Challenges
Challenges in multiphysics simulation
Many of today’s important and challenging problems in science and engineering involve multiple, complex, coupled physical systems requiring advanced modeling techniques. These include, for example, fluid-structure interaction (FSI), multiphase/reactive flows, radiation hydrodynamics, and often involve combustion or other sources of energy release. Large-scale modeling and simulation of multiphysics problems using high performance computing (HPC) has become a crucial component of research and development in academia and the national laboratories where these multi-component systems are routinely encountered.
For high-fidelity multiphysics simulations, accurate models must be employed in each subdomain: structures, materials, fluids, combustion, radiation and heat transfer, electronics and electromagnetics, etc. These models and the software implementing them must be coupled and integrated for multiphyiscs simulation capabilities. Development of these models, the software, and simulation capabilities requires a broad range of expertise rarely present in a single, central environment. Individual user groups or divisions often have a significant investment in their existing domain-specific (uniphysics) simulation capability — computational fluid dynamics (CFD), for example — but could well employ a multiphysics simulation capability wherein their existing simulation will be coupled to one or more simulations of other physical domains — computational structural mechanics (CSM), for example.
According to the Council on Competitiveness, industry and private sector utilization of the nation’s HPC resources has consistently lagged behind that of the academic and government sectors. This underutilization is in large part due to significant technical and economic barriers to the development, ownership and availability of HPC-ready software tools for computer-aided science and engineering. These barriers are especially acute for private sector entities needing multiphysics simulation capabilities.Private sector entities presently have very limited choices when deciding how to address a multiphysics simulation requirement. Current solutions are frequently to invest in expertise for a build-your-own solution, or purchase external expertise and capabilities. Both choices incur costly development and refactoring activities that often result in a single,unique, limited-use capability or project-specific solution.
Today, the research community working in this area relies on the few available multiphysics simulation capabilities implemented by a handful of commercially licensed proprietary, niche, (e.g. MSC MD-Nastran, CD-Adapco Star-CD, COMSOL, to name a few), or in-house applications developed at high cost. The difficulty in establishing a viable multiphysics solution with multiple, disparate commercial-off-the-shelf (COTS) software packages is compounded by prohibitively expensive HPC and/or multiprocessor licensing models employed by the independent software vendors (ISV).
The challenges presented by multiphysics capability development and research are not restricted to the private sector. Researchers and analysts in government and academia spend huge amounts of time, effort, and resources in the development and maintenance of multiphysics capabilities - which often consist of systems integrating multiple, disparate simulation applications. In 2011, Argonne's Institute for Computing in Science (ICiS) hosted a workshop, ICiS - Multiphysics Simulation: Challenges and Opportunities, wherein many of the challenges faced by researchers in the field were discussed in detail.
There are currently no existing standards or protocols for tools and interfaces, commercial or otherwise, which support general software and simulation integration and coupling for multiple parallel simulation applications. According to the Council on Competitiveness, these barriers to the efficient production and testing of high fidelity, high performance multiphysics simulation software present a huge bottleneck for research efforts in this and other areas that now or soon will rely heavily on multiphysics simulation.
