Seminar - Sorbonne Institute for Computational Science and Data

Title: Cost-Effective High-Fidelity CFD Simulations Towards Physics AI

Speaker:

Cetin C. Kiris, PhD.

Volcano Platforms Inc., Palo Alto, CA, USA  (retired from NASA Ames Research Center, Computational Aerosciences Branch)

 

 Abstract:

As Computational Fluid Dynamics (CFD) plays an increasingly important role in design and optimization, there's a significant push to bring high-fidelity numerical modeling to a wider range of industrial applications. We demonstrate how combining automated Cartesian octree grid generation with a precise high-Reynolds number viscous immersed boundary treatment allows for the representation of complex geometries while fully utilizing the computational power of modern GPUs. This approach makes accurate Large Eddy Simulations (LES) achievable with considerably less computational cost. 
 
This talk will present examples where detailed LES investigations of engineering problems involving separated flow aerodynamics (transonic buffet, high-lift, icing and automotive aerodynamics) can be performed on very modest compute resources utilizing 2 to 8 general-purpose GPU cards such the Nvidia RTX-4090 and L40S. This represents a major breakthrough in CFD where LES utilization to study a wide range of off-design characteristics no-longer requires access to DOE-class supercomputers. The last example will be a new open-source high-fidelity dataset for Machine Learning (ML) containing 355 geometric variants of the Windsor body in automotive aerodynamics, to help the development and testing of ML surrogate models for external automotive aerodynamics. 

BIO:

After a distinguished 35-year journey at NASA, Dr. Cetin Kiris is now the CEO and founder of an early-stage technology startup that focuses on providing physics-based simulation capabilities to accelerate digital transformation of physical prototyping to predictive, fast, and cost-effective computing. Prior to his departure from NASA, Dr. Kiris served as the branch chief of the Computational Aerosciences Branch at NASA Ames Research Center. He initiated and orchestrated the development of LAVA, a computational framework for Launch, Ascent, and Vehicle Aerodynamics. He received his master's degree and Ph.D. in Aeronautics and Astronautics from Stanford University. He has published over 200 technical papers and co-authored a book on numerical simulations of incompressible flows. He has received numerous honors and awards, including NASA Outstanding Leadership Medal; NASA Exceptional Achievement Medal; NASA Software of the Year Award; NASA Commercial Invention of the Year Award for co-developing the NASA-DeBakey Ventricular Assist Device.

 

OPEN TO ALL