APDEC - Combustion Applications
Goals
Combustion is one of the cornerstone applications for the APDEC ISIC. Our goal in this application is to develop and apply new simulations methodologies on high-performance parallel architecture. The new tools will be used to understand the interplay between chemistry and fluid dynamics in flames. The focus of our work is on studies combining detailed chemistry and transport where the models for the fluid-dynamics and fluid/chemistry interactions do not incorporate approximations for so-called “subgrid dynamics”. Our core methodology combines a novel low Mach number projection formulation with dynamically adaptive grid technology. The combination provides a substantial improvement in computational efficiency over standard compressible DNS methods.
Accomplishments
Our research in combustion thus far has focused on two target investigations: pollutant production in steady ammonia-seeded methane diffusion flames, and turbulence chemistry interactions in wrinkled premixed methane flames.
We are investigating ammonia-seeded diffusion flames in
collaboration with researchers at the
We have developed a new approach that more closely intertwines simulation and experimental data interpretation. Using a two-dimensional axisymmetric model, we compute steady diffusion flame solutions corresponding to the experimental setup. After validating the computational results against the temperature profiles inferred from the experiment using the two-line approach discussed, we use the simulation results, combined with quantum-dynamical quenching models to generate numerical LIF images. Across the range of experimental parameters, these numerical LIF images showed exceptional agreement with the raw LIF data from the experiment. A comparison of the experimentally-measured and the synthetic computational LIF is presented in Figure 1.
Figure 1: NO A-X(0,2) excitation LIF images obtained (top) from measurement and (bottom) by synthetically processing the results of the flame simulation, for different ammonia seeding concentrations. The experimental data and the synthetic LIF intensities are shown on identical scales.
Additionally, using simulated chemical distributions directly in the quench calculations, we demonstrated very good agreement between the computed and experimentally inferred NO concentrations. This work was presented at the 2002 Combustion Symposium and is documented in Bell et al. [1].
In our second combustion application, we applied our three-dimensional, high-resolution numerical algorithm technology toward understanding the detailed response of a lean premixed methane flame sheet to turbulence in the unburned mixture. These simulations are the largest of their type ever attempted, and were achievable through our combined use of locally-adaptive grid methods and the low Mach number formulation. The flame chemistry for this case is modeled with a 20 species, 84 reaction subset of the GRIMech-1.2 methane chemistry mechanism. The calculations are carried out over the 8x8x16 mm domain on a hierarchical grid structure with an effective resolution of 256x256x512 cells. We considered two different cases corresponding to different levels of turbulent intensity. Flame sheet images from these simulations are depicted in Figure 2.
(a) 
(b) 
Figure 2:
Volume rendered image showing surface of maximum heat release for the weak (a)
and strong (b) turbulence cases.
Simulation results confirm that wrinkling due to turbulence is the dominant factor leading to an increased effective flame propagation speed. Under the turbulence regimes studied thus far, the increase in speed is predominately attributable to the increased flame surface area. This suggests that the time-dependent stretch and flame curvature have little aggregate effect on the overall flame structure in the cases investigated.
However, when we examine local details of the flame, we find a different picture. Even for low turbulence levels, the heat release in the flame correlates strongly with flame curvature. In fact, the strength of many of the reactions in the system are correlated to flame curvature, an observation which is likely related to focusing and defocusing preferential hydrogen diffusion effects. We found that an additional effect of the turbulence was to move apart the regions of production and destruction for a subset of the chemical species, particularly in regions of negative curvature. As a result, the residence times, and computed molar concentrations of these species were correlated to the curvature as well. This work was presented at the 2002 Combustion Symposium and documented in Bell et al. [2].
We are currently extending that work on turbulent premixed combustion to modeling of a laboratory-scale premixed turbulent flame. The configuration for the experiment we are considering consists of a thin rod across the center of a nozzle with a partial obstruction ten centimeters upstream of the rod to generate turbulence. The presence of the rod creates a local stagnation in the flow which leads to the formation of a V-flame. Simulations for this configuration specify flow at the exit of the nozzle using a characterization determined from the experimental data and use that data as boundary conditions for our low Mach number combustion algorithm. As in our earlier work on the turbulent flame sheet, we model the methane chemistry using DRM-19 and we model transport using a mixture model for differential diffusion. Initial simulations show that we can successfully predict a stable turbulent V-flame. Furthermore, mean statistical properties of the flame brush match the experimental data. Comparison of the simulation, presented in Figure 3 with particle-induced-velocimetry (PIV) data show that our simulations also do a good job of predicting the basic flame morphology. In these images, the spreading angle of the “V” is determined by the turbulent enhancement of the flame speed. The agreement between experiment and computation shows that we are accurately capturing the flame dynamics.
(a) 
(b) 
Figure 3: (a) Computed (methane) fuel concentration, and (b) Photograph of illuminated particles seeded in experimental fuel stream.
We are continuing our comparison with experimental data to
provide a more detailed comparison with simulation and experiment and to
identify the key factors controlling the turbulent flame speed in this flow
regime. Initial results for the
turbulent V-flames will be presented at the 19th ICDERS Conference
in July, 2003 (see
Plans
Our short term focus in combustion is to complete the analysis of turbulent V-flames. We are in the process of completing simulations for a variety of inflow turbulence configurations. In the first stage of the analysis we will focus on validation with the experimental data, flame morphology and flame dynamics. Subsequent analyses will examine the details of the turbulence / chemistry interaction.
The next problem we plan to simulate is a turbulent “low-swirl” premixed flame. In the low-swirl configuration, air is injected at high speeds tangential to the interior walls of the nozzle. Above the nozzle exit, the expanding fuel-air mixture creates an axial velocity deficit in the central core capable of stabilizing a premixed flame. Low-swirl nozzles have considerable practical value because the resulting flame has very low pollutant emissions. Although the experimental configuration is similar to that of the V-flame, the interactions between the fluid dynamics and the volumetric expansion of the flame are considerably more delicate and difficult to simulate. Our initial goal for will be to compute a statistically stable flame and analyze the basic flame dynamics.
Our previous work on modeling laboratory-scale premixed turbulent combustion has been based on a description of chemical kinetics given by DRM-19 which contains 21 species and 84 reactions. This mechanism provides a good approximation to the basic carbon pathways in the flame; however, it includes some simplifications for the carbon chemistry and does not include any of the nitrogen chemistry needed to pollutant formation. Our longer term plans are to revisit the two experimental flames we have considered introduce a more complete treatment of chemical kinetics, including the nitrogen chemistry needed to model formation of pollutants in methane flames.
Publications
[1] J. B. Bell, M. S. Day, J. F. Grcar, W. G. Bessler, C. Shultz, P. Glarborg, and A. D. Jensen, "Detailed Modeling and Laser-Induced Fluorescence Imaging of Nitric Oxide in an NH3-seeded non-premixed methane/air flame", LBNL Report LBNL-49333, Proceedings of the Combustion Institute 29 (in press), 2002.
[2] J. B. Bell, M. S. Day, J. F. Grcar, and M. J. Lijewski, "Numerical Simulation of Premixed Turbulent Methane Combustion", LBNL Report LBNL-49331, Proceedings of the Combustion Institute 29 (in press), 2002.
[3] J. B. Bell, M. S. Day, J. F. Grcar, M. J. Lijewski, M. Johnson, R. K. Cheng and I. G. Shepherd, "Numerical Simulation of a Premixed Turbulent V-Flame", 19th International Conference on the Dynamics of Explosions and Reactive Systems (to appear).
For More Information
Contact [email protected],
or visit http://seesar.lbl.gov/CCSE/Research/Combustion