Conference Dates

May 16-21, 2010


One of the targets for coal gasification in the near future is capturing 90% of the carbon with less than a 10% increase in cost of electricity. Aggressive goals like this will require innovative gasifier designs to reach the market place quickly, with less risk, and in an economically viable way. Researchers at the National Energy Technology Laboratory (NETL) are collaborating with industry, academia, and other national labs on multiphase computational models like the legacy code MFIX (Multiphase Flow with Interphase eXchange) which can help design, operate, and scale-up clean coal gasifiers to meet the challenges or a carbon constrained world. In fact, NETL has hosted a series of multiphase workshops which has produced a multiphase flow science technology roadmap to achieve the goal “that by 2015 multiphase science based computer simulations play a significant role in the design, operation, and troubleshooting of multiphase flow devices in fossil fuel processing plants”. In this study, we present our experience of porting MFIX, an open source multiphase computational fluid dynamic model, to a high performance computing platform and how the resulting high fidelity simulations are impacting the design of clean coal gasifiers of tomorrow. Inherent to these gasifiers is the various time and length scales which require very high spatial resolution, large number of iterations with small time-steps to resolve and predict the spatiotemporal variations in gas and solids volume fractions, velocities, temperatures with any associated phase change and chemical reactions. These requirements resulted in perhaps the largest known simulations of gas-solids reacting flows, providing detailed information about the gas-solids flow structure, pressure, temperature and species distribution in the gasifier. From a computational science perspective, we found that global communication has to be reduced to achieve scalability to 1000s of cores and hybrid parallelization can yield substantial improvement in time-to-solution when utilizing thousands of multi-core processors.