Sie sind auf Seite 1von 19

AIAA 2015-4020

Propulsion and Energy Forum


July 27-29, 2015, Orlando, FL
51st AIAA/SAE/ASEE Joint Propulsion Conference

Combustion LES of a Multi-Burner Annular Aero-engine


Combustor using a Skeletal Reaction Mechanism for Jet-
A Air Mixtures

N. Zettervall†,1, E. Fedina†,2, K. Nordin-Bates†,3 E. Heimdal Nilsson‡,4 & C. Fureby†,5



The Swedish Defense Research Agency – FOI,

Div. of Combustion Physics, Lund University, Sweden

In this study we describe combustion simulations of a single sector and a fully annular ge-
neric multi-burner aero-engine combustor. The objectives are to facilitate the understand-
Downloaded by UNIVERSITY OF NEW SOUTH WALES on September 6, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4020

ing of the flow, mixing and combustion processes to help improve the combustor design and
the design process, as well as to show that it is now feasible to perform high-fidelity reacting
flow simulations of full annular gas turbine combustors with realistic combustion chemistry.
For this purpose we use a carefully validated finite rate chemistry Large Eddy Simulation
(LES) model together with a range of reaction mechanisms for kerosene-air combustion.
The influence of the chemical reaction mechanism on the predictive capability of the LES
model, and on the resulting understanding of the combustion dynamics has recently been
proved very important and here we extend this for kerosene-air combustion. As part of this
work a separate study of different kerosene-air reaction mechanism is comprised, and based
on this evaluation the most appropriate reaction mechanisms are used in the subsequent
LES computations. A generic small aircraft or helicopter aero-engine combustor is used,
and modeled both as a conventional single sector configuration and more appropriately as a
fully annular multi-burner configuration. The single-sector and fully annular multi-burner
LES predictions are similar but with the fully annular multi-burner configuration showing
different combustion dynamics and mean temperature and velocity profiles. For the fully
annular multi-burner combustor azimuthal pressure fluctuations are clearly observed, re-
sulting in successive reattachment-detachment of the flames in the azimuthal direction.

I. Introduction
For civilian and military aeropropulsion, including turboshaft engines for small aircrafts and helicopters, turbo-
fan engines for large aircrafts, and afterburning turbojet- and turbofan engines for combat aircrafts there is cur-
rently no practical alternative for gas turbines as power source. Modern aeropropulsion gas turbine engines usu-
ally have an annular combustion chamber with multiple burners sharing a common fuel supply line. Constraints
on such gas turbines include velocity and temperature profiles delivered to the turbine (primarily affecting tur-
bine life), pressure drop across the combustor (affecting thermal efficiency), the capability to withstand flame
extinction, blow-out and pressure oscillations, the ability to relight at high altitudes, unsteady thermal loads and
mechanical vibrations, as well as the ever more stringent emission regulations on CO, CO2, NOX, unburned hy-
drocarbons and smoke. Although the current tendency in gas turbine design is towards fuel lean and premixed
conditions to reduce emissions, most aeropropulsion gas turbines make use of spray flames. The fuel spray is
usually created by passing the liquid fuel through an atomizer, in which a liquid fuel film is dispersed into drop-
lets that are sprayed into the combustor together with air under swirling conditions. Adjacent spray flames in an
annular combustor configuration interact with each other at the edges of the spray cone and at the flame tips,
and with the unsteady pressure field in the combustion chamber. The individual flames are further influenced by
the unsteady fuel flow in the fuel manifold, and by the flow through the mixing, dilution and film cooling holes.

1
Researcher, Defense Security Systems Technology, SE 147 25 Tumba, Stockholm, Sweden
2
Researcher, Defense Security Systems Technology, SE 147 25 Tumba, Stockholm, Sweden
3
Researcher, Defense Security Systems Technology, SE 147 25 Tumba, Stockholm, Sweden
4
Researcher, Div. of Combustion Physics, Lund University, Box 118, SE 221 00, Lund, Sweden
5
Research Director, Defense Security Systems Technology, SE 147 25 Tumba, Stockholm, Sweden, Associate fellow

1
American Institute of Aeronautics and Astronautics

Copyright © 2015 by Zettervall N., Fedina C., Nordin-Bates K., Heimdal-NilssonE. & Fureby C.. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.
The complexity of a normal gas turbine engine, together with the turbulent flow of air, fuel and hot combustion
products, results in a complex flow that is not fully understood and often can be improved.
Accurate observations and quantitative measurements in realistic engine configurations are difficult and
expensive, and are thus in short supply. An alternative to experiments, in particular for studying design altera-
tions and exploration of the fundamental physics, is to use high-fidelity numerical simulations. The rapid devel-
opment in computational capability during the last decade has enabled high-fidelity simulations of sector mod-
els and fully annular models as recently reviewed by Gicquel et al., [1]. The overall complexity and computa-
tional cost of the numerical simulation is determined by the geometric representation of the combustor, the sim-
ulation model, and the complexity and detail of the chemical reaction mechanism.
Current steady-state Reynolds Averaged Navier Stokes (RANS) combustion models, e.g. [2-3], are short
of the accuracy required for reliable predictions of transient phenomena such as vortex shedding, shear layer
mixing, acoustics, combustion instabilities, blow-out, self-ignition and emissions. A more attractive approach is
to use combustion Large Eddy Simulation (LES), [4-7], in which the large scales of the flow are explicitly sim-
Downloaded by UNIVERSITY OF NEW SOUTH WALES on September 6, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4020

ulated and only the small (subgrid) scales are modeled, [8]. It has previously been demonstrated that LES cap-
ture mixing better than RANS, and turbulence chemistry interactions even better, [9], and the combustion dy-
namics, and its effect on mixing and reactions can hence only be captured by combustion LES or even more ad-
vanced models such as Direct Numerical Simulation (DNS), [10]. DNS is however too expensive for studying
actual combustion systems such as aeropropulsion gas turbines. Combustion LES has matured, so that single
sector combustors are routinely computed, [11-12], whereas only a few multi-sector, fully annular, gas turbine
combustor LES, capturing all acoustic modes and interactions, have been performed, [13-15]. Most of these sin-
gle-sector and multi-sector LES gas turbine combustor studies employ global, optimized, one- or two-step reac-
tion mechanisms, [13-15], although it is known, e.g. [16], that incorporation of more detailed reaction mecha-
nisms, such as skeletal reaction mechanisms, is important. The fuel typically used in aeropropulsion gas turbines
is kerosene, which is made up of hydrocarbons with carbon numbers between 8 and 16, having an average mo-
lecular formula between C10H20 and C12H26, depending on the fuel characteristics. Several detailed reaction
mechanisms are available for different kerosene blends, e.g. [17-20], with [20] being the most detailed, includ-
ing 19341 elementary reactions, together with a few global and quasi-global reaction mechanisms, such as [21-
23], but hardly any skeletal reaction mechanisms suited for gas turbine computations are available.
In this paper we examine the influence of the reaction mechanism on the flow, mixing and combustion
dynamics in a generic turboshaft-engine gas turbine combustor burning C12H23 using LES in order to further en-
hance our understanding of gas turbine combustion. Before performing the LES computations, a detailed inves-
tigation of various global, skeletal and detailed C12H23-air reaction mechanisms is carried out with the intent of
down-selecting a few reaction mechanisms suitable for gas turbine combustion LES. A skeletal C12H23-air reac-
tion mechanism is developed as part of this work to provide an appropriate skeletal reaction mechanism. LES
simulations, using global and skeletal reaction mechanisms, are then carried out for a single sector configuration
and for a fully annular multi-burner configuration. These LES predictions are examined and compared in order
to elucidate the detailed combustor physics, and to emphasize the difference between the different global and
skeletal reaction mechanisms tested.

II. Generic Gas Turbine Combustor Model


The gas turbine combustor of interest here is a generic aeropropulsion combustor of reverse-flow type and is
shown in figure 1a. The combustor casing accommodates twelve fuel-air spray nozzles. Air enters through a dif-
fuser and is lead through the air channel into the fuel-air spray nozzles, figure 1b, where the air is mixed with
fuel prior to being discharged into the flame tube. The remaining air fills the space between the flame tube and
the combustor wall, providing cooling of the flame tube walls and dilution of the primary combustion region
through a number of film cooling and dilution holes. The flame tube consists of curved elements with built-in
film cooling holes. Two computational models, of which the first consists of a single 30° sector with periodic
boundary conditions, and the second of the fully annular combustor, have been generated with 6 and 72 million
cells, respectively, figure 1c and 1d. In order to resolve the details of the fuel-air spray nozzles, the dilution and
mixing holes as well as the film cooling holes, unstructured tetrahedral grids with local refinement patches have
been employed throughout the computational domain. Conventional open inflow/outflow boundary conditions
for all dependent variables have been used together with isothermal no-slip wall boundary conditions assuming
a wall temperature provided by experimental data. The initial conditions for the combustion LES were obtained

2
American Institute of Aeronautics and Astronautics
from single sector RANS computations that in the case of the fully annular configurations were rotated and
mapped onto the full annular combustor domain. The nominal Reynolds number is Re≈ 2.20·106, whereas the
nominal combustor swirl number is S≈0.52.

(b)
Downloaded by UNIVERSITY OF NEW SOUTH WALES on September 6, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4020

(a) (c) (c)

Figure 1. (a) Perspective view of the aeropropulsion gas turbine combustor with the twelve fuel-air spray
nozzles, (b) detailed view of a fuel-air spray nozzle, (c) annular combustor grid and (d) detailed view of the
combustor cross-sectional profile with the unstructured grid following the geometry, configuration geometry
of the single sector simulations.

III. LES Model and Numerical Methods


The reactive flow equations are the balance equations of mass, momentum and energy describing convection,
diffusion and chemical reactions, [2]. LES is based on a separation of scales, which is achieved via spatial low-
pass filtering. Physical processes occurring on scales larger than the filter width, ∆, are resolved, whereas phys-
ics occurring on scales smaller than ∆ are modeled, [6]. For a linear viscous mixture with Fourier heat conduc-
tion and Fickian diffusion the LES equations result from filtering the reactive flow equations so that,

( ∂t (ρ)+∇⋅(ρv )=0,
*
* ∂t (ρY i )+∇⋅(ρv Y i )=∇⋅(D i ∇Y
 i −b i )+w  i,
) (1)

∂t (ρv )+∇⋅(ρv ⊗ v )=−∇p+∇⋅(2µD− 3µ(∇⋅v )I−B),
2
*
* 
∂t (ρE)+∇⋅(ρ 
v E)=∇⋅(−p v +2µD  E ),
 v − 23µ(∇⋅v )v +κ∇T−b
+

in which, ρ , v , T and Y  i are the filtered density, velocity, temperature and species mass fractions, respec-
tively, and D=2 (∇v+∇v ) the rate-of-strain tensor. The filtered pressure is p≈ρRT , in which R is the (composi-
1 T

tion dependent) gas constant. The mixture is described by the viscosity µ, and the species and thermal diffusivi-
ties, D i=µ/Sc i and κ=µ/Pr , in which Sci and Pr are the Schmidt and Prandtl numbers, respectively. The total en-
 ε+1 v 2+k is composed of the internal energy ε=Σi (Y
ergy E=  i (h θi,f + ∫ T C p,i (T)dT ))−p/ρ , the kinetic energy 12 v 2 , and
2 T0
the subgrid kinetic energy k. The chemical kinetics of a reaction mechanism, Pjirℑi⇔Pjipℑi , in which ℑi repre-
sents species i, enter equation (1) by means of the low-pass filtered species reaction rate, w  i=M i Pij w
 j , in which
 j=A jT n j e−TA /T Π Nk=1 (ρYk )bk are the Arrhenius reaction rates, Pji=Pjir−Pjip the stoichiometric coefficients, Mi the
w
species molar mass, Aj the pre-exponential factors, TA,j the activation temperatures, nj the temperature exponent
and bj the reaction orders for reaction j.
The subgrid flow physics is hidden in the subgrid stress and flux terms B=ρ(v~
~  Y i )
⊗v−v ⊗v ) , b i=ρ(vYi −v
~
and b E =ρ(vE−v E)  6], that are here modeled by mixed models, [24-25], such that B=ρ(v ⊗v −v ⊗v)−2µ ~  
~ −v Y )− µk ,∇[4,
b i=ρ(v Y Y b =ρ(
~

v 
E −v
  µ k ∇E
E)−
k DD ,

i and in which the subgrid viscosity µ k =c k ρΔk is obtained from an


1/2
i i Sc t E Prt
modeled transport equation for k, [26], and Sct=Prt=0.7. A near-wall model, [27], is used to adjust k and µk at
the wall so that the log-law is satisfied.
The filtered reaction rates are here modeled using a heterogeneous multi-scale method of the first genera-
tion, i.e. the Partially Stirred Reactor LES model (LES-PaSR), [28-31], whereby simplified balance equations
are solved on a subgrid level to model the subgrid mixing and reactions. More specifically, each LES cell is

3
American Institute of Aeronautics and Astronautics
considered to consist of reacting fine structures (*) and surroundings (0), interacting through the simplified bal-
ance equations ρ(Yi*−Yi0 )/τ*=w i (Yi*,T* ) and ρΣi=1
N
(Yi*h*i −Yi0 h 0i )/τ*=Σi=1
N θ
 i (Yi*,T* ) , in which τ* is the residen-
h i,f w
ce time. By introducing the reacting volume fraction, γ* , as the ratio of the volume of the fine structures to the
LES cell volume, Y i=γ*Yi*+(1−γ* )Yi0 and T=γ *T*+(1−γ* )T 0 , these balance equations become,

') ρ(Y* − Y
i
 i )=(1−γ*)τ*w  i (Yi*,T* ),
( (2)
N
)* ρΣi=1  i h i (T))=(1−γ
(Yi*h*i (T* )− Y  *)τ* Σi=1
N θ
 i (Yi*,T* ).
h i,f w

The filtered reaction rates can generally be expressed in terms of a Probability Density Function (PDF), ℘, such
that w  Ψ #)w
 i= ∫℘(Ψ, Ψ,  i (Ψ)dΨ , with Ψ={ρ, T, Yi } , [32]. As a consequence of that the flow is decomposed into
reacting fine structures and surroundings the PDF can be considered as ℘(Ψ)=γ*δ(Ψ * )+(1−γ* )δ(Ψ 0 ) , in which δ
is the Dirac function, resulting in that the filtered species equations (12) can be explicitly expressed as,
Downloaded by UNIVERSITY OF NEW SOUTH WALES on September 6, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4020

 i )+∇⋅(ρv Y
∂t (ρY  i )=∇⋅(D i ∇Y
 i −b i )+ γ* w
 i (ρ,T* ,Y*i ) +(1−γ* )w
 i (ρ,T 0 ,Yi0) . (3)

The reacting volume fraction, γ* , and the associated residence time, τ* , must be modeled to obtain a closed set
of equations. This can be accomplished in many different ways, and here we use the model proposed by Sabel-
nikov & Fureby, [30], which is based on spatio-temporal intermittency, modeled as γ*≈τ c /(τ*+τ c ) . The chemical
time-scale must be representative of the overall combustion reaction, and is modeled by τ c ≈δ u /su , where δ u and
su are the laminar flame thickness and flame speed, respectively. The modeling of τ* is based on the observa-
tion that the fine structure area-to-volume ratio is given by the dissipative length scale,  D=(ν∆/v")1/2 , determined
by the viscosity, ν, and the subgrid velocity stretch, v!/∆ , where v!= 2k/3 , is the subgrid velocity, and that the
velocity influencing these structures is the Kolmogorov velocity, v K , so that τ*= D /v K . By combining the ex-
pressions for  D and v K , utilizing the Kolmogorov length and time scales  K =(ν3 /ε)1/4 and τ K =(ν/ε)1/2 , in which
ε=(v")3/∆ is the dissipation, results in that the residence time is τ*= τ K τ∆ , with τ∆ =∆/v" being the shear time-
scale. This was recently found, [33], to accurately model the first and second order statistics of the dissipation,
and thus also of small-scale mixing, which, in turn, is essential to the onset of chemical reactions.
The resulting LES equations are solved by a finite volume based code developed from the OpenFOAM
C++ library, [34], which is based on an unstructured collocated finite volume method using Gauss theorem to-
gether with a multi-step time-integration method, [35]. Here, the time-integration is performed using a semi-im-
plicit second order accurate two-point backward differencing scheme. Convective fluxes are reconstructed using
multi-dimensional cell limited linear interpolation, whereas diffusive fluxes are reconstructed using a combina-
tion of central difference approximations and gradient face interpolation to minimize the non-orthogonality er-
ror. Here, a compressible version of the Pressure Implicit with Splitting of Operators (PISO) method, [36], is
used to discretize the pressure-velocity coupling, using the thermal equation of state. The combustion chemistry
is integrated separately, based on a Strang-type operator-splitting scheme, [37], using a Rosenbrock time-inte-
gration algorithm, [38]. The scheme is second order accurate in both space and time, and the equations are solv-
ed sequentially, with iteration over the explicit source terms, with a Courant number restriction of 0.5.

IV. Kerosene-Air Combustion Chemistry


The aviation fuel Jet A is a kerosene grade fuel suitable for most aircraft engines. Kerosene is a blend of hydro-
carbons with carbon numbers between 8 and 16, typically with up to 25% aromatic hydrocarbons, 30-35% sat-
urated cyclic hydrocarbons, and the rest linear and branched hydrocarbons, [39], usually resulting in an average
molecular formula of approximately C12H23. Complex reaction mechanisms describing kerosene combustion
typically include a number of hydrocarbon species of different size and functionality, and for the simulations
presented here, a representative mix of these components is used. These surrogate mixtures need to mimic the
chemistry of real fuels, as described by Dooley et al., [40]. Detailed reaction mechanisms for kerosene-air com-
bustion was reviewed by Dagaut & Cathonnet, [17], and later discussed by Ranzi et al., [39]. Recent mecha-
nisms with high level of detail include Westbrook et al., [19], and Wang et al., [41]. Ji et al., [42], use experi-
mental data and the detailed mechanism of Wang et al., [41], to study the laminar flame speed and extinction
strain rate of C5 to C12 n-alkane-air flames and conclude that both these parameters are alike for n-alkane-air
flames under comparable conditions. Mechanism predictions and experimental data for C10H20 might thus be

4
American Institute of Aeronautics and Astronautics
useful for comparison with C12H23, as considered by Ranzi et al., [20], but one should keep in mind that the pre-
sence of cyclic hydrocarbons in real kerosene might under some circumstances play an important role and the
real fuel will then not be comparable with a simple n-alkane. Laminar flame speed data from experimental
measurements are available from Kumar et al., [43], Hui et al., [44], and Chong et al., [45]. All chemical kinetic
simulations described in the following investigation have been performed using the software CANTERA, [46],
and CHEMKIN, [47].
Figure 3 compares experimental data from [43-45] with results from the detailed reaction mechanism of
Ranzi et al., [39], R5738, the two-step global reaction mechanisms of Franzelli et al., [20], 2S-KERO and 2S-
KERO-BFER, the four-step global reaction mechanism of Jones & Tyliszczak, [48], JT4, the 28-step skeletal
reaction mechanism of Kundu, K28, [49], and a novel 57-step skeletal reaction mechanism, Z57, see Appendix
1, proposed to provide an accurate and compact skeletal reaction mechanism appropriate for LES simulations.
The motive for developing a novel, compact, skeletal mechanism for kerosene-air combustion is the lack of ac-
curate, reliable, compact and well-documented kerosene-air mechanisms appropriate for LES modeling in the
Downloaded by UNIVERSITY OF NEW SOUTH WALES on September 6, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4020

open literature. R5738, is referred to as a skeletal mechanism by its authors, but has a very high level of com-
plexity, being composed of 5738 reactions, which in turn is being constructed from an even larger detailed reac-
tion mechanism with 17848 reactions using a lumping procedure. In the present work, R5738 will be used as a
reference mechanism for flame speed, temperature and major species. Global- and smaller skeletal mechanisms
treat the fuel molecule, C12H23, directly, whereas detailed or large skeletal mechanism, such as R5738, consists
of a number of fuel species of different size and functionality. An appropriate surrogate mixture therefore needs
to be defined, having a mixture composition that is similar to that of kerosene and giving the accurate chemical
reactivity. Based on an evaluation by Dooley et al., [40], and considerations from Slavinskaya et al., [56], a sur-
rogate mixture of n-hex-adecane, trimethylbenzene, n-propylbenzene and iso-octane was considered appropriate
for this investigation.
The proposed Z57 skeletal reaction mechanism is developed from a customized version of the Smooke &
Giovangigli, [51], SG25, reaction mechanism for the CH4-O2, CO-O2 and H2-O2 chemistry, which performs sat-
isfactory for CH4-air combustion, [15]. In the proposed Z57 mechanism, most of the aldehyde and low temper-
ature reaction paths are removed as well as all of the alcohol reaction paths. The primary fuel breakdown reac-
tion is an artificial thermal decomposition reaction producing C2H4 and C2H3. Two synthetic fuel-oxidizing re-
actions are comprised in which the fuel reacts with H and OH, respectively, creating intermediate C2-hydrocar-
bons. These intermediate hydrocarbons further react with other major and minor species, producing radicals and
additional major species, which in turn, couple to the underlying CH4-O2 mechanism. As evident from the com-
parisons in figure 3 and 4 the proposed skeletal reaction mechanism compares favorably with the experimental
data and the detailed mechanism.
Figure 2a shows a reaction path diagram of the C species in Z57. As previously mentioned, the fuel
breaks down into smaller intermediate C2-hydrocarbons, which continues to oxidize into gradually smaller spe-
cies. One specie not included in figure 2a is C2H, which results from C2H2, hence completing the C2-chain from
C2H5 down to C2H. C2H4 and C2H2 both play important roles in the structure of the Z57 reaction mechanism.
Note that C2H4 can recombine to form C2H5 but the oxidation of that product does not result in a C2-hydrocarbon
but rather results in two methyl (CH3) radicals, via C-C bond breaking. From CH3 and downwards, the chain is
straightforward, CH3→CH2O→HCO→CO→CO2. In figure 2b a reaction path diagram based on C16H34, which
is one of the four species composing the fuel for the R5738 mechanism, is shown. The number of C-species pre-
sent in the figures 2a and 2b has been reduced to the 15 most common C containing species in the reaction path
from fuel to CO2. Reaction paths of this type show how extremely complex combustion of large hydrocarbons
is, with a large number of viable reaction paths and species, and another 20 or 30 species could easily be added,
making the reaction path diagram even more complex. The final reaction intermediates from C2H5 to CO2 are
common between the Z57 and n-hexadecane of the R5738 mechanisms, but with many more available reaction
paths in the more complex mechanism. It was discussed by Dooley et al., [40] that the most important aspect of
a kerosene mechanism is to have a fuel breakdown part that delivers accurate composition of smaller species in-
to the reactive later part of the mechanism, since the final steps are important for the reactivity and thus for the
combustion properties.
Some of the aforementioned global, skeletal and detailed reaction mechanisms will next be compared for
ignition delays, τig, extinction strain rates, σe, laminar flame speeds, su, flame temperatures, Tmax, and major spe-
cies profiles in flames. In addition, the variation of laminar flame speed with pressure, p, and temperature, T

5
American Institute of Aeronautics and Astronautics
will be discussed. The ignition delay time was computed for atmospheric pressure conditions. Strain rate for
counter-flowing mixture of fuel at 300 K and oxidizer (air) at 559 K, at a pressure of 7.0 atm. Laminar flame
speed, flame temperature and major species such as CO, CO2, H2, H2O were computed for T=400 K and p=1
atm, whereas laminar flame speeds were also computed for p=1 atm and T=400, 470 and 600 K, respectively,
and for T=400 K and p=1, 3 and 10 atm, respectively.
Downloaded by UNIVERSITY OF NEW SOUTH WALES on September 6, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4020

(a) (b)

Figure 2. Reaction path diagram of (a) the proposed 57-step skeletal mechanism, Z57, see Appendix 1, and
(b) of one of four fuel species in the R5738 detailed mechanism, [39].

For su at 1 atm and 400 K, figure 3a, the results of the mechanisms investigated vary significantly, with
the R5837 mechanism and Z57 matching the experimental data very well. JT4 predicts su well up to φ=0.9, and
2S-KERO up to φ=0.8, whereas 2S-KERO-BFER performs up to even moderately fuel rich conditions at φ=1.2,
after which it slowly starts to underpredict su. The K28 reaction mechanism fails to perform over the complete
range of equivalence ratios, underpredicting massively up to highly fuel rich conditions. Figure 3b shows Tflame,
again with significant deviations in performance between the different mechanisms. The best performance,
compared to the detailed mechanism R5738, is again achieved by Z57, albeit with overpredictions over much of
the φ interval, followed closely by K28. 2S-KERO-BFER and 2S-KERO underpredict Tflame up to fuel rich con-
ditions where they start to overpredict Tflame slightly. The JT4 fails to perform with large underpredictions over
the whole φ interval.
Figures 3c and 3d compare how su vary for different temperatures and pressures: figure 3c shows how su
vary with T, for experimental data, [43, 45], and for selected reaction mechanisms. Most of the reaction mecha-
nisms, with the exception of K28, demonstrate correct behavior with increasing T, but with the same individual
variations as observed in figure 3a. 2S-KERO-BFER, Z57, JT4 and R5738 all marginally underpredict su at 470
K compared to the experimental data, whereas 2S-KERO overpredicts su for fuel rich conditions. The behavior
is similar for 1 atm and 600 K, but since experimental data is missing, the primary reference is the R5738 mech-
anism. In figure 3d the pressure dependence for su is presented for T=400 K and p=1, 3 and 10 atm, respectively,
together with experimental data for 1 atm, [43-44], and 3 atm, [43]. Usually, K28 fails to perform whereas 2S-
KERO performs well up to φ=0.8 after which it overpredicts su. 2S-KERO- BFER and Z57 both perform well
compared to experimental data and R5738, albeit with individual deviations, whereas JT4 overpredicts su as the
pressure increases.
τign, shown in 3e, shows large variations, with Z57 having the longest ignition delay times. From R5738,
a slight curvature at lower values of T is seen that is common for long alkane fuels. This trend is not captured by
any of the skeletal and global mechanisms, probably due to the fact that such mechanisms do not include the in-
termediate chemistry of importance in low temperature ignition. As with τign, σe, featured in figure 3f, is com-
monly governed by intermediate chemical species that are usually not included in small mechanisms. R5738
shows an extinction curve that is reasonable in comparison to experimental results of large hydrocarbons. The
Z57 mechanism gives a slightly higher σe curve, however, following the same trend as R5738, whereas the other
global and skeletal mechanisms show unexpected behavior with respect to T. In particular is JT4 problematic,
overpredicting σe and underpredicting T compared to R5738.

6
American Institute of Aeronautics and Astronautics
(a) (b)
Downloaded by UNIVERSITY OF NEW SOUTH WALES on September 6, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4020

(c) (d)

(e) (f)

Figure 3. Comparison of different C12H23-air reaction mechanism predictions, (a) laminar flame speed vs.
equivalence ratio, (b) temperature, (c) laminar flame speed vs. temperature, (d) laminar flame speed vs. pres-
sure, (e) ignition delay time and (f) extinction strain rate. Legend: (—) 2S-KERO by Franzelli et al., [20],
(—) 2S-KERO-BFER by Franzelli et al., [20], (—) JT4 by Jones & Tyliszczak, [53], (—) K28 by Kundu,
[49], (—) Z57 as proposed here, (o) Kumar et al., [43], (¡) Hui et al., [44], (+) Hui et al., [51], (*) Chong et
al., [45] and (—) R5738 by Ranzi et al., [39]. Solid lines in figure 3c correspond to 1atm, 400 K, dot-dashed
to 1 atm, 470 K and dashed to 1atm, 600 K. Solid lines in figure 2d corresponds to 1 atm, 400 K, dot-dashed
to 3 atm, 400 K and dashed to 10 atm, 400 K.

Figure 4 shows the variation of major species, CO2, CO, H2O and H2, from different reacti-on mecha-
nisms, with the equivalence ratio, φ. Figure 4a presents the maximum CO2 concentrati-on, where again Z57
shows the best agreement with R5738, albeit a small overprediction for fuel rich conditions. K28 performs rea-
sonably well up to stoichiometric conditions, φ=1, after which K28 starts to overpredict su. 2S-KERO-BFER
and 2S-KERO both fail to predict the CO2 concentration and are almost as deficient in their predictions as JT4.
Figure 4b shows the CO concentration, and again the prediction of Z57 agrees well with that of R5738. The 2S-
KERO, 2S-KERO- BFER and JT4 reaction mechanisms all fail in predicting CO across the whole equivalence
ratio interval. Also the K28 reaction mechanism fails to predict CO across the whole equivalence ratio interval.
A small underprediction of CO, for φ>1, is found for Z57 that correlates with the overprediction in CO2, show-
ing a potential need for improved CO2-CO-chemistry. Figure 4c shows the H2O concentration, and all mecha-

7
American Institute of Aeronautics and Astronautics
nisms, except JT4, perform up to φ≈1, after which the two two-step global reaction mechanisms overpredict
H2O. Finally, figure 4d presents the H2 concentration, for which Z57 performs well, whereas JT4 and K28 fail.
Downloaded by UNIVERSITY OF NEW SOUTH WALES on September 6, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4020

(a) (b)

(c) (d)

Figure 4. Comparison of different C12H23-air reaction mechanism predictions, (a) CO2-concentration vs.
equivalence ratio, (b) CO-concentration vs. equivalence ratio, (c) H2O-concentration vs. equivalence ratio
and (d) H2-concentration vs. equivalence ratio. Legend: (—) 2S-KERO by Franzelli et al., [20], (—) 2S-
KERO-BFER by Franzelli et al., [20], (—) JT4 by Jones & Tyliszczak, [48], (—) K28 by Kundu, [49], (—)
Z57 proposed here and (—) R5738 by Ranzi et al., [39].

V. Single Sector Gas Turbine Combustor Results


The single sector computational set-up will primarily be used to evaluate the differences between LES predic-
tions using the global and skeletal reaction mechanisms compared in Section 4. Based on these results we find
that the global two-step reactions mechanisms 2S-KERO and 2S-KERO-BFER predict the laminar flame prop-
erties reasonably well, with the 2S-KERO-BFER being the more accurate of the two due to the pre-exponential
factor adjustment. The four-step global reaction mechanism JT4 is the least accurate global reaction mechanism
as it underpredicts the flame temperature, ignition delay time, CO2 and H2O, whilst overpredicting σe, CO and
H2. Regarding the skeletal reactions mechanisms K28 and Z57, K28 fails to predict the laminar flame speed, its
variation with p and T, and the extinction strain rate, whereas it succeeds in predicting the flame temperature as
well as the ignition delay time. In addition, are some major species well predicted whereas others are not so well
predicted. Z57, on the other hand, is found to be astonishingly accurate compared to R5738 and the experi-
mental data. Based on this study, two single sector LES are set up using the 2S-KERO-BFER and Z57 reaction
mechanisms. These simulations are started from the same initial conditions, using similar boundary conditions,
and are continued for 100 ms, with statistical averaging performed over the last 50 ms. To simplify the model-
ing, the fuel spray is replaced by a gaseous kerosene-air mixture with corresponding features.
Figure 5 summarize the results of these LES predictions in terms of (a) to (b) pressure fluctuations to-
gether with an iso-surface of the flame, (c) to (d) velocity vectors together with an iso-surface of the flame, (e)
to (f) temperature distributions together with an iso-surface of the flame, and (g) to (h) mixture fractions (uti-
lizing carbon element balance) together with an iso-surface of the flame. For the global reaction mechanism, 2S-
KERO-BFER, the flame is here represented by an iso-surface of YC12H23, and for the skeletal reaction mecha-
nism, Z57, the flame is represented by semi-transparent iso-surfaces of YC12H23 (in green), YC2H4 (in yellow) and

8
American Institute of Aeronautics and Astronautics
YHCO (in red). Moreover, the velocity vector field is colored by the velocity magnitude and the red solid line in
the mixture fraction distributions denotes the stoichiometric mixture fraction, zs=0.055. For all panels, the same
ranges of values are used to facilitate comparison.

(a) (b) (c)


Downloaded by UNIVERSITY OF NEW SOUTH WALES on September 6, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4020

(d) (e) (f)

(g) (h)

Figure 5. Instantaneous distributions of pressure fluctuations using (a) 2S-KERO-BFER and (b) Z57, veloc-
ity magnitude using (c) 2S-KERO-BFER and (d) Z57, temperature using (e) 2S-KERO-BFER and Z57, (f),
and mixture fraction using (g) 2S-KERO-BFER, and (h) Z57, together with the flame represented by iso-
surfaces of C12H23 (—), C2H4 (—) and HCO (—). In (j) to (l) the solid read line denote the stoichiometric
mixture fraction, zs=0.055.

Regarding the pressure fluctuations in figures 5a to 5b we find that the combustor is populated with alter-
nating patches of high and low pressures with respect to the time-averaged combustor pressure, indicating the
presence of (weak) pressure waves throughout the combustor. In the case of 2S-KERO-BFER a more distinct
radial pressure wave can be observed. From the velocity distributions in figures 5c to 5d we recognize the whirl-
ing annular shear layers discharging from the swirler-injector, a Central Recirculation Zone (CRZ), a Toroidal
Vortex (TV) forming, between the dump plane, swirler-injector and annular shear layers, high-speed air jets dis-
charging through the film-cooling, mixing and dilution holes in the liner, and dividing the flame-tube chamber
into regions, and the high velocities in the diffusor and in the curved end-section of the flame-tube, where hot
combustion products are expelled due to the volumetric expansion caused by the exothermicity. The velocity di-
vides the flame-tube into (i) a flame region located between the dome, swirler-injector and first rows of mixing
holes, (ii) a mixing region in which air is injected into the flame-tube through the liner mixing and dilution holes
to cool the flame, and to provide air for the post-combustion region (iii) that occurs after the second rows of
mixing holes in the liner, in which the remaining unburned hydrocarbons burn with the air injected through the
mixing and dilutions holes. The flame, visualized by iso-surfaces of YC12H23, YC2H4 and YHCO, is located in the
flame region, and takes the shape of a hollow cone, surrounding most of the CRZ. From the temperature distri-
butions in figures 5e to 5f the division into three regions is even more pronounced as is the dilution of cold air
through the mixing and dilution holes. The high-temperature region is confined to the flame region by the car-
bon-containing annular shear layers from the swirler-injector and the discharge of cold air through primarily the

9
American Institute of Aeronautics and Astronautics
first row of mixing holes, whereafter the temperature gradually decrease throughout the remainder of the flame-
tube since even more air is injected through the second row of mixing and dilution holes. The outer (cold) side
of the flame shows marks of the Kelvin-Helmholtz instabilities from the annular shear layers, whereas the in-
trinsic flame structure, defined by the reaction mechanism, figure 2a, reveals a complex pattern of interlaced
patches of different species mass fractions. For Z57, C2H4 is the first oxidation product of C12H23, starting to
form in the pre-heat layer and reaches its peak value in the inner layer, whereas HCO develops in a thin region
between the inner layer and the oxidation layer, in which most of the OH is formed. The mixture fraction, z,
shown in figures 5g to 5h represents the fraction by mass in the mixture of carbon originating from C12H23, and
the stoichiometric value z=zst thus represent the line or surface preferred by a diffusion flame. The flame is lo-
cated close to zst in the most upstream part of the flame region. The difference in p, v, T and z distributions be-
tween the three single sector models is surprisingly small, in particular considering the differences in reaction
mechanism nature described in Section 4. Next we will examine the differences between 2S-KERO-BFER and
Z57 in more detail.
Downloaded by UNIVERSITY OF NEW SOUTH WALES on September 6, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4020

Figures 6a and 6b present the distribution of T versus z, colored by the rate-of-strain, ||D||, for the 2S-
KERO-BFER and Z57 reaction mechanisms, respectively. Overlaid is the mixing line, connecting the C12H23
and air states, and the two equilibrium lines, connecting the C12H23 and air states to the stoichiometric line (ob-
tained from a Burke-Schumann analysis, [52], from a single-step irreversible reaction mechanism). In diffusion
flames, T (and all species mass fractions, Yi) is located between the mixing and equilibrium lines. Mixing lines
are lines along which C12H23 and air would mix without reaction, and points located close to the mixing line are
either close to extinction or not yet ignited. Similarly, points located in the vicinity of the lean and rich equilib-
rium lines indicate vigorous burning. The distributions from 2S-KERO-BFER and Z57 are notably different,
suggesting that the turbulence chemistry interactions and the influence of the combustion chemistry modeling
are important. On the lean size of zst, i.e. on the outside of zst in figures 5j and 5l, the temperature distributions
from the 2S-KERO-BFER and Z57 reaction mechanisms are very similar, and show a lean turbulent flame lo-
cated around the lean equilibrium line, primarily associated with low rate-of-strain. On the rich side of zst, i.e. on
the inside of zst in figures 5j and 5l, in which the turbulent flow is very complex (including the CRZ and fre-
quent reg-ions of high rate-of-strain), the temperature distributions from the 2S-KERO-BFER and Z57 reaction
mechanisms are very different: 2S-KERO-BFER supports a turbulent flame close to the rich equilibrium line for
zst<z<0.3, whereafter the combined effects of decreased laminar flame speed at rich mixtures, turbulence and
quenching due to high rate-of-strain rapidly decrease the temperature along a pseudo-equilibrium line ending
around z≈0.5. Z57 cannot support a flame at high rate-of-strain, resulting in a rapid decrease in temperature on
the rich side. These results are supported and explained by the extinction strain rate results of figure 3f.

(a) (b)

Figure 6. Scatter plots of T versus z colored by ||D|| for the (a) 2S-KERO-BFER reaction mechanism and the
(b) Z57 reaction mechanisms.

Figures 7a and 7b show corresponding scatter-plots of major species, YC12H23, YCO2, YCO, YH2O, and mi-
nor species, YC2H4, YHCO and YOH, versus z, for the 2S-KERO-BFER and Z57 reaction mechanisms, respectively.
Regarding the distribution of major species we find almost a linear relationship between YC12H23 and z, whereas
the other major species show more complicated non-linear distributions. We also find that the distributions from
the Z57 reaction mechanism are associated with a larger scatter than that of the 2S-KERO-BFER reaction
mechanism, which is related to the complexity of the mechanism and the lower extinction strain rate of Z57

10
American Institute of Aeronautics and Astronautics
compared to the 2S-KERO-BFER reaction mechanism. The peak values of CO2 and H2O occur close to zst,
whereas that of CO is shifted towards higher values of z. Regarding the minor species, OH and HCO peak close
to zst as expected, whereas C2H4 shows a more distributed profile, and peaks at higher values of z. According to
the reaction path diagram in figure 2a, C2H4 is the primary fuel breakdown product, which is further broken
down into HCO, CO and CO2, with HCO associated with the heat-release. The high turbulent rates-of-strains are
accountable for most of the broadening of the scatter-plot profiles due to quenching of individual reactions, and
thus are the Z57 profiles broader than those of the 2S-KERO-BFER reaction mechanism.
Downloaded by UNIVERSITY OF NEW SOUTH WALES on September 6, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4020

(a) (b)

Figure 7. Scatter plots of species versus z for the (a) 2S-KERO-BFER and (b) Z57 reaction mechanisms.
Legend: (n) C12H23, (n) CO2, (n) CO, (n) H2O, (n) C2H4, (n) HCO and (n) OH.

Figure 8 shows the distribution of the turbulent Damköhler number, Da t =τ t /τ c , versus the turbulent
Reynolds number, Re t =k 2 /(νε) , colored by YHCO, corresponding to the heat-release occurring between the inner
layer and the oxidation layer. Here, k is the turbulent kinetic energy, ε the turbulent dissipation, τ t =k/ε the tur-
bulence time scale, and τ c=ν/s 2u the chemical time scale. This scatter plot corresponds to the Williams diagram
of turbulent combustion, [53], and allows us to explore which different types of combustion regimes are in-
volved. For Dat1 the turbulent time scale, τt, is much larger than the chemical one, τc, and hence reaction
sheets form, whereas for Dat1 the turbulent scales rapidly mix reactants, which leads to distributed reaction
zones. If both scales are of similar size, Dat≈1, strong turbulence-chemistry interaction effects are expected. For

(a) (b)

Figure 8. Scatter plots of the turbulent Damköhler number, Dat, versus the turbulent Reynolds number, Ret,
for the (a) 2S-KERO-BFER and (b) Z57 reaction mechanisms colored by the heat release.

premixed combustion, τc relates directly to the flame structure and thickness, which permits distinguishing be-
tween wrinkled flamelets, corrugated flamelets, thin reaction zones, broken reaction zones and well-stirred re-
actor regimes. For premixed combustion the influence of the Kolmogorov scales is taken into account using the
Karlovitz number, Ka=τ c /τ K , in which τK denotes the Kolmogorov time scale. For Ka<1 the smallest Kolmogo-
rov scales, are larger than the chemical scales, which limits the interaction between turbulence and chemistry.
The turbulent scales can just alter the shape of the flame front. Depending on the velocity of the turbulent ed-
dies, vrms, compared to su, wrinkled or corrugated flamelets exist. For small vrms the flame front can only be
wrinkled, whereas for larger vrms a strong interaction with the flame front occurs, which leads to convoluted and

11
American Institute of Aeronautics and Astronautics
disturbed flame front shapes. Aero-gas turbine combustion is thus a mix between reaction sheets and distributed
reaction zones experiencing strong turbulence chemistry interactions as suggested by the fact that Dat≈1.

VI. Fully Annular Multi-Burner Gas Turbine Combustor Results


In order to more correctly capture the combustion process occurring in an annular multi-burner combustor, a
simulation including the full annular combustor, and all swirl injectors, is needed since the pressure field may
contain longitudinal, radial and azimuthal pressure waves that can cause the individual flames to interact with
each other and potentially lead to thermoacoustic oscillations, [53]. Such thermoacoustic oscillations naturally
arise when combustion couples with the acoustic eigenmodes of the combustor, but may, under some circum-
stances, result in a self-sustained mechanism which can amplify until failure of the engine. Boudier et al., [54],
investigated thermoacoustic oscillations in an annular multi-burner helicopter combustor utilizing LES based on
a dynamic flame thickening model, [55], a two-step reaction mechanism, and a 42 million cell unstructured grid.
Based on the results of Sections 4 and 5, the reaction mechanism selected for the fully annular multi-burner
Downloaded by UNIVERSITY OF NEW SOUTH WALES on September 6, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4020

combustor LES is the Z57 skeletal reaction mechanism, which previously was found to mimic the detailed be-
havior of R5738 very well.
Figure 9 shows perspective, side and end views of the twelve interacting flames in the annular multi-
burner combustor in terms of semi-transparent iso-surfaces of C12H23 (in green), C2H4 (in yellow) and HCO (in
red), representing the flames, and contours of the temperature (top) and velocity (bottom). Initially, the temper-
ature distribution in the annular multi-burner combustor of figure 9 is similar to that of the single sector model
in figure 5, with the high temperature region lifted away from the fuel-injectors. The individual flames in the
annular multi-burner combustor are, however, different, and that they appears to be interacting with each other.
The single sector model may be a reasonable representative of the actual annular multi-burner combustor in
terms of mean quantities, but will not be able to represent the dynamics caused by interacting flames or by azi-
muthal pressure waves characteristic of thermoacoustic oscillations, known to exist in real gas turbine combus-
tors. The intrinsic flame behavior is however the same as for the single sector model, with a core of C12H23
transforming into a hollow cloud of C2H4, which in turn transforms into an outer cloud of HCO. These species
exist at the upstream edge of the CRZ formed downstream of each fuel injector, and shielded by the swirling
air-flow from the swirler-injector. HCO is also characteristic of the heat release, and downstream of HCO, CO
and CO2 dominate according to the reaction path diagram 2a. The CRZ’s connect to form a single unsteady, to-

(a) (b) (c)

Figure 9. Instantaneous distributions of temperature (top) and velocity magnitude (bottom) in (a) and (b).
Iso-surfaces of C12H23 (—), C2H4 (—) and HCO (—) are shown in all three figures.

pologically complicated, CRZ. In a similar manner are the TV’s, residing between the dump plane, swirler-in-
jector and annular shear layers connected. This results in an even more complex flow field in which all the mix-
ing, dilution and film-cooling air jets rapidly discharge, developing a flame region, a mixing region and a post-
combustion region according to figure 10. The flame dynamics is strongly affected by the equivalence ratio, φ,
the pressure, p, and the velocity gradients, ∇v, in the annular shear layers shielding the flames. φ is controlled
by the fuel-air mixture in the flame, i.e. the amount of air entrained in the fuel discharged through the nozzle,
which in turn is affected by the local flow and the turbulence in the vicinity of the fuel-injectors. These issues
are reasonably well understood, whereas the influence of the pressure is more complicated, having the ability to

12
American Institute of Aeronautics and Astronautics
connect different parts of the combustor with each other.
Figure 10 shows instantaneous distributions of (from left to right) pressure fluctuations, p! , velocity
magnitude, |v|, (top) and temperature, T, (bottom), CO (top) and CO2 (bottom), O2 (top) and H2O (bottom) as
well as OH (top) and C2H2 (bottom). The pressure fluctuations in figure 10a suggest that longitudinal, radial and
azimuthal pressure waves simultaneously cross the annular flame tube, resulting in an unsteady and complex
p! -field. The velocity field in figure 10b (top) is very complex, including multiply connected CRZ’s down-
stream of each swirl-injector, multiply connected TV structures between the dump plane, swirler-injector and
annular shear layers, high-speed air jets discharging through the film-cooling, mixing and dilution holes in the
liner, dividing the flame-tube into a flame region, a mixing region, and a post-combustion region. These flow
features are similar to those found in the single sector model, but for the annular combustor they are also linked
in the azimuthal direction, and thus also more complex. The temperature distribution in figure 10b (bottom) re-
veals the three flame regions, and how they are separated by cold air discharging through the film-cooling, mix-
ing and dilution holes in the liner. The highest temperature is observed in the flame region, where most of the
Downloaded by UNIVERSITY OF NEW SOUTH WALES on September 6, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4020

heat release occurs. CO (top) and CO2 (bottom) concentrations are presented in figure 10c, showing a CO con-
centration limited to the center of the flame-tube whereas CO2 occurs in the whole flame tube, with highest con-
centrations in the mixing region, implying that CO2 is produced in the mixing and post combustion regions. Alt-
hough some regions are fuel rich, the lean character of the combustor implies that almost all CO will be con-
sumed before the combustor exit. Figure 10d presents O2 (top) and H2O (bottom) concentrations from which the
O2 distribution reveals the impact of the cold air jets discharging through the liner, and also the rich flame re-
gion. H2O, in contrast to CO2, is not formed in the mixing and post combustion regions, but in the less oxygen-
ated flame region, and particularly where the temperature is high. OH (top) and C2H2 concentrations (bottom)
are presented in figure 10e, and clearly reveal the post flame behavior of OH, forming a shield of OH around the
core parts of the flame, where CO and C2H2 are formed. C2H2, results from reactions in the middle of the C2-re-
action paths, are located deeper into the flame tube than C2H4, but not as deep as where the higher concentra-
tions of CO2 are located.

(a) (b) (c) (d) (e)

Figure 10. Instantaneous distributions of pressure fluctuations in (a), velocity (top) and temperature (bottom)
in (b), CO (top) and CO2 (bottom) in (c), O2 (top) and H2O (bottom) in (d) and OH (top) and C2H2 (bottom)
in (e). Iso-surfaces of C12H23 (—), C2H4 (—) and HCO (—).

An important objective of this study is thus to examine the unsteady combustion dynamics, for the annu-
lar multi-burner combustor. The pressure fluctuations, p! , are related to the unsteady heat release, Q  , through
2 2 2 
the wave equation ∂t (p")−c ∇ (p")≈(γ−1)∂t Q−p 0 γ||∇v|| , in which the two source terms representing unsteady
2

heat-release and turbulent velocity fluctuations, respectively. By combining this with the linearized momentum
equation, ρ0∂t (v#)≈−∇p# , an equation for the total acoustic energy, E=12 (p!2 /ρ0 c 2+ρ0|v!|2 ) , results that is of the
form, ∂t E+∇⋅F≈(γ−1)(p'Q )/(ρ0 c 2 ) , in which F=p!v! is the acoustic energy flux, whereas the source term denotes
the correlation between p! and Q  . If these are in phase, the amplitude of the acoustic energy will increase,
whereas if the pressure fluctuation field and the unsteady heat release are out of phase the acoustic energy will

13
American Institute of Aeronautics and Astronautics
decrease, [53]. This statement can thus be used to study the sensitivity to thermoacoustic oscillations, which ac-
cordingly will occur if p! and Q  are in phase.
Figures 11a and 11b show contours of the pressure fluctuations, p! , on the inner flame tube wall seen
from the dump plane (figure 11a) and the side (figure 11b), whereas figures 11c and 11d show contours of p! on
the outer flame-tube wall seen from the dump plane. Dark color indicates high pressure whereas light color indi-
cates low pressure, and figures 11a and 11b shows a distinct azimuthal pressure wave traversing the inner liner
wall, whereas this is much less apparent on the outer flame tube wall as seen in figures 11c and 11d. A useful
reference in this respect is figures 9c and 10a showing the individual flames as seen from the flame-tube as well
as from the side at the same instant. The heat release from the individual flames are directly related to the p!
level, and the higher the p! level is, the higher is also the heat-release, Q . In these figures it can be observed
that the flames in the higher p! quadrant are enclosed by a larger volume of HCO, whereas the flames in the
lower p! quadrant are enclosed by a smaller volume of HCO. Numerical integration of Q  with respect to the
azimuthal angle supports this conjecture, and reveals that the total azimuthal variation of Q  is about 25%,
Downloaded by UNIVERSITY OF NEW SOUTH WALES on September 6, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4020

whereas the azimuthal variation of p! is about 20%. The rms pressure and heat-release fluctuations, prms and
Qrms, respectively, from the single sector model using the Z57 reaction mechanism is prms≈ 496 kPa and Qrms≈28
MJ/m3s, whereas the corresponding values for the annular multi-burner combustor are prms≈ 614 kPa and
Qrms≈36 MJ/m3s, respectively. This corresponds to an increase in about 20% over the single sector model, clear-
ly demonstrating the importance of properly incorporating the full annular combustor domain, thereby allowing
for including primarily azimuthal pressure waves and thermoacoustic oscillation. These results also demonstrate
how well the liner damp these pressure waves, prohibiting them from spreading.

(a) (b) (c) (d)

Figure 11. Contours of the pressure fluctuations on the inner flame tube wall seen from the bottom (a) and
side (b), together with contours of the pressure fluctuations on the outer flame tube wall seen from the bot-
tom (c) and side (d).

Figure 12a shows the distribution of T versus z, colored by the rate-of-strain, ||D||, for the annular multi-
burner combustor, and figure 12b shows the corresponding scatter-plots of major species, YC12H23, YCO2, YCO,
YH2O, and minor species, YC2H4, YHCO and YOH, versus z. Figure 12a shows similar results as for the single sector
presented in figure 6b, but revealing a larger scatter due to the presence of the twelve flames compared to a sin-
gle flame in figure 6b. The T distribution in figure 12a suggests a somewhat more vigorous burning in the an-
nular multi-burner combustor compared with the single sector combustor in figure 6b. The corresponding scat-
ter-plots of major, YC12H23, YCO2, YCO, YH2O, and minor species, YC2H4, YHCO and YOH, versus z in figure 12b also
reveal a larger scatter for the annular multi-burner combustor compared with the single sector model in figure
7b. Similar to the single sector case, the peak distributions of the radicals OH and HCO occurs close to zst
whereas CO and C2H4 peak at higher z.
Figure 13 finally describes the circumferentially and time averaged axial velocity, vx, temperature, T, and
species C12H23, O2, H2O, CO2 and CO distributions at the combustor outlet. This section is particularly important
as it is usually also the inlet to the turbine, and the turbines requires well defined inlet profiles for smooth oper-
ation. In particular it is important to have as low fluctuation levels as possible. As can be seen in these profiles,
the temperature is around 1310 K in the core of the duct where also the peak velocity of about 130 m/s is found.

14
American Institute of Aeronautics and Astronautics
(a) (b)

Figure 12. (a) Scatter plots of T versus z colored by ||D|| and (b) scatter plots of different species versus z.
Legend: (n) C12H23, (n) CO2, (n) CO, (n) H2O, (n) C2H4, (n) HCO and (n) OH.
Downloaded by UNIVERSITY OF NEW SOUTH WALES on September 6, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4020

The corresponding temperature and axial velocity fluctuations are about 220 K and 28 m/s, thus being non-neg-
ligible from a design and life-cycle perspective. All the fuel is consumed, and the H2O, CO2 and CO emissions
correspond to a mean equivalence ratio of approximately 0.3, taking film-cooling dilution and mixing air into
account, whereas the equivalence ratio in the flame region varies between approximately 2.0 and 0.7 in order to
support relight at high altitude.

(a) (b)

Figure 13 (a) Temperatuure distribution and (b) circumferentially and time averaged axial velocity, vx, tem-
perature, T, and species C12H23, O2, H2O, CO2 and CO distributions at the combustor outlet. Legend (—)
mean values and (– –) rms-fluctuations.

VII. Concluding Remarks


Here we present combustion LES of a generic aero-propulsion gas turbine combustor, modeled as a fully annu-
lar multi-burner configuration as well as a single sector model. We also present a novel 57-step skeletal kero-
sene-air reaction mechanism and examine its performance for a wide range of parameters, such as laminar flame
speed, flame temperature, ignition delay time, extinction strain rate and major species concentrations, against
detailed and other simplified reaction mechanisms. The performance of the new Z57 skeletal kerosene-air reac-
tion mechanism is then evaluated against an existing two-step global reaction mechanisms in LES of a 30 de-
gree single sector generic gas turbine combustor configuration. After that, the new Z57 skeletal kerosene-air re-
action mechanism is used in LES of the corresponding fully annular multi-burner combustor configuration. The
comparison of the single sector simulations shows similarities in quantities such as flow and pressure, whereas
species concentrations and flame dynamics vary significantly between the two cases, and thus the two global
and skeletal reaction mechanisms. The advantage of using the new Z57 skeletal kerosene-air reaction mecha-
nism is its better ability to properly represent the actual chemical kinetics, and how this is affected by the turbu-
lent flow, thus proving a more realistic description of the combustion process. In particular are the mechanism’s
capabilities in predicting the laminar flame speed, the flame temperature and the extinction strain rate of crucial
importance to represent turbulent combustion, often influenced by the local strain field, sometimes leading to
quenching. The flame dynamics in the annular multi-burner configuration shows similar behavior as the single
sector model, when using the same kinetics, but with larger scatter in data due to the presence of interacting
flames. The annular multi-burner LES also reveals an azimuthal pressure wave progressing in the combustion
chamber not present in the single sector model. The physics associated with thermoacoustic oscillations is of
paramount importance to design and life-cycle analysis of gas turbines engines as it modifies the outflow condi-

15
American Institute of Aeronautics and Astronautics
tions, and in particular the outflow fluctuation levels.

Acknowledgement
The presented work was supported by the Swedish Armed Forces and the Swedish Defense Material Agency.
Additional support was received from the EFFECT program funded by The Swedish Energy Agency Board. Jon
Tegnér is acknowledged for providing the extinction strain rate predictions.

References
[1] Gicquel L.Y.M., Staffelbach G. & Poinsot T.; 2012, “Large Eddy Simulations of Gaseous Flames in Gas
Turbine Combustion Chambers”, Prog. Comb. Energy and Comb. Sci., 38, p 782.
[2] Poinsot T. & Veynante D.; 2001, “Theoretical and Numerical Combustion”, Edwards, Philadelphia,
USA.
[3] Jones W.P., McGuirk J.J., Sodha M.N.; 1989, “Calculation of the Flow in a Sector of an Annular Com-
Downloaded by UNIVERSITY OF NEW SOUTH WALES on September 6, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4020

bustor”, Proc. Inst. Mech. Engs Part A, J. Power and Energy, 203, p 187.
[4] Pitsch H.; 2006, “Large Eddy Simulation of Turbulent Combustion”, Annu. Rev Fluid Mech., 38, p 453.
[5] Fureby C.; 2008, “LES Modeling of Combustion for Propulsion Applications”, Phil. Trans. R. Soc. A,
367, p 2957.
[6] Oefelein J.C., Drozda T.G. & Sankaran V.; 2006, “Large Eddy Simulation of Turbulence-Chemistry In-
teractions in Reacting Flows”, J. Physics Conf. Series, 46, p 16.
[7] Menon S. & Fureby C.; 2010, “Computational Combustion”, In Encyclopedia of Aerospace Engineering,
Eds. Blockley R. & Shyy W., John Wiley & Sons.
[8] Grinstein F.F., Margolin L. & Rider B. (Eds.); 2007, “Implicit Large Eddy Simulation: Computing Tur-
bulent Fluid Dynamics”, Cambridge University Press.
[9] Spencer A. & Adumitroiae V.; 2003, “Large Eddy Simulation of Impinging Jets in Crossflow”, GT2003-
38754. 48th ASME Gas Turbine and Aeroengine Technical Congress, Atlanta, GA, June 16-19.
[10] Chen J.H..; 2011, “Petascale Direct Numerical Simulation of Turbulent Combustion – Fundamental In-
sights Towards Predictive Models”, Proc. Comb. Inst. 33, p 99.
[11] Cannon S.M., Smith C.E. & Anand M.S.; 2003, “LES Predictions of Combustor Emissions in an Aero
Gas Turbine Engine”, AIAA 2003-4521.
[12] di Mare F., Jones W.P. & Menzies K.R.; 2004, “Large Eddy Simulation of a Model Gas Turbine Com-
bustor”, Comb. Flame, 137, p 278.
[13] Staffelbach G., Gicquel L.Y.M. & Poinsot T.; 2007, “Highly Parallel Large Eddy Simulations of Multi-
burner Configurations in Industrial Gas Turbines”, In Lecture Notes in Computational Science and Engineering,
56, p 325.
[14] Staffelbach G, Gicquel L, Boudier G, Poinsot T.; 2009, “Large Eddy Simulation of Self-Excited Azi-
muthal Modes in Annular Combustors”, Proc Comb. Inst., 32, p 2909.
[15] Fureby C.; 2012, “Combustion LES of a Multi-Burner Annular Aero-Engine Combustor”, Invited
presentation at the SME-CMD International Computational Mechanics Symposium, Kobe, Japan.
[16] Bulat G., Fedina E., Fureby C., Meier W. & Stopper U.; 2015, “Reacting Flow in an Industrial Gas Tur-
bine Combustor: LES and Experimental Analysis”, Proc. Comb. Inst., 35, p 3175.
[17] Dagaut P. & Cathonnet M.; 2006, “The Ignition, Oxidation, and Combustion of Kerosene: A Review of
Experimental and Kinetic Modeling”, Prog. Energy Comb. Sci., 32, p 48.
[18] Luche J., Reuillon M., Boettner J.-C. & Cathonnet M.; 2004, “Reduction of Large Detailed Kinetic
Mechanisms: Application to Kerosene-Air Combustion”, Comb. Sci. Tech., 176, p 1935.
[19] Westbrook C.K., Pitz W.J., Herbinet O., Curran H.J. & Silke E.J.; 2008, “A Comprehensive Detailed
Chemical Kinetic Reaction Mechanism for Combustion of n-alkane Hydrocarbons from n-octane to n-hexade-
cane” Comb. Flame, 156, p 181.
[20] Ranzi E., Frassoldati A., Grana R., Cuoci A., Faravelli T., Kelley A.P. & Law C.K.; 2012, “Hierarchical
and Comparative Kinetic Modeling of Laminar Flame Speeds of Hydrocarbon and Oxygenated Fuels”, Progress
Energy and Comb. Sci., 38, p 468.
[21] Franzelli B., Riber E. Sanjosé & Poinsot T.; 2010, “A Two-Step Chemical Scheme for Kerosene-Air Pre-
mixed Flames”, Comb. Flame, 157, p 1364.
[22] Choi J.-Y.; 2011, “A Quasi Global Mechanism of Kerosene Combustion for Propulsion Applications”,
AIAA 2011-5853.
[23] Frassoldati A., Cuoci A., Faravelli T. & Ranizi E.; 2011, “Global Kinetic Mechanism of Kerosene Com-
bustion for CFD Applications”, Proc. European Comb. Meeting.

16
American Institute of Aeronautics and Astronautics
[24] Layton W.J. & Lewandowski R.; 2009, “Residual Stress of Approximate Deconvolution Models of Tur-
bulence”, J. Turb., 7, p N46.
[25] Bensow R. & Fureby C.; 2007, “On the Justification and Extension of Mixed Models in LES”, J. Turb. 8,
N54, p. 1.
[26] Schumann U.; 1975, “Subgrid Scale Model for Finite Difference Simulation of Turbulent Flows in Plane
Channels and Annuli”, J. Comp. Phys., 18, p 376.
[27] Fureby C.; 2007, “On LES and DES of Wall Bounded Flows”, Ercoftac Bulletin, March issue.
[28] Berglund M., Fedina E. Tegnér J. Fureby C. & Sabelnikov V.; 2010, “Finite Rate Chemistry LES of Self
Ignition in a Supersonic Combustion Ramjet”, AIAA.J. 48, p 540.
[29] Fedina E. & Fureby C.; 2010, “A Comparative Study of Flamelet and Finite Rate Chemistry LES for an
Axisymmetric Dump Combustor”, J. Turb. 12, N24.
[30] Fureby C.; 2012, “A Comparative Study of Flamelet and Finite Rate Chemistry LES for a Swirl Stabi-
lized flame”, ASME J. Engineering for Gas Turbines & Power, 134, 041503-1.
[31] Sabelnikov V. & Fureby C.; 2013, “LES Combustion Modeling for High Re Flames using a Multi-Phase
Downloaded by UNIVERSITY OF NEW SOUTH WALES on September 6, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4020

Analogy”, Comb. Flame, 160, p 83.


[32] Pope S.B.; 1985, “PDF Method for Turbulent Reacting Flows”. Prog. Energy Comb. Sci., 11, p 119.
[33] Yeung P.K., Pope S.B. & Sawford B.L.; 2006, “Reynolds Number Dependence of Lagrangian Statistics
in Large Numerical Simulations of Isotropic turbulence”, J. Turb., 7, p 1.
[34] Weller H.G., Tabor G., Jasak H. & Fureby C.; 1997, “A Tensorial Approach to CFD using Object Ori-
ented Techniques“, Comp. in Physics, 12, p 629.
[35] Drikakis D., Fureby C., Grinstein F.F. & Liefendahl M.; 2007, “ILES with Limiting Algorithms”, In Im-
plicit Large Eddy Simulation: Computing Turbulent Fluid Dynamics, Eds. Grinstein F.F., Margolin L. & Rider
B., Cambridge University Press, p 94.
[36] Issa R.I.; 1986, “Solution of the Implicitly Discretized Fluid Flow Equations by Operator Splitting”, J.
Comp. Phys., 62, p 40.
[37] Strang G.; 1968, “On the Construction and Comparison of Difference Schemes”, SIAM J. Numerical
Analysis, 5, p 506.
[38] Rosenbrock H.H.; 1963, “Some General Implicit Processes for the Numerical Solution of Differential
Equations”, The Computer Journal, 5, p 329.
[39] Ranzi E., Frassoldati A., Stagni, A., Pelucchi, M., Cuoci, A. & Faravelli, T.; 2014, “Reduced Kinetic
Schemes of Complex Reaction Systems: Fossil and Biomass-Derived Transportation Fuels”, Int. J. Chem. Kin.,
46, p 512.
[40] Dooley S., Won S.H., Heyne J., Farouk T. I., Ju Y.G., Dryer F.L., Kumar K., Hui X., Sung C.J., Wang,
H.W., Oehlschlaeger M.A., Iyer V., Iyer S., Litzinger T.A., Santoro R.J., Malewicki T. & Brezinsky K.; 2012,
“The Experimental Evaluation of a Methodology for Surrogate Fuel Formulation to Emulate Gas Phase Com-
bustion Kinetic Phenomena”, Comb. Flame, 159, 1444.
[41] Wang H, Dames E., Sirjean B., Sheen D.A., Tango R., Violi A., Lai J.Y.W., Egolfopoulos F.N., Da-
vidson D.F., Hanson R.K., Bowman C.T., Law C.K., Tsang W., Cernansky N.P., Miller D.L. & Lindstedt R.P.;
2010, “A High-Temperature Chemical Kinetic Model of n-Alkane (up to n-Dodecane), Cyclohexane, and Me-
thyl-, Ethyl-, n-Propyl and n-Butyl-Byclohexane Oxidation at High Temperatures”, JetSurF version 2.0, Sep-
tember 19. (http://web.stanford.edu/group/haiwanglab/JetSurF/JetSurF2.0/index.html).
[42] Ji C., Dames E., Wang Y.L., Wang H. & Egolfopouos F.N.; 2010, “Propagation and Extinction of Pre-
mixed C5-C12 n-Alkane Flames”, Comb. Flame, 157, p 277.
[43] Kumar K., Sung C.-J. & Hui X., 2011, “Laminar Flame Speeds and Extinction Limits of Conventional
and Alternative Jet fuels”, Fuel, 90, p 1004.
[44] Hui X. & Sung C. J.,; 2013, “Laminar Flame Speeds of Transportation-relevant Hydrocarbons and Jet
Fuels at Elevated Temperatures and Pressures”, Fuel, 109, p 191.
[45] Chong C. T. & Hochgreb S.; 2011, “Measurements of Laminar Flame Speeds of Liquid Fuels: Jet-A1,
Diesel, Palm Methyl Esters and blends using Particle Imaging Velocimetry (PIV)”. Proc. Comb. Inst., 33, p 979.
[46] http://www.cantera.org/docs/sphinx/html/index.html
[47] http://www.reactiondesign.com/products/chemkin/chemkin-pro/
[48] Jones W.P. & Tyliszczak A.; 2010, “Large Eddy Simulation of Spark Ignition in a Gas Turbine Combus-
tor”, Flow Turb. Comb., 85, p 711.
[49] Molnar M. & Marek C.J.; 2003, “Reduced Equations for Calculating the Combustion Rates of Jet-A and
Methane Fuel”, NASA 2003-212702.
[50] Slavinskaya N.A., Zizin A. & Ridel U.; 2010, “Towards Kerosene Reaction Model Development:
Propylcyclohexane, cyC9H18, n-Dodecane, C12H26, and Hexadecane C16H34 Combustion”, AIAA 2010-0605.
[51] Smooke M.D. & Giovangigli V.; 1991, “Formulation of the Premixed and Nonpremixed Test Problems”,

17
American Institute of Aeronautics and Astronautics
in Lecture Notes in Physics: Reduced Kinetic Mechanisms and Asymptotic Approximations for Methane-Air
Flames, Smooke M.D. (Ed.), 384, p 1, Springer-Verlag, New York.
[52] Burke S.P. & Schumann T.E.W.; 1928, “Diffusion Flames”, Ind. Eng. Chem., 20, p 998.
[53] Rayleigh L.; 1878, “The Explanation of certain Acoustic Phenomena”, Nature,18, p 319.
[54] Boudier G., Lamarque N., Staffelbach G., Gicquel L.Y.M. & Poinsot T.; 2009, “Thermo-Acoustic Stabil-
ity of a Helicopter Gas Turbine Combustor using Large Eddy Simulation”, Aeroacoustics, 8, p 69.
[55] Colin O., Ducros F., Veynante D. & Poinsot T.; 2000, “A Thickened Flame Model for Large Eddy Sim-
ulations of Turbulent Premixed Combustion”, Phys. Fluids, 12, p 1843.
Downloaded by UNIVERSITY OF NEW SOUTH WALES on September 6, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4020

18
American Institute of Aeronautics and Astronautics
Appendix 1

Table 1. Summary of Z57 kerosene (C12H23)-air reaction mechanism.


Reaction Reaction Reaction Reaction
C12H23 ⇒ 5 C2H4+C2H3 C2H2+H ⇒ C2H+H2 CH3O+H ⇒ CH2O+H2 O+H2O ⇒ OH+OH
C12H23+OH ⇒ 6 C2H4+O C2H2+OH ⇒ C2H+H2O CH3O+M⇒CH2O+H+M H+O2 + M ⇒ HO2+M(*)
C12H23+H ⇒ 6 C2H4 C2H+H2O ⇒ C2H2+OH CH2O+H ⇒ HCO+H2 H+HO2 ⇒ OH+OH
C2H4+H+M ⇒ C2H5+M C2H2+O ⇒ C2H+OH CH2O+OH⇒HCO+H2O H+HO2 ⇒ H2+O2
C2H5+H ⇒ CH3+CH3 C2H+OH ⇒ C2H2+O HCO+H ⇒ CO+H2 OH+HO2 ⇒ H2O+O2
C2H4 ⇒ C2H3+H C2H+O2 ⇒ HCO+CO HCO+M ⇒ CO+H+M HO2+HO2 ⇒ H2O2+O2
C2H4+H ⇒ C2H5 HCO+CO ⇒ C2H+O2 CO+OH ⇒ CO2+H H2O2+M ⇒ OH+OH+M
HCO+CH3 ⇒ C2H4+O CH4(+M) ⇒ CH3+H(+M)(**) CO2+H ⇒ CO+OH OH+OH+M ⇒ H2O2+M
C2H4+OH ⇒ C2H3+H2O CH3+H(+M) ⇒CH4(+M)(**) H+O2 ⇒ OH+O H2O2+OH ⇒ H2O+HO2
C2H3+H2O ⇒ C2H4+OH CH4+H ⇒ CH3+H2 OH+O ⇒ H+O2 H2O+HO2 ⇒ H2O2+OH
Downloaded by UNIVERSITY OF NEW SOUTH WALES on September 6, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2015-4020

C2H4+CH3 ⇒ C2H3+CH4 CH3+H2 ⇒ CH4+H O+H2 ⇒ OH+H OH+H+M ⇒ H2O+M


C2H4+H ⇒ C2H3+H2 CH4+OH ⇒ CH3+H2O OH+H ⇒ O+H2 H+H+M ⇒ H2+M
C2H2+H + M ⇒ C2H3+M CH3+H2O ⇒ CH4+OH H2+OH ⇒ H2O+H
C2H3+H ⇒ C2H2+H2 CH3+O ⇒ CH2O+H H2O+H ⇒ H2+OH
C2H+H+M ⇒ C2H2+M CH3+O2 ⇒ CH3O+O OH+OH ⇒ O+H2O
(*). Third body efficencies: CH4:6.5, CO:0.75, CO2:1.5, H2:1, H2O:6.5, N2:0.4, O2:0.4
(**). Third body efficencies: C12H23:3, CH4:6.5, CO:0.75, CO2:1.5, H2:1, H2O:6.5, N2:0.4, O2:0.4

19
American Institute of Aeronautics and Astronautics

Das könnte Ihnen auch gefallen