Skip to main content
NASA Advanced Supercomputing (NAS) Division World-class modeling, simulation, and supercomputing in support of the NASA mission

On the Potential for High Throughput High Fidelity Simulations

AMS Seminar Series

Speaker: Sanjeeb Bose, Cascade Technologies

Presentation

Seminar Slide Deck (PDF-11MB)

Abstract

It has been established that high fidelity flow simulations (e.g., large eddy simulations) can offer accurate prediction for a number of complex flow phenomena, including separating boundary layers, heat transfer, or flow generated sound. The use of these simulations in engineering design has remained limited due to their relative computational cost and obstacles associated with their setup. We revisit the potential for the use of LES in-design environments that require high throughput on two fronts. First, recent developments in numerical methods for unstructured grids, wall modeling for high Reynolds number flows, and mesh generation for complex geometries have reduced the computational cost for the prediction of salient quantities of interest in practical engineering flows. We quantify this reduction using a few selected benchmark flows, including simulations of high lift aircraft configurations and turbomachinery applications. Second, we assess the computational cost impact due to the use of accelerated computing architectures (e.g., GPUs) both in terms of time and dollar costs using performance data collected from a variety of architectures. The synthesis of these data points suggest that we may be reaching a tipping point for more routine use of high-fidelity simulations in the engineering design process.

Biography

Photo of Sanjeeb Bose
Sanjeeb Bose is the Chief Technology Officer at Cascade Technologies and an Adjunct Professor in the Institute of Computational and Mathematical Engineering at Stanford University. He received his B.S. in Mechanical Engineering from Caltech in 2007 and his Ph.D. in Mechanical Engineering from Stanford University in 2012. His research interests include numerical methods for multi-physics flow environments and subgrid/wall modeling for large eddy simulation. He is also one of the principal developers of Cascade’s charLES flow solver.