Over at Argonne National Lab, John Spizzirri writes that GE researchers are using the Mira supercomputer to learn how noise propagates from jet propulsion and wind turbine systems. The team, lead by Umesh Paliath and Giridhar Jothiprasad, is studying the complex behavior of air as it passes through jet exhaust nozzles and over wind turbine blades.
The GE team is using large eddy simulations (LES) to understand and predict turbulent flow from exhaust nozzles and around wind turbine blades. These turbulent flows contain unsteady vortices of many scales that interact with each other over a large range of time scales. By leveraging LES, the GE team can accurately characterize key flow physics of multi-scale turbulent mixing phenomenon that occurs in real industrial applications.
Where it once took two to three months to get the level of predictions for which Paliath’s team is striving, now takes only one to two weeks given the power and precision of ALCF’s 10-petaflops supercomputer, Mira, an IBM Blue Gene/Q. “These are what we call high-fidelity LES, which are very accurate. You are directly predicting the jet engine noise, without modeling the turbulent flow noise sources as in a traditional RANS (Reynolds Averaged Navier Stokes) approach,” he says.