SciDAC Institutes SciDAC Partnerships Basic Energy Sciences Biological and Environmental Research Fusion Energy Sciences High Energy Physics Nuclear Physics Participating Orgs SciDAC Overview (2000-2011) Logos DOE Office of Advanced Scientific Computing Research (ASCR)

The Applied Partial Differential Equations Center (APDEC): Combustion Applications

Investigators: J. B. Bell, M. S. Day and J. F. Grcar.

Combustion is one of the major science applications areas of the APDEC ISIC Center. The focus of our combustion work is on developing and applying tools for high fidelity simulation of flames with detailed chemistry and transport. The simulation software provides an end-to-end capability that includes data analysis tools for interpreting simulation data.

Although combustion is one of our oldest technologies, it remains a critical mission area for the Department of Energy. Combustion of fossil fuels continues to provide most of the energy required for transportation and for stationary power generation. In spite of its fundamental scientific and technological importance our knowledge of the combustion process is surprisingly incomplete. Laboratory measurements are difficult to make and often limited in the level of detail they can provide about a flame. The multiple-scale character of combustion problems leads to daunting computational requirements for traditional simulation methodologies.

The adaptive methodologies being developed in APDEC are ideally suited to the multi-scale, multi-physics simulations typified by combustion. We have developed a simulation capability based on a low Mach number formulation for modeling time-dependent combustion phenomena in two and three dimensions with detailed chemistry and transport. The low Mach number formulation allows the method to take time step that are two orders of magnitude larger than standard approaches based on a compressible formulation. The algorithm is implemented in an adaptive projection framework and combines high-resolution finite difference methods with local adaptive mesh refinement. The parallel implementation of this methodology is based on domain decomposition; however, special care is required in load balancing because of the heterogeneous character of the underlying physics.

In conjunction with the development of the simulation capability, we have also developed a data analysis framework based on the scripting language Python that allows us to apply a variety of mathematical tools to our simulation data to answer key questions about the basic flame physics.

Science Applications

In one application of this methodology we have performed a simulation of a turbulent methane flame sheet using a detailed chemical mechanism and a mixture model for preferential species diffusion. Methane chemistry and transport is modeled using the DRM-19 (21-species, 84-reaction) mechanism derived from GRIMech-1.2 along with its associated thermodynamics and transport databases. An image of the flame sheet is shown in Fig. 1.

Fig. 1. Turbulent methane flame sheet

One of the major scientific goals in this type of simulation is to understand the mechanisms by which turbulent modulates the flame chemistry. In Fig. 2 we show mole fractions of two of the species in a vertical slice through the flame illustrating the variability of the solution along the flame.

Fig. 2. Mole fraction of CH3O and C2H2 in a vertical slice through the turbulent flame

Joint probability distributions that illustrate the relationship between chemical variability and flame curvature are shown in Fig. 3.

Fig. 3. Joint probability distributions of CH3O and C2H2 with curvature

Another application of our simulation methodology has been to the formation of pollutants in diffusion flames. We have performed detailed simulations of ammonia enriched methane diffusion flames using a chemical mechanism with 65 species and 447 reactions that includes the necessary chemical kinetics to accurately model NOx formation. These simulations help to understand the dependence of emissions on fuel bound nitrogen. In Fig. 4, we show a comparison of NO mole fraction measured using Planar Laser Induced Fluorescence to simulations results for various levels of ammonia enrichment.

Fig. 4. NO mole fraction comparison of experiment (top) and computation (bottom) for varying levels of ammonia enrichment

Future Plans

Recent advances in both algorithm technology and computer hardware have brought us to the point where it will be possible to simulate laboratory-scale turbulent flames with detailed chemistry and transport. Our plan, in conjunction with experimentalist in BES, is to simulate premixed turbulent methane flames and perform detailed comparisons between the computational results and the laboratory measurements. We anticipate using 3.0M cpu-node hours for FY2003 and 5.0M for FY2004.

For further information on this subject contact:
Dr. John Bell
Lawrence Berkeley National Laboratory
Phone: 510-486-5391
jbbell@lbl.gov

back to project page