- Spectacular visualization showcase at SC15.
- Argonne Lab team visualization will optimize fuel efficiency.
- Visualization lends insight, revealing unknown information to domain scientists.
You can always count on the scientific visualization showcase at the Supercomputing conference to give you a thrill, and this year did not disappoint. The showcase offered a tasty blend of photorealistic potential and earthy applied science — all with an artistic flair.
One of the presenters was Joseph Insley, data wrangler for the Argonne Leadership Computing Facility (ALCF) at Argonne National Laboratory outside of Chicago. Insley is part of a new breed of computer scientists who hold advanced degrees in art and computer science. He finds creative satisfaction in scientific visualizations, and is rewarded by the exposure to new knowledge brought by scientific domain experts. "For instance, who would’ve thought that you could run a diesel engine on gasoline without a spark plug," Insley says.
Insley’s work with Janardhan Kodavasal, Kevin Harms, Marta Garcia, and Sibendu Som from Argonne, and Xiaochuan Chai from Convergent Science, Inc, is helping move combustion engines into more efficient and environmentally friendly waters. For the SC15 visualization showcase, Insley presented his rendering of the inner workings of an internal combustion engine. Not just any engine though: Their simulation was of a diesel engine burning regular gasoline as fuel.
The objective is to combine the high efficiency of diesel engines with the lower emission of gasoline. To meet this goal, scientists are exploring a spark-plug free combustion mode called Gasoline Compression Ignition (GCI), a technique that starts an engine ignition by compressing the fuel. Simulations were needed to determine the perfect moment to inject fuel; knowing the right time helps optimize performance and thus increase efficiency, lower operating costs, and reduce noxious emissions.
To run the simulations, Argonne researchers used 256 nodes of a high-performance compute cluster, working on problems with about 25 million cells for each computation. Four simulations were employed to confirm the coveted goldilocks point of ignition — a point found but not previously understood in lab experiments since researchers couldn’t see what was occurring inside the engine block.
Insley showed these researchers fuel loss occurring as a piston advances in a cylinder. As the piston moved up, fuel was directed to an area outside the combustion bowl, referred to as the ‘squish.’ Without the visualization, researchers were unaware of wasted fuel.
“To be able to explain to the scientist something they couldn’t see otherwise, that was pretty rewarding,” Insley says. “They could tell there was an optimal spot but they couldn’t tell why.”
This theme of art enabling scientists was a continual refrain from the SC15 showcase, with visualizations repeatedly teaching scientists something they didn’t know about their research.
For Insley, this is the hallmark of a good visualization. “I think a good visualization is anything that’s going to give the scientist insight. Though this particular visualization was not very flashy, it was very informative, and the scientists learned something from it.”
Scientists and data wranglers are starting to see the value of collaboration. "It’s an iterative process, and it's really important to work closely with the scientists," Insley concludes. "Data visualization experts like myself need to learn what scientists are looking for, and the scientists need to talk to us about what it is they are looking for — together we can figure out ways to find it. Maybe there’s some visualization technique that we bring to the game that could help them get insight, so keeping the lines of communication open and working closely with scientists is one of the big keys to making visualization and science successful."
