BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Europe/Stockholm
X-LIC-LOCATION:Europe/Stockholm
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20190719T085744Z
LOCATION:HG F 3
DTSTART;TZID=Europe/Stockholm:20190613T111500
DTEND;TZID=Europe/Stockholm:20190613T114500
UID:submissions.pasc-conference.org_PASC19_sess120_msa319@linklings.com
SUMMARY:Combination of High-Performance Computing and Machine Learning in 
 Computational Physics: Application to Aeronautical Flows
DESCRIPTION:Minisymposium\nComputer Science and Applied Mathematics, Emerg
 ing Application Domains, Chemistry and Materials, Climate and Weather, Phy
 sics, Solid Earth Dynamics, Life Sciences, Engineering\n\nCombination of H
 igh-Performance Computing and Machine Learning in Computational Physics: A
 pplication to Aeronautical Flows\n\nBauerheim\n\nOver the past decades, th
 e numerous improvements of flow solvers combined with the prodigious incre
 ase of CPU resources have made Computational Fluid Dynamics (CFD) an essen
 tial tool to design and optimize aeronautical vehicles. Yet, aeronautical 
 flows usually exhibit turbulence, acoustic waves and shocks, for which the
  large diversity of scales and amplitudes is still challenging CFD methods
 . To step further, breakthrough in CFD codes and flow modeling is necessar
 y, as well as proposing alternative ways to use flow simulations. In that 
 context, machine learning has the potential to renew the way CFD is perfor
 med. This presentation will show how Deep Neural Networks (DNN) can be bui
 lt from flow simulations to create data-driven surrogate models. Technique
 s employed to enforce physical laws (symmetries, conservation laws etc.)&n
 bsp;into DNN will be discussed. It will be illustrated with applications o
 f increasing complexity: (1) the substitution of the expansive Poisson&rsq
 uo;s equation resolution by a DNN in incompressible solvers, and (2) a ful
 l DNN-based surrogate simulator for the propagation of acoustic waves
 . Those examples will highlight the benefits of combining machine learning
  with simulations, as well as new challenges for both the HPC and AI 
 communities.
END:VEVENT
END:VCALENDAR