Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Fuel
journal homepage: www.elsevier.com/locate/fuel
h i g h l i g h t s
A multi-objective optimization of a gas-turbine combustor is performed.
An efficient model is proposed for the design optimization process.
Entropy generation minimization is considered with other conventional targets.
The Swirl number design criterion is revisited based on the new analysis.
a r t i c l e
i n f o
Article history:
Received 29 March 2016
Received in revised form 28 May 2016
Accepted 7 September 2016
Keywords:
Swirl number
Entropy generation minimization
Computational optimization
Combustion chamber
Gas turbine
a b s t r a c t
In this study, the non-premixed swirl-stabilized flame in a gas turbine combustor is simulated. For this
purpose, first, an efficient model for the simulation of combusting flow in real combustor liners is proposed which is appropriate for the computationally expensive design optimization process. Then, this
model is used to perform a multi-objective optimization of a real combustor based on the combustion
efficiency, pattern factor, pollutant (CO and NO) emission, and entropy generation minimization targets.
After investigating the effect of the Swirl number on each objective through response surfaces and sensitivity analysis, the optimal case is introduced by the decision making process. The results show that the
base case Swirl number, chosen based on a well-known correlation, is the best case if only the combustion efficiency is maximized. However, the overall optimal case which is the trade-off among the above
four objectives possesses a Swirl number 44 percent smaller than the base case.
2016 Elsevier Ltd. All rights reserved.
1. Introduction
Most of the energy produced is driven from fossil fuel resources.
One of the shortcomings of these sources of energy is pollutant
emission which is the researchers main concern because of its
effects on the global warming and the climate changes of the earth.
However, natural gas power generation facilities have been considered as more environmentally friendly power sources compared to
other hydrocarbon-fueled power plants. Gas turbines are high efficiency energy producing systems which can work with natural gas,
but still have emission such as NOx, CO2, and CO. Therefore, to alleviate this harmful effects researchers are working to reduce the
emission from gas turbines by changing the geometry or working
conditions of the gas turbine combustion chamber.
Corresponding author at: Mechanical Engineering Dept., Amirkabir University
of Technology (Tehran Polytechnic), 424 Hafez Avenue, P.O. Box: 15875-4413,
Tehran, Iran.
E-mail addresses: m.m.torkzadeh@aut.ac.ir (M.M. Torkzadeh), fbolourchi@aut.
ac.ir (F. Bolourchifard), eamani@aut.ac.ir (E. Amani).
http://dx.doi.org/10.1016/j.fuel.2016.09.022
0016-2361/ 2016 Elsevier Ltd. All rights reserved.
735
Method
Model (Turbulence/
Combustion)
Application
Findings
Computation
LES/Cold flow
Experiment
Experiment
Premixed kerosene-fueled
combustor using lean staged
combustion
Computation
Realizable k-e/Mixture
fraction-Probability
density function (PDF)
Non-premixed gas-fired
combustor with hydrogencontaining fuel
Shanbhogu et al.
(2016) [13]
Experiment
Taamallah et al.
(2016) [14]
Experiment
Experiment
Computation
LES/Conditional moment
closure
Swirl-stabilized non-premixed
combustor
Krieger et al.
(2015) [17]
Computation
Experiment
and
Computation
Realizable k-e/
Equilibrium nonpremixed flamelet
Non-premixed kerosene-fueled
swirl laboratory combustor
Santhosh and
Saptarshi
(2015) [19]
Experiment
Experiment
and
Computation
Kerosene-fueled swirl-cup
combustor
Mahesh and
Mishra (2015)
[21]
Experiment
and
Computation
SST k-x/non-premixed
flamelet
Computation
LES/Transported filtered
density function
Experiment
Colorless Distributed
Combustion under both swirling
and non-swirling conditions
Experiment
and
Computation
Realizable k-e/
Equilibrium presumed
PDF
Non-premixed kerosene-fueled
model combustor
Computation
LES/Eddy dissipation
concept
Hermeth et al.
(2013) [26]
Computation
LES/Dynamic thickened
flame
Experiment
Koyama and
Tachibana
(2013) [28]
Experiment
Experiment
and
Computation
LES/G-equation flamelet
Steinberg et al.
(2013) [30]
Experiment
Computation
LES/Cold flow
736
Table 1 (continued)
Authors (year)
Method
Model (Turbulence/
Combustion)
Application
Findings
De and Acharya
(2012) [32]
Computation
LES/Thickened flame
Franzelli et al.
(2012) [33]
Computation
LES/2S_CH4_BFER
Methane-air premixed
laboratory-scale combustor
Computation
LES/Dynamically
thickened flame
2. Mathematical model
Swirl number is an important parameter in the design and operation of gas-turbine combustors. Swirl number is the ratio of the
tangential momentum to the axial momentum and is defined by
@
quj 0
@xj
r
@P @ sij
qu u
sij
@ui @uj
2
@ul
leff
leff
dij
3
@xj @xi
@xl
leff l lt
3
4
where l and lt are the molecular and turbulent viscosity, respectively. lt is determined by a turbulence closure model. Here, the
4-equation transition k x SST model [54] which solves transport
R Ro
UWr2 dr
R
S Ri Ro 2
U rdr
Ri
where Ro and Ri are the outer and inner radius of the swirler, respectively, U is the axial velocity and W is the tangential velocity of inlet
air flow. In this study, we aim to carefully investigate the effects of
swirl velocity of inlet air on several important objectives in the
design process of a real gas-turbine combustion chamber and determine optimal Swirl number to satisfy all targets.
Three conventional objectives in the design of combustors are
combustion efficiency, pollutant emission, and Pattern Factor
(PF). Here, combustion efficiency is calculated by
gc
P
_ s
inlets mh
_ FQF
m
outlet mhs
737
Design parameters
Optimization objectives
Applications
Findings
Chmielewski and
Gieras (2016)
[35]
Combustion efficiency,
NOx, and CO emissions
averaged in a range of
operations
Entropy generation,
maximum temperature
and flame height
5 Geometry parameters,
equivalence ratio, reactant mass
flow rate, and injection strategy
Normalized efficiency
and fuel consumption
Makhanlall et al.
(2013) [6]
Equivalence ratio
Entropy generation
Kerosene-fueled
single-stage burner
with flameless/MILD
combustion
Hajitaheri (2012)
[38]
Combustion efficiency
14 geometry parameters,
equivalence ratio, and reactant mass
flow rate
Normalized efficiency
and fuel consumption
Wankhede et al.
(2011) [40]
NO emission
Premixed model
combustor
Burmberg and
Sattelmay (2011)
[41]
Swirl number
Flashback constraint
Premixed low
emission combustor
Motsami et al.
(2010) [42]
Kulshreshtha et al.
(2010) [43]
Kerosene-fueled can
type combustor
Lean Premixed
prevaporized
combustion chamber
Pattern factor, NO
emissions, and pressure
loss
Lean premixed
methane-fueled
tabular combustor
Duchaine et al.
(2009) [46]
Combustion efficiency
and pattern factor
Mixing quality
Non-premixed
methane combustion
in a cylindrical
chamber
Maximum temperature
and flamlet uniformity
Methane-fully
premixed burner
Combustion efficiency
and pressure loss
Non-premixed
methane micro
combustor
738
Table 2 (continued)
Authors (year)
Design parameters
Optimization objectives
Applications
Findings
CO emission
Laminar burner
Combustor weight,
vibration frequency,
and low-cycle fatigue
life constraints
Annular aircraft
engine combustor
Kerosene-fueled BSE
Ltd. YT-175 engine
(annular reversal flow)
combustor
Jyothishkumar and
Ganesan (2005)
[53]
T outlet;max T outlet;ave
T outlet;ave T inlet
PF
where T outlet;max and T outlet;ave are the maximum and average temperature at the outlet section of combustor, respectively, and T inlet is
the inlet air temperature.
To account for both CO and NOx emissions, a merit function is
usually defined by
f
CO
CObase
1000
NOx
NOx
base
where CO and NOx in this equation stand for CO and NOx mass flow
rates out of the combustor outlet. Also, the subscript \base" refers
to the values corresponding to the base case design which is
described in Section 3.
Here, we consider another objective, i.e. entropy generation
minimization, to account for the quality of energy conversion
and reduce the energy loss. For this purpose, the total entropy generation rate in the combustor is calculated by
I
Sgen
q
Aoutlet
walls
!
I
X
Y i si uj dAj
i
dQ
T
Ainlets
!
X
Y i si uj dAj
LRZ 2Ro S
10
739
Fig. 1. The geometry of the liner of the combustor studied in this work (top), sectional view of the liner (bottom). All dimensions are in millimeters.
Table 3
Natural gas composition (mole fractions).
Species
Fuel
CH4
C2H6
CO2
N2
0.9
0.085
0.005
0.01
Table 4
The specification of the combustion chamber. FOA indicates the
fraction of the total air entering the combustor from each section.
a round groove around the liner whose inlet area equals the area of
the original holes (see Fig. 4), i.e.
npD2hole =4 pDliner wr
11
Design parameter
Value
Liner diameter
Fuel flow rate
Inlet fuel temperature
Number of primary holes
Diameter of primary holes
Number of secondary holes
Diameter of secondary holes
Number of dilution holes
Diameter of dilution holes
Total inlet air flow rate
Primary holes FOA
Secondary holes FOA
Dilution holes, 1st row, FOA
Dilution holes, 2nd row, FOA
Swirler FOA
Cooling air, 1st slot, FOA
Cooling air, 2nd slot, FOA
Cooling air, 3rd slot, FOA
Cooling air, 4th slot, FOA
Inlet air temperature
Inlet air pressure
Inlet air Swirl number (swirler)
245 mm
0.0712 kg/s
430 K
20
18 mm
20
14 mm
8 (two rows)
34 mm
4.288 kg/s
20%
15%
15%
15%
12%
8%
5%
5%
5%
461 K
3.899 bar
1.87
given Swirl number. The inlet flow through cooling slots is also
normal to the boundary face, parallel to the liner walls. This flow
forms an air layer on the inner liner wall to protect it from hot
combustion gaseous. It is known that the inflow from primary, secondary, and dilution holes is not normal to the hole surface. Here,
740
Rcond
ttbc tliner
ktbc kliner
17
where t and k are the thickness and thermal conductivity, respectively. The thermal barrier coating (TBC) protects combustor liner
from overheating and reduces the heat loss across the walls. The
radiation resistance is calculated by
Rrad
CD
C D1
13
where h is the angle between the inflow and main flow direction
and C D the discharge coefficient. An estimation of the discharge
coefficient for round sharp-edged holes is provided by [68]
1:25K 1
4K K2 b2
2
Dan
Nu kf
19
K!1
CD
Rconv
12
C D1 lim C D
0:5
0:8
T an
Tf
b
b0
Tf
T an T w2
;
2
ReD;F
_ an
m
qan Aan
Dan
tf
14
15
16
20
where b is the ratio of the hole mass flow rate to annulus, between
liner and casing, mass flow rate, and K is the ratio of the jet dynamic
pressure to the annulus dynamic pressure upstream of the holes.
The coefficient K is approximated by the correlation [68]
q
K 1 0:64 2w2 4w4 1:56w2 4b b2
18
sin h
Fig. 3. The 2D axisymmetric geometry of the model combustor liner (figure has been plotted with the scale shown at the bottom).
21
741
Fig. 5. Left: the configuration of the liner wall (e=Dan 0:0275). TBC stands for thermal protection of the liner. Right: the thermal circuit model for the wall boundary
condition.
_ an and T an are local air mass flow rate and bulk air temperature
m
through the annulus and tf and Pr f are dynamic viscosity and
Prandtl number of air evaluated at T f .
(a)
(b)
(c)
(d)
(e)
Fig. 6. Temperature versus radial direction in different cross sections with different distances from the inlet nozzle (a) x = 6.8 mm, (b) x = 20 mm (c) x = 40 mm (d) x = 70 mm
(e) x = 125 mm. Comparison between experimental measurements [70] (symbols) and present computations (lines).
742
(b)
(a)
(c)
(d)
(e)
Fig. 7. Axial velocity versus radial direction in different cross sections with different distances from the inlet nozzle (a) x = 6.8 mm, (b) x = 20 mm (c) x = 40 mm (d) x = 70 mm
(e) x = 125 mm. Comparison between experimental measurements [70] (symbols) and present computations (lines).
(a)
(b)
(c)
(d)
(e)
Fig. 8. Swirl velocity versus radial direction in different cross sections with different distances from the inlet nozzle (a) x = 6.8 mm, (b) x = 20 mm (c) x = 40 mm (d) x = 70 mm
(e) x = 125 mm. Comparison between experimental measurements [70] (symbols) and present computations (lines).
tioned above, this region is also the region of high turbulent kinetic
energy which promotes mixing and reaction rates (in nonpremixed regime). Therefore, the flame surface is located in this
region, near walls, and special care must be taken to prevent the
dome walls from burning. This is the reason why larger cooling
flow rates has been assigned to the dome cooling slot (see Table 4).
Figs. 18 and 19 illustrate pollutant emissions, including NO and
CO. It can be observed in these figures that CO forms first but oxidizes to CO2 due to the increase of temperature beyond 1700 K. In
this case, due to the relatively large length of the combustor the
level of CO is low at the combustor exit. However, this is not the
case for NOx emission (see Fig. 19). NOx formation is considerable
in this case due to the high temperatures. Maximum NOx exists in
the middle of combustor near the centerline, but further downstream the level of NOx decreases a bit due to the effect of air
entering from dilution holes.
743
(c)
(d)
(e)
Fig. 9. CO2 mass fraction versus radial direction in different cross sections with different distances from the inlet nozzle (a) x = 6.8 mm, (b) x = 20 mm (c) x = 40 mm (d)
x = 70 mm (e) x = 125 mm. Comparison between experimental measurements [70] (symbols) and present computations (lines).
(a)
(b)
(c)
(d)
(e)
Fig. 10. CO mass fraction versus radial direction in different cross sections with different distances from the inlet nozzle (a) x = 6.8 mm, (b) x = 20 mm (c) x = 40 mm (d)
x = 70 mm (e) x = 125 mm. Comparison between experimental measurements [70] (symbols) and present computations (lines).
Fig. 11. The computational grid used in this study. Top: full view. Bottom left: near inlet. Bottom right: near dilution holes.
744
(d)
(c)
(b)
(a)
Fig. 12. Comparing temperature profiles using main (solid line) and fine (dashed line) grid in different cross sections with different distances from the inlet (a) x = 0.09 m, (b)
x = 0.11 m (c) x = 0.145 m (d) x = 0.18 m.
W (m/s):
30
55
80
105
130
155
180
205
Fig. 13. Contour of the swirl velocity for the base case (S = 1.87). The contour level W = 38 (m/s) emphasized by the black line surrounds the high swirl region.
Fig. 14. Contour of the axial velocity for the base case (S = 1.87).
Fig. 15. Contour of the turbulent kinetic energy for the base case (S = 1.87).
745
Fig. 17. Contour of the temperature for the base case (S = 1.87).
Fig. 18. Contour of the mass fraction of CO for the base case (S = 1.87).
Fig. 19. Contour of the mass fraction of NO for the base case (S = 1.87).
7800
2500
7600
2300
Temperature (K)
Merit funcon
7400
7200
7000
6800
6600
6400
DOE points
Response Curve
6200
0.5
1900
Tmax outlet
Tmax chamber
1700
6000
0
2100
1.5
2.5
1500
0
0.5
Swirl number
1.5
2.5
Swirl number
Fig. 20. Left: Response Surface (RS) of the merit function, f, as a function of the Swirl number. Right: Maximum temperature of outlet and chamber versus Swirl number.
1.6
1.5
1.4
Paern factor
1.3
1.2
1.1
1
DOE points
Response Curve
0.9
0.8
0
0.5
1.5
2.5
Swirl number
Fig. 21. RS of the Pattern Factor, PF, as a function of the Swirl number.
746
Combuson eciency
0.99
0.98
0.97
0.96
0.95
DOE points
Response Curve
0.94
0
0.5
1.5
2.5
Swirl number
Fig. 22. RS of the combustion efficiency, gc , as a function of the Swirl number.
4900
DOE points
Response Curve
Sgen (W/K)
4800
4700
4600
4500
4400
0
0.5
1.5
2.5
Swirl number
Fig. 23. RS of the total entropy generation rate, Sgen , as a function of the Swirl
number.
0.1
Eciency
Sgen
0
-0.1
Merit
funcon
Paern
factor
-0.2
-0.3
-0.4
-0.5
factor will decrease which indicates the better mixing and more
uniform outlet temperature profile induced by the larger Swirl
numbers.
Table 5
Candidate points obtained from present multi-objective optimization.
Parameter
Swirl number
Combustion efficiency
Merit function
Pattern factor
Candidate point 1
Candidate point 2
Candidate point 3
1.053
0.937
0.821
0.9731
0.9682
0.9638
7229.09
7437.93
7568.98
1.335
1.384
1.421
4507.80
4501.31
4498.94
747
Table 6
The results of the single-objective optimization.
(Single) objective
Swirl number
Combustion efficiency
Merit function
Pattern factor
Pattern factor
Combustion efficiency
Merit function
Entropy generation
2.200
1.841
0.757
0.766
0.9896
0.9980
0.9624
0.9625
6373.17
6441.07
7589.63
7589.14
0.959
1.057
1.434
1.433
4823.74
4836.17
4498.65
4498.64
case value (S = 1.87). This indicates that the base case (original criterion for the Swirl number [65]) can be acceptable when only the
first-law efficiency is considered.
Fig. 25. Temperature contours of the case listed in Table 6. From top to bottom: minimum PF (S = 2.20), maximum gc (S = 1.841), multi-objective optimal (S = 1.053) and
maximum f (S = 0.757).
Fig. 26. Contour of CO (up) and NO (down) mass at the condition of maximum merit function (S = 0.757). For direct comparison of the colors of this figure with those of
Figs. 18 and 19, exactly the same legends are used for all these figures.
748
749