Closing the Gap between HPC and AI

Keyvisual image main
Figure representing the general architecture of Relexi. For more details, see the HPE Developer Blog post at the link below. Image courtesy of the Relexi development team.

New software that incorporates reinforcement learning into a computational fluid dynamics solver could make it easier to address common engineering problems.

A multidisciplinary collaboration involving scientists at the Institute of Aerodynamics and Gas Dynamics (IAG) of the University of Stuttgart, Hewlett Packard Enterprise (HPE), and the High-Performance Computing Center Stuttgart (HLRS) has implemented a large-scale simulation framework called Relexi that uses machine learning to optimize turbulence models within computational fluid dynamics simulations. Tested on HLRS’s Hawk supercomputer, this approach could potentially enable the development of other hybrid applications that combine traditional high-performance computing with artificial intelligence methods.

Read the full story at the HPE Developer Blog.