15064 gas turbine combustor - eddy characteristics
TRANSCRIPT
8/9/2019 15064 Gas Turbine Combustor - Eddy Characteristics
http://slidepdf.com/reader/full/15064-gas-turbine-combustor-eddy-characteristics 1/6
AN EDDY CHARACTERI ST I C T IME MODELI NG IN LES FOR GAS TURBI NE COMB USTOR
Mitsuru Yaga
Tohoku U niversity
Department of Chem ical Engineering
07, Aoba , Aramaki, Aoba-ku
Sendai 980-8579
Japan
TEL:+81-22-217-7252, FAX:+81-22-217-6165
Tsuyoshi Yamamoto
Tohoku University
Department of Chemical
Engineering
07, Aoba, Aram aki, Aoba-ku
Sendai 980-8579
Japan
TEL:+81-22-217-7252,
FAX:+81-22-217-6165
Hideyuki Aoki
Tohoku University
Department of Chemical
Engineering
07, Aoba, Aramaki, Aoba-ku
Sendai 980-8579
Japan
TEL: +81-22-217-7251,
FAX:+81-22-217-6165
Kousuke Sasada
Tohoku University
Department of C hemical Engineering
07, Aoba, Aramaki, Aoba-ku
Sendai 980 8579
Japan
TEL: +81-22-217-7252, FAX:+81-22-217-6165
Takatoshi Miura
Tohoku U niversity
Department of Chemical
Engineering
07, Aoba, Aramaki, Aoba-ku
Sendai 980 8579
Japan
TEL:+81-22-217-7250,
FAX:+81-22-217-6165
ABSTRACT
For calculating turbulent flame characteristics such as
temperature and chem ical species in the gas turbine combustor,
an effort applying Large Eddy Simulation (LES) to a
combustion simulation has been made in recent years.
However, there is n o established method of estimating reaction
rate on LES comb ustion simulation. In this paper, we construct
an eddy characteristic tim e derived from large-scale motio n to
calculate the combustion reaction rate using an eddy
dissipation concept model (EDC) and estimate combustion
characteristics (temperature and chemical species distribution)
in the combustor for the purpose of solving fundamental
problems of gas turbine combustor such as swirling effect or
formation of pollutant products. As a result, it is shown that
the combustion simulation using LES with EDC model is
effective method to calculate the characteristics of turbulent
diffusion flame su ch as g as turbine combustor.
KEYWORDS
Large Eddy Simulation, Combustion, Eddy Dissipation
Concept Model, Eddy Characteristic Time, Three Step Global
Reaction Mechanisms
INTRODUCTION
In order to construct a new combustor, scaling up of the
furnace usually advances in order o f laboratory-, batch-, pilot-
and commercial-scale. Since this met hod needs much cost and
time, a method o f using sim ulations adjusted on the
commercial furnace is recommended to predict the
characteristics of combustion. On the other hand, it is usually
said that experiments are better than th e simulations to predict
characteristics of combustion. However, it is difficult to
measure the characteristics o f combustion such as temperature
and chemical compounds even in the case of experiment and
the data include unavoidable errors, because the characteristics
of combustion are complicated in high temperature condition.
So the need of established combustion simulation model is
earnest ly desired. The simulatio n mod el is available for 1)
scaling up of the furnace, 2) development of new com bustor
purposing reduction of pollutant formation and adjustment on
various fuels and 3) s earch o f reasonable operation condition.
The sim ulation model is, furthermore, expected as an effective
technology to adjust various needs in future. Nowadays, from
the view of saving energy and reducing amount of discharged
environmental pollution substances, the importance of
controlling combustion phenom ena hasincreased. We may not
Copyright © 2000 by ASM E
Proceedings of2000 International Joint Power Generation Conference
Miami Beach, Florida, July 23-26, 2000
IJPGC2000-15064
8/9/2019 15064 Gas Turbine Combustor - Eddy Characteristics
http://slidepdf.com/reader/full/15064-gas-turbine-combustor-eddy-characteristics 2/6
8/9/2019 15064 Gas Turbine Combustor - Eddy Characteristics
http://slidepdf.com/reader/full/15064-gas-turbine-combustor-eddy-characteristics 3/6
¢~ : genera lized variable
or equivalence ratio [-]
<subscripts>
A r t : Arrhenius
e f f : effective
e d d y : eddy mixing
f u : fuel
i n : inlet
o x
: oxidant
<superscripts>
: spatial average
S I M U L A T I O N M E T H O D
G o v e r n i n o E a u a t i o n
All transport equations in this study can be expressed for
cylindrical three-dimensional geometry as:
( , o : ) + O --- / ; : l + ± l + ' - ±
O x - - r o t r O 0 ~
: o +
(')
oxk o~) fo rk or) ro ok ~oo)
where ff represents the dep endent variables which denote the
mass (1), momentum (u, v, w), enthalpy (h) and mass fraction
(m~; i=CI-h, O~, CO~, 1-120, N~, C O a nd 1-12). F + is th e
exchange coefficient, S# is the source term in the gas phase, p
is density, x, r and 0 are axial, radial and tangential coordinate,
respectively. Over bar means a spatial average. The source
terms and exchange coefficients are shown in Table 1. The
wall function model (Gosman and Pun, 1974) is applied to
calculate the flow near the wall. Radiative heat transfer is
calculated by 6-flux model (Gosman and Lockwood, 1972).
Mathemat i ca l Mode l ino
Large eddy simulation (LES) is a method of solving
spatial-averaged Navier-Stokes equation. In this method,
dependent variables are divided into resolved scale and
subgrid-scale (SGS) by a fil tering procedure. This procedure
is shown by following equation as:
f f f o , (2)
~ t J i = l
Table 1 Source terms and diffusion coefficients for
govcrining Equations.
Mass 1 0 0
Axial
u /teff
momentum
m
Radial v /~eff
momentum
m
Tangential w tt¢ff
momentum
o f d ; ) l o 0 ; 3 i o o g ) o-~
o f o : ] l o ( o ;3 l o f o ¢= ,3 1
7 t,'~'~
7 j , or
o ( o ; )
s < . n f
o(~,)
l o ; /
, ~ , { t , o : , J J , o : { - : , f
~w OP
r rO
Ma ss m / t~eff ~/
fraction am
Enthalpy ~ , u e f f 2 a R ( F x+ F r
F o -
3 E )
oh
/ 5 f 1 o ;
o;)]' f,(o= oZ312 h ( o ; o=
= , ~ ? l \
\ + fo ;q :+ V ;Lf o= +=)2
/
\ t~-J t T t ,~ ; j /
#eft --- # + /2t
O ' m O h
0.4 0.4
where G i ( x i , x ' i ) is a fil tering function and dash means a
fluctuation. In this study, the Gaus sian filter is used.
G(x,,x,0__ f-g7_ x p( - 6(x - x' ,Y )
V W --~ C T ~ 7 ) , (3)
where A is grid length. The dependent variable can be
decompose d as following equation:
= 7 + ¢ ' , 4 )
After the filtering procedure operated on Navier-Stokes
equations, the Reynolds term R O. is appear. The Reynolds term
is modeled as follows (written in Cartesian geometry):
l r r
Ru = - ~ u , u : oS:- , (5)
1 (0~ ag)
Copyright © 2000 by ASME
8/9/2019 15064 Gas Turbine Combustor - Eddy Characteristics
http://slidepdf.com/reader/full/15064-gas-turbine-combustor-eddy-characteristics 4/6
where o~0 s K ronecker's delta, vt is turbulent viscosity and S O.
is strain rate. V iscosity v t is further mode led as follows:
where Cs represents the Smagorinsky constant and w e assume
it as 0.2 (Smagorinsky, 1963).
React ion Model ino
We use three step global reaction mechanisms to express
methane-air com bustion reaction as:
C H 4 + 0 . 5 0 2 ~ C O + 2 H 2 ,
(9)
CO+0.502~::~CO2,
(10)
H2+0.502~::~H20, (11
Because the methane-air (oxygen) overall reaction (CI-L+202
CO2+2H20), often adopted to the turbulent combustion
simulation for saving calculation time, can not represent CO
and HE which are formed in local fuel-rich region. To consider
the interaction between eddy motion and chemical reaction,
the eddy dissipation concept (EDC) model is used to express
the reaction rate (Magnussen and Hjertager, 1976) as:
p . ( m - o x - - 1
= 4.0__ mm/_ _,mr, oq (12)
J '
where ~ is a stoichiometric oxidant requirement to burn lkg
of fuel, Reddy s eddy mixing rate o f fuel and oxygen and ~ is
eddy characteristic time. The eddy characteristic time z is
estimated with considering K olmogorov scale as below:
where e is eddy dissipation rate. From the assumption that
turbulence energy generation and dissipation are locally equal,
eddy dissipation rate is written as below (written in Cartesian
geometry):
= ~ t.O x - ~ x , -~-xj ' (14 )
Eq. (14) is substituted to Eq. (13), the eddy characteristic time
is finally expressed as below:
I =
( 1 5 )
Considering both the chemical reaction and the eddy motion,
reaction rate is finally expressed a s :
ff~ = - m i n ~ , a ~ , ~ ) , 1 6 )
where A is pre-exponential factor of rate constant, E is
Activation energy, R is universal constant of gases and T is
temperature. The value of coefficients A, E, a and b are
referred from Jones (Jones and Lindstedt, 1988).
N u m e r i c a l S o l u t i o n
The SIMPLE algorithm with TDMA method (Patanker,
1980) is used to solve the partial differential equations sh own
in (1). In this simulation, equatio ns of continuity, mom entum ,
enthalpy and gas species mass fractions are discretized in
space by a control volume method. The convective terms are
differenced in space with the quadratic upstream interpolation
for convective kinematics m ethod (Leonard, 1979) and the
diffusive terms are differenced in space w ith the second-order
central difference scheme. The fully implicit scheme is used
for time marching and time step is 0.05s. The schematic
diagram of computational domain is shown in Fig. 1. The
coaxial combustor has 200m m in internal diam eter and
800mm in length. The inner pipe diameter is 6 5mm and the
annulus pipe diameter is ~ 23mm. A computational grid
number in axial, radial and tangential direction are 120 × 40 x
50, respectively.
~
800 ~[
ir . . . . . . . ~ . . . . ~ . l ~
o - . . . . . J J . . . . . .
Fig. 1 Schematic diagram of computational domain.
The Schem atic diagram of experimental setup is shown in
Fig. 2. This combustor has 200ram in diameter and 800mm in
length. The inner pipe diameter is ~ 5ram and the annulus
pipe diameter is ~ 23 mm The experimental condition is
shown in Table 2. A flame temperature is measured by using
suction pyrometer probe equipped with R-type thermocouple.
Gases are sucked through a water-cooled sampling probe made
of stainless steel. The tip diameter of gas sampling probe is O
4.32mm. Intensive cooling starts at the joint o f the probe tip
Copyright © 2000 by ASM E
8/9/2019 15064 Gas Turbine Combustor - Eddy Characteristics
http://slidepdf.com/reader/full/15064-gas-turbine-combustor-eddy-characteristics 5/6
with a water jacket cooling system where chemical reaction
halts. Sampling gases are analyzed by a gas chromatograph.
. a m p . i n . e W a l l o . o o
f \
ii2 c o m p ro
Methane cylinders
Combustor
/
Gas chromatograph
Fig 2 Schematic diagram of experimental setup.
Table 2 Experimetal condition.
0.1 lml
Air
CI%
Fig. 4 Calculated instantaneous eddy characteristic time
distribution in computational domain.
Figure 5 shows the radial distribution of temperature in
the equivalence ratio ~ = 1.0 at x= 0.1m dow nstream from the
burner. From the Fig. 5, good agreements are seen except for
the centedine. The reason why thermocouple system indicates
exceed temperature is that the effect of radiation from suction
pyrom eter tip is large.
1800
1600
[-] 1.0 140o
CI-h [Nm3 h l] 0.2 0 ~ 12oo
Air [Nm3 h a] 1.90 ~ 100o
Temp erature [K] 293.15 ~ 800
6 0 0
ReD [-] 33228 w
4 0 0
2OO
0
RESULTS AND DISCUSSION
Figure 3 shows the predicted time mean velocity of
combustion gas flow. The line that axial velocity is zero is
plotted on Fig. 3. The strength of the vortex shedding is much
weaker by heat release, large-scale eddies are broken up
because thermal expansion occurs in the vortex core (Fureby
and L/ffstr0m, 1994).
Figure 4 shows the distribution of eddy characteristic time
calculated by the present model. The eddy characteristic time
in the region whe re CI-h and 0 2 are mixed is short and it is
expected that LES combustion simulation is carried out with
the strict assumption o f eddy dissipation concept model. The
combustion reaction occurs at the moment that fuel and
oxygen are fully mixed in micro-scale eddy.
0.1 [ml 0.4 [m|
i
A i r ~ ~ ~
~ u / U ~ = l . O [-]
Fig. 3 Predicted time mean velocity vectors
in computational domain.
• Experiment ]
.......... LES
0 0.1 0.2 0.3 0.4 0.5
r/D [-I
Fig. 5 Radial distribution of temperature at 0. lm from
burner at ~ =1.0.
Figure 6 shows the radial distribution of CFL mole fraction
in the equivalence ratio 6=1.0 at x=0.1m downstream from
the burner. This calculatio n result ,agrees well with the
measured results of CH4 mole fraction.
0.8
O
0.6
0 .4
0 .2
0
[ o ExperimentLEs1
. . . . . . . . . v v
0 0.1 0.2 0.3 0.4 0.5
rid [-]
Fig. 6 Radial distribution of CI-h mole fraction at 0. lm
fro m bu rne r a t 4~ -- 1.0.
Copyright © 2000 by ASM E
8/9/2019 15064 Gas Turbine Combustor - Eddy Characteristics
http://slidepdf.com/reader/full/15064-gas-turbine-combustor-eddy-characteristics 6/6
i
Figure 7 shows the radial distribution of CO m ole fraction
in the equivalence ratio ~ =1.0 at x=0.1m downstream from
the burner. This calculation result overestimates experim ental
result near the centerline. The reason is because the CO
consum ption reaction is too simple to estimate the distribution
of CO mole fra ction correctly.
The selection of a reac tion mech anism seriously affects on
the calc ulated results of the distribution o f chemical species.
o.1
0 .08
0
0 .06
0 .04
0 .02
o
• LEsEXperiment I
, O o , o ; • • • • .
0 0.1 0.2 0.3 0.4 0.5
r/D [-]
Fig. 7 Radial distribution of CO mole fraction at 0. lm
fro m burn er at 4~ =1.0.
CONCLUSION
A three-dimensional Large eddy simulation turbulent
combustion simulation in a coaxial combustor is carried out.
We construct an eddy characteristic time model derived from a
large-scale motion to estimate the combustion reaction rate
using eddy dissipation concept model for saving a
computational time. An estimation of eddy characteristic time
considering Kolmogorov scale is effective for knowing subgrid
scale motion of flame region and the present model is more
suitable for the assumption of eddy dissipation concept model
than a model using time averaged turbulent energy and eddy
dissipation rate. In this study, the calculation results of
temperature and CI-h mole fraction distribution agree well with
the m easured results, however the calculated result of CO mole
fraction distribution does no t well agree with m easured result
for the sake of using simple CO reaction mechanism. If we
want to apply LES combustion simulation on engineering
applications such as gas turbine combustor, w e should
thoroughly check the reaction model with reaction mechanism
whe ther it could be suitable f or LES com bustion simulation or
not.
REFERENCES
Cook, A. W. and Bush¢, W. K, 199 9, A Subgrid-Sca le
Model for the Scalar Dissipation Rate in nonpremixed
combustion, P h y s i c s o f F lu i d s , V ol. ll, pp. 746-748.
Fureby, C. and L6fstr6m, C., 1994, Large Eddy
Simulation of Bluff Body Stabilized Flames,
T wen ty -F i f th
S i m p o s i u m ( I n t e r n a t i n a l ) o n C o m b u s t i o n , The Combustion
Institute, Pittsburgh, PA, pp. 1257-1264.
Gosman, A. D. and Lockwood, E C., 19 72 , In
Corporation of a Flux Model for Radiation into a Finite-
Difference Procedure for Furnace Calculation, Fo ur teen th
S i m p o s i u m ( I n t e r n a t i n a l ) o n C o m b u s t i o n , The Combustion
Institute, Pittsburgh, PA, pp. 661-671.
Gosman, A. D. and Pun, W. M., 1974, Lecture Notes for
Course Entities Computation of Recircu lating Flows, Imp erial
College Hea t Transfe r See. Report HTS/74/2.
Jones, W. P., and Lindstedt, R. P., 198 8, Global Reaction
Schemes for Hydrocarbon Combustion, C o m b u s t i o n a n d
F l a m e . V ol. 73, pp, 233-249.
Launder, B. E., 1972, The Prediction of Laminariz ation
with a Two-Equation Model of Turbulence , I n t. J . H e a t M a s s
Transfer, V ol 15, pp. 301-314.
Leonard, A., 197 4, On tile Energy Cascade in Large-
Eddy Simulations of Turbulent Flows, A d v . G e o p h y s . , V ol.
18A, pp. 237-248.
Leonard, B. E, 1979 , A stable and accurate convective
modeling procedure based on quadratic upstream
interpolation,
C o m p u t . M e t h . A p p l . M e c h . E n g . ,
V ol. 19, pp.
59-98.
Magnussen, B. E, and Hjertager, B. H., 19 76 , On '
Mathematical Modeling o f Turbulent C ombustion with Special
Emphasis on Soot Formation and Combustion,
S ix teen th
S im p o s ium ( I ntern a t in a l) o n Co m b us t io n , The Combustion
Institute, Pittsburgh, PA, pp. 719-729.
M611er, S. L., Lundgren, E a nd Fureby. C., 19 96, Large
Eddy Simulation of Unsteady Combustion, T wen ty -S i x th
S i m p o s i u m ( I n t e r n a t i n a l ) o n C o m b u s t i o n , The Combustion
Institute, Pittsburgh, PA, pp. 241-248.
Patankar, S. V ., 1980, Num erical Heat Transfer and Fluid
Flow, Hemis. pub. co., Washington.
Smagorinsky, J., 1963, Gene ral Circulation Expe rimen ts
with the Primitive Equations, Mo rt. W eath . Rev. , V ol. 91, No.
3, pp. 99-164.
6 Copyright © 2000 by ASM E