Multiphysics Simulations: Challenges and Opportunities
Scientific Overview
Increasingly, computational science communities are confronting questions about complex physical and engineered systems characterized by multiple, interacting physical processes that have traditionally been considered separately. Solutions that can accurately capture nonlinear coupling among dynamically relevant phenomena while stepping over fast waves or rapid adjustments must be in place before we can realize the potential of next-generation computational capabilities for high-impact policy and decision-making. In particular, consistent implicit discretizations, both spatial and temporal, are paramount to accurate solutions.
Approaches that preserve stability while accommodating multiple physics at differing time scales are essential. Constraints must be included in a manner consistent with discretization and physics. Implicit nonlinear solvers must be made robust to changing dominance of physical processes as time evolves within a model. In addition, automatic strategies for scaling both variables and equations must be incorporated into linear and nonlinear solvers in order to accommodate variables of differing scales that change over the simulation time. Moreover, the rapid increase in the levels of parallelism and heterogeneity of emerging computer architectures bring to the forefront issues of programming models and the practical realities of working with complex codes developed by interdisciplinary teams.
To address these issues, we recently hosted an ICiS workshop on multiphysics simulations. The workshop participants, from both the applied mathematics and computational science communities, joined forces to explore both the challenges and the opportunities in large-scale, PDE-based multiphysics simulations. Targeted areas included the following:
· Consistent temporal and spatial discretizations for coupling physics, including adaptivity and subcycling
· Stability analysis
· A posteriori error estimation for multiphysics applications
· Coupling approaches that maintain accuracy and efficiency
· Scalable nonlinear solver algorithms
· Software design issues for supporting multiphysics
· Identification of common scientific coupling problems and code design patterns across multiple application domains.
Organizing Committee
David Keyes (KAUST, Columbia University)
Lois Curfman McInnes (Argonne National Laboratory)
Carol Woodward (Lawrence Livermore National Laboratory)
This workshop was held during July 30 - August 6, 2011, in Park City, Utah, under the auspices of the Institute for Computing in Science (ICiS).
A few words about ICiS: An ICiS workshop is intended as an opportunity for intensive but relaxed brainstorming on challenging topics that can take advantage of emerging petascale and exascale computing technology. An ICiS workshop is not a place to promote your latest research or to make the case for increased funding in some area. Come instead with questions and wild ideas. The workshop is a success if participants come away with new thoughts, collaborations, and projects.
Logistics: Please see the ICIS Multiphysics Workshop Website for more details.
Report from this ICiS workshop: Multiphysics Simulations: Challenges and Opportunities, D. E. Keyes, L. C. McInnes, C. Woodward, W. D. Gropp, E. Myra, M. Pernice, J. Bell, J. Brown, A. Clo, J. Connors, E. Constantinescu, D. Estep, K. Evans, C. Farhat, A. Hakim, G. Hammond, G. Hansen, J. Hill, T. Isaac, X. Jiao, K. Jordan, D. Kaushik, E. Kaxiras, A. Koniges, K. Lee, A. Lott, Q. Lu, J. Magerlein, R. Maxwell, M. McCourt, M. Mehl, R. Pawlowski, A. P. Randles, D. Reynolds, B. Riviere, U. Ruede, T. Scheibe, J. Shadid, B. Sheehan, M. Shephard, A. Siegel, B. Smith, X. Tang, C. Wilson, and B. Wohlmuth. International Journal of High Performance Computing Applications, 27(1): 4--83, 2013. Special issue. DOI 10.1177/1094342012468181.
Photo of attendees of the 2011 ICiS Workshop on Multiphysics Simulations: Challenges and Opportunities