Sie sind auf Seite 1von 18

This article was downloaded by: [Ryerson University]

On: 10 May 2013, At: 07:41


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Combustion Science and Technology


Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/gcst20

Experimental and Numerical Study of


Flameless Combustion in a Model Gas
Turbine Combustor
a a a b b
C. Duwig , D. Stankovic , L. Fuchs , G. Li & E. Gutmark
a
Division of Fluid Mechanics, Department of Energy Sciences, Lund
University, Lund, Sweden
b
Department of Aerospace Engineering and Engineering Mechanics,
University of Cincinnati, Ohio, USA
Published online: 11 Jan 2008.

To cite this article: C. Duwig , D. Stankovic , L. Fuchs , G. Li & E. Gutmark (2007): Experimental and
Numerical Study of Flameless Combustion in a Model Gas Turbine Combustor, Combustion Science and
Technology, 180:2, 279-295

To link to this article: http://dx.doi.org/10.1080/00102200701739164

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation
that the contents will be complete or accurate or up to date. The accuracy of any
instructions, formulae, and drug doses should be independently verified with primary
sources. The publisher shall not be liable for any loss, actions, claims, proceedings,
demand, or costs or damages whatsoever or howsoever caused arising directly or
indirectly in connection with or arising out of the use of this material.
Combust. Sci. and Tech., 180: 279–295, 2008
Copyright # Taylor & Francis Group, LLC
ISSN: 0010-2202 print/1563-521X online
DOI: 10.1080/00102200701739164

EXPERIMENTAL AND NUMERICAL STUDY OF


FLAMELESS COMBUSTION IN A MODEL GAS
TURBINE COMBUSTOR

C. Duwig1, , D. Stankovic1, L. Fuchs1, G. Li2, and E. Gutmark2


1
Division of Fluid Mechanics, Department of Energy Sciences, Lund University,
Lund, Sweden
2
Department of Aerospace Engineering and Engineering Mechanics, University
of Cincinnati, Ohio, USA
Downloaded by [Ryerson University] at 07:41 10 May 2013

Flameless combustion is an attractive solution to address existing problems of emissions and


stability when operating gas turbine combustors. Theoretical, numerical and experimental
approaches were used to study the flameless gas turbine combustor. The emissions and
combustion stability were measured and the limits of the flameless regime are discussed.
Using experimental techniques and Large Eddy Simulation (LES), detailed knowledge
of the flow field and the oxidation dynamics was obtained. In particular the relation between
the turbulent coherent structures dynamics and the flameless oxidation was highlighted. A
model for flameless combustion simulations including detailed chemistry was derived. The
theoretical analysis of the flameless combustion provides 2 non-dimensional numbers that
define the range of the flameless mode. It was determined that the mixture that is ignited
and burnt is composed of  50% of fresh gases and  50% vitiated gases.

Keywords: Flameless combustion; Gas turbine; Large eddy simulation; PIV; Tabulated complex chemistry

INTRODUCTION
Due to new emission regulations, environmental issues of power generation
play an important role in the economic viability of modern power plants. To reduce
harmful emissions, the current trend is to design industrial combustion devices to
operate at premixed fuel lean conditions. Consequently, the peak temperature and
NOx emissions are reduced but high CO=UHC emission levels and combustion
instabilities (leading to flash back or flame blow out) may occur when operating
traditional combustors. The control of turbulent premixed flames is particularly
difficult since it is a multi-scale and non-linear four-dimensional problem.
To eliminate temperature peaks, the aim is to have a ‘‘homogeneous’’ tempera-
ture distribution. The distributed combustion mode can be achieved using different
engineering solutions. ‘Flameless’ or ‘Mild’ combustion consists of operating devices
under very lean but well preheated conditions (Wunning and Wunning, 1997). The

This work was partially financed by the Swedish energy agency STEM within the program
‘‘termiska processer för el-produktion.’’ The simulations were run on HPC2N and LUNARC facilities
within the allocation program SNAC.

Address correspondence to Christophe.Duwig@vok.lth.se

279
280 C. DUWIG ET AL.

reaction zone is distributed within a very large volume compared to traditional


combustors. Since the local reaction rates are low, thermo-acoustic wave generation
is avoided but flame blow off becomes an issue. The latter problem is overcome by
keeping the combustion temperature high enough to maintain the flame. Practically,
one designs the device for having a large amount of hot products mixing with the
incoming fresh reactants.
A key point necessary for efficient design of a flameless combustor is to ensure
good mixing between the incoming fresh fuel=air mixture and the hot burnt gases.
This can be achieved by designing combustor aerodynamics with strong recirculation
that redirects the hot products upstream, towards the fuel=air mixture injection nozzle.
Flameless combustion modeling requires different tools compared to tra-
ditional combustion simulation. Unlike Lean Premixed Pre-vaporized burners, the
reactive region is not restricted to a thin layer (i.e., as small as the Kolmogorov
length scale) with high reaction rate but rather looks like a thick brush (typically
Downloaded by [Ryerson University] at 07:41 10 May 2013

several millimeters) with distributed low reaction rates. Another difference is that
the reaction dynamics are governed by a stirring controlled process since fresh gases
have to mix with burnt products prior to ignition and reaction. Consequently, the
simulation tools should include accurate models for describing both turbulent
stirring=mixing and chemical reactions.
In this article, the combustion process in a flameless burner is investigated
experimentally and by using Computational Fluid Dynamics (CFD). The burner
is assumed to operate in a perfectly premixed mode so that only constant fuel=air
equivalence ratio mixtures are discussed. The paper consists of the following sec-
tions: (a) theoretical aspects of flameless combustion; (b) overview of the closure
problem for CFD equations and presentation of the closure; (c) description of the
experimental apparatus and techniques; (d) presentation of the experimental and
numerical results to assess the limits of the flameless regime and to determine the
efficiency of the present combustor in terms of mixing and emissions.

OVERVIEW OF FLAMELESS COMBUSTION


Reduced NOx and improved fuel efficiency have been successfully demon-
strated in industrial furnaces using flameless combustion (Wunning and Wunning,
1997; Katsuki and Hasegawa, 1998). This success motivated the extension of its
application to other fields such as gas turbine engines for power generation. Flamme
(2004) demonstrated single-digit NOx emissions at atmospheric conditions and Levy
et al. (2004) obtained low NOx at elevated pressure but with poor combustion
efficiency. Nevertheless, more investigations are needed in order to adapt flameless
combustion process to gas turbine applications.
Lückerath et al. (2005) demonstrated, using a FLOX burner under high press-
ure conditions, the spatial uniformity of the temperature and reaction zone. They
found that high velocity of the premixed jets is beneficial both for the emissions
and the operating range. Furthermore, when the burner is operated in flameless
regime, NOx is not increasing with pressure. Li et al. (2006a, 2006b) used a similar
design to investigate the effect of pressure drop, confinement, and air preheating
temperature on the emissions, stability of operation and the regime of flameless
combustion.
FLAMELESS COMBUSTION IN A GAS TURBINE COMBUSTOR 281

They emphasized the importance of confinement for this type of burner by


showing that if the combustion chamber is not closely adjacent to the fuel=air
injectors, the central recirculation zone disappears. This implies that in order to
use this burner in gas turbines, the combustion chamber should be converted from
the annular shape to tubular kind. They also pointed that close to the flameless
regime, unsteady heat release is easily coupled with the acoustic modes of the com-
bustion chamber leading to large pressure fluctuations.
Flameless combustion modeling has been primarily done within the
Reynolds Averaged Navier-Stokes (RANS) context. Coelho and Peters (2001)
and Dally et al. (2004) applied an interactive flamelet model to mild combustion.
Christos and Dally (2005) compared different combustion models and showed
that mixture fraction based models fail to capture mild combustion while an
Eddy Dissipation concept performed better. Kim et al. (2005) showed that Con-
ditional Moment Closure technique is applicable to mild combustion. A common fea-
Downloaded by [Ryerson University] at 07:41 10 May 2013

ture of all successful simulations of mild=flameless combustion is the use of detailed


chemistry schemes. The reason for that is inherent to the flameless combustion process;
the fresh gases are diluted with hot exhaust gases and the local temperature rises with-
out reaction until the mixture is hot enough for auto-ignition. The two key processes
are the mixing=dilution and the auto-ignition prediction, and the latter requires the use
of detailed chemistry.
A main limitation of current modeling works based on RANS techniques is
their inability to simulate unsteady stirring or mixing. Large Eddy Simulation
(LES) based techniques overcome this difficulty but also increase the computational
cost (Poinsot and Veynante, 2001). Using LES for an accurate description of mixing
implies that the number of scalar transport equations has to be reduced below 10.
One possibility is to reduce the detailed complex chemical scheme to a few steps
while maintaining a correct simulation of the auto-ignition process. Another method
is to use detailed chemistry schemes for tabulation of the reaction rates thus describ-
ing the problem with only a few representative scalars. Among other techniques,
ILDM (Mass and Pope 1992) has been used successfully for generating tables map-
ping combustion as manifolds into the composition space. Recently, Duwig and
Fuchs (2006) proposed another tabulation technique based on stirred reactor con-
cept that is used in the present investigation.

EXPERIMENTAL TECHNIQUES AND TEST RIG


The experimental combustion test rig consists of a 36 kW heater to preheat the
air, a flow conditioning section that is composed of a series of 5 fine screens and a
honeycomb flow straightener, a plenum chamber, the flameless burner and a quartz
combustion chamber. The combustor, shown in Figure 1, consists of a 0.3 m long
and 0.1 m diameter straight tube.
The flameless burner has 12 equally spaced injectors (of diameter 0.006 m)
distributed on a 0.076 m diameter with each injector having 4 fuel injection holes.
The air flow enters into the burner from the upstream plenum through radial holes
positioned at the outer and inner surfaces of the burner’s premixing chamber. The air
and the fuel injected from the inner fuel injectors are premixed within the 12 straight
coaxial tube nozzles over a distance of 0.025 m before injected into the combustion
282 C. DUWIG ET AL.
Downloaded by [Ryerson University] at 07:41 10 May 2013

Figure 1 Left: schematic drawing of the estimated flow pattern in the flameless gas turbine combustor with
details of the fuel=air nozzle. Right: Half-cut representation of the combustor and the jets (presented as
snapshot of the iso-surfaces cT  0.8).

chamber. The operating conditions were variable propane=air equivalence ratio in


the range of U ¼ 0.31–0.63, Tu ¼ 823 K with a total mass flow of fuel,
mf ¼ 0.01 kg=s. The characteristic length is the diameter of the combustor
D ¼ 0.1 m. The characteristic velocity is a single jet bulk velocity U0. The data are
presented time averaged (denoted by the operator <  > ).
The test rig was instrumented with thermocouples and pressure transducers to
monitor the process parameters. An emission sampling probe was mounted at a dis-
tance of 0.025 m from the inner wall of the combustor chamber at the exit plane of
the combustor. The sampling gas was analyzed by individual gas analyzers for
CO=CO2=O2=NOx=UHC. Accuracy of the emission measurements is about 1% of
the measuring range. The exhaust gas temperature was measured by a type B thermo-
couple positioned at the center of the combustor exit plane.
A LaVision Particle Image Velocimetry (PIV) system was used to investigate
the 2D velocity field on the streamwise plane of non-reacting and reacting flows.
The flame OH chemiluminescence was imaged by an Intensified CCD (ICCD)
camera with a narrow band pass filter centered at 308 nm.
Since the fresh mixture is injected into the combustor very close to the wall
there is a tendency for fast deposition of PIV seeding on the wall of the combustion
chamber. Within seconds from the beginning of the measurements, combustor walls
would become nontransparent and cleaning of the walls must be performed. With
only a couple of collected images, in order to obtain a good statistical estimate of
the velocity field, process of image collection and cleaning must be tediously
repeated. The particle deposition prevented PIV measurements in the reacting cases.
Nevertheless, the global uncertainty of the velocity measurement is around 5% tak-
ing into account the uncertainty introduced through the measurement of the mass
flow, distortions by the quartz combustion chamber and the minimum particle
displacement that PIV algorithm distinguishes (Huang et al., 1997).
FLAMELESS COMBUSTION IN A GAS TURBINE COMBUSTOR 283

TABULATED COMPLEX CHEMISTRY FOR FLAMELESS COMBUSTION


‘‘Flameless’’ combustion involves a kinetically controlled oxidation of a hot
diluted mixture. We seek a description of the detailed chemical oxidation that
requires few (1–2) scalars. Duwig and Fuchs (2006) suggested that the flameless com-
bustion can be represented using an ensemble of elementary reactors. One considers
a fluid parcel containing fuel=air mixture that enters a flameless burner. A typical
scenario is described as follows:

. Fluid parcel enters the burner with a temperature that is too low for auto-ignition.
. It mixes with vitiated gases. Several steps may elapse before the mixture is hot
enough to ignite.
. Combustion starts and proceeds into the parcel. It might interact with neighboring
parcels. Since the gradients in the reaction zone are smooth, the neighboring
Downloaded by [Ryerson University] at 07:41 10 May 2013

parcels are expected to be in the same state (temperature and composition).

According to this scenario, one may propose non-dimensional criteria to


describe the flameless combustion process. Starting at the nozzle exit, we require that
the fresh gases are diluted with the vitiated mixture. Consequently, the local turbu-
lence should be strong enough to achieve distributed reaction (i.e., the turbulent
eddies are smaller than the reacting layer). A Karlowitz number (Poinsot and
Veynante, 2001) can be defined based on the reacting layer thickness dr: Kar ¼
(dr=gK)2 where gK denotes the Kolmogorov length scale. Flameless combustion
corresponds to Kar > 1. The second condition requires that after mixing with the
vitiated gases, the mixture ignites before leaving the combustor, i.e., the ignition
delay of the mixture sI is shorter than the residence time sR. This condition is quan-
tified by introducing an ignition number NI ¼ (sR=sI) > 1. sI depends strongly on
the local composition and temperature of the mixture, i.e., the amount of vitiated
gas that has been mixed with the fresh gases.
One may model the distributed oxidation by an unsteady Perfectly Stirred
Reactors (PSR) and assume that the 2 processes (mixing and reaction) are consecu-
tive. a gram of burnt gases are diluted with (1  a) gram of fresh gases at time t ¼ 0.
The resulting mixture burns in an unsteady PSR. The initial conditions for n species
and temperature are:

X ðt ¼ 0Þ ¼ ½Y1 ; . . . ; Yi ; . . . ; Yn ; T ¼ aXb þ ð1  aÞXu ð1Þ

where Yi denotes the species i’s mass fraction, T is the temperature and the subscripts
u and b denotes the unburnt and burnt states of the vector X, respectively. The set of
equations describing the idealized process is ðdx=dtÞ ¼ xðX Þ where x is the vector
source of X. This system is solved using standard chemical software packages.
The present approach requires 2 variables a and t to describe the system (i.e.,
X(a, t) and x(a, t)) while using detailed oxidation mechanisms. The variable a con-
tains the information related to the dilution by vitiated gases (i.e., temperature
increase due to mixing) while the time t describes the advancement of the reactions.
284 C. DUWIG ET AL.

(a, t) are oxidation coordinates. However, since both (a, t) are not known, it is better
to use physical variable such as temperature and mass fractions.
The new oxidation coordinates must lead to a one-to-one correspondence with
(a, t). Duwig and Fuchs (2006) suggested to use the temperature and the fuel mass
fractions as coordinates for the flameless oxidation. Consequently, one may convert
X(a, t) and x(a, t) obtained using detailed chemistry to X(cT, cF) and x(cT, cF) where
cZ denotes the progress variable (the normalized quantity is 0 when unburnt and 1
when burnt) based on the physical variable Z (temperature or fuel mass fraction).
In addition, the table gives a(cT, cF), which quantifies the local mixing=dilution.
In the present work, the freeware Cantera (2005) is used together with a detailed
oxidation mechanism containing 46 species and 235 reactions (San Diego, 2005). The
burnt state (i.e., the state of the gases trapped in the recirculation zone) is assumed to
be at chemical equilibrium. Since the residence time in the recirculation zone is long
compared to the chemical reaction time scale (obtained from a PSR computation), the
Downloaded by [Ryerson University] at 07:41 10 May 2013

approximation of these vitiated gases by a mixture at chemical equilibrium is valid.


The computations were performed for a lean propane=air mixture with equivalence
ratio of 0.35. The fresh gas temperature was set to 823 K. PSR calculations were per-
formed varying a from 0 to 1. The resulting profiles (temperature, species or reaction
rates) were obtained in the (a, t) plane and mapped into the (cT, cF) plane. Figure 2

Figure 2 Tabulated chemistry. Propane=air, U ¼ 0.35, Tu ¼ 823 K.


FLAMELESS COMBUSTION IN A GAS TURBINE COMBUSTOR 285

shows an example of the lookup tables generated by this procedure. One can follow
the oxidation of a fluid parcel in the (cT, cF) plane as iso-a lines.
The results show that the lower part of the plots (cF < cT) is a forbidden state:
one cannot have heat released and low cF (i.e., combustion without fuel consump-
tion). A second forbidden sate lies in the quarter cF > cT, expressing that fuel com-
bustion generates heat and increases cT. The allowed states lie between the curve
a ¼ 0 and a ¼ 1 (pure mixing line, cF ¼ cT). The reaction rates present a peak within
the quarter cF > 0.5 and cT > 0.5. The fuel oxidation rate is higher than the heat
release rate. The heat release results from consecutive oxidation of propane into
CO and CO into CO2. It ensures that cF > cT. We also notice that the reaction rates
are very low if cT < 0.5. Consequently, a suitable reaction path would follow the
mixing line (cF ¼ cT) up to cF ¼ cT  0.5–0.6 where the mixture is hot enough so
that fuel oxidation can start.
Regarding intermediate species like CO, one should point out that high CO
Downloaded by [Ryerson University] at 07:41 10 May 2013

concentrations are obtained for low a values and remain low close to the mixing line.
Any oxidation trajectory following the mixing line up to cF ¼ cT  0.5–0.6 prior to
combustion will ensure relatively low intermediate species concentration.
The tables enable an easy lookup procedure for obtaining the reaction rates for
cF and cT also in addition to different species like CO, OH or the dilution fraction a.
It is also possible to access sI(a) for estimating NI. The method follows the oxidation
in the (cT, cF) plane as the trajectory depends on the interaction between chemical
oxidation and turbulence=combustor aerodynamics.

EQUATIONS, CLOSURES AND NUMERICS


The equations describing turbulent combustion are the reactive Navier-Stokes
Equations (rNSE), or the balance equations of mass, momentum and energy, supple-
mented with constitutive equations and state-equations (Poinsot and Veynante,
2001). For low Mach numbers and reduced chemical systems, it is possible to sim-
plify the rNSE. In LES, the dependent variables are decomposed into resolved
and unresolved (subgrid) components by a spatial filtering operation f ¼ f~ þ f 00 with
f~ ¼ qf =
q, where overbars denote spatial filtering.
Applying the filtering to the rNSE yields,

@q
þ r  ðq~uÞ ¼ 0 ð2Þ
@t

@q~u 
þ r  ðq~
u~uÞ ¼ r p þ r  ðquu þ q~u~u þ lr~uÞ ð3Þ
@t

@q~ci
þ r  ðq~
u~ci Þ ¼ r  ðquci þ q~u~ci þ qDi r~ci Þ þ wi ð4Þ
@t

where q is the density, u the velocity, ci a progress variable, Di the Di diffusion, xi the
ci reaction rate, p the pressure, and l is the dynamic viscosity. In addition, the
subgrid stress and flux terms require closure modeling. The thermodynamic model
286 C. DUWIG ET AL.

typically consists of the equation-of-state giving the density as function of the


progress variables.
The fluid dynamics subgrid model provides closure models for the subgrid
stress and flux terms. The momentum and c-equations use gradient type closure:
the Filtered Structure Function (FSF) model (Ducros et al., 1996) with a turbulent
Schmidt number of 0.7.
The unclosed reaction rate is modeled using a presumed filtered density func-
tion (FDF) closure (Poinsot and Veynante, 2001). It gives:
Z Z
xZ ðcT ; cF Þ ¼ xZ ðcT ; cF Þ  Pð~cT ; ~cF ; cT ; cF ; dT ; dF Þdci ð5Þ

where P represents the presumed joint filtered density function. We choose a top hat
shape so that:
Downloaded by [Ryerson University] at 07:41 10 May 2013

(
1
; if j~cT  cT j  0:5  dT and j~cF  cF j  0:5  dF
P ¼ dT dF ð6Þ
0; otherwise

The variable dX is computed from the resolved field variables following Pierce and
Moin (1998):
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
dX ¼ CD D2 r~cX  r~cX ð7Þ

where CD is a model constant (taken here to be 0.25) and D denotes the spatial filter
length. Note that CD can be obtained dynamically (Pierce and Moin, 1998).
The reaction rates are read in lookup tables: xi  xTAB ci ðcT ; cF Þ where the
superscript TAB denotes the look-up table discussed in the previous section. The
LES code is a high-order finite difference Cartesian code that solves the semi-com-
pressible NSE with variable density on Cartesian grids (Gullbrand et al., 2001). The
spatial discretization is done using a fourth order centered scheme. Stability is
ensured by including hyper-viscosity. For the scalar equations, the convective terms
are discretized using a 5th-order Weighted Essentially Non-Oscillatory (WENO)
scheme (Jiang et al., 1996) enabling an accurate capture of steep gradients. A second
order finite difference scheme is used for time discretization and the time integration
is done implicitly. Multi-grid iterations are used to solve the implicit parts of the sys-
tem and local refinements are used to capture high gradients. More details can be
found in (Gullbrand et al., 2001).
The vortex core has been visualized using a criteria based on the second largest
eigenvalue of the second invariant of the velocity derivative tensor proposed by
Jeong and Hussain (1995) (the so-called k2 technique). In the following figures the
vortex core corresponds to a region where the eigenvalue is negative.
The computations presented here were performed on two different grids of
 2,300,000 (medium grid) and  3,900,000 (fine grid) points. Local refinements
were used close to the nozzles to have  20 points per nozzle diameter for both grids
which is commonly used for LES of turbulent jets e.g., (Olsson and Fuchs, 1996;
DaSilva and Métais, 2002). The size of the local refinements was varied so that
FLAMELESS COMBUSTION IN A GAS TURBINE COMBUSTOR 287

Table 1 Summary of operating conditions

Air mass Air preheat


flow temperature Measurement Equivalence Reynolds
Flow type [g=s] [K] technique ratio U number Re LES

Reacting 10–30 823 OH chemiluminescence 0.31–0.63 5 000–15 000 U ¼ 0.31


Re ¼ 10 000
Non-reacting 10–35 298 PIV – 10 000–35 000 Re ¼ 10 000

Based on one of the jets.

the fine grid has a better resolution between the jets and in the recirculation zone.
The Reynolds number based on the nozzle exit is  10000 and the estimated Taylor
micro-scale is of the order of the grid spacing.
Downloaded by [Ryerson University] at 07:41 10 May 2013

Elsewhere in the combustor, the grid spacing was 4 (  2 for the fine grid) times
coarser yielding  80 (  160 for the fine grid) grid points per combustor diameter.
The Dirichlet inflow boundary conditions (supplemented by a 12% seemingly turbu-
lent velocity fluctuation) and zero gradient outflow boundary conditions, were used.
As presented in Table 1, 3 computations were performed; two non-reacting for com-
parisons with velocity measurements and grid dependency study and one reacting in
the flameless mode of U ¼ 0.31, Tu ¼ 823 K, Tb ¼ 1556 K and total fuel mass flow of
mf ¼ 0.01 kg=s. Note that the non-reacting case was performed with air only as fluid.

RESULTS AND DISCUSSION


Burner Performance
The reaction zone of the flameless combustor was visualized at one typical
operating condition by OH chemiluminescence images as shown in Figure 3. When
the equivalence ratio U was gradually decreased from 0.6 to 0.3, the flame trans-
formed from discrete flame regime (U  0.6–0.5), through distributed flame regime
(U  0.5–0.31), to lean blow off at U < 0.31. At the high fuel equivalence ratios, each
of the nozzles produced a well defined individual flame. As the fuel equivalence ratio
decreased, the reaction zone spread in both radial and axial directions while simul-
taneously shifting downstream. When U was further reduced, the reaction zone
spread such that the flame from each jet merged into a single reaction zone.

Figure 3 OH Chemiluminescence images at different equivalence ratios. The LES visualization was
obtained by averaging over the direction normal to the picture.
288 C. DUWIG ET AL.

Figure 4 Limits of the flameless regime vs. equivalence ratio. Left: Measurements: NOx, CO, exhaust tem-
perature, and p0 versus fuel equivalence ratio. Right: Theoretical: Kar (solid line) and NI (dashed lines;
U ¼ 0.2–0.8) vs. equivalence ratio.

Figure 4 shows that NOx emissions below 10 ppm were reached at U below 0.5.
Downloaded by [Ryerson University] at 07:41 10 May 2013

Relatively strong pressure fluctuations were measured at U  0.42 as the reaction


excited resonant acoustic modes in the combustion chamber (quarter-wave mode).
It indicates that the heat-release was still concentrated in thin surfaces that act as
thermo-acoustic sources. Consequently, despite of low-emissions, it can not be seen
as a flameless mode.
The images indicate that the strong reaction zone, i.e. high temperatures
regions, located in the separate flames are responsible for higher emissions of
NOx. It was also noticed that the pressure oscillations, p0 , decreases as the combus-
tion transformed from discrete flames (U > 0.4) into the flameless mode (U < 0.4).
Although the discrete flames (U > 0.4) are a potential source of thermo-acoustic
instability, the pressure fluctuations were only observed at U  0.42 when the
response of the flame was in phase with a resonant mode of the cavity.
When the equivalence ratio was further reduced, the reaction zone became
evenly distributed across the combustor and was lifted. With the transformation
of the flame mode, the exhaust temperature gradually decreased and so were the
NOx emissions (Figure 4). CO was also reduced with the temperature until the tem-
perature drop resulted in lean blow out (U < 0.31). Figures 1 and 3 shows LES
visualizations of the reaction zone for a flameless-like operation. Figure 3 emulates
the OH pictures by integration of the OH concentration in the direction normal to
the picture and by time averaging. The LES picture (Figure 3 right) shows a large
central reaction zone that is lifted to the middle of the combustor. A snapshot of
the reduced temperature (Figure 1) indicates that the 12 jets inject cold gases into
the combustor. The jets merge creating the large distributed reaction zone presented
in Figure 3 containing a very fuel lean, highly diluted hot mixture.
Figure 4 (right) shows the non-dimensional numbers Kar and NI as function of
the equivalence ratio and a. The reacting layer thickness is evaluated at the CH peak
mid-height thickness in a 1D laminar detailed simulation. The turbulence quantities
and residence time are estimated using the CFD simulation results. The flameless or
distribute reaction regime corresponds to Kar > 1 and NI > 1. Figure 4 shows that
the boundaries of this regime depend strongly on a. For a < 0.4, it is not possible to
reach the flameless mode. For large a values ( > 0.6), the flameless mode can be easily
reached when the equivalence ratio falls below  0.42. In between, there is a U-
window of flameless operation.
FLAMELESS COMBUSTION IN A GAS TURBINE COMBUSTOR 289

The experimental data shown in Figure 4 show that the outlet CO concen-
tration increases as U decreases below U  0.42. It can be explained by a too-low
oxidation rate or a too-short residence time. In other words, at the lean limit
(U < 0.31) the reactions are not fast enough to complete combustion within the resi-
dence time, i.e., NI < 1. The condition NI < 1 for U < 0.31 provides an estimate for
a global value a  0.5 in the present burner. At the rich limit (U > 0.4) the reaction is
not distributed but concentrated in thin sheets and Kar < 1.
Consequently, the heat-release and its fluctuations are locally higher, becoming
a potential source of thermo-acoustic instabilities. Note that both Kar and NI are
global estimations of the reaction=turbulence interaction. The accuracy of the
scaling approach in ‘flamelet regime’ analysis (e.g. Poinsot and Veynante, 2001) is
limited by the unsteady and chaotic nature of the turbulent flow. Yet, the present
criteria enable to estimate the boundaries of the premixed flameless regime and to
identify the key design parameters.
Downloaded by [Ryerson University] at 07:41 10 May 2013

Flow and Combustion in the Burner


Figures 5 and 6 present the velocity field in the combustion chamber determ-
ined by both experimental PIV and modeling LES for non-reacting conditions.
The PIV data do not cover the region where the jets are strong but do capture the
recirculation zone. Figure 5 shows that the 12 jets injected in the combustion

Figure 5 LES (lines) and PIV (symbols) results of the axial velocity in the flameless combustor during
non-reacting operation. (a) Time averaged field (b) Root Mean Squared (RMS) of the velocity fluctuation.
290 C. DUWIG ET AL.
Downloaded by [Ryerson University] at 07:41 10 May 2013

Figure 6 LES (lines) and PIV (symbols) results of the radial velocity in the flameless combustor during
non-reacting operation. (a) Time averaged field (b) Root Mean Squared (RMS) of the velocity fluctuation.

chamber induce a central recirculation zone. The jets entrain mass and a central
recirculation is formed. The strength of the recirculation zone (or jet entrainment)
is maximum along the axis at X=D  0.5. Further downstream, the jets are bending
toward the centerline and small recirculation zones are formed along the walls of the
combustion chamber. It results in a weakening of the central recirculation zone. In
addition, the turbulent fluctuations are rather high both in the recirculation zone and
in the jet shear layers. In particular at X=D ¼ 0.6, the turbulence intensity is close to
50% indicating that the present device is well stirred and suitable for flameless
operations.
The agreement between the numerical results and the measurements is good
verifying that the present LES tools captures accurately the flow physics. Figure 5
shows that both grids capture well the recirculation both in term of size and strength.
Both grids also capture the jet spreading and entrainment. The agreement of the LES
results with the PIV data is also good regarding the fluctuation. Both grids predict
reasonable fluctuations levels. The maximums are seen in the jet shear layers for
x=D < 0.6. Further downstream, the fluctuation profile is rather flat. The discrepan-
cies between the LES and PIV data are mainly seen in the jet region. It can be
explained by looking at Figure 6 that shows the crosswise velocity statistics. The
LES results are symmetric but the PIV show a net flow toward the negative values
of y=D. This net flow is seen for x=D < 0.5 and disappears further downstream. It
can be explained by an un-even distribution of the mass flow through the different
FLAMELESS COMBUSTION IN A GAS TURBINE COMBUSTOR 291

jet nozzles. This effect is dumped by turbulent mixing in the combustor so that
symmetry is recovered for x=D > 0.5. The overall agreement between the numerical
and experimental results is good and the use of a finer grid did not lead to any improve-
ments indicating that the medium grid is suitable for simulation of this device.
Figure 7 shows the axial flux density statistic for both the reacting and non-
reacting cases. Unlike the velocity, the flux density is not affected by thermal expan-
sion and therefore is suitable for comparing the flow fields between cold and reacting
cases. Figure 7a shows that the combustor aerodynamics is not strongly affected by
combustion. The jets as well as the central recirculation zone are relatively
unchanged. The jet spreading rate is lower for the reacting case with a slightly longer
recirculation zone. The thermal expansion across the jet shear layer is known to
reduce the transfer of momentum from the jet to the environment and has been
observed both numerically and experimentally (Chen et al., 1996; Duwig et al.,
2007). Figure 7b shows the axial flux density fluctuations for the cold and reacting
Downloaded by [Ryerson University] at 07:41 10 May 2013

cases. The shapes of the profiles are similar but the cold case presents higher fluctu-
ation levels. It follows the previous analysis since the jet shear layers are the major
source of turbulence. Note that because of thermal expansion, the velocity fluctua-
tions are higher by a factor of 2 (graphs not presented here) in the reacting case com-
pared to the cold case.
Figure 8 presents the recirculation rate (Wünning and Wünning, 1997) versus
time. The recirculation rate (RR) compares the mass of (hot) products that are

Figure 7 LES (lines) results of the axial flux density in the flameless combustor during non-reacting and
reacting operation. (a) Time-averaged field (b) Root Mean Squared (RMS) of the flux density fluctuation.
292 C. DUWIG ET AL.

Figure 8 Recirculation rate (RR) versus time at 4 different locations downstream of the jet nozzle for
non-reacting and reacting case.
Downloaded by [Ryerson University] at 07:41 10 May 2013

transported in the recirculation zone with the mass of incoming (fresh) gases. Mass
conservation implies that it also represents the amount of products entrained by the
jets. It is the integral across a combustion sector of the flux density. Figure 8 shows
that RR varies with x=D, increasing up to x=D  0.5 and decreasing further down-
stream. It follows the remarks drawn previously on the recirculation strengths.
Figure 8 shows that RR is not constant in time and exhibits relatively strong fluctua-
tions ( 25%) in the isothermal case. These fluctuations are of low frequency f
(St ¼ f*D=U0 < 0.05). The maximum recirculation rate is  1.3 at x=D  0.5. For
the reacting case, the RR are found to be much lower. The maximum RR is  1
at x=D  6. It can be explained by the effect of thermal-expansion that lowers the
jet entrainment rate. In the reacting case, the RR fluctuates around a mean value
but with a much lower amplitude ( < 10%). The heat-release suppresses the large
slow fluctuations.
Figure 9 shows the oxidation trajectories at different axial locations into the
combustor. Once the fresh gases are injected in the device, they mix and react, form-
ing a well-distributed reaction zone as already shown on Figure 3. The oxidation his-
tory can be followed by plotting the physical variables as function of cT. Figure 9
shows that at low temperatures (cT < 0.6); the evolution of the two variables is cor-
related with very low OH levels. cT and cF increase simultaneously as the fresh gases
mix with the hot burnt products. The change is due to mixing while no reaction is
occurring. For cT > 0.5–0.6, the evolution of the 2 scalars is different. Here cF
increases faster than cT, indicating the start of the oxidation reactions with higher
levels of OH.
Fuel disappears (cF ¼ 1) before CO is completely oxidized and the final tem-
perature (cT ¼ 1) is reached. The plots also show that the oxidation trajectories
change as one travels downstream of the jet nozzles. At x=D  0.2, the trajectories
match closely the mixing line. Small departure corresponding to high OH levels is
seen cT  0.9. At x=D  0.4, the departure is large from cT  0.6. At x=D  0.6,
the occurrence of cT < 0.2 is very rare indicating that the incoming gases have been
mixed with hot products. Similarly at x=D  0.8, most of the fluid parcels are at
cT > 0.4. The trajectories indicate that parcels of fluid at cT  0.5 start burning so
that no occurrences are found on the mixing line. Consequently, the mixing rate a
FLAMELESS COMBUSTION IN A GAS TURBINE COMBUSTOR 293
Downloaded by [Ryerson University] at 07:41 10 May 2013

Figure 9 Left: cF vs. cT (each point represents a location in the combustor). Right: OH mass fraction YOH
vs. cT (each point represents a location in the combustor). The dashed-line represents the pure mixing.

at the start of reactions is estimated to be  0.5 (cF ¼ cT  0.5). It is consistent with


the value derived in the theoretical section.
Figure 10 shows a snapshot of one of the jets discharging into the combustor.
At x=D < 0.2, vortex rings are formed in the jet shear layer. These vortices entrain

Figure 10 Snapshot of one of the jet; 2D axial cut of the field cT and visualization of the vortex core (3D
iso-k 2 surface in grey).
294 C. DUWIG ET AL.

some fluid and promote large scale mixing of burned gases into the incoming fresh
mixture. Downstream of x=D  0.2, the vortex rings merge into spiraling structures.
These structures breakdown in the range x=D in [0.3 0.5] ensuring small scale mixing.
It explains that Figure 9 showed mixing without reaction for x=D < 0.4 with a strong
combustion at x=D > 0.5.

CONCLUDING REMARKS
Flameless combustion has many advantages over more conventional combus-
tion processes. It provides a stable and quiet oxidation process while keeping the
pollutant emissions very low. It is an attractive solution for the design of new gas
turbine burners. In the present paper, the performance of a model gas turbine flame-
less burner was investigated using theoretical, experimental and numerical tools.
Downloaded by [Ryerson University] at 07:41 10 May 2013

The numerical results provide an insight of the combustion process. A large


central recirculation zone transports vitiated hot gases upstream towards the jet-noz-
zles. In the first section of the combustor (x=D < 0.2), rings type of structures are
formed in the shear-layer of the jets. These structures are responsible for entrainment
of vitiated gases by the fresh gases jets. In a second section (0.2 < x=D < 0.5), the
vortex rings merge and breakdown into smaller structures. These structures promote
small scale mixing blending fresh and vitiated gases. The resulting mixture is hot
enough (cT > 0.5) at locations x=D > 0.4 so that combustion can start. Combustion
proceeds as the last fresh gas or not fully burnt parcels mix with vitiated gases.
The theoretical and experimental analysis enables the determination of the
boundaries of the flameless operation regime in term of fuel=air equivalence ratio.
The Large Eddy Simulation results in conjunction with the theoretical and experi-
mental analysis allow measuring the global dilution rate a which was found to be
 0.5, in the present case. In other words, the present device (operating in the flame-
less mode) dilutes the fresh gases with the same amount of vitiated gases before
combustion.
Although this level allowed flameless operations in the present setup, it is
unlikely to be adequate for practical gas turbine operations since the flameless oper-
ation range is relatively narrow. Future work will focus on improving the combustor
aerodynamics making the flameless operation range larger.

REFERENCES

CANTERA. (2005) Object-Oriented Software for Reacting Flows, http://www.cantera.org


Chen, Y.-C., Peters, N., Schneemann, G.A., Wruck, N., Renz, U., and Mansour, M.S. (1996)
The detailed flame structure of highly turbulent premixed methane-air flames. Combust.
Flames, 107, 233–244.
Christo, F.C. and Dally, B. (2005) Modeling turbulent reacting jets issuing into a hot and
diluted coflow. Combust. Flame, 142(1–2), 117–129.
Coelho, P.J. and Peters, N. (2001) Numerical simulation of a mild combustion burner.
Combust. Flame, 124, 503–518.
Dally, B., Riesmeier, E., and Peters, N. (2004) Effect of fuel mixture on moderate and intense
low oxygen dilution combustion. Combust. Flame, 137, 418–431.
FLAMELESS COMBUSTION IN A GAS TURBINE COMBUSTOR 295

DaSilva, C. and Métais, O. (2002) Vortex control of bifurcating jets: A numerical study.
Phys. Fluids, 14, 3798–3819.
Ducros, F., Comte, P., and Lesieur, M. (1996) Large-eddy simulation of transition to turbu-
lence in a boundary layer spatially developing over a flat plate. J. Fluid Mech., 326, 1–36.
Duwig, C. and Fuchs, L. (2006) Modeling of flameless combustion using large eddy simula-
tion. In Proc. TurboExpo 2006, ASME Paper GT-2006-90063.
Duwig, C., Fuchs, L., Griebel, P., Siewert, P., and Boschek, E. (2007) Study of a confined tur-
bulent jet: Influence of combustion and pressure on the flow. AIAA J., 45(3), 624–639.
Flamme, M. (2004) New combustion systems for gas turbines (NGT). App. Thermal Eng., 24,
1551–1559.
Gullbrand, J., Bai, X.S., and Fuchs, L. (2001) High-order Cartesian Grid Method for Calcu-
lation of Incompressible Turbulent Flows. Int. J. Num. Meth. Fluids, 36, 687–709.
Huang, H., Dabiri, D., and Gharib, M. (1997) On errors of digital particle image velocimetry.
Meas. Sci. Technol., 8, 1427–1440.
Jeong, J. and Hussain, F. 1995. On the identification of a vortex. J. Fluid Mech., 285, 69–94.
Jiang, G.S. and Shu, C.W. (1996) Efficient implementation of weighted ENO schemes.
Downloaded by [Ryerson University] at 07:41 10 May 2013

J. Comp. Phys., 126, 202.


Katsuki, M. and Hasegawa, T. (1998) The science and technology of combustion in highly
preheated air. Proc. Combust. Instit., 27, 3135–3146.
Kim, S.H., Huh, K.Y., and Dally, B. (2005) Conditional moment closure modeling of turbu-
lent nonpremixed combustion in diluted hot coflow. Proc. Combust. Instit., 30, 751–757.
Levy, Y., Sherbaum, V., and Arfi, P. (2004) Basic thermodynamics of FLOXCOM, the low-
NOx gas turbines adiabatic combustor. App. Thermal Eng., 24, 1593–1605.
Li, G., Gutmark, E.J., Stankovic, D., Overman, N., Cornwell, M., Fuchs, L., and
Milosavljevic, V. (2006a) Experimental study of flameless combustion in gas turbine com-
bustors. AIAA paper 2006-546.
Li, G., Gutmark, E.J., Overman, N., Cornwell, M., Stankovic, D., Fuchs, L., and Milosavl-
jevic, V. (2006b) Experimental study of a gas turbine combustor. In Proc. TurboExpo
2006, ASME paper GT-2006-91051.
Lückerath, R., Schütz, H., Noll, B., and Aigner, M. (2005) Experimental Investigations of
FLOX Combustion at High Pressure. Proceedings of the Flameless Workshop, Lund
2005 (available upon request: Dragan.Stankovic@vok.lth.se).
Maas, U. and Pope, S.B. (1992) Simplifying chemical kinetics: Intrinsic low-dimensional mani-
folds in composition space. Combust. Flame, 88, 239–264.
Olsson, M. and Fuchs, L. (1996) Large eddy simulation of the proximal region of a spatially
developing circular jet. Phys. Fluids, 8, 2125–2137.
Pierce, C. and Moin, P. 1998. A dynamic model for subgrid scale variance and dissipation rate
of a conserved scalar. AIAA J., 36(7), 1325–1327.
Poinsot, T. and Veynante, D. (2001) Theoretical and Numerical Combustion. RT Edwards,
Philadelphia.
San Diego (2005) Chemical oxidation mechanism release 2005=06=15, http://maemail.ucsd.
edu/combustion/cermech/sandiego20050615.mec.
Wünning, J.A. and Wünning, J.G. (1997) Flameless oxidation to reduce thermal no-forma-
tion. Prog. Energy Combust. Sci., 23, 81–94.

Das könnte Ihnen auch gefallen