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:20230831T095754Z
LOCATION:Dischma
DTSTART;TZID=Europe/Stockholm:20230626T140000
DTEND;TZID=Europe/Stockholm:20230626T160000
UID:submissions.pasc-conference.org_PASC23_sess162@linklings.com
SUMMARY:MS1A - Machine Learning Techniques for Modeling Under-Resolved Phe
 nomena in Massively Parallel Simulations: Algorithms/Frameworks/Applicatio
 ns (Part 1/2)
DESCRIPTION:Minisymposium\n\nOver the last ten years there has been a prof
 usion of scalable packages, such as TensorFlow, Pytorch, and ONNX, that al
 so run on heterogeneous computing platforms. The ability of these tools to
  ingest massive amounts of training data and make predictions, makes them 
 an obvious choice as tools for scientific machine learning (SciML). In our
  presentations we demonstrate how researchers in diverse areas of scientif
 ic inquiry employ these tools, creatively, to model small scale phenomena 
 in coarse grain simulations, which are then used as predictive tools. Also
 , with the availability of exa-scale computing platforms, it is becoming c
 lear that storing several petabytes of training data, for machine (ML) mod
 els, is not a viable option. We present ongoing research in the area of in
 -situ ML, where the simulation code and ML and deep learning (DL) framewor
 k are run together, to generate, and use, the streaming simulation data, t
 o train the model, and make predictions using coarser simulations. Our pre
 sentations also explore the performance of in-situ machine learning framew
 orks, and the portability of the generated ML models to simulation codes t
 hat are different from the one that was used to train the model.\n\nGraph 
 Neural Networks for Interpretable Data-Based Modeling of Fluid Flows\n\nRe
 duced-order modeling strategies based on neural networks can accelerate tr
 aditional computational fluid dynamics simulations for rapid design optimi
 zation and prediction of a wide range of fluid flows. To realize this visi
 on of improved modeling, key limitations -- namely, incompatibility with u
 ns...\n\n\nShivam Barwey (Argonne National Laboratory), Varun Shankar and 
 Venkatasubramanian Viswanathan (Carnegie Mellon University), and Venkatram
  Vishwanath and Romit Maulik (Argonne National Laboratory)\n--------------
 -------\nDeep-Reinforcement-Learning-Based Drag Reduction in Turbulent Cha
 nnel Flows\n\nWe introduce a reinforcement-learning (RL) environment to de
 sign and benchmark control strategies aimed at reducing drag in turbulent 
 fluid flows enclosed in a channel. The environment provides a framework fo
 r computationally-efficient, parallelized, high-fidelity fluid simulations
 , ready to interfa...\n\n\nRicardo Vinuesa and Luca Guastoni (KTH Royal In
 stitute of Technology), Jean Rabault (Norwegian Meteorological Institute),
  and Hossein Azizpour (KTH Royal Institute of Technology)\n---------------
 ------\nOnline Learning of Sub-Grid Stress Models for Large Eddy Simulatio
 n of Wall Bounded Turbulent Flows\n\nData-driven approaches for the develo
 pment of sub-grid stress (SGS) closure models for large eddy simulation (L
 ES) have been gaining popularity as they offer encouraging results for imp
 roved predictive capacity over traditional models. Due to the time and sca
 le-resolving nature of LES, these models ...\n\n\nRiccardo Balin (Argonne 
 National Laboratory), Aviral Prakash and Basu Parmar (University of Colora
 do Boulder), Filippo Simini (Argonne National Laboratory), and John A. Eva
 ns and Kenneth E. Jansen (University of Colorado Boulder)\n---------------
 ------\nA Scale Separation Approach: A Machine-Learned Surrogate Model for
  an Injector Coupled with Spray Simulations\n\nIn this work, we use machin
 e learning to develop a surrogate model for an expensive portion of a comp
 utational fluid dynamics simulation. In particular, we learn a surrogate m
 odel for fuel injection, predicting the flowfields at the exit of an autom
 otive injector. The output of the surrogate model i...\n\n\nSudeepta Monda
 l, Gina M. Magnotti, Bethany Lusch, Romit Maulik, and Roberto Torelli (Arg
 onne National Laboratory)\n\nDomain: Computer Science, Machine Learning, a
 nd Applied Mathematics 
\n\nSession Chairs: Timothy C Germann (Los Alamos 
 National Laboratory) and Ramesh Balakrishnan (Argonne National Laboratory)
END:VEVENT
END:VCALENDAR