MFIX-Exa: Using the Next Generation of Supercomputers for Multiphase Flow Research

23 Aug 2017 - The US National energy Laboratory (NETL) is collaborating with Lawrence Berkley National Laboratory (LBNL) and the University of Colorado Boulder to develop MFIX-Exa, the next generation of NETL's internationally acclaimed Multiphase Flow with Interphase Exchanges (MFIX) - a suite of specialized computational fluid dynamic codes (CFDs) that help researchers study the simultaneous flow of gases, liquids, or solid materials.

For example, a variety of flow regimes are encountered as particles traverse a chemical looping reactor (CLR) - from a relatively dilute riser to a dense bubbly particle bed. CFD helps researchers gain a greater understanding of the gas-solids interactions and device performance.

The work is ongoing because computational science evolves rapidly. Faster and more complex computing systems appear every day. Supercomputers that broke ground just a few years ago will soon be replaced by systems operating at the exascale (1018calculations each second) - a thousand-fold increase. This surge of processing power will open the doors to scientific breakthroughs across the scientific community, including fossil energy research into advanced power generation systems that are more efficient and release less carbon dioxide into the atmosphere.

Reactors based on multiphase flow are difficult to design, scale up, and operate. Physical experiments to guide this process are increasingly expensive and sometimes impossible to perform. So, it is crucial to develop a high-fidelity computational tool that can accurately simulate systems with multiphase flows. Such a tool could drastically shorten the time and cost associated with developing new energy conversion technologies.

Currently, MFIX-DEM (the discrete element model) can track millions of particles, accounting for the mass, momentum, and energy transfer between the particles and gas phase. The MFIX-Exa project will advance this capability so that billions of reacting particles can be tracked in a full-loop reactor, making it feasible to simulate a pilot-scale chemical looping reactor (CLR) such as the 50kW CLR at the Morgantown NETL site. This kind of simulation will allow researchers to affect the design of CLR systems early enough in the development process to help control costs and reduce risks.

NETL Researcher Dr. Jordan Musser explained the goals of the programme.

"It's well known that traditional scale-up methods do not work well for multiphase flow reactors, so if we can take advantage of the high-performance exascale computing systems, we can outperform existing MFIX code by orders of magnitude, creating a step change in the fidelity of fossil fuel reactor simulations", he said. "These new models will result in considerable savings in development and cost and could be applied to other multiphase systems as well, including resource extraction and fuel utilization."

The project has both short- and long-term goals. As a 3-year target, the team proposes to use MFIX-Exa to model a cold-flow CLR like NETL's 50kW unit with 107particles. The 10-year challenge will then be to scale the NETL CLR simulation to a small pilot 1MWe system containing 1011 particles with 10 minutes of operation in less than 24 wall clock hours using an exascale system. Deployment of MFIX-Exa will begin by integrating LBNL's open source AMReX infrastructure - a well-tested and widely used software framework.

The collaboration features a team with more than 90 years of experience in high-performance computing, with over 60 years of experience in multiphase modeling and developing large-scale multiphysics applications. The project is funded in part by the Exascale Computing Project (ECP), a collaborative effort of two U.S. Department of Energy organisations - the Office of Science and the National Nuclear Security Administration. ECP was established to develop a capable exascale ecosystem, encompassing applications, system software, hardware technologies and architectures, and workforce development to meet the scientific and national security mission needs of DOE in the mid-2020s timeframe.

Ad Emmen