Scientist simulates complex diesel injection mist with supercomputer
11 April 2016
The ability to study complex interactions inside a diesel engine gives army scientists a new perspective that may help design more fuel efficient and effective engines.
Using supercomputers, Army scientist Dr. Luis Bravo led efforts to create simulations highlighting the first year of his project, allowing them to investigate the jet fuel spray breakup process with microsecond time fidelity and sub-millimetre resolution, while generating petabytes of data.
"The simulation captures the spray primary atomisation in the near nozzle dense region using a realistic complex diesel injector," Bravo said. "As diesel sprays are mostly optically thick, the measurements in the dense region are restricted, and the present simulation provides a novel diagnostic method."
As a research mechanical engineer with the U.S. Army Research Laboratory, Bravo is the principal investigator for a five-year Department of Defence Frontier project called, "Peta-scale High-fidelity Simulation of Atomisation and Spray/Wall Interactions and High-temperature and Pressure Conditions."
Frontier projects receive DOD funding and involve the teaming of High Performance Computing Modernisation Program expertise and supercomputers with DOD laboratory and test centre domain expertise.
Bravo said researchers conduct concurrent studies in well-controlled laboratory facilities using high-speed video imaging; however this is the first time in the Army that we've created a high-resolution 3D simulation of the diesel injection process with this level of precision. This precision requires access to distributed memory parallel computers and millions of CPU hours, only possible with Frontier resources.
When the team started, there was not a visualisation functionality to export the geometry of the datasets to a format that could be used efficiently in their specialised software program, Bravo said.
Computer visualisation scientists from the Data Analysis and Assessment Centre, known as DAAC, helped develop a novel pipeline to work around the issue, exporting the data in one format and then converting it to another. The results impressed Bravo.
"It created much higher quality renderings of the data revealing important process insights," he said. "Using this tool, our animators were able to smooth out the normals and use ray tracing technology on the surface of the spray geometry to capture the realistic attributes of liquid."
The researchers used a Cray XE6 Garnet Supercomputer from the DOD Supercomputing Resource Centre located at the U.S. Army Engineer Research and Development Centre in Vicksburg, Mississippi.
"The simulation used 80 million cell nodes, distributed over 5,000 processors, generating petabytes of information," Bravo said. "The simulations allow us to interrogate the spray process in ways that were not previously possible, while revealing critical insights about its behaviour. The bottom line is that the enhanced simulation and visualisation capability, enabled by our DAAC collaborators, is helping us solve critical Army bottlenecks and accelerating the development of next generation engine platforms."
More information can be found on the US Army Research Laboratory website.