The supercomputers allow scientists like Arthur and his colleague Michal Osusky, a lead thermosciences engineer at GE Research, to precisely simulate how heat flows inside jet engines and gas turbines working inside power plants. The work may seem like an arcane corner of engineering, but the benefits can be huge. Even tiny improvements in turbine efficiency yield enormous savings in fuel consumption and cut emissions.
But getting an accurate grasp of what heat is doing inside an engine is a tough task. It’s like trying to discern patterns in the swirls of hot air shimmering over the highway in the summer. That’s why GE Research recently partnered with the U.S. Department of Energy’s Office of Science and gained access to the Summit supercomputer at Oak Ridge National Laboratory in Tennessee. “It opens up a whole new area of predictions we never would have been able to do,” says Osusky, who is leading a project seeking to improve the design of high-pressure turbines. “It wasn’t that the methodology wasn’t there, it’s more that the computing resources weren’t there at the necessary scale.”
The supercomputer will allow the researchers to create realistic simulations of turbulent heat flows coursing through the engine, allowing them to effectively observe the “large-scale violent freestream turbulence” in real time. Older computer models of turbines couldn’t process data fast enough to handle the complexity of the chaotic swirls of heat, and instead had to settle for using estimates based on average temperatures in different parts of the engine.
Engineers like to run their engines as hot as possible to maximize efficiency, but those crude averages sometimes disguised the complicated flows of heat, which might intensify enough in one spot and could start damaging parts inside the engine.
In their application to use the supercomputer, the GE engineers estimated that it will allow them to identify opportunities to increase “turbine aerothermal efficiency by 2% to 4% and extend hot-gas-path durability,” which they estimate will end up combined cycle efficiency gains of 0.4% to 0.8%. If those savings sound small, they scale up quickly when multiplied by the size of the global oil and gas industries.
While the early simulations may lead to some unexpected design breakthroughs, both Osusky and Arthur say the supercomputers will certainly create a “gold standard data set” that will allow scientists to analyze previously hidden problems and illuminate potential solutions. Arthur says that “such high-value data is becoming even more essential as training sets as we explore novel application of machine learning in science and engineering.”
Adds Osusky: “Simulating it more accurately you get better efficiency, better performance, better cost. It’s really a win-win-win situation, simply from understanding the flow of heat better.”