Fluid dynamics experts take gas turbines for a spin on NCSA's TeraGrid cluster, concentrating on features that help keep the blades cool.

Only someone like Danesh Tafti, who's been taking advantage of NCSA's computers for almost 15 years, would consider his work of three years ago starting out small. But consider these details about his models of the airflow around tiny ribs inside gas turbine blades.

Back then, he was using 32 processors per run. Now, he's averaging about 150. Then, he was looking at three ribs. Now, he's up to 10. Then, the models were covered in a grid with two million individual zones. Now, we're talking about 10 or 20 million. Then, the calculations were taking a month on a Pentium-based cluster. Now, he blasts through much larger calculations in a week on NCSA's Itanium-based TeraGrid cluster.

"We've made a really big jump," says Tafti, who is an associate professor of mechanical engineering at Virginia Tech. They aren't big runs for the sake of big runs, though. "We've taken prediction technology in gas turbines a step higher--maybe more than a step. This is a big jump in resolution and quality." He and graduate students Evan Sewall, Aroon Viswanathan, and Samer Abdel-Wahab work with engineers at General Electric and the South Carolina Institute for Energy Studies at Clemson University, which is supported by the Department of Energy. Together, they make sure these improved simulations translate into better turbines in the real world.

"While we do use massively parallel computations of external flows and in other areas of the engine, very little simulation is currently done with these advanced techniques when it comes to internal turbine passages," according to Andy Smith, a mechanical engineer for GE Global Research's fluid mechanics lab in Niskayuna, NY. "But turbine seizure and failure happen at the local level," so the degree of resolution and quality delivered by the team is highly valuable.

Go To Page 2


Access Online | Posted 1-25-2005






Snapshot of the heat transferred from the wall (colors) and turbulence (white iso-surfaces) in a rotating bend.