CCSE Home

Home

Research

Publications

People

Software

Links

Downloads


Low-swirl Burner


Low-swirl burners (LSB) are emerging as an important technology for meeting design requirements in terms of both reliability and emissions for next generation combustion devices. Continued development of these types of burners, particularly for alternative fuels, depends on improving our understanding of basic flame structure, stabilization mechanisms, emissions and response to changes in fuel. Numerical modeling has the potential to address some of these issues, but simulation of these types of burners has proven to be difficult because of the large range of spatial and temporal scales in the system; the bulk of the analysis to date has been experimental. However, detailed numerical simulation of laboratory-scale flames that are stabilized on a low-swirl burner has been ongoing research in our group, closely collaborative inverstigations between computational and experimental combustion scientists.

Physical Configuration of LSB

The low-swirl burner concept is extremely simple: premixed fuel exits a pipe after passing through a turbulence generation plate and an annular set of curved vanes. The vanes impart a swirl component to the flow over a narrow layer near the pipe wall, and a detached premixed flame anchors in the diverging flow above the pipe exit. Turbulence in fuel stream wrinkles the flame, which enhances the overall rate of combustion in the device; the flames stabilize where the mean burning speed matches the axial flow velocity.

(a) Low-swirl nozzle, showing vanes and turbulence generation plate. (b) typical turbulent low-swirl methane flame stabilized in divergent flow above the low-swirl nozzle. More details regard the low-swirl burner experimental configuration can be found at LBNL's EETD laboratory

Computational Setup of LSB

There are two possible approaches for how to treat the nozzle in a simulation of this sort. One possiblility is to include the flow in the nozzle directly as part of the simulation; the alternative is to use measured data to prescribe the flow at the exit to the nozzle. Considerably better agreement was found using large eddy simulation between simulated and measured profiles when the flow inside the nozzle was included (Nogenmyr et al.). In particular, the simulations correctly predicted large-scale structures emanating in their configuration from the swirl vanes inside the nozzle, and these structures ultimately played a key role in stabilizing their flame. Our simulations consider both approaches. We included the nozzle in the simulation so to provide the inflow boundary (see below two figures) for the simulation of methane flame. For the hydrogen flame, we incorporate an experimental characterization of the flow at the nozzle exit. Typical profiles for this purpose were provided by Petersson et al. and detailed boundary conditions are prescribed in Day et al..

Fuel/air inside swirl nozzle
Fuel [orange] and air [blue] inside nozzle.
(Mouse over for animation/click for QuickTime)
Vertical velocity at nozzle exit
Axial velocity at nozzle exit plane.
(Mouse over for animation/click for QuickTime)

In the figures above, an animation shows the simulated evolution of the premixed fuel upward though the vertical midplane of the nozzle. The blue colored activity is dilution due to the high-speed air jets. The animation was taken well after nozzle had flushed its initial volume of air. The animation on the right depicts the magnitude of the vertical velocity at the exit plane of the swirl nozzle as a function of time, where red represents strong upward flow. It is apparent that the swirling flow remains rather confined to the outside edges of the nozzle, and the inner core of fuel is moving at relatively constant, low rate.

Simulations of LSB Flames

The simulations presented here are based on a low Mach number formulation (Rehm and Baum ) of the reacting Navier-Stokes equations. The methodology treats the fluid as a mixture of perfect gases. We use a mixture-averaged model for differential species diffusion, which is critical in the lean hydrogen flames to capturing the thermodiffusive behavior. We ignore Soret, Dufour and radiative transport processes. The hydrogen flame, phi=0.37, was based on the hydrogen sub-mechanism of GRI-Mech 2.11. The transport coefficients and thermodynamic relationships are obtained from EGLib. A complete description of our numerical algorithm can be found in Day and Bell.

Hydrogen Flame

We show a cross-section through the middle of the simulation that provides a picture of the overall structure of the flame. See the picture below (on the left half page) for a typical slice from a snapshot in time of the computed flame solution: panel represents a vertical slice centered on the symmetry axis of the nozzle; (a) the mole fraction of hydrogen over a (25 cm x 25 cm); and (b)-(e) enlargements of the white-boxed region (approx. 2.5 cm wide) in (a); (b) hydrogen mole fraction; (c) destruction rate of hydrogen; (d) mole fraction of OH; and (e) temperature. All scales are relative to the peak values over the domain at that instant.

we also show the T= 1144 K isotherm (pictures above on the right side of the page), colored by local fuel consumption in the core of the burner. For comparison, we show the analogous image for a freely propagating hydrogen flame. Both images are based on the same color map, and the length scales are as indicated in the figure. Comparing these images, one can see that in the turbulent flame, there are finer structures than in the non-turbulent case. The turbulent flame shows more ridge-like structures (picture above on the right side of the papge labelled by "a") compared to the freely propagating flame (labelled by "b") in which the features are more bulbous. Finally, the burning rate over most of the flame surface is considerably higher in the turbulent configuration, particularly where the flame is tightly folded.

One of the principal diagnostics used in the experiments is OH-PLIF (planar laser-induced fluorescence) based on imaging the flouresence of OH radicals excited by a tuned laser sheet. Below, we show a typical vertical slice of the OH concentration from the simulation alongside typical PLIF images from the experiment. In the figure (a) shows the profile of OH over the (25 cm x 25 cm) slice, while (b) and (d) show progressive enlargements of the data corresponding to the field of view of the typical OH-PLIF data shown in (c) and (e). The experimental data are representative images taken from the low-swirl experiment. The figure show that the simulation captures features of these flames, with remarkably similar sizes, shapes and global structure. The simulation also captures the observed variability of the OH signal (brightness on the experimental images) along the flame surface.

More interesting results are prescribed in the "cellular burning in lean premixed turbulent hydrogen-air flames: coupling experimental and computational analysis at the laboratory scale" (see Day et al.).

Interaction between Computational and Experimental Combustion Scientists

This work lays the foundation for a program of closely collaborative inverstigations between computational and experimental combustion scientists. Practical difficulties remain, including gathering suitable inlet profile specifications from the experiments, and extremely long simulation times which complicate the gathering of converged statistics of the quasi-steady flow, in spite of the highly optimized solution strategy. Nevertheless, the resulting simulations can be used to explore experimental diagnostic assumptions and to gather higher-dimensional statistics and correlations that are simply unavailable from the experiment. Therefore, experimentalists are interested in the simulation data and involve us in the design of experiments and the analysis of experimental data. With our recent detailed flame computations, we have provided them with their first 3D, time-dependent picture of the detailed structure of their flames for a range of fuels and turbulence scenarios.

Future Work on LSB

Future work will involve the incorporation of emission chemistry and further characterization of the burner configurations under a variety of fuels relevant to ultra-low emission burner scenarios. Related studies currently underway are related to turbulence/chemistry interactions across a much more broad range of fluctuation intensities, and in high-pressure environments. The studies provide valuable insight that will enable the development of reduced models for turbulent flame propagation.