-clock hours.

[1] Bravo, L., Kim, D., Tess, M., Kurman, M., Ham, F., Kweon, C., High Resolution Numerical Simulations of Primary Atomization in Diesel Sprays with Single Component Reference Fuels, 27th Annual Conference on Liquid Atomization and Spray Systems, Rayleigh, NC, 2015.

[2] Bravo, L., Ivey, CB., Kim, D., Bose, ST., Highfidelity simulation of atomization in diesel engine sprays, Proceedings of the Standord turbulence summer program, pp. 89, 2014.

Visualization Details:

The simulation results were visualized using EnSight by Dr. Luis Bravo, the principal investigator of the Army FRONTIER project. The results of four simulation runs were generated and submitted to the DAAC for higher quality visualizations. All four runs consists of a grid cell size of 77 million, and the total CPU time ran for ~1.5 months. The number of timesteps for each run ranged from 324 to 526 timesteps.

During this time, there was not yet a functionality in EnSight to export the geometry of the datasets to a format that could be used efficiently with Adobe's 3D Studio Max. A pipeline was developed to work around the issue, whereby the data would be exported from EnSight in one format and then converted to another. After the diesel spray geometry was visualized, it was then exported out of EnSight as ASCII formatted VRML files. A custom script was then developed from within the DAAC to convert the VRML files to PLY formatted files.

When the conversion was completed, the PLY files were brought into 3D Studio Max to create higher quality renderings of the data. Using this tool, our animators were able to smooth out the normals and used ray tracing on the surface of the spray geometry to mimic the realistic attributes of liquid.

An animation was created for each of the four datasets provided to the DAAC. Another animation was created of all four datasets progressing side by side for comparison. A final conceptual animation was created to be used for a promotional video for the ARL DSRC. A few of these animations can be viewed from this page.