Research Article Numerical Analysis of Turbulent ...downloads.hindawi.com/journals/jc/2016/2572035.pdf · Research Article Numerical Analysis of Turbulent Combustion in ... is model

Research ArticleNumerical Analysis of Turbulent Combustion ina Model Swirl Gas Turbine Combustor

Ali Cemal Benim1 Sohail Iqbal1 Franz Joos2 and Alexander Wiedermann3

1Center of Flow Simulation (CFS) Department of Mechanical and Process EngineeringDusseldorf University of Applied Sciences Josef-Gockeln-Str 9 40474 Dusseldorf Germany2Laboratory of Turbomachinery Helmut Schmidt University Holstenhofweg 85 22008 Hamburg Germany3Engineering Gas Turbines MAN Diesel and Turbo SE Steinbrinkstr 1 46145 Oberhausen Germany

Correspondence should be addressed to Ali Cemal Benim alicemalbenimhs-duesseldorfde

Received 30 November 2015 Revised 28 July 2016 Accepted 17 August 2016

Academic Editor Hong G Im

Copyright copy 2016 Ali Cemal Benim et al This is an open access article distributed under the Creative Commons AttributionLicense which permits unrestricted use distribution and reproduction in any medium provided the original work is properlycited

Turbulent reacting flows in a generic swirl gas turbine combustor are investigated numerically Turbulence is modelled by aURANSformulation in combination with the SST turbulence model as the basic modelling approach For comparison URANS is appliedalso in combination with the RSM turbulence model to one of the investigated cases For this case LES is also used for turbulencemodelling For modelling turbulence-chemistry interaction a laminar flamelet model is used which is based on the mixturefraction and the reaction progress variable This model is implemented in the open source CFD code OpenFOAM which hasbeen used as the basis for the present investigation For validation purposes predictions are compared with the measurements fora natural gas flame with external flue gas recirculation A good agreement with the experimental data is observed Subsequentlythe numerical study is extended to syngas for comparing its combustion behavior with that of natural gas Here the analysis iscarried out for cases without external flue gas recirculation The computational model is observed to provide a fair prediction ofthe experimental data and predict the increased flashback propensity of syngas

1 Introduction

Modern gas turbines are to provide high efficiency reliabilityand stability while meeting strict low emission requirementswith emerging additional requirements such as fuel flex-ibility In that respect the combustor is obviously a corecomponent and a detailed understanding of the complexflow heat and mass transfer processes in the flame is ofgreat importance Experimental investigation of gas turbinecombustion is difficult and can provide only limited infor-mation due to practical limitations Numerical simulationscan provide detailed insight and reduce the number of costlyexperiments Nevertheless the highly complex processesin the combustor are difficult to model and the simula-tions are afflicted with inaccuracies Thus development ofmathematical and numerical models for GTC and theirexperimental validation have been a continuous endeavorto which the present work is aimed to provide a contribu-tion

Fureby [1] applied an EDC-type combustion model incombination with LES to analyze GTC Lorstad et al [2]analyzed the reacting flow in the Siemens SGT-800 burnerexperimentally and numerically applying RANS and LESapproaches along with an EDC-type combustion modela focus of the work being on the effect of burner fueldistribution on flame dynamics A recent study on URANSand LES modelling of GTC for a Siemens scaled combustorwas presented byGoldin et al [3] using the flamelet generatedmanifold model as combustion model in combination withLES and turbulent flame speed models in combination withURANS They found that LES predictions of the mean andRMS axial velocity mixture fraction and temperature fieldsdo not show an improvement over the RANS Again recentlyALSTOMrsquos reheat combustor was successfully analyzed byKulkarni et al [4] applying a novel combustion modelbased on a composite reaction progress variable along witha tabulated chemistry approach and the stochastic fieldsturbulence-chemistry interaction model

Hindawi Publishing CorporationJournal of CombustionVolume 2016 Article ID 2572035 12 pageshttpdxdoiorg10115520162572035

2 Journal of Combustion

In the simulation of GTC one of the main challengesis turbulence modelling This is caused by the highly non-isotropic turbulence structure which is created by the highswirl levels applied to induce a flame-stabilizing vortexbreakdown In the previous work of present authors [5ndash7] it was shown that a three-dimensional and unsteadyformulation that can resolve the unsteadiness of coherentstructures is necessary for achieving sufficient accuracy insuch flows (as also confirmed by the recent work of otherauthors [1ndash4]) This requirement is fulfilled (at differentlevels) by URANS and LES approaches In the present appli-cation a URANS formulation is used as the basic modellingapproach in combination with a two-equation turbulencemodel In modelling turbulent swirling flows the RSM hasgenerally been found to be more accurate compared to two-equation turbulence models (while convergence difficultiesmay impede its application in some cases) We have foundhowever that this superiority of the RSM is given basically forthe RANS that is steady-state formulations In an unsteadyformulation such as URANS we quite often observed thattwo-equation models perform similarly well In this casethe use of a two-equation turbulence model is reasonablesince the additional computational overhead of the RSM (5additional transport equations and less superior convergencebehavior) does not seem to be justified Still for comparisonURANS is applied also in combination with the RSM to oneof the investigated cases For this case LES is also used forturbulence modelling

Modelling of the turbulence-chemistry interaction isthe further main challenge of course A method which isfound to be adequate in modelling turbulence-chemistryinteraction is the EDC [8] as successfully applied to GTCby several authors [1] A drawback of EDC is however thatan individual transport equation needs to be solved for eachspecies which increases the computational demand propor-tionally with the considered number of species For meetingthe current demands of combustion technology reactionmechanismswith always increasing level of sophistication arerequired that incorporate a rather large number of speciesIn combination with computationally demanding turbulencemodelling approaches such as URANS and LES whichare necessary for achieving sufficient accuracy as discussedabove the computational costs become extremely high espe-cially for real applications in the industrial development envi-ronment On the other hand the laminar flamelet method(LFM) [9] provides a very efficient way of consideringdetailed reaction kinetics in turbulent combustion where acompletely detailed reactionmechanism can be incorporatedvia a few variables that describe flamelet characteristicsAlthough the validity of the LFM for GTC was questionedin the past based on Damkohler number arguments it wasshown later on [10 11] that purely dimensional argumentsneglecting the transient effects may bemisleading and typicalGTC operation conditions are rather in the validity range ofthe LFMThus the LFM is used as the turbulent combustionmodel in the present study It is argued that a mixturefraction-reaction progress variable characterization of theflamelets is convenient for the present partially premixedflame as will be described in more detail below

Oxidizer

Oxidizer

Symmetry

Evaluationline

axis

30

x

Figure 1 Combustor geometry

The computational model is applied to predict turbulentcombustion in a model gas turbine combustor firing naturalgas (NG) and syngas (SG) First a validation study is per-formed for the NG flame with flue gas recirculation (FGR)Numerical results are compared with the experimental datawhere a reasonably fair agreement is observed Then thevalidatedmodel is applied to predict the combustion behaviorof SG in comparison to NG without FGR

The motivation of the present work has been the assess-ment of the performance of the advocated laminar flameletmethod in predicting gas turbine swirl flames consideringthe realistic fuel injection configuration of an industrial swirlburner Moreover the assessment of the predictive capabilityof the approach for the CO and NO emissions under flue gasrecirculation and using syngas as fuel (instead of natural gas)has been a further motivationThe coherent consideration ofthe all abovementioned aspects on the same rig and withinthe same modelling framework is a novelty of the presentinvestigation

2 The Test Case

Experiments were performed at the Turbomachinery Labo-ratory of Helmut Schmidt University Hamburg The CADdrawing of the atmospheric model combustor which isequipped by a single swirl burner with 12 channels isillustrated in Figure 1 The oxidizer stream passes through ahood (not displayed in the figure) and enters radially into thechannels of the radial swirler

Fuel is injected into the cross-flowing air by 12 injectionholes each located at the wall of each swirler channel Afterpassing through the swirler channels each of which havingan inclination of 45∘ the fuel-air mixture enters the centralconverging-diverging burner nozzle and subsequently intothe octagonal main combustion chamber A convergingexhaust gas nozzle is attached to the combustor exit to avoidreverse flow Based on the local unburnt bulk axial velocityand the diameter at the exit of the converging-divergingburner nozzle the Reynolds number turns out to be about34000 Assuming a perfect guidance by the swirler channelsand conservation of angular momentum within the burnerand defining the swirl number as the ratio of the maximumswirl velocity to the bulk axial velocity a swirl number ofabout 09 can be estimated at the burner nozzle outlet A

Journal of Combustion 3

Table 1 Composition (vol ) of fuel stream

CH4

C3

H8

CO2

CO H2

O2

N2

NG 925 52 13 mdash mdash mdash 1SG 10 22 4 22 40 mdash 2

Table 2 Composition (vol ) of oxidizer stream

O2

H2

O CO2

Ar N2

FGR0 206 1 mdash 09 775FGR20 179 23 22 09 767

detailed description of the setup and measurement methodsare provided in [12]

The combustor was designed for premixedpartially pre-mixed operation In the experiments [12] different fuelcompositions were considered The compositions of the NGand SG that are considered in the present work are presentedin Table 1

The measurements [12] were performed for investigatingthe effect of external flue gas recirculation (FGR) Thusdifferent oxidizer compositions were investigated [12] Theoxidizer compositions that are considered in the presentstudy are shown in Table 2 (FGR0 corresponding to ldquozerordquoflue gas ratio ie to pure air)

In the present analysis totally three cases are analyzedNG with FGR (NG-FG20) NG without FGR (NG-FGR0)and SG without FGR (SG-FGR0) For the NG-FG20 casethe mixture composition was adjusted to have an adiabaticflame temperature of about 1525K (corresponding to anequivalence ratio of about 05 fuel and oxidizer mass flowrates being 00009217 kgs and 003575 kgs and combustorinlet temperature 573K) For SG-FGR0 case an operationpoint was chosen corresponding to a slightly lower adiabatictemperature of 1450K

3 Modelling

31 Grid Boundary Conditions and Outline of the Mathemat-icalModel Ablock structuredmesh consisting of 12millioncells is used A detailed view of the surfacemesh is illustratedin Figure 2 At the fuel and oxidizer inlets constant profilesare prescribed for velocities temperature and mass fractionsthat result from the global combustor data At the outlet zero-gradient boundary conditions are applied for convective-diffusively transported variables along with a constant staticpressure The no-slip walls are assumed to be adiabatic Nearsolid walls the turbulencemodelling is augmented by the useof standard logarithmic wall-functions The near-wall gridresolutionwas such that the119910+ values [13] were not exceeding120 and had an average value of approx 50 in the burnerAlthough the applied near-wall resolution is not very fine(to resolve the wall boundary layers) we assume that thiswould not lead to a serious deterioration of the predictivecapability since the governing processes of the problem suchas jet mixing or vortex breakdown are of rather free-shearlayer type and thus not much affected by the wall boundarylayers

Figure 2 The mesh (detail view of surface mesh)

For the computational investigation the finite volumemethod based open source CFD code OpenFOAM [13] isused where a pressure-correction scheme (PIMPLE whichis a combination of PISO and SIMPLE schemes) was appliedfor treating the velocity-pressure coupling As turbulencemodel the Shear Stress Transport (SST) model [13] is appliedwithin a URANS context as the main approach For scalarsa gradient-diffusion approximation is used along with theassumption of constant turbulent Schmidt numbers of value07 For comparison the Reynolds Stress Model (RSM) [14] isalso used in comparison with URANS for one of the casesAdditionally the Large Eddy Simulation (LES) [15] approachis also employed as turbulence modelling strategy for thiscase using the Smagorinsky-Lilly subgrid-scale model [15]

A second-order upwind scheme was applied to discretizethe convective terms in the transport equations for all thevariables A first-order Euler scheme was used for time step-ping since stability problems were quite often encounteredwith a second-order time discretization The time step sizeis chosen in such a way that cell Courant numbers do notexceed unity Starting from an initial field the numericalsimulations were performed for a time period which is longenough to allow the development of a quasi-periodic flowfield that is nomore dependent on the initial conditions Afterthis state the time-averaging of the results was started whichwas continued until the time-averaged fields did not show anysubstantial change in time

Along with the threemomentum equations the pressure-correction equation and two equations for the turbulencemodel three additional differential transport equations (fourequations if NO is included) are solved for combustionmodelling which are discussed in the following section

32The CombustionModel The turbulence-chemistry inter-actions necessitate the use of a combustion model if flowturbulence is not directly simulated but modelled In thepresent work the laminar flamelet method (LFM) is usedAccording to the usual assumptions of the LFM [9] for asteady one-dimensional adiabatic laminar diffusion flameall thermochemical dependent variables (120593) can be expressedas unique functions of the mixture fraction 119885 and thestoichiometric value of the scalar dissipation rate 120594st as

120593 = 120593 (119885 120594st) (1)


The mixture fraction 119885 represents the mass fraction of fuelstream locally in the unburnt mixture For 120594st an assumedfunctional dependence on 119885 (eg as the one suggested byPeters [9]) is quite commonly used Such functional relation-ships (see (1)) are obtained by performing 1D laminar flamecalculations once for all and made available in tabulatedform (flamelet libraries) for the subsequent CFD analysisNote that the laminar flame extinguishes for a sufficientlylarge value of the scalar dissipation rate This is the so-called ldquoquenching limitrdquo For a quenched laminar flamerelationships of type (1) describe the nonburning state Inthe turbulent flow the average values of the thermochemicaldependent variables such as the species mass fractions andthe static mixture temperature can be obtained by the follow-ing expression utilizing presumed PDFs (the expression iswritten for a Favre-averaged quantity but holds analogouslyfor a Reynolds-averaged quantity)

= int

infin

0

int

1

0

120593 (119885 120594st) 119875 (119885) 119875 (120594st) 119889119885119889120594st (2)

In the above expression a statistical independence of 119885

and 120594st is assumed along with the assumption of a single-delta PDF for 120594st The presumed PDF for 119885 is typicallycontrolled by its twomoments that is the average value andthe variance

119885101584010158402 The modelled time-averaged differential

transport equations of these variables [16] are

120597120588

120597119905+120597 (120588119895)

120597119909119895

minus120597

120597119909119895

[(120583

120590+120583t120590t)

120597

120597119909119895

] = 0 (3)

120597120588119885101584010158402

120597119905+

120597 (120588119895

119885101584010158402

)

120597119909119895

minus120597

120597119909119895

[(120583

120590+120583t120590t)120597119885101584010158402

120597119909119895

]

= 2120583t120590t

120597

120597119909119895

120597

120597119909119895

minus 119888984858120598

119896

119885101584010158402

(4)

where 119888 is a model constant (119888 = 20 is used) The ldquosteadyrdquoLFM outlined above is known to perform successfully for alarge class of nonpremixed flames However it is known [9]that it cannot accurately describe phenomena such as localextinction reignition and flame lift-offThus it is principallynot adequate for premixedpartially premixed flames like thepresent one

Therefore in the present study a flamelet model based onthe mixture fraction and the reaction progress variable (119862) isadopted which is more suitable for premixedpartially pre-mixed flames The model was originally proposed by Pierceand Moin [17] within an LES framework who suggesteda parameterization of the flamelets based on the reactionprogress variable instead of the scalar dissipation rate Thecomplete locus of solutions of the flamelet equations resultsin a so called S-shaped curve an example of which is shownin Figure 3 for the NG-FGR20 caseThe upper branch repre-sents the stable burning flamelets till the turning point whichcorresponds to the quenching limit After the quenching limitthe curve continues to decrease scalar dissipation rate anddescribes the unstable flamelets whereas the lower branch

0

500

1000

1500

2000

2500

Max

imum

flam

elet

tem

pera

ture

(K)

01 1 10 100001Stoichiometric scalar dissipation rate (1s)

Figure 3 S-shaped curve for the NG-FGR20 flame

corresponds to nonburning flamelets Pierce and Moin [17]parameterized the flamelets based on the so-called reactionprogress variable by projecting the flame states horizontallyalong the S-shaped curve (Figure 3)

The reaction progress variable can be defined in differentways In the present work a temperature-based definition ispreferred (the alternative species-based definitions are ratherdifficult to handle in the present case with FGR due to theexistence of combustion products at the oxidizer inlet)

119862 =119879max120594 minus 119879u

119879b minus 119879u (5)

In (5) 119879max120594 is the maximum static temperature locallyprevailing within the stretched 1D laminar flame 119879u and119879b denote the unburnt and burnt equilibrium temperaturesof the mixture Please note that 119879max120594 also depends onthe scalar dissipation rate incorporating its effect indirectlyThus within this concept the laminar flamelet functionalrelationships are established as

120593 = 120593 (119885 119862) (6)

Assuming 119885 and 119862 are independent with the help ofpresumed PDFs the average values of the thermochemicalvariables are obtained in a similar fashion to (2) In thepresent work a single-delta PDF is assumed for 119862 (as it wascommonly assumed in the previous applications of themodelincluding the original work of Pierce and Moin [17]) For 119885a beta PDF (120573) is used Thus the average values of the staticmixture temperature and all species mass fractions (exceptthat of NO) are obtained from

= int

1

0

120593 (119885 ) 120573 (119885 119885101584010158402

)119889119885 (7)

What remains is the field information on the averagedprogress variable to close the system For this a differential


transport equation is solved which is derived following Brayrsquosapproximation [18] of the chemical source term reading as

120597120588

120597119905+120597 (120588119895)

120597119909119895

minus120597

120597119909119895

[(120583

120590+120583t120590t)

120597

120597119909119895

] = 120579119862 (8)

In (8) 120579119862is defined as

120579119862=

119876

119888119875(119879b minus 119879u)

(9)

where 119876 and 119888119875denote volumetric heat release rate and the

mean isobaric heat capacity respectivelyThe volumetric heatrelease rate119876 is obtained from the 1D laminar flame libraries

Since NO reactions are very slow compared to the maincombustion reactions the extraction of NO mass fractionsout of the flamelet data (see (6)) is inappropriate Thusfor NO an additional transport equation is solved Thisapproach does not create a serious conflict with the detailedcombustion mechanism underlying the LFM due to theextremely small NO mass fractions having negligibly smalleffects on the transport processes and heat release in theflame The modelled NO transport equation reads as

120597120588119884NO120597119905

+120597 (120588119895119884NO)

120597119909119895

minus120597

120597119909119895

[(120583

120590+120583t120590t)120597119884NO120597119909119895

]

= 119878NO

(10)

The critical issue here is the determination of the sourceterm (119878NO) As previously demonstrated [11] a quite effectivemeans for this purpose is the extraction of its instantaneousvalue out of the flamelet libraries like many other thermo-chemical variables (via (6)) and obtaining the time-averagedvalue (for closing (10)) by means of PDFs (see (7)) Since thesource term stems from a detailed reaction mechanism allNO formation paths are simultaneously considered whereasthe thermal NO formation is expected to be the dominatingone in the present application

The flamelet libraries (see (6)) are constructed by1D steady adiabatic laminar flame calculations using theFlameMaster code [19] before the field calculations of theturbulent reacting flow by means of CFD As the underlyingreactionmechanism theGRIMech 30 [20] is used assuminga Lewis number of unity for all speciesWithin the subsequentCFD calculations the local values of the Favre-averagedspeciesmass fractions and temperature are obtained from (7)which relies on the solution of three differential transportequations for (see (3)) 119885101584010158402 (see (4)) and (see (8)) IfNO prediction is required (10) is additionally solved for119884NOThe above-described combustion model is implemented inOpenFOAM [13]

33 On the Adequacy of the Grid Resolution A formalgrid independence study was not performed The grid isconstructed based on our previous experience on similarflames [6 7] The following analysis of the turbulent scales ofthe present results indicates that the applied grid resolution

is reasonably fine based on the suggestions of Celik et al [21]and our previous experience [6 7] Within the framework ofLES different measures were proposed for assessing the gridresolution such as the Grid Index (GI) defined as ratio of thelocal grid size (assumed to be given by the third root of thecell volume) to the Kolmogorov length scale

GI = Δ

119897119870

(11)

with

119897119870= (

]3

120576)

14

(12)

According to Celik et al [21] GI should be smaller than 25for achieving sufficient accuracy for LES In our previousLES modelling work [7] on swirling flows we found that theresults obtained for GI le 50 do not remarkably differ fromthose obtained for GI le 25 This led to the conclusion thatthe criterion GI le 50 was sufficient for good accuracy atleast for the present class of flow problems In the presentwork besides the URANS approach as the main turbulencemodelling strategy LES is also used We assume that theLES grid resolution criteria can be regarded to be useful forURANS too since URANS principally has less stringent gridresolution requirements compared to LES Figure 4 presentsthe distribution of GI (see (11)) in a plane through the swirler(at the channelmid-height) and in a plane through themiddleof the combustor at a time step (for the NG-FGR20 flame)

One can see that there are large regions fulfilling GI lt 25In the remaining regions the values remainmostly within theband 25 lt GI lt 50 Thus the calculated GI values (Figure 4)can be interpreted to indicate a sufficiently fine grid for thepresent purposes

4 Results and Discussion

The results will be presented in two parts in the first partthe results for the NG flame with FGR (NG-FGR20) willbe discussed In the second part the results for the NGand SG flames without FGR (NG-FGR0 SG-FGR0) will bepresented

41 NG Flame with FGR The present combustor is designedto operate in premixedpartially premixed mode [12] Thefuel jets mix into the oxidizer in the swirler channels andthen along the converging-diverging burner nozzle (Figure 1)Given the high swirl level of the flow the sudden areaexpansion at the exit of the burner nozzle induces a vortexbreakdown that is an inner recirculation zone which shallact as an aerodynamic flame holder at which the flamefront shall be anchored Depending on the degree of mixingachieved in the burner a premixed or partially premixedflame can result

Although it was obvious that the LFM based on themixture fraction and the scalar dissipation rate (see (1)ndash(4)) isnot convenient for the present case it was still applied to seeits performance in the specific application and to demonstrate


80 10020 60400GI(a)

20 40 60 80 1000GI

(b)

Figure 4 Distribution of GI in planes through (a) swirler and (b) combustor

210075

0

500

1250

1500

1750

2000

1000

T (K)

Figure 5 Predicted temperature field at a time step in a planethrough combustor calculated using mixture fraction-scalar dissi-pation rate based LFM (see (1)ndash(4)) for NG-FGR20 flame

the improvement of the method by the introduction of thereaction progress variable (see (3)ndash(9)) Figure 5 displaysthe predicted temperature field at a time step in a planethrough the combustor which was calculated using themixture fraction-scalar dissipation rate based LFM (see (1)ndash(4)) for the NG-FGR20 flame The predicted flame burnsin diffusion mode (Figure 5) and a flame lift-off could notbe predicted The rather low temperatures (lt1000K) in theswirler channels are caused by locally high strain rates andquenching effects in the close vicinity of fuel injection holesDownstream the swirler channels where the local strain rateis lower locally near-stoichiometric combustion leads to highflame temperatures about 2100K

In the present work the LFM based on the mixturefraction and the reaction progress variable (see (5)ndash(9)) isused as the combustion model In the following the resultsobtained by this model will be presented

Distributions of the axial velocity predicted by the mix-ture fraction-reaction progress variable LFM for the NG-FGR20 flame at two different time steps in a plane throughthe combustor are presented in Figure 6 using URANS-SST URANS-RSM and LES approaches The predictedinner recirculation zone due to vortex breakdown can beseen in Figure 6 which exhibits a quite unsteady andthree-dimensional structure As expected LES (Figure 6(c))resolves finer structures compared to URANS (Figures 6(a)and 6(b)) whereas URANS-RSM (Figure 6(b)) also seems tocapture finer structures than URANS-SST (Figure 6(a)) Itis interesting to note that the recirculation zone is attachedto the burner back plate In the present plots in a two-dimensional plane (Figure 6) the extension of the three-dimensional recirculation zone to the burner back plate

450minus10minus20 4010 20 30

Ux (ms)

(a)

45400 20minus20

Ux (ms)

(b)

45400 20minus20

Ux (ms)

(c)

Figure 6 Predicted fields of axial velocity at an arbitrary timestep in a plane through combustor by the mixture fraction-reactionprogress variable based LFM forNG-FGR20 flame (a) URANS-SST(b) URANS-RSM and (c) LES

cannot directly be seen but deduced for example fromFigure 6(a)

Time-averaged predictions of the axial velocity compo-nent and the velocity magnitude as predicted by URANS-SST URANS-RSM and LES for the NG-FGR20 flameare shown in a plane through the combustor in Figure 7One can see that the time-averaged axial velocity fieldexhibits a bubble-shaped vortex breakdown recirculationzone (Figures 7(a) 7(c) and 7(d)) The slender negativeaxial velocity regions that were precessing inside the burner(Figure 6(a)) disappear in the time averaging forURANS-SST


4520minus20 400

Ux (ms)

(a)

6040200

U (ms)

(b)

45200 40minus20

Ux (ms)

(c)

45200 40minus20

Ux (ms)

(d)

Figure 7 Predicted time-averaged velocity fields in a plane throughcombustor for NG-FGR20 flame (a) axial velocity URANS-SST (b)velocity magnitude URANS-SST (c) axial velocity URANS-RSMand (d) axial velocity LES

(Figure 7(a)) URANS-RSM predicts a quite intensive innerrecirculation zone with higher negative axial velocities anda deeper extension of the time-averaged recirculation zoneinto the burner practically up to the burner back plate (Fig-ure 7(c)) The velocity magnitude plot (Figure 7(b)) indicatesthe very strong vortex core in the burner which expandsdownstream and extends along the combustor length

As an indication of the flow turbulence the distributionof the representative RMS value of the velocity fluctuations(119880rms) normalized by a reference velocity (119880ref) is presentedin Figure 8 for a plane through the combustor The LESresults are used for this purpose The representative 119880rms isobtained from the calculated turbulence kinetic energy (119896)from 119880rms = radic(23)119896 Doing so the turbulence kineticenergy 119896 is calculated from 119896 = (11990610158402+V10158402+11990810158402) where 11990610158402 V10158402

001 07060402UrmsUref

Figure 8 Predicted 119880rms119880ref in a plane through the combustor(LES)

and 11990810158402 represent the RMS of the three velocity componentsresolved by LES As the reference velocity the bulk axialvelocity at the throat of the burner nozzle is used (119880ref =23ms) One can see that quite high turbulence intensitiesprevail especially in the burner nozzle and in its downstreamwhere the vortex breakdown occurs

The predicted time-averaged fields of temperature andCO mass fraction for the NG-FGR20 flame resulting fromURANS-SST URANS-RSM and LES calculations are dis-played in Figure 9 As it can be deduced from the distributionsgiven in Figure 9 the predictions indicate that the flame isnow anchored at the burner nozzle exit as confirmed bythe experiments and no high temperature zones exist nearburner walls (Figure 5) which is important for the integrityof the hardware One can also observe that URANS-SST(Figure 9(a)) and LES (Figure 9(b)) predict a flame frontwhich is practically positioned just downstream the burnernozzle whereas URANS-RSM indicates a more stronglyinclined flame front reaching deeper into the burner on itsaxis (Figure 9(c)) LES (Figure 9(d)) predicts a slightly thickerflame brush than the both URANS solutions (Figures 9(a)and 9(c))The combustion takes place in the premixedmodewhere a maximum temperature of about 1500K is achievedbehind the curved flame brush (Figures 9(a) 9(c) and 9(d))The COmass fraction quickly attains a local maximum in thereaction zone (Figure 9(b)) and gets depleted downstream

The predicted and measured profiles of time-averagedtemperature along the ldquocombustor axisrdquo and along the ldquoevalu-ation linerdquo (Figure 1) are compared in Figure 10 Note that thedisplayed domain in Figure 10 covers the combustor domaindownstream the burner nozzle exit (119909 gt 0 Figure 1)

Both experiments and predictions show (Figure 10) aquite rapid increase of the temperature across the relativelythin flame front as already indicated by the temperatureplot presented in Figure 9 The evaluation line touches theedge of the burner nozzle exit at 119909 = 0 (Figure 1) Sincethe flame front is curved and rather thick near the edge ofthe burner nozzle exit especially for URANS-SST and LES(Figure 9) the evaluation line (Figure 1) crosses through thewhole reaction zone On the combustor axis the displayedpart in Figure 10(a) (119909 ge 0) covers the rather thin reactionzone of URANS-SST and LES only partially (only the rearpart since the flame brush is located slightly within theburner nozzle) The reaction zone of URANS-RSM is notcovered at all on the combustor axis shown in Figure 10(a)since the flame is located at a substantially upstream positioncompared to 119909 = 0 Thus the temperature increase observed


573

160075

0

1250

1000

1500

T (K)

(a)

002

50

002

001

Mass fraction CO (mdash)

(b)

573

160075

0

1250

1000

1500

T (K)

(c)

573

160075

0

1250

1000

1500

T (K)

(d)

Figure 9 Predicted time-averaged fields of (a) temperatureURANS-SST (b) CO mass fraction URANS-SST (c) temperatureURANS-RSM and (d) temperature LES in a plane through com-bustor

along the evaluation line (Figure 10(b)) is larger and takesplace in a larger distance compared to the temperaturerise along the combustor axis (Figure 10(a)) This trend isqualitatively the same in all predictions as well as in theexperiments On the evaluation line which encompasses thewhole flame front the measured temperatures exhibit a localmaximum (Figure 10(b)) Calculations agree quite well withthe measurements in predicting this temperature peak in theflame zone (Figure 10(b)) This is predicted slightly better byLES compared to the other models URANS-SST predictionsare quite close to those of LES The temperature gradientpredicted by URANS-RSM occurs at a slightly upstreamposition compared to the experiments and other predictions

(Figure 10(b)) due to the difference in the predicted flamefront shape as discussed above Similar trends are generallyobserved for the combustor axis (Figure 10(a)) Due to thequite upstream position of the flame front predicted byURANS-RSM on the combustor axis as discussed above theURANS-RSM results do not exhibit a temperature gradientfor the profile shown in Figure 10(a) for the combustor axisDownstream the initial peak the measured temperaturesdecline whereas the predicted values do not decrease Theburnt gas temperature is overpredicted by about 100ndash150K(Figure 10) This may be due to the assumption of no heatloss to the environment (adiabatic combustor walls) in thepredictions

Thepredicted andmeasured profiles of time-averagedCOmass fraction along the combustor axis and along the ldquoevalua-tion linerdquo (Figure 1) are compared in Figure 11The evaluationline displayed in Figure 11(b) cuts through the whole flamefront (as discussed in the preceding paragraph) Thus peakvalues produced in the flame zone are well captured Onecan see in Figure 11(b) that quite large CO mass fractions arepredicted by all models at the beginning of the evaluationline which agree quite well with the measurements LESseems to predict even higher maximum values compared toURANS-SST and URANS-RSM (Figure 11(b)) Downstreamof this peak the measurements show a quite sharp decayThe calculations agree well with this gradient only for a shortdistance (for 119909 le approx 005m) Beyond this region themeasured values continue to decay sharply up to approx 119909 =

01m where the predictions exhibit a rather milder decaywhich may be due to an underprediction of the CO burn-out rate in this region (Figure 11(b)) This results in a localoverprediction of the CO mass fractions beyond the initialregion (119909 ge 005m) (Figure 11(b)) Compared to the othermodels the decay of the CO mass fractions is predicted atbest by URANS-RSM which however overpredicts the exitvalue (Figure 11(b)) URANS-SST and LES agree better withthe experiments at the exit (Figure 11(b)) For the combustoraxis (Figure 11(a)) the predicted peak values occur at amore upstream position compared to Figure 11(b) due tothe curved shape of the flame front Due to similar effectsan overprediction of the CO mass fractions for a regionalong the centreline is observed (Figure 11(a)) The predictedcombustor outlet values are quite close to the measurements(Figure 11(a)) For the evaluation line (Figure 11(a)) andespecially for the combustor axis (Figure 11(a)) the shapeof the experimental curve is predicted at best by URANS-RSM In comparison LES generally predicts a quite gradualdecay (instead of an initial sharp decay followed by a ratherflat curve as observed in experiments) whereas URANS-SST results may be seen to be qualitatively between those ofURANS-RSM and LES (Figure 11)

The predicted and measured profiles of time-averagedNO mass fraction along the combustor axis and along theldquoevaluation linerdquo (Figure 1) are compared in Figure 12 ForNOmass fraction the calculations overpredict the experimentsthroughout (Figure 12) However the discrepancy is not toolarge and this is at least partially due to the overpredictedtemperature field (Figure 10)The relative performance of thedifferent models correlates with the temperature predictions


300

600

900

1200

1500

1800Te

mpe

ratu

re (K

)

010 02 04 0603 05Axial position (m)

ExperimentURANS-SST

URANS-RSMLES

(a)

300

600

900

1200

1500

1800

Tem

pera

ture

(K)

010 02 04 060503Axial position (m)

ExperimentURANS-SST

URANS-RSMLES

(b)

Figure 10 Predicted and measured profiles of time-averaged temperature along (a) combustor axis and (b) evaluation line for NG-FGR20flame

CO p

pm

15

O

2

01

1

10

100

1000


ExperimentURANS-SST

URANS-RSMLES

(a)

CO p

pm

15

O

2

01

1

10

100

1000


ExperimentURANS-SST

URANS-RSMLES

(b)

Figure 11 Predicted andmeasured profiles of time-averagedCOmass fraction along (a) combustor axis and (b) evaluation line forNG-FGR20flame

LES seems to perform slightly better than the other modelsin the initial parts where the NO mass fractions sharplyincrease URANS-SST also shows a rather good agreementthere The URANS-RSM predictions show a slight upstreamshift compared to the other results since the flame frontwas predicted at a slightly upstream position in comparison(Figure 9) Overall a fair agreement between the predictionsand the measurements is observed (Figure 12)

42 NG and SG Flames without FGR All results presentedin this section are obtained by URANS-SST The predicted

time-averaged temperature fields for the NG-FGR0 and SG-FGR0 flames are displayed in Figure 13

As seen in Figure 13 theNG-FGR0 case exhibits a shorterless curved flame brush (Figure 13(a)) compared to NG-FGR20 (Figure 9(a)) In the central part of the burner theflame front shows a slightly higher penetration into theburner nozzle for NG-FGR0 (Figure 13(a)) compared to NG-FGR20 (Figure 9(a)) However there is no ldquoflashbackrdquo asthe burner walls are not exposed to high temperatures Onthe contrary in the SG-FGR0 flame although the adiabaticflame temperature is lower a flashback tendency is predicted


NO

xpp

m

15

O

2

0

4

8

12

16

20


ExperimentURANS-SST

URANS-RSMLES

(a)

NO

xpp

m

15

O

2

0

4

8

12

16

20


ExperimentURANS-SST

URANS-RSMLES

(b)

Figure 12 Predicted and measured profiles of time-averaged NO mass fraction along (a) combustor axis and (b) evaluation line for NG-FGR20 flame

16e+03

750

500

1250

1500

1000

T (K)

(a)

573

1000

1250

150075

0

1600

T (K)

(b)

Figure 13 Predicted time-averaged temperature fields in a planethrough combustor for (a) NG-FGR0 flame and (b) SG-FGR0 flame

(Figure 13(b)) as the flame propagates deep upstream intothe burner with an accompanying overheating of the burnerwalls around the edges of the nozzle exit There are severalmechanisms that can trigger a flashback [22] Flashback dueto combustion induced vortex breakdown is less likely to bethe cause since the expansion ratio which is observed tocorrelate to this phenomenon [22] is even smaller for theSG (where the flashback occurs) compared the NG flameThus it likely that the flashback is due to flame propagation in

Table 3 Combustor exit CO and NOmass fractions for NG-FGR0

Measured PredictedCO ppm 15 O

2

012 342NO ppm 15 O

2

302 795

the core or in the boundary layers due to increased laminarflame speed by the higher hydrogen content of SG The testrig was not designed to make detailed measurements in theburner Thus it is not clear how well the predicted flashbacktendency corresponds to the experiments for this operationpoint However although not quantified and documented agenerally strong flashback propensity of the used SG fuel wasobserved during the experimental investigation [12]

Figure 14 shows the contours of the time-averaged mix-ture fraction with respect to the flame front for both (NG-FGR0 and SG-FGR0) flamesThe flame front is characterizedby the = 05 isoline of the reaction progress variable whichis indicated by the thick red line in the figure Thinner blacklines indicate the isolines of the mixture fraction where thecorresponding values are typed nearbyThe fully premixedvalues of are 0025 and 0065 for the NG-FGR0 and SG-FGR0 flames respectively One can see that the NG-FGR0flameburns in fully premixed regime (similar toNG-FGR20)For the SG-FGR0 flame the mixture fraction varies withinthe range about 0035ndash0075 indicating a partially premixedcombustion (Figure 14)

For NG-FGR0 and SG-FGR0 cases measurements werenot available within the combustor but at the combustoroutlet Tables 3 and 4 compare the predicted and measuredCO and NO mass fractions at the combustor outlet for NG-FGR0 and SG-FGR0 flames respectively

For CO the prediction for the SG-FGR0 flame is quiteclose to the measurement (Table 3) For the NG-FGR0 flame


0025

00260030 C = 05

(a)

003 004 006

007

0065

005C = 05

(b)

Figure 14 Predicted time-averaged isolines of the mixture fraction with respect to the time-averaged flame front ( = 05) for (a) NG-FGR0flame and (b) SG-FGR0 flame (the premixed values are 0025 and 0065 for NG-FGR0 and SG-FGR0 flames resp)

Table 4 Combustor exit CO and NOmass fractions for SG-FGR0


2

104 174NO ppm 15 O

2

041 261

the deviation between the prediction and measurement islarger but still not too large (Table 4) For NO there is amoderate overprediction for both flames (Tables 3 and 4)This is affected by the overprediction of the flame temperaturedue to the assumption of adiabatic walls as discussed in theprevious section Overall a fair agreement of the predictionswith experiments can be observed (Tables 3 and 4)

5 Conclusions

Turbulent flames in a generic swirl gas turbine combustordesigned to operate in premixedpartially premixed modeare investigated numerically Turbulence is modelled by aURANS formulation using the SST turbulence model asthe basic modelling approach For comparison URANS isapplied also in combination with the RSM to one of theinvestigated cases For this case LES is also used for tur-bulence modelling For modelling the turbulence-chemistryinteraction a laminar flamelet model based on the mixturefraction and reaction progress variable is used coupledwith a presumed probability density function approachNatural gas and syngas flames with and without external fluegas recirculation are investigated Comparing the predictiveperformances of different turbulence models for one of thecases (NG-FGR20) it was observed that LES predicts aslightly thicker flame brush compared to URANS-SST whileURANS-RSM predicts a slightly sharper and more inclinedflame front compared to URANS-SST extending deeperinto the burner on the axis The numerical results obtainedby different turbulence models are observed to show acomparable overall performance and a fair overall agreementwith the experimental data The slight overprediction ofthe combustor exit temperature is assumed to be affectedby the assumption of adiabatic walls in the mathematicalmodel The model will be improved in the future to include

nonadiabatic effects which is also expected to lead to amore accurate prediction of NO emissions For syngas anincreased flashback propensity could be predicted whichqualitatively agrees with the experimental observations

Nomenclature

119862 Reaction progress variable (mdash)119888119901 Mean specific heat capacity at constant pressure(J kgminus1 Kminus1)

119896 Turbulence kinetic energy (m2 sminus2)119897119870 Kolmogorov length scale (m)

119901 Static pressure (Pa)119875 Favre or Reynolds probability density function (mdash)119876 Volumetric heat release rate (Jmminus3 sminus1)119878119895 Source term of transport equation for species 119895(kgmminus3 sminus1)

119905 Time (s)119879 Static temperature (K)119909119894 Space coordinates (m)

119906119895 Velocity vector (msminus1)

119880119909 Axial velocity (msminus1)

119880 Velocity magnitude (msminus1)119884119895 Mass fraction of species 119895 (mdash)

119910+ Dimensionless wall distance (mdash)

119885 Mixture fraction (mdash)

Greek Symbols120573 Beta probability density function (Favre or Reynolds)

(mdash)Δ Local finite volume cell size (m)120576 Dissipation rate of 119896 (mminus2sminus3)120579119862 Source term for progress variable equation (kgmminus3 sminus1)

120583 Viscosity (kgmminus1 sminus1)120588 Density (kgmminus3)120590 Schmidt number (mdash)120593 Thermochemical variable to be extracted from flamelet

libraries (mdash)120594 Scalar dissipation rate (sminus1)120596 Turbulence eddy frequency (sminus1)


Superscriptssim Favre-averaged value Reynolds-averaged value

10158401015840 Favre fluctuational value

Subscripts119887 Burntst Stoichiometric119905 Turbulent119906 Unburnt

Abbreviations

EDC Eddy dissipation conceptFGR Flue gas recirculationGI Grid IndexGT Gas turbineGTC Gas turbine combustioncombustorLES Large Eddy SimulationLFM Laminar flamelet methodNG Natural gasPDF Probability density functionRANS Reynolds Averaged Numerical SimulationRMS Root Mean SquareRSM Reynolds Stress ModelSG SyngasURANS Unsteady RANSSST Shear Stress Transport

Competing Interests

The authors declare that they have no competing interests

References

[1] C Fureby ldquoLarge eddy simulation modelling of combustion forpropulsion applicationsrdquo Philosophical Transactions of the RoyalSociety A Mathematical Physical and Engineering Sciences vol367 no 1899 pp 2957ndash2969 2009

[2] D Lorstad A Lindholm N Aklin et al ldquoExperimental andLES investigation of a SGT-800 burner in a combustion rigrdquo inProceedings of the ASME Turbo Expo 2010 Power for Land Seaand Air Paper no GT2010-22688 Glasgow UK June 2010

[3] G Goldin F Montana and S Patil ldquoA comparison of RANSand LES of an industrial lean premixed burnerrdquo in Proceedingsof the ASME Turbo Expo 2014 Turbine Technical Conferenceand Exposition Paper No GT2014-25352 Dusseldorf GermanyJune 2014

[4] R Kulkarni B Bunkute F Biagioli M Duesing andW PolifkeldquoLarge eddy simulation of ALSTOMs reheat combustor usingtabulated chemistry and stochastic-fields combustion modelrdquoin Proceedings of the Turbo Expo 2014 Turbine Technical Con-ference and Exposition Paper no GT2014- 26053 DusseldorfGermany June 2014

[5] A C Benim A Nahavandi and K J Syed ldquoURANS and LESanalysis of turbulent swirling flowsrdquo Progress in ComputationalFluid Dynamics vol 5 no 8 pp 444ndash454 2005

[6] A C Benim M P Escudier A Nahavandi A K Nickson K JSyed and F Joos ldquoExperimental and numerical investigation of

isothermal flow in an idealized swirl combustorrdquo InternationalJournal of Numerical Methods for Heat amp Fluid Flow vol 20 no3 pp 348ndash370 2010

[7] A C Benim S Iqbal A NahavandiWMeier AWiedermannand F Joos ldquoAnalysis of turbulent swirling flows in an isother-mal gas turbine combustor modelrdquo in Proceedings of the ASMETurbo Expo 2014 Turbine Technical Conference and ExpositionPaper No GT2014-25008 Dusseldorf Germany June 2014

[8] B F Magnussen ldquoOn the structure of turbulence and gen-eralized eddy dissipation concept for chemical reactions inturbulent flowrdquo in Proceedings of the 19th Aerospace SciencesMeeting Aerospace Sciences Meetings pp 1ndash6 St Louis MoUSA January 1981

[9] N Peters Turbulent Combustion Cambridge University PressCambridge UK 2002

[10] T Poinsot D Veynante and S Candel ldquoDiagrams of premixedturbulent combustion based on direct simulationrdquo Symposium(International) on Combustion vol 23 no 1 pp 613ndash619 1991

[11] A C Benim and K J Syed ldquoLaminar flamelet modelling of tur-bulent premixed combustionrdquoAppliedMathematicalModellingvol 22 no 1-2 pp 113ndash136 1998

[12] S Fischer D Kluss and F Joos ldquoExperimental investigation ofa fuel flexible generic gas turbine combustor with external fluegas recirculationrdquo in Proceedings of the ASME Turbo Expo 2014Turbine Technical Conference andExposition PaperNoGT2014-25388 Dusseldorf Germany June 2014

[13] httpwwwopenfoamcom[14] B E Launder G J Reece and W Rodi ldquoProgress in the

development of a Reynolds-stress turbulent closurerdquo Journal ofFluid Mechanics vol 68 no 3 pp 537ndash566 1975

[15] M Lesieur O Metais and P Comte Large-Eddy Simulations ofTurbulence Cambridge University Press New York NY USA2005

[16] T Poinsot and D Veynante Theoretical and Numerical Com-bustion R T Edwards Inc Philadelphia Pa USA 2nd edition2005

[17] C D Pierce and P Moin ldquoProgress-variable approach forlarge-eddy simulation of non-premixed turbulent combustionrdquoJournal of Fluid Mechanics vol 504 pp 73ndash97 2004

[18] K N C Bray ldquoThe interaction between turbulence and com-bustionrdquo in Proceedings of the 17th International Symposiumon Combustion pp 223ndash233 The Combustion Institute Pitts-burgh Pa USA 1978

[19] H Pitsch ldquoA C++ computer program for 0-D and 1-D laminarflamelet calculationsrdquo Tech Rep RWTH Aachen Germany1998

[20] httpcombustionberkeleyedugri-mech[21] I B Celik Z N Cehreli and I Yavuz ldquoIndex of resolution qual-

ity for large eddy simulationsrdquo Journal of Fluids Engineering vol127 no 5 pp 949ndash958 2005

[22] A C Benim and K J Syed Flashback Mechanisms in LeanPremixed Gas Turbine Combustion Academic Press New YorkNY USA 2015

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Active and Passive Electronic Components

Control Scienceand Engineering

Journal of



RotatingMachinery


Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics


Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of


Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation



DistributedSensor Networks



In the simulation of GTC one of the main challengesis turbulence modelling This is caused by the highly non-isotropic turbulence structure which is created by the highswirl levels applied to induce a flame-stabilizing vortexbreakdown In the previous work of present authors [5ndash7] it was shown that a three-dimensional and unsteadyformulation that can resolve the unsteadiness of coherentstructures is necessary for achieving sufficient accuracy insuch flows (as also confirmed by the recent work of otherauthors [1ndash4]) This requirement is fulfilled (at differentlevels) by URANS and LES approaches In the present appli-cation a URANS formulation is used as the basic modellingapproach in combination with a two-equation turbulencemodel In modelling turbulent swirling flows the RSM hasgenerally been found to be more accurate compared to two-equation turbulence models (while convergence difficultiesmay impede its application in some cases) We have foundhowever that this superiority of the RSM is given basically forthe RANS that is steady-state formulations In an unsteadyformulation such as URANS we quite often observed thattwo-equation models perform similarly well In this casethe use of a two-equation turbulence model is reasonablesince the additional computational overhead of the RSM (5additional transport equations and less superior convergencebehavior) does not seem to be justified Still for comparisonURANS is applied also in combination with the RSM to oneof the investigated cases For this case LES is also used forturbulence modelling

Modelling of the turbulence-chemistry interaction isthe further main challenge of course A method which isfound to be adequate in modelling turbulence-chemistryinteraction is the EDC [8] as successfully applied to GTCby several authors [1] A drawback of EDC is however thatan individual transport equation needs to be solved for eachspecies which increases the computational demand propor-tionally with the considered number of species For meetingthe current demands of combustion technology reactionmechanismswith always increasing level of sophistication arerequired that incorporate a rather large number of speciesIn combination with computationally demanding turbulencemodelling approaches such as URANS and LES whichare necessary for achieving sufficient accuracy as discussedabove the computational costs become extremely high espe-cially for real applications in the industrial development envi-ronment On the other hand the laminar flamelet method(LFM) [9] provides a very efficient way of consideringdetailed reaction kinetics in turbulent combustion where acompletely detailed reactionmechanism can be incorporatedvia a few variables that describe flamelet characteristicsAlthough the validity of the LFM for GTC was questionedin the past based on Damkohler number arguments it wasshown later on [10 11] that purely dimensional argumentsneglecting the transient effects may bemisleading and typicalGTC operation conditions are rather in the validity range ofthe LFMThus the LFM is used as the turbulent combustionmodel in the present study It is argued that a mixturefraction-reaction progress variable characterization of theflamelets is convenient for the present partially premixedflame as will be described in more detail below

Oxidizer

Oxidizer

Symmetry

Evaluationline

axis

30

x

Figure 1 Combustor geometry

The computational model is applied to predict turbulentcombustion in a model gas turbine combustor firing naturalgas (NG) and syngas (SG) First a validation study is per-formed for the NG flame with flue gas recirculation (FGR)Numerical results are compared with the experimental datawhere a reasonably fair agreement is observed Then thevalidatedmodel is applied to predict the combustion behaviorof SG in comparison to NG without FGR

The motivation of the present work has been the assess-ment of the performance of the advocated laminar flameletmethod in predicting gas turbine swirl flames consideringthe realistic fuel injection configuration of an industrial swirlburner Moreover the assessment of the predictive capabilityof the approach for the CO and NO emissions under flue gasrecirculation and using syngas as fuel (instead of natural gas)has been a further motivationThe coherent consideration ofthe all abovementioned aspects on the same rig and withinthe same modelling framework is a novelty of the presentinvestigation

2 The Test Case

Experiments were performed at the Turbomachinery Labo-ratory of Helmut Schmidt University Hamburg The CADdrawing of the atmospheric model combustor which isequipped by a single swirl burner with 12 channels isillustrated in Figure 1 The oxidizer stream passes through ahood (not displayed in the figure) and enters radially into thechannels of the radial swirler

Fuel is injected into the cross-flowing air by 12 injectionholes each located at the wall of each swirler channel Afterpassing through the swirler channels each of which havingan inclination of 45∘ the fuel-air mixture enters the centralconverging-diverging burner nozzle and subsequently intothe octagonal main combustion chamber A convergingexhaust gas nozzle is attached to the combustor exit to avoidreverse flow Based on the local unburnt bulk axial velocityand the diameter at the exit of the converging-divergingburner nozzle the Reynolds number turns out to be about34000 Assuming a perfect guidance by the swirler channelsand conservation of angular momentum within the burnerand defining the swirl number as the ratio of the maximumswirl velocity to the bulk axial velocity a swirl number ofabout 09 can be estimated at the burner nozzle outlet A



CH4

C3

H8

CO2

CO H2

O2

N2



O2

H2

O CO2

Ar N2

FGR0 206 1 mdash 09 775FGR20 179 23 22 09 767





3 Modelling







120593 = 120593 (119885 120594st) (1)



= int

infin

0

int

1

0

120593 (119885 120594st) 119875 (119885) 119875 (120594st) 119889119885119889120594st (2)





120597120588

120597119905+120597 (120588119895)

120597119909119895

minus120597

120597119909119895

[(120583

120590+120583t120590t)

120597

120597119909119895

] = 0 (3)

120597120588119885101584010158402

120597119905+

120597 (120588119895

119885101584010158402

)

120597119909119895

minus120597

120597119909119895

[(120583

120590+120583t120590t)120597119885101584010158402

120597119909119895

]

= 2120583t120590t

120597

120597119909119895

120597

120597119909119895

minus 119888984858120598

119896

119885101584010158402

(4)



0

500

1000

1500

2000

2500

Max

imum

flam

elet

tem

pera

ture

(K)





119862 =119879max120594 minus 119879u

119879b minus 119879u (5)


120593 = 120593 (119885 119862) (6)


= int

1

0

120593 (119885 ) 120573 (119885 119885101584010158402

)119889119885 (7)




120597120588

120597119905+120597 (120588119895)

120597119909119895

minus120597

120597119909119895

[(120583

120590+120583t120590t)

120597

120597119909119895

] = 120579119862 (8)


120579119862=

119876

119888119875(119879b minus 119879u)

(9)




120597120588119884NO120597119905

+120597 (120588119895119884NO)

120597119909119895

minus120597

120597119909119895

[(120583

120590+120583t120590t)120597119884NO120597119909119895

]

= 119878NO

(10)





GI = Δ

119897119870

(11)

with

119897119870= (

]3

120576)

14

(12)








80 10020 60400GI(a)

20 40 60 80 1000GI

(b)


210075

0

500

1250

1500

1750

2000

1000

T (K)





450minus10minus20 4010 20 30

Ux (ms)

(a)

45400 20minus20

Ux (ms)

(b)

45400 20minus20

Ux (ms)

(c)





4520minus20 400

Ux (ms)

(a)

6040200

U (ms)

(b)

45200 40minus20

Ux (ms)

(c)

45200 40minus20

Ux (ms)

(d)











573

160075

0

1250

1000

1500

T (K)

(a)

002

50

002

001


(b)

573

160075

0

1250

1000

1500

T (K)

(c)

573

160075

0

1250

1000

1500

T (K)

(d)








300

600

900

1200

1500

1800Te

mpe

ratu

re (K

)


ExperimentURANS-SST

URANS-RSMLES

(a)

300

600

900

1200

1500

1800

Tem

pera

ture

(K)


ExperimentURANS-SST

URANS-RSMLES

(b)


CO p

pm

15

O

2

01

1

10

100

1000


ExperimentURANS-SST

URANS-RSMLES

(a)

CO p

pm

15

O

2

01

1

10

100

1000


ExperimentURANS-SST

URANS-RSMLES

(b)







NO

xpp

m

15

O

2

0

4

8

12

16

20


ExperimentURANS-SST

URANS-RSMLES

(a)

NO

xpp

m

15

O

2

0

4

8

12

16

20


ExperimentURANS-SST

URANS-RSMLES

(b)


16e+03

750

500

1250

1500

1000

T (K)

(a)

573

1000

1250

150075

0

1600

T (K)

(b)





2

012 342NO ppm 15 O

2

302 795






0025

00260030 C = 05

(a)

003 004 006

007

0065

005C = 05

(b)




2

104 174NO ppm 15 O

2

041 261


5 Conclusions



Nomenclature


















Abbreviations


Competing Interests


References


























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











CH4

C3

H8

CO2

CO H2

O2

N2



O2

H2

O CO2

Ar N2

FGR0 206 1 mdash 09 775FGR20 179 23 22 09 767





3 Modelling







120593 = 120593 (119885 120594st) (1)



= int

infin

0

int

1

0

120593 (119885 120594st) 119875 (119885) 119875 (120594st) 119889119885119889120594st (2)





120597120588

120597119905+120597 (120588119895)

120597119909119895

minus120597

120597119909119895

[(120583

120590+120583t120590t)

120597

120597119909119895

] = 0 (3)

120597120588119885101584010158402

120597119905+

120597 (120588119895

119885101584010158402

)

120597119909119895

minus120597

120597119909119895

[(120583

120590+120583t120590t)120597119885101584010158402

120597119909119895

]

= 2120583t120590t

120597

120597119909119895

120597

120597119909119895

minus 119888984858120598

119896

119885101584010158402

(4)



0

500

1000

1500

2000

2500

Max

imum

flam

elet

tem

pera

ture

(K)





119862 =119879max120594 minus 119879u

119879b minus 119879u (5)


120593 = 120593 (119885 119862) (6)


= int

1

0

120593 (119885 ) 120573 (119885 119885101584010158402

)119889119885 (7)




120597120588

120597119905+120597 (120588119895)

120597119909119895

minus120597

120597119909119895

[(120583

120590+120583t120590t)

120597

120597119909119895

] = 120579119862 (8)


120579119862=

119876

119888119875(119879b minus 119879u)

(9)




120597120588119884NO120597119905

+120597 (120588119895119884NO)

120597119909119895

minus120597

120597119909119895

[(120583

120590+120583t120590t)120597119884NO120597119909119895

]

= 119878NO

(10)





GI = Δ

119897119870

(11)

with

119897119870= (

]3

120576)

14

(12)








80 10020 60400GI(a)

20 40 60 80 1000GI

(b)


210075

0

500

1250

1500

1750

2000

1000

T (K)





450minus10minus20 4010 20 30

Ux (ms)

(a)

45400 20minus20

Ux (ms)

(b)

45400 20minus20

Ux (ms)

(c)





4520minus20 400

Ux (ms)

(a)

6040200

U (ms)

(b)

45200 40minus20

Ux (ms)

(c)

45200 40minus20

Ux (ms)

(d)











573

160075

0

1250

1000

1500

T (K)

(a)

002

50

002

001


(b)

573

160075

0

1250

1000

1500

T (K)

(c)

573

160075

0

1250

1000

1500

T (K)

(d)








300

600

900

1200

1500

1800Te

mpe

ratu

re (K

)


ExperimentURANS-SST

URANS-RSMLES

(a)

300

600

900

1200

1500

1800

Tem

pera

ture

(K)


ExperimentURANS-SST

URANS-RSMLES

(b)


CO p

pm

15

O

2

01

1

10

100

1000


ExperimentURANS-SST

URANS-RSMLES

(a)

CO p

pm

15

O

2

01

1

10

100

1000


ExperimentURANS-SST

URANS-RSMLES

(b)







NO

xpp

m

15

O

2

0

4

8

12

16

20


ExperimentURANS-SST

URANS-RSMLES

(a)

NO

xpp

m

15

O

2

0

4

8

12

16

20


ExperimentURANS-SST

URANS-RSMLES

(b)


16e+03

750

500

1250

1500

1000

T (K)

(a)

573

1000

1250

150075

0

1600

T (K)

(b)





2

012 342NO ppm 15 O

2

302 795






0025

00260030 C = 05

(a)

003 004 006

007

0065

005C = 05

(b)




2

104 174NO ppm 15 O

2

041 261


5 Conclusions



Nomenclature


















Abbreviations


Competing Interests


References


























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











= int

infin

0

int

1

0

120593 (119885 120594st) 119875 (119885) 119875 (120594st) 119889119885119889120594st (2)





120597120588

120597119905+120597 (120588119895)

120597119909119895

minus120597

120597119909119895

[(120583

120590+120583t120590t)

120597

120597119909119895

] = 0 (3)

120597120588119885101584010158402

120597119905+

120597 (120588119895

119885101584010158402

)

120597119909119895

minus120597

120597119909119895

[(120583

120590+120583t120590t)120597119885101584010158402

120597119909119895

]

= 2120583t120590t

120597

120597119909119895

120597

120597119909119895

minus 119888984858120598

119896

119885101584010158402

(4)



0

500

1000

1500

2000

2500

Max

imum

flam

elet

tem

pera

ture

(K)





119862 =119879max120594 minus 119879u

119879b minus 119879u (5)


120593 = 120593 (119885 119862) (6)


= int

1

0

120593 (119885 ) 120573 (119885 119885101584010158402

)119889119885 (7)




120597120588

120597119905+120597 (120588119895)

120597119909119895

minus120597

120597119909119895

[(120583

120590+120583t120590t)

120597

120597119909119895

] = 120579119862 (8)


120579119862=

119876

119888119875(119879b minus 119879u)

(9)




120597120588119884NO120597119905

+120597 (120588119895119884NO)

120597119909119895

minus120597

120597119909119895

[(120583

120590+120583t120590t)120597119884NO120597119909119895

]

= 119878NO

(10)





GI = Δ

119897119870

(11)

with

119897119870= (

]3

120576)

14

(12)








80 10020 60400GI(a)

20 40 60 80 1000GI

(b)


210075

0

500

1250

1500

1750

2000

1000

T (K)





450minus10minus20 4010 20 30

Ux (ms)

(a)

45400 20minus20

Ux (ms)

(b)

45400 20minus20

Ux (ms)

(c)





4520minus20 400

Ux (ms)

(a)

6040200

U (ms)

(b)

45200 40minus20

Ux (ms)

(c)

45200 40minus20

Ux (ms)

(d)











573

160075

0

1250

1000

1500

T (K)

(a)

002

50

002

001


(b)

573

160075

0

1250

1000

1500

T (K)

(c)

573

160075

0

1250

1000

1500

T (K)

(d)








300

600

900

1200

1500

1800Te

mpe

ratu

re (K

)


ExperimentURANS-SST

URANS-RSMLES

(a)

300

600

900

1200

1500

1800

Tem

pera

ture

(K)


ExperimentURANS-SST

URANS-RSMLES

(b)


CO p

pm

15

O

2

01

1

10

100

1000


ExperimentURANS-SST

URANS-RSMLES

(a)

CO p

pm

15

O

2

01

1

10

100

1000


ExperimentURANS-SST

URANS-RSMLES

(b)







NO

xpp

m

15

O

2

0

4

8

12

16

20


ExperimentURANS-SST

URANS-RSMLES

(a)

NO

xpp

m

15

O

2

0

4

8

12

16

20


ExperimentURANS-SST

URANS-RSMLES

(b)


16e+03

750

500

1250

1500

1000

T (K)

(a)

573

1000

1250

150075

0

1600

T (K)

(b)





2

012 342NO ppm 15 O

2

302 795






0025

00260030 C = 05

(a)

003 004 006

007

0065

005C = 05

(b)




2

104 174NO ppm 15 O

2

041 261


5 Conclusions



Nomenclature


















Abbreviations


Competing Interests


References


























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











120597120588

120597119905+120597 (120588119895)

120597119909119895

minus120597

120597119909119895

[(120583

120590+120583t120590t)

120597

120597119909119895

] = 120579119862 (8)


120579119862=

119876

119888119875(119879b minus 119879u)

(9)




120597120588119884NO120597119905

+120597 (120588119895119884NO)

120597119909119895

minus120597

120597119909119895

[(120583

120590+120583t120590t)120597119884NO120597119909119895

]

= 119878NO

(10)





GI = Δ

119897119870

(11)

with

119897119870= (

]3

120576)

14

(12)








80 10020 60400GI(a)

20 40 60 80 1000GI

(b)


210075

0

500

1250

1500

1750

2000

1000

T (K)





450minus10minus20 4010 20 30

Ux (ms)

(a)

45400 20minus20

Ux (ms)

(b)

45400 20minus20

Ux (ms)

(c)





4520minus20 400

Ux (ms)

(a)

6040200

U (ms)

(b)

45200 40minus20

Ux (ms)

(c)

45200 40minus20

Ux (ms)

(d)











573

160075

0

1250

1000

1500

T (K)

(a)

002

50

002

001


(b)

573

160075

0

1250

1000

1500

T (K)

(c)

573

160075

0

1250

1000

1500

T (K)

(d)








300

600

900

1200

1500

1800Te

mpe

ratu

re (K

)


ExperimentURANS-SST

URANS-RSMLES

(a)

300

600

900

1200

1500

1800

Tem

pera

ture

(K)


ExperimentURANS-SST

URANS-RSMLES

(b)


CO p

pm

15

O

2

01

1

10

100

1000


ExperimentURANS-SST

URANS-RSMLES

(a)

CO p

pm

15

O

2

01

1

10

100

1000


ExperimentURANS-SST

URANS-RSMLES

(b)







NO

xpp

m

15

O

2

0

4

8

12

16

20


ExperimentURANS-SST

URANS-RSMLES

(a)

NO

xpp

m

15

O

2

0

4

8

12

16

20


ExperimentURANS-SST

URANS-RSMLES

(b)


16e+03

750

500

1250

1500

1000

T (K)

(a)

573

1000

1250

150075

0

1600

T (K)

(b)





2

012 342NO ppm 15 O

2

302 795






0025

00260030 C = 05

(a)

003 004 006

007

0065

005C = 05

(b)




2

104 174NO ppm 15 O

2

041 261


5 Conclusions



Nomenclature


















Abbreviations


Competing Interests


References


























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










80 10020 60400GI(a)

20 40 60 80 1000GI

(b)


210075

0

500

1250

1500

1750

2000

1000

T (K)





450minus10minus20 4010 20 30

Ux (ms)

(a)

45400 20minus20

Ux (ms)

(b)

45400 20minus20

Ux (ms)

(c)





4520minus20 400

Ux (ms)

(a)

6040200

U (ms)

(b)

45200 40minus20

Ux (ms)

(c)

45200 40minus20

Ux (ms)

(d)











573

160075

0

1250

1000

1500

T (K)

(a)

002

50

002

001


(b)

573

160075

0

1250

1000

1500

T (K)

(c)

573

160075

0

1250

1000

1500

T (K)

(d)








300

600

900

1200

1500

1800Te

mpe

ratu

re (K

)


ExperimentURANS-SST

URANS-RSMLES

(a)

300

600

900

1200

1500

1800

Tem

pera

ture

(K)


ExperimentURANS-SST

URANS-RSMLES

(b)


CO p

pm

15

O

2

01

1

10

100

1000


ExperimentURANS-SST

URANS-RSMLES

(a)

CO p

pm

15

O

2

01

1

10

100

1000


ExperimentURANS-SST

URANS-RSMLES

(b)







NO

xpp

m

15

O

2

0

4

8

12

16

20


ExperimentURANS-SST

URANS-RSMLES

(a)

NO

xpp

m

15

O

2

0

4

8

12

16

20


ExperimentURANS-SST

URANS-RSMLES

(b)


16e+03

750

500

1250

1500

1000

T (K)

(a)

573

1000

1250

150075

0

1600

T (K)

(b)





2

012 342NO ppm 15 O

2

302 795






0025

00260030 C = 05

(a)

003 004 006

007

0065

005C = 05

(b)




2

104 174NO ppm 15 O

2

041 261


5 Conclusions



Nomenclature


















Abbreviations


Competing Interests


References


























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










4520minus20 400

Ux (ms)

(a)

6040200

U (ms)

(b)

45200 40minus20

Ux (ms)

(c)

45200 40minus20

Ux (ms)

(d)











573

160075

0

1250

1000

1500

T (K)

(a)

002

50

002

001


(b)

573

160075

0

1250

1000

1500

T (K)

(c)

573

160075

0

1250

1000

1500

T (K)

(d)








300

600

900

1200

1500

1800Te

mpe

ratu

re (K

)


ExperimentURANS-SST

URANS-RSMLES

(a)

300

600

900

1200

1500

1800

Tem

pera

ture

(K)


ExperimentURANS-SST

URANS-RSMLES

(b)


CO p

pm

15

O

2

01

1

10

100

1000


ExperimentURANS-SST

URANS-RSMLES

(a)

CO p

pm

15

O

2

01

1

10

100

1000


ExperimentURANS-SST

URANS-RSMLES

(b)







NO

xpp

m

15

O

2

0

4

8

12

16

20


ExperimentURANS-SST

URANS-RSMLES

(a)

NO

xpp

m

15

O

2

0

4

8

12

16

20


ExperimentURANS-SST

URANS-RSMLES

(b)


16e+03

750

500

1250

1500

1000

T (K)

(a)

573

1000

1250

150075

0

1600

T (K)

(b)





2

012 342NO ppm 15 O

2

302 795






0025

00260030 C = 05

(a)

003 004 006

007

0065

005C = 05

(b)




2

104 174NO ppm 15 O

2

041 261


5 Conclusions



Nomenclature


















Abbreviations


Competing Interests


References


























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










573

160075

0

1250

1000

1500

T (K)

(a)

002

50

002

001


(b)

573

160075

0

1250

1000

1500

T (K)

(c)

573

160075

0

1250

1000

1500

T (K)

(d)








300

600

900

1200

1500

1800Te

mpe

ratu

re (K

)


ExperimentURANS-SST

URANS-RSMLES

(a)

300

600

900

1200

1500

1800

Tem

pera

ture

(K)


ExperimentURANS-SST

URANS-RSMLES

(b)


CO p

pm

15

O

2

01

1

10

100

1000


ExperimentURANS-SST

URANS-RSMLES

(a)

CO p

pm

15

O

2

01

1

10

100

1000


ExperimentURANS-SST

URANS-RSMLES

(b)







NO

xpp

m

15

O

2

0

4

8

12

16

20


ExperimentURANS-SST

URANS-RSMLES

(a)

NO

xpp

m

15

O

2

0

4

8

12

16

20


ExperimentURANS-SST

URANS-RSMLES

(b)


16e+03

750

500

1250

1500

1000

T (K)

(a)

573

1000

1250

150075

0

1600

T (K)

(b)





2

012 342NO ppm 15 O

2

302 795






0025

00260030 C = 05

(a)

003 004 006

007

0065

005C = 05

(b)




2

104 174NO ppm 15 O

2

041 261


5 Conclusions



Nomenclature


















Abbreviations


Competing Interests


References


























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










300

600

900

1200

1500

1800Te

mpe

ratu

re (K

)


ExperimentURANS-SST

URANS-RSMLES

(a)

300

600

900

1200

1500

1800

Tem

pera

ture

(K)


ExperimentURANS-SST

URANS-RSMLES

(b)


CO p

pm

15

O

2

01

1

10

100

1000


ExperimentURANS-SST

URANS-RSMLES

(a)

CO p

pm

15

O

2

01

1

10

100

1000


ExperimentURANS-SST

URANS-RSMLES

(b)







NO

xpp

m

15

O

2

0

4

8

12

16

20


ExperimentURANS-SST

URANS-RSMLES

(a)

NO

xpp

m

15

O

2

0

4

8

12

16

20


ExperimentURANS-SST

URANS-RSMLES

(b)


16e+03

750

500

1250

1500

1000

T (K)

(a)

573

1000

1250

150075

0

1600

T (K)

(b)





2

012 342NO ppm 15 O

2

302 795






0025

00260030 C = 05

(a)

003 004 006

007

0065

005C = 05

(b)




2

104 174NO ppm 15 O

2

041 261


5 Conclusions



Nomenclature


















Abbreviations


Competing Interests


References


























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










NO

xpp

m

15

O

2

0

4

8

12

16

20


ExperimentURANS-SST

URANS-RSMLES

(a)

NO

xpp

m

15

O

2

0

4

8

12

16

20


ExperimentURANS-SST

URANS-RSMLES

(b)


16e+03

750

500

1250

1500

1000

T (K)

(a)

573

1000

1250

150075

0

1600

T (K)

(b)





2

012 342NO ppm 15 O

2

302 795






0025

00260030 C = 05

(a)

003 004 006

007

0065

005C = 05

(b)




2

104 174NO ppm 15 O

2

041 261


5 Conclusions



Nomenclature


















Abbreviations


Competing Interests


References


























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










0025

00260030 C = 05

(a)

003 004 006

007

0065

005C = 05

(b)




2

104 174NO ppm 15 O

2

041 261


5 Conclusions



Nomenclature


















Abbreviations


Competing Interests


References


























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation













Abbreviations


Competing Interests


References


























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation









Documents

Research Article Numerical Analysis of Turbulent ...downloads.hindawi.com/journals/jc/2016/2572035.pdf · Research Article Numerical Analysis of Turbulent Combustion in ... is model