Analysis of reacting flows in an aero-engine afterburner usingcomputational fluid dynamics

dian Journalof Engineering & Materials Sciences01.11,February2004, pp. 31-37

-- ---

to simulate the characteristics of practical afterburnercombustion system. Kumar'? proposed a method toestimate pressure loss across flameholders in highvelocity streams. Experiments were carried out in a 20inch diameter afterburner fitted with an anti-screechliner and spray bars for fuel injection. Ganesan" pre-dicted the turbulent flow behind a conical baffle usingfinite difference prediction procedure. The flow wasassumed to be 2D axisymmetric, steady, turbulent andnon-reacting. Governing differential equations, to-gether with boundary conditions were solved by fi-nite-difference method based on SIMPLE algorithm.Zhang and Chiul6 developed a numerical method andcomputer code for modeling afterburner spray com-bustion and gas phase combustion processes, based onthe Navier-Stokes equations and finite differencemethod. Chuang and Jiangl? analyzed diffusion flamein an afterburner as a function of air-fuel ratio by em-ploying SIMPLE-C algorithm and k-t: model for tur-bulence. In order to simulate practical afterburnerflow, two open-mouth-type V-gutters were consideredinstead of a solid cone as employed by Zhang andChiul6

.

25 30

Sunil V Unaune & V GanesanInternalCombustion Engines Laboratory, Department of Mechanical Engineering, Indian Institute of Technology,

Chennai 600 036, Indiav]

01 feed rate, materi1vibrations to the rl1ll

Received 4 February 2003; accepted 16December 2003

In this paper, reacting flows in an aero-engine afterburner are analyzed using computational fluid dynamics (CFD).A computationalprocedure is described for calculating the three-dimensional reacting flow fields in a gas turbine after-burner.The computations are based on numerical solution of time-averaged transport equations for mass, momentum, tur-bulentkineticenergy and dissipation rate using a finite volume formulation. The numerical calculations are performed usingSIMPLE(Semi Implicit Method for Pressure Linked Equations). The RNG (Re-normalization Group Theory) k-s model isusedfor turbulence modeling. Combustion is modeled using PDF (Probability Density Function). The results for air-fuelratioof 30 and 46 are obtained and analyzed.

he amplitude an'ace and helps inship between,urface roughnes"

In aircraft applications, larger thrust for a small dura-tion is required sometimes. This can be achieved intwo ways. One way is to increase the mass flow rate.

'E Magazine MarchiThe other way is to add energy by burning additional, fuel in the tail pipe between turbine exhaust and the

-387. entrance section of the exhaust nozzles. This way ofind nanometro!ogy, t~rust. augmentation is called afterburning and the ~e-adelphia),2003. vice IS called the afterburner. Due to the ever in-itni G, Steinert J & creasing demand for thrust, especially in military air-I. crafts, afterburner seems to be the most practicalJ C, Opt Eng, 24 method during such flight operations. Some investi-

gations'? provide details about the flow and flamefjineering, Vol. x I studies of systems applied to afterburners and flameNyant, (Academic stabilization that are close to afterburning systems. In

modern theoretical analysis, CFD plays a major partand its details are also known'<'",

Useller" presented the results of several full-scaleturbojets-engine afterburner investigations that havebeen conducted by NASA. Considering combustion

Vlanuf, 41 (2001 f efficiency as a criterion of the afterburner perform-) ance, effect of combustion chamber length on after-

:ingX Z & Xiao I burner combustion performance was experimentallydetermined. Macquisten and Dowling'? investigated

-;IRP, 34 (1895) , low frequency combustion oscillations in a modelafterburner. An experimental rig with a conical gutterflame stabilizer was run for a range of inlet Machnumbers and temperatures representative of an engineafterburner. Baxter and Lefebvre'< conducted experi-ments on bluff body flame stabilization to obtainweak extinction limit data from an apparatus designed

) 201-208.

r, Poduje N &1.

'87) 17-20.

125-130.0) 77-87.

'C Eng, 26 (2002)

TheoreticalSchematic of an afterburner shown in Fig. 1 con-

sists of an exhaust diffuser with struts, fuel manifolds,flame stabilizer, combustion chamber with anti-screech holes and cooling rings and variable area noz-zle. Liquid fuel is added through fuel manifolds and

32 INDIAN J. ENG. MATER. SCI., FEBRUARY 2004

ann-screech hOles

nozzle

Fig. I - Geometry of the afterburner (600 sector)

the combustion process is initiated in the wakes cre-ated by the flame stabilizer. The core flow leaving theturbine is de-swirled and is decelerated in the diffuser.The mixing further depends on turbulence created byv-gutter, the distance between the v-gutter and fuelmanifolds, the method of injection of fuel etc. Further,the entrainment within the re-circulation zone plays akey role in the sustained combustion process. So, it isnecessary to carry out a detailed study involving vari-ous parameters affecting mixing and combustion.

Governing equationsThere are seven unknowns namely pressure, tem-

perature, density, velocities along the three co-ordinate axes and internal energy for fluid. Now,seven governing equations are required to solve theproblem. These governing equations are continuity,momentum equations in three co-ordinate axes, en-ergy equation, two equations of state.

Continuity equation

where u, is the mean velocity component and Sm is thesource of the mass added to the continuous phase dueto the vaporization of fuel droplets.

Momentum equation in x-direction

a ap ar. .a 2-(puu.)=....:_+_'J ....:-(-pk8.)+Fax . I J ax. ax . ax. 3 IJ IJ I J J "

[ (au, au . 2 au)]where r ..= Jl -; +_J --0.-;'J 'If ax ax 3 'J ax

J' ,

The effective viscosity J.1ejf is the sum of turbulentviscosity J.11 and laminar viscosity J.1b p is the staticpressure, k" the turbulent kinetic energy and F, is theexternal body force that arises from interaction withthe dispersed phase in the i direction. The turbulentviscosity is obtained from the turbulent kinetic energyk and dissipation rate E.

k2

J.11 = pCf.l-E

where CJ.l=0.0845. k and E are obtained from the so-lution of their transport equations which are derivedfrom the RNG theory.

Energy equation

where H is the total enthalpy and Sh is the source ofenergy due to chemical reaction.

Combustion modelingCombustion phenomena have been modeled using

probability density function (PDF) approach. PDFapproach is good for turbulentdiffusion flarnes'". Fuelis injected in the primary zone where it mixes with theoxidizer and burns. So, the flame is just like the tur-bulent diffusion flame. In', PDF approach transportequations for two conserved scalars (the mean' mix-ture fraction and its variance) are solved and individ-ual component concentrations are derived from thepredicted mixture fraction distribution. Here the tur-bulence effects are accounted for with the help of aprobability density function. Equilibrium chemistry

UN

model is usedtry model assuexists at mole,minimization cspecies mole j

Thus value ofdensity, etc.) dvariance and tl

For non-aditaneous enthaltransport equaPDF look-upscalars for dilfraction, its Val

Turbulence mod,Flow throng

aged Navier-Ssix additionalif the turbulenterms of any •be calculated.tional procediReynolds stre,terburner hasmodel is usedthe swirl effecReynolds num

Pressure-velociuSemi Impli

tion (SIMPLEcoupling. SI.Nhas been SUCCI

procedures. Itder -relaxationgence. By trifactors are sek

For solvingintegration ovtion domain isgral equationorder to solvecalculate the tthe first orderOne of the mencing schermrection. VelocHence flow diculating the 1

which the up

,

turbulentthe staticF; is the

tion withturbulentic energy

n the so-: derived

+Sh

ource of

ed using.h. POP:SIS. Fuelwith the.the tur-rarisportan 'mix~individ-rom thethe tur-elp of aiernistry

UNAUNE & GANESAN: ANALYSIS OF REACTING FLOW IN AN AERO-ENGINE AFTERBURNER ' 33

model is used during POF calculations. This chemis-try model assumes that chemical equilibrium alwaysexists at molecular level. An algorithm based on theminimization of Gibbs free energy is used to computespecies mole fractions from mean mixture fraction.Thus value of any averaged scalars (e.g. temperature,density, etc.) depends upon mean mixture fraction, itsvarianceand the chemistry model'".

For non-adiabatic system, it depends upon instan-taneous enthalpy also. Hence the solver solves thetransport equation for instantaneous enthalpy. also.PDF look-up table contains the values of differentscalars for different combinations of mean mixturefraction,its variance and enthalpy.

Turbulence modelingFlow through afterburner is highly turbulent. Aver-

aged Navier-Stroke equation for turbulent flow hassixadditional stresses called Reynolds stresses. Now,if the turbulent viscosity is calculated or expressed interms of any known variable, Reynolds stresses canbe calculated. Turbulent models develop computa-tional procedures to predict turbulent viscosity (orReynolds stresses directly). As the flow through af-terburner has swirl component at inlet, RNG k-emodelis used for turbulence modeling to take care ofthe swirl effect. RNG k-e model is also good for lowReynolds number flows IS.

Pressure-velocity couplingSemi Implicit Method for Pressure Linked E~a-

tion (SIMPLE) algorithm is used for pressure-velocitycoupling. SIMPLE-algorithm is straightforward andhas been successfully implemented in numerous CFOprocedures. It produces correct velocity field2o. Un-der-relaxation factors are used to avoid the diver-gence. By trial and error method under-relaxationfactors are selected for convergence. '

For solving the governing equations of! fluid flow,integration over all the control volumes ~ithe _~olu-tion domain is taken. Discretisation convdrts the inte-gral equation into a system of algebraic equations. Inorder to solve these algebraic equations 'we need to -calculate the transport property at the cell faces:'! So,the first order upwind differencing scheme is used.One of the major inadequacies of the centraldiffer-encing scheme is its inability to identify the flow di-rection. Velocities in the afterburner are quite high ..Hence flow direction should be considered while-cal-culating the transport properties at the cell faces,which the upwind differencing scheme does. The

analysis has been carried out at sea level operatingconditions.

Grid generationPre-processor GAMBIT has been adopted for geo-

metric modeling of afterburner. Only 60° sector of theafterburner is modeled for analysis so as to exploit the'symmetry in the geometry. An unstructured grid ofsize'8,30,678 cells has been used for meshing. F(g~2shows the radial and ring V -gutters used for flamestabilization. There are 18 top and 6 bottom radial V-gutters.

The main objective of present work is to study re-acting flows in afterburner. Results for air-fuel ratiosof 30 and 46 are obtained and analyzed.

Fig. 3 shows different boundaries of the domainunder consideration. The boundary conditions used atabove boundaries are given in Table 1.

0° 100

Fig, 2 - Radial and ring V-gutter

outlet

main inletbottom face

Fig. 3 - Boundaries of the domain

34 INDIAN J. ENG. MATER. SCI., FEBRUARY 2004

The main input conditions are: Air-fuelratio (inprimary zone) (30 and 46); Sauter mean diameter(18.5 u); Fuel injection velocity (30 m/s); Tempera-ture of fuel (300 K).

Results and DiscussionFigs 4 and 5 show the velocity contours at 0° plane

for air-fuel ratio of 30 and 46 respectively from inletto' exit. A low velocity region can be observed.behindthe top gutter while the velocity is seen increasingnear the top region where the bypass air entersthrough screech rings. The dark blue (or black ifprinted in black and white) regions indicate the effectof the wall in reducing the velocity while maximumvelocity can be seen to be at the exit of the nozzle.Velocity is reduced from 250 m/s to 170 m/s in the

4.64 ••• 024.45 ••• 024.27 ••• 024p8&+023.00 •••0£3.71 •••0£3.53 •••.0£3.34 •••.0£3.15 •••.0£2.97 e+0£2.78&+0£2.60 •••.0£2.41&+0£2.23 •••.0£2.04&+0£1.86&+0£1.67&0-0£1.48&+0£1.31••• 0£1.11••• 0£9.28 •••017.42 ••• 015.51•••013.71 ••• 011.86&+010.00 ••• 00

Fig. 4 - Contours of velocity magnitude for air-fuel ratio of 30

<1;~2

1.~1

.oo~o

Fig. 5 - Contours of velocity magnitude for air-fuel ratio of 46

diffuser. The bypass along the nozzle liner can be ob-served from the figure. Velocity is seen to be in-creasing as the flow approaches the nozzle. This isdue to the expansion of gases resulting from hightemperature.

The flow field velocity vectors behind the top gut-fer are shown in Fig. 6. The top portion of the guttershows the formation of eddy with clockwise rotationdue to the separation of flow at the top end of thegutter resulting in a sharp change in direction of flow.In the rest of the gutter the flow seems to have moreor less anticlockwise motion. However, the flow phe-nomenon is quite complex in the middle region wherethe mixing of the flow from top and bottom portion ofthe gutter occurs. The complexity of the flow phe-nomenon in and around the gutter can be attributed tothe volume expansion due to heat addition and theproduction of species due to combustion.

Fig. 7 shows the anchoring of flame at the gutter.High temperature regions are found in the top and

.,.-3:."2..02

$.4;_

SJli...0.2

iii 2,tr2~

2.1'0A+02

1.1~

1.'1~

•.1....01

4.39..01'r~'"

,,08 •.. 01

Fig. 6 - Re-circulation zone behind top gutter

bottom radall the grflame. Initgutter canattributed'fuel strearfrom the fiair much tconducivestarts corrbustion m:ter where(2246K).

Fig. 8 ~with fueldence timtop and 1which getthe gap be

The reahead of

c ooe-co

Fig. 8

Fig. 7 - Total temperature contours at gutter cross section

20Se~03

1,759"03

1 6Oe~03

i.ose-oa

Fig. 9

j

an be ob-a be in-., This is'om high

: top gut-he gutter: rotationld of theof flow.

ive morelow phe-m whereortion ofow phe-buted toand the

e gutter.top and

, ,~~~-,..,.~~ ----...>.--......,. -.-..-....--..-....•--..-......•.•.

..•.....•-...•"

.tion

UNAUNE & GANESAN: ANALYSIS OF REACTING FLOW IN AN AERO-ENGINE AFTERBURNER

bottom radial gutters along with the ring gutter. Thusall the gutters are instrumental in stabilizing thef1ame. Initiation of combustion just upstream of thegutter can be observed from the figure. This can beattributed to the fact that some amount of fuel in thefuel streams during their travel towards the gutterfrom the fuel manifolds evaporates and mixes with theair much before passing through the gutter. Since, theconducive temperature field is available the mixturestarts combusting. However, completeness in com-bustion may be noticed in the wake region of the gut-ter where the temperature reaches maximum value(2246K).

Fig. 8 shows three-dimensional view of the gutterwith fuel particle streams colored by particle resi-dence times. The ring gutter and some portions of thetop and bottom gutters intercept the fuel streams,which get diverted while the remaining pass throughthegap between the gutters.

The residence time of the streams is maximumahead of the gutter in the RC zone thus resulting in

41 ••..1).3

97.03

~

,,.....3 0l'I+0)

~t'-()!

22011-03

'l6t-03

132 •...03

".•..I(x.I.CW

""·00

Fig. 8 - Gutter with particle streams colored by particleresidence time

218e"03

lO3e·03

18ge*03

\ 75t"03

, eoe~03

1 468"OJ

, 32e~Ol

I 11e·Q3

1 03or03

888&"02

14SE1"02Y

601s·02\.x-Z

" 596·01

Fig. 9 - Total temperature contours (gutter back surface)

anchoring of flame. In Fig. 9 it can be seen that thestreams striking the gutter result in cooling of guttersurface due to latent heat of vaporization of fuel,while the hot spots on the gutter back surface are theresult of initiation of combustion of the mixture up-stream of the gutter, which is shown by red spots.Thus, it may be seen that the gutter is successfullyblocking the high velocity stream of air-fuel mixtureto stabilize the flame.

Figs 10 and 11 show the static temperature con-tours for air-fuel ratio of 30 and 46 from entry to exitof the afterburner.

It is observed that flame anchors well at the gutterand the high temperature region is continuous till theexit. However, the liners are seen to be at much lowertemperatures, due to the cooling effect produced bythe bypass air. Some hot spots are observed on thefuel manifolds. But the maximum temperature iswithin the limit. On the top and bottom gutters, some

2.28 ••• 03

2.10&>03

1.02&>03

1.73&>03

1.55 ••• 03

1.37 •••.03

1.19 ••• 03

1.01 •••.03 ••8.23&+02

6.41&+02

4.m •••Q2

Fig. 10 - Contours of static temperature for air-fuel ratio of 30

22 •• "OS

•••••••10 ••.•••

11:10<0$

I-

1.3?..tO!

1.1 •• ..03

1 ......,.

~021...0>

••••m

•••••••

Fig. 11 - Contours of static temperature for air-fuel ratio of 46

35

36 INDIAN J. ENG. MATER. SCI., FEBRUARY 2004

2.19•••01

1.97•••01

1.78•••01

1.540-01

1.35•••01

1.12•.•01

9.02 •. 02

L$.87•.. 02

4.75•.. 02

2.S. •.. 02

4.tIl •.. OS

Fig. 12 - Contours of CO2 mass fraction

1.17•.. 01

1.88•.• 01

1.49•••01

1.31••..01

1.12••• 01

9.S40-02

7.41•.• 02

5.80 ••..02 LS.1'SI •. 02

1.87•.• 02

0.00..00

Fig. 13 - Contours of O2 mass fraction

regions of maximum temperature are observed. Thatmay be due to flame stabilization. As compared to theradial V-gutters, ring gutters seem to experience lowertemperature. Due to the cooling air, liner temperatureis also within limit. Some cooling air is bypassed tonozzle liner. The average temperature at nozzle outletis 1500 K, while the maximum temperature of nozzleliner is 1000K, which shows the effect of bypass air.Fig. 12 shows the contours of CO2 mass fraction atthe central plane whereas Fig. 13 shows the contoursof O2 mass fraction for an air-fuel ratio of 30.

Contours of static temperature and CO2 mass frac-tion are nearly complementary. This is becausemaximum temperature is the indication of completecombustion causing higher CO2 production. So CO2

mass fraction will be maximum where temperature ismaximum. Regions of maximum temperature and

CO2 mass fraction have minimum O2 mass fraction.This is due to utilization of O2 for combustion. Thus,analyzing species fraction, one can obtain the state ofcombustion. From the total pressure contours (notgiven in this paper) it is seen that increasing air-fuelratio decreased total pressure losses.

ConclusionsDuring the present investigations, CO2 and O2

fraction is found to be as expected at outlet. This indi-cates that combustion is complete.

Temperature of liners has been found to be withinpermissible limit. Also, temperature of nozzle liner isfound to be less than average temperature at exit. Thisreveals that bypass air is cooling the nozzle liner ef-fectively. Maximum temperature at the exit is lessthan the average temperature variation for the two air-fuel ratios studied.

Air-fuel ratio of 13 gives better afterburner per-formance as the pressure loss is within the designlimit. Since the system is quite complex, experimentalvalidation was not possible. However, to gain moreconfidence validation must be attempted.

References1 Ravichandran M & Ganesan V, IE (I) J-MC, 77 (1996) 67-

75.2 Frolov S M, basevich V Ya & Belyaev A A, Modeling of

bluff-body stabilized combustion with detailed chemistry(N.N.Semenov Institute of Chemical Physics, Moscow),1999.

3 Bakrozis A G, Papailiou D D & Koutmos P, CombustFlame, 119 (1999) 291-306.

4 Suresh D, Modeling of isothermal and reacting flows in gasturbine afterburners, M.S. Thesis, I.C. Engines Laboratory,I.I.T. Madras, 2001.

5 FUJENT 5, User's Guide, Vols 2& 3 (Fluent Incorporated,India), 1998.

6 Henderson R E & Blazowski W S, Turbo propulsion com-bustion technology: Comprehensive study of aircraft gasturbine engines, Part II, Chapter 20, AFAPL-TR-78-52, AirForce Aero Propulsion Laboratory, Wright-Patterson AirForce Base, Ohio, 1978.

7 Cohen H, Rogers G F C & Saravanamuttoo H I H, Gas tur-bine theory (Longman Group Ltd, USA), 1998.

8 Versteeg H K & Malalasekera W, An introduction to com-putational fluid dynamics: The finite volume method(Longman Group Ltd, UK), 1995.

9 Shaw C T, Using computational fluid dynamics, (PrenticeHall, USA), 1992.

10 Lefebvre A H, Gas turbine combustion (Hemisphere Pub-lishing Corporation, New York), 1993.

11 James W Useller, Effect of combustor length on afterburnercombustion, (Lewis Research Center, NASA), 1959.

12 Macquisten M A & Dowling A P, Combust Flame, (1993)253-264.

U

13 Baxter M114 (1992)

14 Kumar R]15 Ganesan V16 Zhang X ~

262.17 Chuang S

(1990) 30:

raction.!. Thus,state ofrs (notair-fuel

UNAUNE & GANESAN: ANALYSIS OF REACTING FLOW IN AN AERO-ENGINE AFfERBURNER 37

13 Baxter M R & Lefebvre A H, J Eng Gas Turbine Power,114 (1992), 776-781.

14 Kumar R K, Combustion Sci Technol, 21 (1980) 199-203.15 Ganesan V, Indian J Technol, 18 (1980),447-450.16 Zhang X & Chiu H, Int J Turbo Jet Engines, 4 (1987) 251-

262.17 Chuang S H & Jiang J S, Int J Numer Methods Fluids, 11

(1990) 303-316.

18 Ravichandran M, Experimental and theoretical investiga-tions of afterburner, Ph.D. Thesis, IC Engines Laboratory,Indian Institute of Technology Madras, Chennai, India,1993.

19 Lee D & Lin J S, Numer Heat Trans], 20 (1991) Part A, 65-79.

20 Takashi Tamaru, JSME Int J, 39 (1996) Series B, No.1.

md O2

is indi-

Combust

withinliner isit. Thismer ef-is lesswo air-

er per-design

imentaln more

996) 67-

deling ofihemistryl1oscow),

ws in gasboratory,

rporated,

'ion com-craft gas8-52, Airrson Air

Gas tur-

I to com-method

(Prentice

iere Pub-

terburner)

e, (1993)