Evaluation of Optimized 3-step Global Reaction Mechanism

for CFD Simulations on Sandia Flame D

Abdallah Abou-Taouk and Lars-Erik Eriksson
Department of Applied Mechanics,
Chalmers University of Technology
Gothenburg, Sweden
Abstract. The aim of this paper is to evaluate a new optimized 3-step global reaction mechanism (opt) [1] for a methaneair mixture for industry purpose. The global reaction mechanism consists of three reactions corresponding to the fuel
oxidation into CO and H2O, and the CO  CO2 equilibrium reaction. Correction functions that are dependent on the local
equivalence ratio are introduced into the global mechanism. The optimized 3-step global reaction scheme is adapted into
the Computational Fluid Dynamics (CFD) analysis of a partially-premixed piloted methane jet flame. The burner consists
of a central nozzle (for premixed fuel/air), surrounded by a premixed pilot flame, and an annular co-flow stream. Both
steady-state RANS (Reynolds Averaged Navier Stokes) and time-averaged hybrid URANS/LES (Unsteady RANS/Large
Eddy Simulation) results have been computed and compared with experimental results obtained from the Sydney burner
at Sandia National Laboratories, Sandia Flame D [2]. The CFD results with the optimized 3-step global reaction
mechanism show reasonable agreement with the experimental data based on emission, velocity and temperature profiles,
while the 2-step Westbrook Dryer (WD2) [3] global reaction mechanism shows poor agreement with the emission
profiles.
Keywords: CFD, PSR, global mechanism, combustion, methane-air mixture, WD2
PACS: 47.70.Pq

INTRODUCTION
A demand on reduced emissions and improved efficiency for the gas turbine combustors implies that reliable and
accurate modelings of chemical kinetics are crucial. In reality detailed reaction mechanisms for hydrocarbon
combustion are extremely complex since they involve around 100 species and 10000 reactions. Some detailed
mechanisms of methane-air combustion involve more than 300 elementary reactions and over 30 species [4]. From a
CFD perspective it is still too expensive to include all species and reactions. The use of global reaction mechanisms
is one way to go for CFD simulations, since they are easily implemented in commercial software. Several different
reduced reaction mechanisms of methane-air mixtures exist in the literature [5-7]. The drawback of most of the
published global mechanisms is that they are not flexible enough to cope with a wide range of equivalence ratios.
The presently used optimized 3-step global reaction mechanism (opt) has been developed for equivalence ratios in
the range 0.5-1.8 and at atmospheric pressure [1].
The WD2 global reaction mechanism [3] is commonly seen in the literature and is an old industry standard for CFD
simulations. The drawback with the WD2 is due to the poor emission prediction at rich conditions since it produces
too much CO2 and not enough carbon monoxide compared to the detailed reaction mechanism. LES simulations by
Pitsch et al. [8] show very good agreement with the experimental data. However, the mesh size is three times larger
and the time step is ten time smaller than the SAS-SST model, which implies that the computational time is very
expensive. The aim of the present work is to improve, validate and evaluate current standard industrial CFD tools. It
is too expensive for the industry to run LES in their daily work. In the CFX validation report [9] the conclusion is
that the RNG k- model together with the WD2 model works well for industry purpose. CFD simulation with the
same settings as in the CFX validation report [9] has been performed and used for comparisons in the present work
(RNG k- WD2). The results obtained agree well with those given in the CFX validation report. So, no
improvements or modifications have been done for the RNG k- WD2. The improvements of the predictions of the
emissions are seen by the 3-step global reaction mechanism since the optimized mechanism increases the reaction
rate on the first reaction to increase the CO production. Simultaneously, the second reaction rate is reduced, thus
oxidizing less CO and therefore less CO2 is produced.
Numerical Analysis and Applied Mathematics ICNAAM 2011
AIP Conf. Proc. 1389, 66-69 (2011); doi: 10.1063/1.3636672
2011 American Institute of Physics 978-0-7354-0956-9/$30.00

66

KINET
TIC MODE
ELING
The opptimized 3-stepp global reacttion mechanism
m consists of the reactions seen in table 1. The first reaction is the
oxidation oof methane intto CO and H2O and the seccond reaction is the oxidatio
on of CO intoo CO2. The baackward rate
for the seccond reactionn is based on
n an equilibriium assumptiion. Table 1 also shows tthe optimized
d Arrhenius
coefficientss (activation energy,
e
pre-ex
xponential facttor and tempeerature coefficcient) that are used in the 3-step
3
global
reaction meechanism.
TABLE 1. Activation ennergy Ea, pre-ex
xponential factoor A and temperrature coefficient  used for thhe optimized sch
heme.
Reaction
CH4O24H2
O2 CO2

Ea [J
J/kmol]




The backw
ward rate for the
t second reeaction is bas ed on an equ
uilibrium assu
umption and tthe reaction rates
r
for the
forward reaactions are thee following:


(1)


(2)

y, R is the gaas constant, T is the temperrature and f1
where A is the pre-exponnential factor,, Ea is the acttivation energy
T aim of th
these correctio
on functions is to ensure good agreem
ment for rich
and f2 are the correction functions. The
a [10] optimiized similar coorrection funcctions, but for kerosene fuell ant thereforee one expect
conditions. Franzelli at al.
me differences in the shape of these funcctions for meethane air-mix
xture, which ccan be seen to
o the left in
to see som
Figure 1. T
The 3-step gloobal reaction mechanism iss optimized against
a
a detailed referencee mechanism (GRI Mech
3.0) for peerfectly stirredd reactor (PSR
R) calculationns [1]. The CANTERA
C
software has beeen used for the detailed
mechanism
m simulations and an in-ho
ouse PSR codde for the global reaction mechanism. In a comparrison of the
optimized 33-step global reaction mechanism with tthe reference detailed reaction mechanissm, the resultts show that
the gas tem
mperature and emissions are reasonably w
well predicted for lean and rich
r conditionss [1]. Figure 1 shows also
the temperaature predictioons at differen
nt equivalencee ratios.


FIGURE 1. Left: Plot shoowing the correection functionss f1 and f2 betw

ween Franzelli at al. [8] (for kerosene fuell) and Abought: Plot showinng temperature comparisons of a detailed meechanism (Gri Mech
M
3.0) and
Taouk et al. [1] (methane-aair mixture). Rig
mixture at equiv
valence ratios of 0.7, 0.9 and 11.2, Tin=295K
optimized 3--step global mechanism for meethane-air gas m

C
CFD Analyssis
The Annsys CFX soft
ftware packagee [11] was useed as a solverr. The RNG k--epsilon turbuulence model was
w selected
for the steaady-state sim
mulations and SAS-SST moodel for the transient
t
simu
ulation. Two different glob
bal reaction
schemes w
were used for the CFD sim
mulations, the optimized 3--step global reaction
r
mechhanism [1] an
nd the WD2
global reacction mechanism [3]. The combined turrbulence-chem
mistry interacttion model, thhe Finite Ratee Chemistry/
Eddy Dissiipation Modell (FRC-EDM)), in Ansys C FX [11], was chosen for all
a CFD analyyses. The com
mbined FRCEDM modeel gives two different
d
reacttion rates for each reaction
n, one from th
he EDM modeel and one fro
om the FRC
model. Thee minimum ratte for each reaaction is then chosen. The EDM
E
model is based on thee work of Mag
gnussen and
Hjertager [12].

67

Geometry Dom
main and Boun
ndary Conditio
ons
The Sanndia Flame D consists of a main jet withh a mixture off 25% of meth
hane and 75%
% of air. This jet is located
in a co-floowing gas streeam of air an
nd the flame is stabilized by a pilot. Experimental
E
he Technical
data from th
University of Darmstadt [2] is used to
o set temperatuure, species mass
m fractions, turbulence annd velocity prrofiles at the
FD simulations were done oon a 360o mo
odel. The mesh
h size for thee RANS simullations were
inlet bounddaries. All CF
0.5M cells and 3.5M cellls for the transsient simulatioon.
Results
C 2 and CO
Figure 2 shows plotts of axial prrofiles of masss fraction (computed and experimentaal) of CH4, CO
g
reaction
n mechanism over-predictss CO2 and und
der-predicts C
CO at all positions. The 3respectivelyy. The WD2 global
step optimiized global reaction
r
mech
hanism showss reasonable agreement
a
beetween the exxperimental data
d
and the
simulationss considering the
t CH4 and CO
C 2. Howeverr, the CO leveel is over-predicted..

FIG
GURE 2. Axiaal profiles of CH
H4 mass fractionn, CO2 mass fraaction and CO mass
m fraction pllotted at center line

d experimentaal) of CH4, CO
C 2 and CO
Figure 3 shows plotts of radial prrofiles of maass fraction (ccomputed and
ons. Similar ass before, the WD2
W
global reaction
r
mechhanism over-p
predicts CO2
respectivelyy for different axial positio
and under--predicts CO at radial possitions. The 33-step optimiized global reeaction mechhanism showss reasonable
agreement considering thhe CH4 and CO
O2. However,, similar to thee previously plot, the CO leevel is over-prredicted.

FIGURE 3. Radial profiiles of CH4 masss fraction plotteed at axial positions 0.0144m and 0.108m, CO
O2 mass fractio
on plotted at
axial positionns 0.0144m and
d 0.108m, CO m
mass fraction pllotted at axial positions
p
0.144m
m and 0.108m

68

Figure 5 shows radial/axial temperrature profiless and axial pro

ofile of the axial velocity. It can be seen that there is
generally reeasonable agrreement betweeen the experiimental data and
a the simulations. The teemperatures are
a generally
over prediccted close too the inlets of
o the burnerr and well prredicted down
nstream of thhe burner. A reasonable
explanationn of this is duee to that the raadiation is nott included in the
t CFD-simu
ulations. The aaxial velocity
y is also well
predicted. T
The transient simulation (SAS-SST) capttures the temp
perature and the
t velocity prrofiles better downstream
than the steeady-state sim
mulations.

FIGURE
E 4. Left: Radiial profile of staatic temperaturee [K] plotted at axial position 0.3240m,
0
Midddle: Axial profile of static
temperature [K]
[ plotted at ceenter line, Righht: Axial profilee of axial velociity [m/s] plottedd at center line

CO
ONCLUSIO
ONS
An optimized 3-steep global reacction mechaniism for meth
hane-air mixtu
ures is evaluaated and appllied in CFD
The Sandia Flaame D has beeen modeled w
with the optim
mized 3-step sccheme and WD
D2 using CFD
D tools. The
analyses. T
CFD resultts with the 3-step global reaaction mechannism show reaasonable agreeement with thee experimentaal data based
a temperatu
ure profiles, w
while the 2-sstep WD2 glo
obal reaction mechanism shows poor
on emissioon, velocity and
agreement w
with the emission profiles.

ACKNO
OWLEDGM
MENTS
This reesearch has beeen funded by
y the Swedishh Energy Agen
ncy, Siemens Industrial Tuurbomachinery
y AB, Volvo
Aero Corpooration, and the
t Royal Insttitute of Technnology throug
gh the Swedissh research prrogram TURB
BOPOWER,
the supportt of which is gratefully
g
ackn
nowledged.

RE
EFERENC
CES
1. A. Abouu-Taouk and L.E Eriksson, 2011, Optimizeed Global Mecchanisms For CFD
C
Analysis Of Swirl-Stabiilized Syngas
Burner F
For Gas Turbinees, ASME Turrbo Power Expoo, Power for lan
nd, Sea and Air, GT2011-458553.
2. Sandia N
National Laboraatories, Turbulent Diffusion Fllame Laboratory
y Web Site, http
p://www.ca.sanndia.gov/tfd
3. Charles K. WestBrookk and Frederick
k L. Dryer, 19 84, Chemical Kinetic Modeeling of Hydroocarbon Combu
ustion, Prog.
Energy C
Combustion Scii. 1984, Vol. 10
0, pp. 1-57.
4. Y. Liu, K
K. S. Lau, C. K.. Chan, Y. C. Guo,
G W. Y. Lin,, Structures off Scalar Transpo
ort in 2D Transsitional Jet Difffusion Flames
by LES, Int. J. Heat Mass
M
Tran. 46, 38413851
3
(20003).
5. Karl V. M
Meredith and David
D
L. Black, 2006, Automaated Global Meechanisms Geneeration for use iin CFD Simulations, AIAA
2006-11668.
6. N. Slavinnskaya, M. Braaun-Unkhoff and P.Frank, 20088, Reduced Reeaction Mechan
nisms for Methaane and Syngass Combustion
in Gas T
Turbines, Journnal of Engineeriing for Gas Turrbines and Poweer, Vol.130
7. M P. Gookulakrishnan, S.
S Kwon, A.J. Hamer,
H
M.S. K
Klassen and R..JJ. Roby, 2006, Reduced

Kineetic Mechanism
m for Reactive
Flow Sim
mulation of Synngas/Methane Combustion
C
at G
Gas Turbine Co
onditions, ASM
ME Turbo Powerr Expo, Power for land , Sea
and Air, GT2006-905733
8. H. Pitschh and H. Steinerr, 2000, Largee-eddy simulati on of a turbulent piloted meth
hane/air diffusioon flame (Sand
dia flame D),
Phys. Fluuids 12, 2541-22555.
9. D. Gobby, Piloted metthane jet flame, 2009.
M Sanjose and
d T. Poinsot, 20010, A two-steep chemical sch
heme for keroseene-air flames, Combustion
10. B. Franzzelli, E. Riber, M.
and Flam
me 175:1364-13373.
11. Commeercial software Ansys CFX, http://www.ansyss.com/default.aasp
12. B. F. M
Magnussen and B. H. Hjertageer. On mathem
matical modelss of turbulent combustion
c
witth special emph
hasis on soot
formatioon and combustiion. 16th Symp. (Int'l.) on Coombustion. The Combustion In
nstitute, 1976.

69