Content uploaded by Iskender Gökalp
Author content
All content in this area was uploaded by Iskender Gökalp on May 22, 2015
Content may be subject to copyright.
Revue des Energies Renouvelables CISM’08 Oum El Bouaghi (2008) 227 - 237
227
TFC modeling of hydrogenated methane premixed combustion
A. Mameri1*, A. Kaabi2 and I. Gökalp3
1 Institut des Sciences Technologiques, Centre Universitaire Larbi Ben M’Hidi, Oum El Bouaghi, Algérie
2 Département du Génie Climatique, Université des Frères Mentouri, Constantine, Algérie
3 Institut de Combustion, Aérothermique, Réactivité et Environnement, ‘ICARE’, CNRS, Orléans, France
Abstract - The use of hydrogenated fuels shows considerable promise for applications in
gas turbines and internal combustion engines. Hydrogen addition to methane will have an
important role to reach a fully developed hydrogen economy. The effects of this addition
on the flame structure is evaluated in three fuels with the following compositions at
constant global equivalent ratio: 100 % CH4, 10 % H2 + 90 % CH4, 20 % H2 + 80 %
CH4. The turbulence is modeled by the standard k -
ε
model which is improved by the
Pope correction in order to better predict round jet spreading. Combustion is modeled by
the turbulent flame closure (TFC) model which is used with flamelet to give detailed
chemistry. Computations were achieved by the ANSYS CFX code. A good agreement with
experiments was found, it was noted that one can replace a significant fraction of basic
fuel by hydrogen without making recourse to major modifications on the installations.
Résumé - L’utilisation de combustibles hydrogénés est très prometteuse pour des
applications dans les turbines à gaz et les moteurs à combustion interne. L’hydrogène en
addition avec le méthane aura un rôle important pour arriver à une économie en
hydrogène pleinement développé. Les effets de cet ajout sur la structure de la flamme est
évalué à trois carburants à la suite de compositions à taux constant global équivalent:
100 % CH4, 10 % H2 + 90 % CH4, 20 % H2 + 80 % CH4. La turbulence est modélisé par
le modèle standard k -
ε
, qui est inspiré par la correction de Pope, afin de mieux prédire
la propagation du jet. La combustion est modélisée par le modèle de la fermeture de la
flamme turbulente (TFC), qui est utilisé avec des flamelets de façon détaillée en chimie.
Les calculs ont été réalisés par le code ANSYS CFX. Un bon accord a été trouvé avec les
expériences. Il est à noter que l’on peut remplacer une fraction importante du carburant
de base par de l’hydrogène sans recourir à des modifications importantes sur les
installations.
Keywords: Premixed flames - Hydrogenated fuels - Burning velocity - Turbulent flame
speed.
1. INTRODUCTION
Internal combustion engine and gas turbine manufacturers are faced with stricter
anti-pollution regulations. Lean premixed combustion is a well established technique to
achieve low emissions while maintaining high efficiency. According to the thermal
x
NO generation mechanism; low flame temperatures given by lean conditions, results
in low x
NO emissions. However, close to lean flammability limits, the stability of the
flame decreases and flame extinction phenomena may occur.
Since further reduction of x
NO will require even leaner mixtures, schemes for lean
stability extension must be considered. A solution to increase the flame stability at lean
condition is to add small amounts of hydrogen into the mixtures [1-4]. Several studies
* mameriabdelbaki@yahoo.fr
A. Mameri et al.
228
have been performed to estimate the impact of 2
H addition on the stability, on the
reactivity and on the pollutant emissions of the methane-air flames.
It has been shown that hydrogen addition, at constant global equivalence ratio,
extends the lean operating limit of natural gas engines, leading to a potential decrease in
pollutant formation. The origin of this effect is that the stretch resistance of these flames
is considerably increased by hydrogen blending, while other properties are
comparatively little modified.
In spark ignited engine [5], the authors show that mixture of natural gas blended
with hydrogen improves thermal efficiency, specific fuel consumption, reduces CO
and x
NO emissions, extend lean operating limit to lower equivalence ratios and finally
allows stable engine operation with lower pollutant emissions.
Hydrogen addition was experimentally tested in several atmospheric flame burners’
configurations. It was observed a decreasing in the height of the blue cone with
hydrogen addition. The increase in laminar burning velocity was identified as the main
effect of the behavior of this parameter. Also a significant reduction in CO emission
was obtained [6].
Experimental investigation on the flame stability of hydrogenated mixtures was also
performed for swirl stabilized flame configuration [7].
We perform a numerical study to understand and to complete F. Halter et al. [8]
measured data. The authors conducted experiments on a lean methane combustor at
several pressures with hydrogen enrichment.
2. EXPERIMENTAL SETUP
The ICARE (Institut de Combustion, Aérothermique, Réactivité et Environnement)
high pressure turbulent flame facility is composed from stainless steel cylindrical
combustion chamber (Fig. 1). The inner chamber diameter is 300 mm. The chamber is
composed of two superposed vertical portion each of 600 mm height, and each equipped
with four windows of 100 mm diameter for optical diagnostics.
Fig. 1: ICARE combustion installation Fig. 2: Bunsen burner
An axisymmetric bunsen type turbulent burner (Fig. 2) is located inside the
chamber. The burner internal diameter is 25 mm. It is fed by methane/hydrogen/air
mixtures.
CISM’2008: TFC modeling of hydrogenated methane premixed combustion
229
The global equivalence ratio is 0.6, the bulk mean flow velocity of the reactants, at
the exit is 2 m/s. The reactants flow exhibits a turbulence level about 10 %. The integral
length scale is about 3 mm.
3. COMPUTATIONAL MODEL
To gain a more complete understanding of the impact of 2
H addition to lean 4
CH
flames, the Ansys CFX 11 code is used to compute the multi component turbulent
reacting flow. The Reynolds averaged equations are given by:
Continuity
0)U(
t=ρ•∇+
∂
ρ∂ (1)
Momentum
{}
M
'' S)uu()UU(
t
U+⊗ρ−τ•∇=⊗ρ•∇+
∂
ρ∂ (2)
Energy
h
''
tot
tot S)huT()hU(
t
p
t
h+ρ−∇λ•∇=ρ•∇+
∂
∂
−
∂
ρ∂ (3)
Where kU
2
1
hh 2
tot ++=
The other equations are similar and they can be cast in the following general form:
φφ +φρ−φ∇Γ•∇=φρ•∇+
∂
φρ∂ S)u()U(
t
'' (4)
Where φ is a general scalar.
The perfect gas state equation is given by:
∑
ρ=
jj
j
W
Y
TRp (5)
The Reynolds stresses '' uu ⊗ρ and fluxes ''
uφρ represent the convective effect
of turbulent velocity fluctuations. These terms need to be modelled.
3.1 ε−k turbulence model
Today, even with the successful development of DNS and LES for turbulent flows,
the most popular models for industrial modeling are the two-equation Reynolds
averaged Navier-Stokes (RANS) models. Of these, the
ε
−
k two equations model
accounts for 95 % or more of the industrial use at the present [9]. This form of model is
easy to solve, converges relatively quickly, is numerically robust and stable, is able to
solve large domains and high Reynolds numbers and requires minimal computational
expense, which is important for industrial models. The standard
ε
−
k model with the
standard constants predicts the velocity field of a two dimensional plane jet quite
accurately, but results in large errors for axisymmetric round jets. Although the standard
ε−k model matches the spreading rate of the round jet more accurately than other two
equations models, it still overestimates it by 15 % [10]. In this model, the Reynolds
stresses are given by a Newtonian type closure which looks like:
A. Mameri et al.
230
)U.k(
3
2
))U.(U.(uu t
T
t
'' ∇µ+ρδ−∇+∇µ=⊗ρ− (6)
Where t
µis the turbulence ‘viscosity’ (also called the eddy viscosity). By analogy with
the turbulence viscosity, the turbulence diffusivity is defined, and the Reynolds fluxes
are given by:
φ∇Γ=φρ− ..u t
'' (7)
Where t
Γ is the turbulence ‘diffusivity’, it is related to the turbulence viscosity by:
t
t
tPr
µ
=Γ (8)
Where t
Pr is the turbulent Prandtl number.
In the
ε
−k model, the turbulent viscosity is computed by the relation:
ε
ρ=µ µ
2
t
k
..C (9)
The
k
and
ε
equations are given by:
ερ−+
∇
σ
µ
+µ•∇=ρ•∇+
∂
ρ∂
k
k
tPk.)kU(
t
k (10)
()
ερ−
ε
+
ε∇
σ
µ
+µ•∇=ερ•∇+
∂
ερ∂ εε
ε2k1
tCPC.
k
.)U(
t (11)
To tailor the ε−k model for solving round jet flows, McGuirk and Rodi [11],
Morse [12], Launder et al. [13], and Pope [10] suggested modified turbulence model
constants. The best correction in our case is the Pope’s one; it is a new source term
added to the ε equation. It is given by the following relation:
p
2
3pope .
k
..CS χ
ε
ρ= ε (12)
With
∂
∂
−
∂
∂
∂
∂
−
∂
∂
ε
=χ y
v
x
u
.
x
v
y
u
.
k
4
12
3
p (13)
for two dimensional axisymmetric geometry.
In this work, a limited version of the Pope correction introduced by Davidenko [14]
is adopted and incorporated in the CFX code. Its expression is:
()
)(sign,min.
k
..CS plimp
2
3pope χχχ
ε
ρ= ε (14)
With lim
χ, a limiting value of p
χ
.
3.2 Combustion model
The model for premixed or partially premixed combustion can be split into two
independent parts:
• Model for the progress of the global reaction: Burning Velocity Model (BVM),
also called Turbulent Flame Closure (TFC);
CISM’2008: TFC modeling of hydrogenated methane premixed combustion
231
• Model for the composition of the reacted and non-reacted fractions of the fluid:
Laminar Flamelet with PDF.
Reaction Progress
A single progress variable c
~
is used to describe the progress of the global reaction:
iburndifreshi Y
~
.c
~
Y
~
.)c
~
1(Y
~
+−= (15)
The reaction progress variable is computed by solving a transport equation:
c
jc
t
jj
i
x
c
~
.D.
xx
)c
~
u
~
(
t
c
~ω+
∂
∂
σ
µ
+ρ
∂
∂
=
∂
ρ∂
+
∂
ρ∂ (16)
The burning velocity model (BVM), also known as turbulent flame closure (TFC), is
used to close the combustion source term for reaction progress.
∂
∂
ρ
∂
∂
−∇ρ=ω
jj
Tuc x
c
~
.D.
x
c
~
..S. (17)
Where the turbulent flame velocity is given by:
4/l
t
4/1
u
2/1
L
4/3'
Tl..S.u.G.AS
−
λ= (18)
And the stretch factor:
σ
+
ε
ε
σ
−= 2
ln.
2
1
erfc
2
1
Gcr
(19)
The integral and Kolmogorov length scales are given by:
ε
=2/3
t
k
l and 4/1
4/3
v
ε
=η
Flamelet libraries
Under flamelet regime hypothesis [15], the species transport equations are
simplified to:
k
2
k
2
k
lk
Z
Y
Le2t
Yω=
∂
∂χρ
−
∂
∂
ρ (20)
A detailed chemical mechanism of 64 species and 752 equations was adopted.
The simplified energy equation is:
k
N
1k
k
p
2
2
l.h.
C
1
Z
T
2t
Tω=
∂
∂χρ
−
∂
∂
ρ∑
=
(21)
With the laminar scalar dissipation:
2
l)Z(.D2 ∇=χ (22)
An external program CFXRIF solves these equations to obtain a laminar flamelet
table, which is integrated using a beta PDF to have the turbulent flamelet library.
This library provides the mean species mass fractions as functions of mean mixture
fraction Z
~
, variance of mixture fraction 2''
Z
~ and turbulent scalar dissipation rate χ
~
:
(
)
st
2''
ii ~
,Z
~
,Z
~
Y
~
Y
~
χ= (23)
A. Mameri et al.
232
On the other hand two transport equations are solved in the CFD code, the first gives
mixture fraction:
∂
∂
σ
µ
+µ
∂
∂
=
∂
ρ∂
+
∂
ρ∂
jz
t
jj
j
x
Z
~
xx
)Z
~
u
~
(
t
Z
~
(24)
And the second gives the mixture fraction variance:
χρ−
∂
∂
σ
µ
+
∂
∂
σ
µ
+µ
∂
∂
=
∂
ρ∂
+
∂
ρ∂ ~
x
Z
~
2
x
Z
~
xx
)Z
~
u
~
(
t
Z
~2
jz
t
j
2''
z
t
jj
j
2''
(25)
The turbulent dissipation scalar is modelled by:
2''
Z
~
.
k
~
~
.C
~ε
=χ χ (26)
To interpolate species mass fractions from the turbulent flamelet table, the CFD
program use the mixture fraction, mixture fraction variance and the turbulent scalar
dissipation computed above.
4. FLAME GEOMETRY AND NUMERICAL PROCEDURE
The computational domain is the half of the chamber with 3 degrees thick (Fig. 3).
200 × 300 × 1 grid nodes are taken inside the domain. The grid is refined near walls and
in the high gradients regions (Fig. 4). Boundary conditions for the main flame, pilot
flame, at wall, exit and symmetry axis are taken from experiments; the most important
are shown for the flame in figure 5.
Fig. 3: Combustion chamber geometry
Fig. 4: Near burner meshing
CISM’2008: TFC modeling of hydrogenated methane premixed combustion
233
Fig. 5: Inlet velocity and turbulence kinetic energy
A double precision computation is done with the CFX code. The high resolution
scheme with an automatic time scale control was used. To reach the target maximum
residual of 10-7 for all equations, the computation takes about 3000 to 5000 iterations to
converge.
5. COMPUTATION RESULTS
5.1 Cold jet test
Before computing the reacting flow, a cold jet test case is performed with the same
boundary conditions. The Pope correction constants 3
Cε and lim
χ
were adjusted to
have the best agreement with the experimental axial turbulent kinetic energy variation
and mean velocity decay.
The potential core length is well reproduced, the figures 6 and 7 show this
agreement. We note that the core length is 75 mm or three times the jet diameter
(3D/X =).
5.2 Reacting flow computations
Three cases are taken into consideration; the first is the methane-air combustion
using TFC with a detailed chemical mechanism. The second and third cases are the
blended hydrogen-methane combustion using the same model with a chemical
A. Mameri et al.
234
mechanism of 64 species and 752 reactions. The fuel compositions, for the second and
third cases, are: 10 % H2 + 90 % CH4 and 20 % H2 + 80 % CH4.
The following figures showing general field are taken for the case of 10 % H2. The
other cases are similar, and the difference resides in the flame length and maximum
variables values. Figure 8 shows the flow structure near the burner. Streamlines are
parallel in the potential core, and they are deviated in the preheat zone and the velocity
increases under the gas expansion effect.
Fig. 6: Axial velocity
Fig. 7: Axial turbulent kinetic energy
Fig. 8: Near burner flow field
CISM’2008: TFC modeling of hydrogenated methane premixed combustion
235
Fig. 9: Near burner temperature field
In figure 9, we have the temperature distribution for the fuel with 10 % H2. In this
case, the maximum temperature is about 1668 K (the adiabatic one is 1674 K). For the
case of the pure methane, we obtain 1664 K which is the same as the adiabatic one. The
last fuel (20 % H2) has an adiabatic temperature of 1680 K.
Fig. 10: Top- Turbulent flame velocity; Bottom- Laminar flame velocity
The figure 10 shows the difference between laminar and turbulent flame velocity.
The laminar flame velocity is a fuel property; it depends only on the chemical
composition. The turbulent flame velocity depends also on the flow conditions. It
represents the interaction of the flame with the turbulence. In this case, it is nearly thirty
times the laminar one.
Fig. 11: Axial reaction progress
A. Mameri et al.
236
We can see the good agreement between measured and computed axial reaction
progress variable c for all fuels (Fig. 11). We note that maximum reaction depends on
fuel composition, it becomes close to the burner when we add more hydrogen to the
fuel. This means that flame velocity becomes more important.
6. CONCLUSION
In this study turbulent premixed combustion of pure methane and hydrogenated
methane is computed. We have used a detailed chemistry with a turbulent flame closure
model. The most important result found is that hydrogen enrichment with small
amounts doesn’t increase significantly combustion temperature. On the other hand it
increases flame velocity and stabilises combustion process. Also hydrogen doesn’t
contain carbon, it is a clean fuel. The replacement of a fraction of the methane by
hydrogen results in lower pollutant emissions. Turbine gas and IC engines can be fed
with hydrogenated fuels without recourse to major modification in installations.
NOMENCLATURE
A: Zimont model constant 2
Cε:
ε
−
k model coefficient (= 1.92)
1
Cε: ε−k model coefficient (= 1.44) µ
C:
ε
−
k model coefficient (= 0.09)
3
Cε: POPE correction coefficient (= 0.79) I: Chemical species I
h, tot
h: Static and total enthalpy k
N: Number of species in the mixture
k
: Turbulence kinetic energy k
R: Production rate of species
k
t
Pr :Turbulent Prandtl number I
S: Source term of the I species equation
M
S, E
S: Source terms of the momentum and energy equations
R
: Universal gas constant t: Time
U: Velocity vector U(u, v, w) T: Static temperature
u: Axial velocity component v: Radial velocity component
I
W: Species I molar weight i
Y: Species I mass fraction
Greek symbols
α: Hydrogen mass fraction t
Γ
: Turbulent diffusivity coefficient
δ: Kronecker delta c
σ
: Transport coefficient for c (=0.9)
σ: Standard deviation – Distribution of
ε
µ
: Dynamic viscosity
k
σ, ε
σ: Transport coefficient for
k
and
ε
(= 1.0 and 1.3)
z
σ, ''z
σ: Transport coefficient for Z
~
and 2''
Z
~
(= 0.9 and 0.9)
t
µ: Turbulent viscosity c
ω
: Rate of production of c
k
ω: Rate of production of species
k
p
χ, lim
χ: Vortex stretching rate invariant and limited invariant (Pope correction)
CISM’2008: TFC modeling of hydrogenated methane premixed combustion
237
REFERENCES
[1] I. Yamaoka and H. Tsuji, ‘An Anomalous Behavior of Methane–Air and Methane–Hydrogen–
Air Flames Diluted with Nitrogen in a Stagnation Flow’, Process of the Energy Combustion
Science, Vol. 24, pp. 145 – 52, 1992.
[2] I. Wierzba and B.B. Ale, ‘Rich Flammability Limits of Fuel Mixtures Involving Hydrogen at
Elevated Temperatures’, International Journal of Hydrogen Energy, Vol. 25, pp. 75 – 80,
2000.
[3] G.S. Jackson, R. Sai, J.M. Plaia, C.M. Boggs and K.T. Kiger, ‘Influence of H2 on the
Response of Lean Premixed CH4 Flames to High Strained Flows’, Combustion Flame, Vol.
132, pp. 503 – 511, 2003.
[4] A. Mameri, I. Gökalp et D. Boukeffa, ‘Simulation Numérique de la Stabilisation d’une
Flamme Turbulente de Méthane en Régime Pauvre par Ajout d’Hydrogène’, Revue des
Energies Renouvelables, Vol. 10, N°1, pp. 39 – 48, 2007.
[5] S.R. Bell and M. Gupta, ‘Extension of the Lean Operating Limit for Natural Gas Fueling of a
Spark Ignited Engine Using Hydrogen Blending’, Combustion Science and Technology, Vol.
123, pp. 23 – 48, , 1997.
[6] H.J. Burbano, A.A. Amell and J.M. Garcıa, ‘Effects of Hydrogen Addition to Methane on the
Flame Structure and CO Emissions in Atmospheric Burners’, International Journal of
Hydrogen Energy, Vol. 33, pp. 3410 – 3415, 2008.
[7] R.W. Schefer, ‘Combustion of Hydrogen-Enriched Methane in a Lean Premixed Swirl
Burner’, In: Proceedings of the 2001 U.S. DOE Hydrogen Program Review, NREL/CP-570-
30535, 2001.
[8] F. Halter, C. Chauveau and I. Gökalp, ‘Characterization of the Effects of Hydrogen Addition
in Premixed Methane/Air Flames’, International Journal of Hydrogen Energy, Vol. 32, pp.
2585 – 2592, 2007.
[9] W.K. George, H. Wang, C. Wollblad and T.G. Tohansson, 14th AFMC, Adelaide, Australia,
pp. 41 – 48, 2001.
[10] S.B. Pope, AIAA Journal, Vol. 16, pp. 3, 1978.
[11] J.J. McGuirk and W. Rodi, 1st Symposium on Turbulent Shear Flows, Eds. F. Durst, B.E.
Launder, F.W. Schmidt and J.H. Whitelaw, pp. 79 – 83, 1979.
[12] A.P. Morse, Ph.D. Thesis, London University, 1977.
[13] B.E. Launder, A.P. Morse, W. Rodi and D.B. Spalding, Report NASA SP-311, 1972.
[14] D. Davidenko, Ph.D. Thesis, Orleans University, 2005.
[15] N. Peters, ‘Turbulent Combustion’, Edition Cambridge University Press, 2000.
[16] Document of Ansys CFX, ‘Combustion Theory, User guide’, 2006.
[17] A. Mameri, I. Fedioun et M. Boumaza, ‘Simulation Numérique d’une Flamme d’Hydrogène
dans l’Air, Confrontation avec l’Expérience’, Revue des Energies Renouvelables, Vol. 9,
N°3, pp. 229 – 236, 2006.