Held in conjunction with the International Supercomputing Conference (ISC) High Performance 2021, July 2, 2021
Note: This year the workshop will be held virtually
ISC 2021 : International Supercomputing Conference (ISC) CFDML21 proceedings are now available online.
WORKSHOP SCOPE
The combination of computational fluid dynamics (CFD) with machine learning (ML) is a recently emerging research direction with the potential to enable the solution of so far unsolved problems in many application domains. Machine learning is already applied to a number of problems in CFD, such as the identification and extraction of hidden features in large-scale flow computations, finding undetected correlations between dynamical features of the flow, and generating synthetic CFD datasets through high-fidelity simulations. These approaches are forming a paradigm shift to change the focus of CFD from time-consuming feature detection to in-depth examinations of such features, and enabling deeper insight into the physics involved in complex natural processes.
The workshop is designed to stimulate this research by providing a venue to exchange new ideas and discuss challenges and opportunities as well as expose this newly emerging field to a broader research community. It brings together researchers and industrial practitioners working on any aspects of applying ML to the CFD and related domains, in order to provide a venue for discussion, knowledge transfer, and collaboration among the research community.
WORKSHOP FORMAT
Live Keynote will follow by pre-recorded video presentations. All speakers will be available for a virtual Q&A at their designated times. Link to connect to the event will be posted here shortly.
PROGRAM
The list of activities occurring on July 2, 2021
Live Keynote: Discovering hidden fluid mechanics using PINNs and DeepONets — George Karniadakis, Brown University
14:00-14:30 (Central Europe Time) / 08:00-08:30 (U.S. Eastern Time)
Session chair: Eloisa Bentivegna, IBM Research Europe, UK
Keynote Q&A
14:30-14:40 (Central Europe Time) / 08:30-08:40 (U.S. Eastern Time)
Session 1: Fluid mechanics with turbulence, reduced models, and machine learning (Q&A with the authors)
14:40-16:00 (Central Europe Time) / 08:40-10:00 (U.S. Eastern Time)
Session chair: Ashley Scillitoe, The Alan Turing Institute, UK
Nonlinear mode decomposition and reduced-order modeling for three-dimensional cylinder flow by distributed learning on Fugaku — Kazuto Ando, Keiji Onishi, Rahul Bale, Makoto Tsubokura, Akiyoshi Kuroda and Kazuo Minami
Watch the presentation | View the presentation slides (PDF)
Reconstruction of mixture fraction statistics of turbulent jet flows with deep learning — Michael Gauding and Mathis Bode
Watch the presentation | View the presentation slides (PDF)
Reservoir computing in reduced order modeling for chaotic dynamical systems — Alberto Costa Nogueira Junior, Felipe de Castro Teixeira Carvalho, João Lucas de Sousa Almeida, Andres Codas, Eloisa Bentivegna and Campbell D Watson
Watch the presentation | View the presentation slides (PDF)
A data-driven wall-shear stress model for LES using gradient boosting decision trees — Sarath Radhakrishnan, Lawrence Adu-Gyamfi, Arnau Miró, Bernat Font and Joan Calafell
Watch the presentation | View the presentation slides (PDF)
Session 2: Novel methods development in machine learning and fluid simulation (Q&A with the authors)
16:00-17:00 (Central Europe Time) / 10:00-11:00 (U.S. Eastern Time)
Session chair: Alberto Costa Nogueira Junior, IBM Research Brazil, Brazil
Lettuce: PyTorch-based lattice Boltzmann framework — Mario C. Bedrunka, Dominik Wilde, Martin Kliemank, Dirk Reith, Holger Foysi and Andreas Krämer
Watch the presentation | View the presentation slides (PDF)
Novel DNNs for stiff ODEs with applications to chemically reacting flows — Thomas Brown, Harbir Antil, Rainald Lohner, Fumiya Togashi and Deepanshu Verma
Watch the presentation | View the presentation slides (PDF)
Machine-learning-based control of perturbed and heated channel flows — Mario Rüttgers, Moritz Waldmann, Wolfgang Schröder and Andreas Lintermann
Watch the presentation | View the presentation slides (PDF)
Session 3: Confluence of machine learning and fluid simulation applications (Q&A with the authors)
17:00-18:00 (Central Europe Time) / 11:00-12:00 (U.S. Eastern Time)
Session chair: Charalambos Chrysostomou, The Cyprus Institute
Physics informed machine learning for fluid-structure interaction — Qiming Zhu, Jinhui Yan
Watch the presentation | View the presentation slides (PDF)
Film cooling prediction and optimization based on deconvolution neural network — Yaning Wang, Shirui Luo, Wen Wang, Guocheng Tao, Xinshuai Zhang and Jiahuan Cui
Watch the presentation | View the presentation slides (PDF)
Turbomachinery blade surrogate modeling using deep learning — Shirui Luo, Jiahuan Cui, Vignesh Sella, Jian Liu, Seid Koric and Volodymyr Kindratenko
Watch the presentation | View the presentation slides (PDF)
WORKSHOP CO-CHAIRS AND PROGRAM COMMITTEE
- Volodymyr Kindratenko, National Center for Supercomputing Applications, USA
- Andreas Lintermann, Jülich Supercomputing Centre, Forschungszentrum Jülich, Germany
- Charalambos Chrysostomou, The Cyprus Institute, Cyprus
- Jiahuan Cui, Zhejiang University, China
- Eloisa Bentivegna, IBM Research Europe, UK
- Ashley Scillitoe, The Alan Turing Institute, UK
- Morris Riedel, University of Iceland, Iceland
- Jenia Jitsev, Jülich Supercomputing Centre, Forschungszentrum Jülich, Germany
- Seid Koric, National Center for Supercomputing Applications, USA
- Shirui Luo, National Center for Supercomputing Applications, USA
- Alberto Costa Nogueira Junior, IBM Research Brazil, Brazil
- Jeyan Thiyagalingam, Science and Technology Facilities Council, UK
- Nikos Savva, The Cyprus Institute, Cyprus