Seminars

Events Calendar

Fluid Mechanics Seminar

HPC and AI – Accelerating Simulations for Piston Engine and Gas Turbine Flows

Thursday, November 11, 2021
3:30 pm

Hybrid event: Attend in person at ASE 1.126 or virtually via zoom (link sent in email announcement)

The Department of Energy’s National Labs are home to some of the world’s most advanced supercomputing capabilities, consistently holding positions in the list of the top ten fastest in the world.  With exascale computing scheduled to come online at Argonne National Lab soon, the U.S. will maintain its leadership in supercomputing.  These computing resources are being used to solve problems on a “very small scale” (e.g., research on subatomic particles), a “very large scale” (e.g., research on the origins of the universe), and everything in between. This is an opportune moment to tackle some previously computationally intractable research questions. To leverage the hybrid nature of the computing architecture, at Argonne we are developing the next-generation spectral element based highly scalable Nek5000 code for Direct Numerical Simulations (DNS) and wall-resolved Large Eddy Simulations (LES). These high-fidelity simulations when fused with unique experimental measurements from Advanced Photon Source at Argonne are enabling us to develop better understanding and improved models for propulsion applications. Several examples of the above approach will be discussed related to gas turbines, for e.g., liquid and gaseous jets in vitiated cross flow, liquid film cooling, lean blow-out, high-altitude relight, etc. New insights gained from high-fidelity simulations will be discussed and how these insights are being used to improve models that are being implemented in faster running design codes will be shown. Finally, in collaboration with several industrial partners, we are reducing time-to-design by using active learning based machine learning techniques in conjunction with fast running simulations.

Contact  Fabrizio Bisetti, fbisetti@utexas.edu