7/23/2019 Application of Statistical Narrowband Model to Three-Dimensional Absorbing-Emitting-Scattered Media
1/7
JOURNAL OFTHERMOPHYSICS ANDHEATTRANSFERVol. 13 , No. 3, JulySeptember 1999
Application of Statistical Narrowband Model toThree-Dimensional Absorbing-Emitting-Scattering Media
Fengshan Liu, Gregory J. Smallwood, andOmer L. Gulder
National Research Council, Ottawa, Ontario K1A 0R6, Canada
A novel method for implementing the statistical narrowband model into the radiative transfer equation was
devised to perform nongreyradiative transfer calculationsin three-dimensionalabsorbing,emitting, and scattering
media. In this new method the radiation intensity is split into two parts: the nonscattered part and the scattered
part. The nonscattered part is solved accuratelyusing the statisticalnarrowband modeland a ray-tracingtechnique
and does not require iteration. The scattered part is solved with approximation by using the gray-band model to
estimate the band correlation between the scattered intensity and the gas absorption coefcient. The accuracy
of this method for narrowband radiation intensity prediction was evaluated in a homogeneous H 2O-N2-Al2 O3mixture under both isothermal and nonisothermal conditions by comparing its results with those of the statistical
narrowband correlated-K method. This new implementation method alleviates to some extent the difculty of the
band model-scattering incompatibility and offers good results provided the particle loading is not too high.
Nomenclature
f = species molar fractionInm = nonscattered part of spectral radiation intensity,
W m2 cm sr1
Ism = scattered part of spectral radiation intensity,
W m2 cm sr1
Im = spectral radiation intensity, W m2 cm sr1
NIm = mean radiation intensity over a narrowband,W m2 cm sr1
kj m = absorption coefcient associated with thejth
quadrature point,m1
Nkm = mean line intensity to spacing ratio, cm1
atm1
L = path length,mlm = local mean path length, mn = particle number density, m3
n = unit normal vector of a wall surface pointing toward
the gas sidep = pressure, atmS = surface area of a computationalgrid, m2
s, s 0 = position variables along a line of sight, mT = temperature, KV = volume of a computational grid, m3
wj m = weight parameter associated with the j th quadrature
pointx , y ,z = Cartesian coordinates,m
Nbm = mean line width to spacing ratioNcm = mean half width of an absorption line, cm
1
D m = wave-number interval of a narrowband,cm1
Ndm = equivalent line spacing, cm1
w
= wall surface emissivityjgm = gas spectral absorption coefcient, m
1
Njgm = narrowbandaveraged gas absorption coefcient, m1
Received 13 July 1998; revision received 28 December 1998; acceptedfor publication 29 March 1999. Copyright c 1999 by the Government ofCanada. Published by the American Institute of Aeronautics and Astronau-
tics, Inc., with permission. Associate Research Ofcer, Combustion Research Group, Institute for
Chemical Process and Environmental Technology, Building M-9, Montreal
Road; [email protected]. Associate Research Ofcer, Combustion Research Group, Institute for
Chemical Process and Environmental Technology, Building M-9, MontrealRoad; [email protected].
Group Leader, Combustion Research Group, Institute for ChemicalProcess and Environmental Technology, Building M-9, Montreal Road;[email protected].
jpm = particulatespectral absorption coefcient, m1
jsm = particulatespectral scattering coefcient, m1
m = wave number, cm1
n,g,l = direction cosinestgm = gas spectral transmissivitytn m = medium spectral transmissivity associated withIn mU = scatteringphase function\ = solid angle, srX = direction of radiation propagation
Subscripts
b = blackbodyg = gasi = spatial discretization (along a line of sight)indexn = angulard iscretization indexp = particulatew = wallm = spectral
I. Introduction
E XTENSIVE research attention has been paid to develop accu-rateyet computationallyefcient nongreygas radiationmodelsinrecentyears becauseof theimportantrole ofgas radiationinmanycombustion-related sciences and technologies. These models in-
clude the spectral line based weighted-sum-of-grey-gases (WSGG)
model developed by Denison and Webb,13 the exponential wide-
band model,4,5 the statisticalnarrowband (SNB) model,57 and var-ious K-distribution based methods.812 As far as total(spectrally
integrated)quantities are concerned,the spectral line based WSGGmodel of Denison and Webb,the exponentialwide-bandmodel, andcorrelated-K (CK)based wide-band models9,10 offer both accuracyand computational efciency. However, these models do not pro-
vide adequate spectral information for certain applications wherelow-resolution(at a resolution of 525 cm1 bandwidth)spectral
intensities are required instead of total quantities. One example ofsuchapplicationsis the predictionofinfraredsignaturefroma rocket
plume,which is a nonisothermalmixtureof radiating gases (mainlyCO2 and H2O)and particles(soot and Al2O3 ). Although the line-
by-line(LBL)method provides the exact results, its application to
three-dimensional scattering problems is computationally infeasi-ble at the present time. Therefore, approximate methods have tobe developed for these applications. The SNB model is naturally
a good candidate because it provides the required spectral resolu-tion. However, the SNB model suffers from two major drawbacks.
First, the straightforwardimplementationof the SNB model intothe
285
7/23/2019 Application of Statistical Narrowband Model to Three-Dimensional Absorbing-Emitting-Scattered Media
2/7
286 LIU, SMALLWOOD, AND GULDER
radiative transfer equation(RTE)leads to complicated correlationterms,5 and the solution of the resultant equation requires a ray-tracing technique in multidimensions.7 These two factors dramati-cally increase the computingeffortsof the SNB model calculations.
Secondly, it encountersthe so-called band model-scatteringincom-patibilitywhen applied to scatteringmedia.13,14 The incompatibility
constitutes the major obstacle to the application of the SNB modelto multidimensionalsc attering problems.
To overcome the difculties of the SNB model, Kim and Song15applied the WSGG concept to a narrowband (WNB) to maintain
the compatibility of the WSGG model with the RTE in the stan-dard form as well as the spectral resolution of the SNB model.
More recently, Yang a nd Song16 presented an improved version ofthe WNB model for CO2 using the scaling approximation of the
CK method.17 The CK-based WNB(CKWNB)model of Yang andSong16 is conceptually very similar to the statistical narrowband
CK(SNBCK)method described by Lacis and Oinas12 except for
how the gas absorption coefcients are obtained. These two meth-ods represent an alternative approach for low-resolution spectral
intensity prediction. The SNBCK method extracts gas absorptioncoefcients from band-averaged gas transmissivity using inverse
Laplace transformation12 and is also a narrowband WSGG method.
The advantageof theSNBCK method is thatthe efcientand populardiscrete-ordinatesmethod(DOM)can be used to solve the transferequation. The seven-point GaussLabatto quadratureis commonlyused to obtain band-averaged quantities,8,16 which is equivalent tousing seven gray gases at each narrowband; therefore,the RTE has
to be solvedseventimes in each nonoverlappingnarrowband.Hencethe applicationof these models could also be quite time-consuming
for scattering problems where solutions have to be obtained itera-tively for each gray gas. At an overlapping band, which arises for
radiativetransfer in a mixture containingtwo radiatinggases such asCO2 and H2O, the radiative transfer has to be representedby 7 7(49)gray gases,12 assuming each gas componentis represented byseven gray gases, because the product of two Malkmus bands is no
longer a Malkmus band. This quadratic increase in the number ofgray gases at overlappingbands constitutesthe disadvantageof the
SNBCK method for overlapping band calculations.The mostsimple implementationmethodof the SNB modelis the
gray narrowbandmodel of Kim et al.5 and Liu et al.18 in which onlyone gray gas is usedin eachnarrowband.Althoughthe graynarrow-
band model of Liu et al.18 in generalo ffers much better results thanthe version of Kim et al.,5 it p redicts unreliable narrowband radia-
tion intensitiesin some spectral regions in nonisothermalCO2-H2Ogas mixtures.19 Extension of the gray-band model to include radia-
tion scattering is straightforward and has been done by Liu et al.20
Through an evaluation study they found that the inclusion of scat-teringto the gray-bandgas radiationmodeldoes notintroduceextraerrors.20 This is indeed expected because the radiative propertiesof
particulate are weakly dependenton wave number in a narrowbandof 25 cm1 width and can be treated as constants.
The present study made an attempt to alleviate the band model-scatteringincompatibility by devising a variable splittingtechnique
to implement the SNB model into the RTE in three-dimensionalscattering media. Because benchmark narrowband results in three-
dimensionalscattering media are typically lacking in the literature,the accuracy of this method was evaluatedby comparing its results
with those obtained using the SNBCK method in a homogeneousH2 O-N2 -Al2O3 mixture under both isothermal and nonisothermal
conditions.This mixture contains only a single radiating gas and ischosento avoidthe difcultyof the SNBCK methodfor calculations
at overlapping bands.
II. Formulation
The spectral radiative transfer equation in absorbing, emitting,and scattering media can be written as
Im
sD (jgm C jpm C jsm)Im C (jgm C jpm)Ibm C
jsm
4p
Z4p
ImUm d\
(1)
The correspondingboundary condition at a diffuse wall is given as
Im,w D m,wIbm,w C1 m,w
p
ZnX< 0
Im,wjnj d\, for n X > 0
(2)
The formal solution of Eq.(1)is21
Im
(s) D Im
(sw
)tm
(sw
! s) C Z ssw
Sm
(s 0)t(s 0 ! s) ds 0 (3)
wheretm is the spectral medium transmissivity given as
tm(s1 ! s2 ) D exp
Z s2
s1
(jgmC jpmC js m) ds (4)
and Smis the source function dened as
Sm(s) D (jgm C jpm)IbmCjs m
4p
Z4p
ImUmd\ (5)
The direct implementation of the statistical narrowband modelrequires Eq.(1)to be averaged over a narrowband. This procedure
resultsin severalcorrelationterms. The one that deserves specialat-tention is the correlationbetween the radiation intensityand the gas
absorption coefcientjgmIm. For radiative transfer in nonscattering
media, a closed form ofjgmImis obtained by using Eq.(3) (Ref. 5).However, such a closed expression for jgmImno longer exists be-
cause of the dependence ofImin Eq.(3)on the entire radiation eldthrough the in-scattering term [the integral term on the right hand
side of Eq.(5)]. This difculty of applying a narrowband model toscattering problems was termedband model-scattering incompati-
bilityby Reed et al.14 The incompatibility is inherent for scatteringproblems when using the band model methodology and cannot beeliminated.
To alleviate the incompatibility of using a narrowband model to
scattering problems, an alternative method to implement the SNB
model into Eq.(1) is suggestedin the present study. This alternativemethod makes use of the linear property of Eq.(1)by splitting theradiation intensity into two parts: Inm and Ism, representing respec-
tively nonscatteredradiation and scattered radiation,
Im D In m C Ism (6)
The nonscattered part Inm is set to satisfy the following transfer
equation:
In m
sD (jgmC jpm)InmC (jg m C jpm)Ibm (7)
and its boundary condition is
In m,w D m,wIbm,w (8)
The scattered partIsmhas then to satisfy
Ism
sD (jgmC jpmC jsm)Ism js mInm
Cjsm
4p
Z4p
InmUmd\Cjsm
4p
Z4p
IsmUmd\ (9)
and its boundary condition is
Ism,w D 1 m,w
p
ZnX< 0
(Inm,w C Ism,w)jnj d\, for n X> 0
(10)
The presenceof the nonscatteredpart of the intensityInmin Eqs.(9)and (10) represents the coupling between the scattered intensity
and the nonscattered intensity. The scattered part Ism is in generalnegative as a result of using this splitting scheme.
Based on the facts that the blackbody function and the radiativeproperties of particulate are weakly dependent on wave number in
7/23/2019 Application of Statistical Narrowband Model to Three-Dimensional Absorbing-Emitting-Scattered Media
3/7
LIU, SMALLWOOD, AND GULDER 287
a narrowband, the average of Eqs. (7)an d(9)over a narrowbandleads to
NInm
sD (jgmC jpm)InmC ( Njg m C Njpm) NIbm (11)
NIsm
sD ( NjpmC Njsm) NIsm jgmIsm Njsm NInm
CNjsm
4p
Z4p
NInmNUmd\CNjsm
4p
Z4p
NIs m NUmd\ (12)
Equation(11)can be solved rigorously without approximation by
using the SNB model in the same way as that for gas radiationdescribedby Kim etal.5 The onlymodication requiredis to replace
the gas transmissivity by the following transmissivity:
Ntnm(s1 ! s2) D Ntgm(s1 ! s2) exp
Z s2
s1
Njpmds (13)
to account for the contributionby particulate absorption and emis-
sion. Following Kim et al.,5 the discretized form of Eq.(11)alonga line of sight is written as
NIm,i C 1 D NIm,i C (1 Ntnm, i ! i C 1) NIbm,i C 12
C NCm,i C 12
(14)
where
NCm,i C
12
D NIwm,1 ( Ntn m,1 ! i C 1 Ntn m, 1 ! i )
C
i 1XkD 1
( Ntn m,kC 1 ! i C 1 Ntn m, kC 1 ! i )
( Ntn m, k! i C 1 Ntn m,k ! i ) NIbm,kC 12
(15)
which arises from the correlation of the sum of spectral gas absorp-tion coefcient and particulate absorption coefcient and the non-
scattered radiation intensity. The spatial discretization indexi D 1corresponds to the starting point of the line of sight under consid-eration on a wall boundary. A simpler form of the discretized RTE
along a line of sight can be obtained by starting from the integralform, rather than the differential form, of the RTE. The boundary
condition associated with Eq.(11)is
NInm,w D m,w NIbm,w (16)
Equation(12)has to be solved approximately because there is
no closed form for the correlation term between the gas absorptioncoefcient and the scattered radiation intensity jg mIs m. The gray-
band model developed by Liu et al.18 was used to approximate thisterm as
jgmIsm D Njgm NIsm (17)
with the mean gas absorption coefcient estimated from
Njgm D Ntgm(lm )
lm(18)
where lm D 3.6 V/S is the mean path length of the local com-
putational control volume.22 Using the gray-band approximation,Eq.(12)then takes a similar form as the standard RTE written in
terms of absorption and scattering coefcients. Note that terms in-volving NIn min Eq.(12)are treated as source terms because they are
known once Eq.(11)is solved. Therefore, Eq. (12)can be solvedby using an arbitrarysolution method developedin the literature.In
this work the popular discrete ordinates method was used to solve
Eq. (12). The boundary condition for the band-averaged scatteredradiation intensity is given as
NIsm,w D 1 m,w
p
ZnX< 0
( NIn m,w C NIsm,w)jnj d\ , for n X> 0
(19)
Although other splitting schemes ofIm are possible, the splittingjust discussed offers at least two major advanta ges. First, solution ofthe nonscatteredpartis noniterativebecauseit is decoupled fromthescatteredpart and itsboundaryconditionis in general known.There-
fore, the computationallyintensive ray-tracingproceduredescribedby Liu7 can be used to solve Eq. (14). Second, the nonscattered
part In m, which is in general greaterin magnitudethan the scatteredpart, is calculated accurately. The role of radiation scattering is to
redistribute but not to alter the radiant energy. Therefore, the ap-proximation made to the correlation term in Eq.(17)to close the
transferequationof the scatteredportion of theradiationintensityistherefore unlikelyto reducesignicantly the overall accuracy of the
present method provided that the scattered part is relatively smallcompared to the nonscatteredpart. In addition, the implementation
of the variable splitting strategy just discussed is not restricted tomodels( SNB, SNBCK) and solvers(ray-tracing, DOM) used in
the present study. It allows the nonscattered part to be solved moreaccurately with a less accurate treatment for the scattered part.
The SNB model was used in the present study to calculate thenarrowbandaveragedga s transmissivities.In this modelthe narrow-
band averaged transmissivity for an isothermal and homogeneouspath js0 ! sj is given as23
Ntgm(s0 ! s) D exp
"
Nbm
p
1 C
2 pf pjs 0 ! sj Nkm
Nbm 1
!# (20)
where the average line width to spacing ratio is given as Nbm D
2p Ncm/Ndm. For a nonisothermal and/or inhomogeneous path, theCurtisGodsonapproximation17,24 is used. The updatedSNB modelparameters provided by Souani(A. Souani, private communica-
tion, Ottawa, 1997)were used in the calculationsof this study. ThisdatasetcontainsSNB model parametersforCO,CO2 ,andH2O with
a constantspectral interval of 25-cm1 width. The covered temper-ature and spectral ranges are 3002900 K and 1509300 cm1,
respectively. Further details of the updated data set were given by
Souani and Taine.
24
III. Results and Discussions
The calculationof the nonscatteredpart, Eq.(14), was conducted
rst without iteration.Once the nonscatteredpart was obtained, thescattered part was calculated with terms containing NIn macting as
known source terms. The transfer equation for nonscattered radia-tion,Eq. (14), was solvedusing the ray-tracingmethoddescribedby
Liu.7 Ray tracing was performedat thecenterof each computationalgrid along all the directions dened by the T4 quadrature set(128
directions). The SNB model with theupdated modelparameterswasused to calculate the gas transmissivity. The transfer equation for
scattered radiationwas solved using DOM. The spatial and angulardiscretizationswere achieved using the positive scheme and the T 4
set, respectively,for all DOM calculations conducted in this study.The test problem is a rectangularplume of 2 2 8 m. The ob-
ject is to predict the infrared signature in the wave-number r ange
between 2000 and 4000 cm1 from this plume. The plume was di-vided into 11 11 40 control volumes. A nonuniform grid wasused along the length of the plume with ner grids placed aroundthe peak temperature for the nonisothermalcase described next and
uniform grid was used in the other two directions. The speciedspatial resolutionwas found to be adequatefor the results presented
next, i.e., further grid renement does not alter signicantly thenumerical results. The radiating gas in the plume was assumed to
be a mixture of 0.2H2 O C 0.8N2 (mole basis)at 1 atm. A singleradiating gas was c onsidered in the p resent calculations to avoid
the time-consuming treatment of overlapping bands encounteredin the SNBCK method. The assumption was also made that the
plumecontainsuniformlydistributedaluminumoxide particleswitha number density of 2 109 m3 . The diameter of these particles
is uniform at 11.6lm. In addition, these particles are assumed tobe nonabsorbingand have a refractive index ofm D 1.74. Two sets
of thermal conditions were used to perform the calculations.In therst case the plume was assumed to be isothermal at 1800 K. In
7/23/2019 Application of Statistical Narrowband Model to Three-Dimensional Absorbing-Emitting-Scattered Media
4/7
288 LIU, SMALLWOOD, AND GULDER
the second case the gas temperature is nonuniform but symmetri-cal about the centerline of the plume and is specied in terms ofTD (Tc Te)f(r/R) C Te. In this equationTc is the gas tempera-ture along the centerline of the plume, and Te is the exit temper-
ature at z D 8 m. Inside the circular region of the cross section ofthe plume, the variation of gas temperatureis dened by f(r/R) D
1 3(r/R )2 C 2(r/R)3 , where ris thedistancefromthe plumecen-terline andR is the radius of the circularregion (R D 1 m). The gas
temperature outside the circular region is assumed to be uniformand at the value of the exit temperature. The centerlinetemperature
is assumed to increase linearly from 400 K at the inlet (z D 0)to2400 K at z D 0.75 m, then decreases linearly to 800 K at the exit.
The boundaries of the plume are assumed to be black and cold.The computational conditions just described constitute a rather
severe test to evaluate the accuracy of the variable splitting methodfor two reasons. First, the particlesare assumed to be nonabsorbing
with relatively large number density, conditions that will increasethe singlescattering albedo.Second,the nonisothermaltemperature
eld possesses large gradients.Radiative properties of particles were calculated using the
BHMIE code based on Mie theory.25 The scattering phase func-
tion was approximated by the HenyeyGreenstein function26 with
the asymmetry factor calculated from Mie theory. For calculationsinvolvingradiationscattering,iterationswerestoppedwhen therela-tive error of the solidangle integratedintensity (gray-band intensityin the gray-band model, the scattered part of the intensity in thepresent method, or the intensity of the j th gray gas in the SNBCK
method)is less than 1 102.Because narrowband results in three-dimensionalscattering me-
diaare not availablein the openliterature,at least to ourknowledge,results of the variable splitting method are compared with those
obtained using the SNBCK method12 and the gray-band model.20
Results of the SNBCK method cannot be treated as a benchmark
solution because of its approximate nature. However, they serve asa good reference solution to evaluate the accuracy of the variable
splitting method.A detailed description of the SNBCK method can be found in
Lacis and Oinas.12 The RTE associated with the j th quadraturepoint takes the following form:
Ij m
sD (kj m C jpmC jsm)Ij mC (kj mC jpm)Ibm
Cjsm
4p
Z4p
Ij mUmd\ (21)
where kj m is the gas absorption coefcient at the j th quadrature
point. The narrowband intensity is calculated as
NIm D
7
Xj D1
xj mIj m (22)
wherexj mis the weight parameter associated with the j th quadra-
ture point. The seven-point GaussLabatto quadrature points andcoefcients are given in Riviere et al.8 The relative errors of the gastransmissivitiesover isothermal and homogeneous columns calcu-lated using the seven-point GaussLabatto quadrature are less than
2%, compared to the transmissivities of the SNB model. The RTEof the SNBCK method, Eq.(21), was solved using DOM with the
positive spatial differencing scheme.Because the objective of the present paper is to demonstrate the
accuracy and capability of the variable splitting method, numeri-cal calculations were conducted in the spectral region of 2000 to
4000 cm1 instead of the entire infrared range. To evaluate the per-formance of the present variables plittingmethod for scatteringcal-
culations, numerical calculationswere rst performed for the non-scattering isothermal case(T D 1800 K)by setting the particulate
scattering coefcient to zero. Comparing the performance of thevariable splitting methodand the gray-bandmodel of Liu et al.20 for
scatteringproblems as they bear some similarity in the treatment ofthe correlation term yet differ fundamentally is also interesting.In
Fig. 1 Comparisons of narrowband integrated radiation intensitiesat (1 m, 1 m, 7.9 m) and along the direction (0.0990147, 0.0990147,0.990147)for the isothermal case without scattering.
thegray-bandmodelof Liu etal.,20 thecorrelationbetweenthe spec-
tral radiation intensity and the gas absorption coefcient was eval-uated approximately using the noncorrelated expression, Eq.(17).
As a result, this modelproducessignicant errors in the calculationsof narrowband intensities in a nonisothermal CO2 -H2O mixture.
19
In contrast,the variablesplitting method treats the gas radiation(thenonscatteredp art of the intensity)rigorously, which is the difcult
partof the problem,and makes use of thegray-bandapproximation,Eqs. (17) and (18), only to thescatteredpartof theintensity,which is
in general smaller in magnitude compared to the nonscatteredpart.A direct comparison of the results of the variable splitting methodand the gray-band model serves to illustrate the improvement in ac-curacy achieved by the variable splitting strategy with the SNBCK
results as a reference.Figure 1 shows the infrared signature (narrowbandintegratedra-
diation intensity)viewed at (x , y ,z) D (1 m, 1 m, 7.9 m)and alongthe direction (0.0990147, 0.0990147, 0.990147)for this isothermal
and nonscattering case. This direction is the rst discrete ordinatein the T4 set and is almost in parallel with the plume centerline.
Results of the SNB model are in good agreement with the resultsof the SNBCK method, and the difference between them is within
10%. The discrepancy between the results of these two methods iscaused by using two different solvers, i.e., ray tracing for the SNB
model and DOM for the SNBCK method. The ray-tracing methodis numerically more accurate thanDOM as faras radiationintensity
along a line of sight is concerned. This is because in the ray-tracingalgorithm the RTE is discretized along the direction of radiation
propagation, which is one-dimensional. The gray-band results arequalitativelycorrectbut signicantlyerroneousat mostof the bands.
Numerical calculations were then carried out for the isothermalandsc atteringcase,and theresultsare comparedin Fig. 2 forinfrared
signature along the same line of sight as in Fig. 1. The agreementbetween the results of the present variable splitting method and theSNBCK method is found to be similar to and even better for somebands than what was observed in Fig. 1. This apparent improved
agreement between the results of the two methods found in Fig.2, compared to those in Fig. 1, is actually caused by the errors of
the variable splitting method for scatteringcalculations.On the onehand,theray-tracingalgorithmforthenonscatteredparttendstopre-
dict higherintensities than the SNBCK method using DOM, Fig. 1.On theotherhand,the gray-bandapproximationusedin thecalcula-
tion of the scattered part in general overestimatesthe magnitude of
the intensity based on our previous studies.19,20 Note that the scat-teredpart is negativein the presentcontext.Therefore,the combined
effects of using ray tracing for the nonscattered part and gray-bandapproximationfor the scatteredpart lead to improvedagreementbe-
tween the results of the variable splitting and the SNBCK. Resultsshown in Figs. 1 and 2 indicate that the variable splitting method
7/23/2019 Application of Statistical Narrowband Model to Three-Dimensional Absorbing-Emitting-Scattered Media
5/7
LIU, SMALLWOOD, AND GULDER 289
Fig. 2 Comparisons of narrowband integrated radiation intensitiesat (1 m, 1 m, 7.9 m) and along the direction (0.0990147, 0.0990147,0.990147) for the isothermal case with scattering. Particle number den-
sity is 2 109 m 3.
Fig. 3 Comparisons of narrowband integrated radiation intensities
at (1 m, 1 m, 7.9 m) and along the direction (0.0990147, 0.0990147,0.990147)for the nonisothermal case without scattering.
offers similar accuracy to the SNBCK method in this isothermal
scattering case. Again, results of the gray-band model are signi-cantly erroneous as shown in the nonscatteringcase, Fig. 1. Com-
parisonbetween results shown in Figs. 1 and 2 shows that the effectof particle scattering under these conditions is to lower radiation
intensity along this line of sight.Figure 3 shows the narrowband intensities along the line of sight
dened in Fig. 1 forthe nonisothermalandnonscatteringcase. Goodagreement was found between the results of the SNB model andthe SNBCK method, except in the wave-number region between
2800 and 3300 cm1 where the discrepancy reaches about 30%.Part of such a large discrepancy is attributed to difference caused
by using two different solvers as mentioned earlier. However, thediscrepancy is mainly caused by the two different approximations
made in the SNB model and the SNBCK method for nonisother-mal media. The CurtisGodson approximation was used to obtain
equivalentband parametersfor the calculationof gas transmissivityalong a nonisothermal path17 in the SNB model, whereas the scal-
ing approximationwas used in the SNBCK method for applicationto nonisothermal media.17 The question of which model is more
accurate for radiative transfer calculation in nonisothermal mediacan onlybe answeredwith the helpof more accurateresults, suchas
fromline-by-linecalculations,andis beyondthe scopeof the presentstudy. Nevertheless,the results shown in Fig. 3 serve as a reference
Fig. 4 Comparisons of narrowband integrated radiation intensitiesat (1 m, 1 m, 7.9 m) and along the direction (0.0990147, 0.0990147,0.990147)for the nonisothermal case with scattering. Particle number
density is 2 109 m 3.
Fig. 5 Nonscattered and scattered portions of the narrowband inte-gratedintensities at (1 m , 1 m , 7 . 9 m) andalongthe direction (0.0990147,
0.0990147,0.990147) for the nonisothermalcasewith scattering.Particlenumber density is 2 109 m 3 .
to evaluate the performance of the variablesplitting method for thenonisothermalscatteringca se to be shown in Fig. 4. The gray-band
model yields large errors at most of the bands calculated.Results of the nonisothermal scattering case along the same line
of sight areshownin Fig. 4. A comparison betweenthe results of theSNB and theSNBCK shownin Fig. 4 andthosein Fig. 3 impliesthat
the variable splitting method predicts unsatisfactoryband-averagedintensities for this nonisothermal scattering case because the levelof agreement between the results of the variable splitting and the
SNBCK method in Fig. 4 deterioratessignicantlycomparedto thatin Fig. 3. The results of the variable splitting method are in general
still better than the gray-band results.As a direct consequenceof variable splitting,the present method
naturallyseparatesthe scatteredand the nonscatteredportions of theradiation intensity. The effects of scatteringcan be readily analyzed
from the calculationsusing this method. For the nonisothermalandscattering case the scattered and the nonscattered portions of the
narrowband intensities along the line-of-sight dened in Fig. 1 areshown in Fig. 5 along with the total(sum of the scattered and the
nonscattered portions)intensities. The importance of scattering ineach band are clearly seen. The scattered portion of the intensity is
in general negative because in the present variable splittingschemethe transferequationof the scatteredpart,Eq.(9), containsall terms
7/23/2019 Application of Statistical Narrowband Model to Three-Dimensional Absorbing-Emitting-Scattered Media
6/7
290 LIU, SMALLWOOD, AND GULDER
Table 1 CPU times(s)of the three methods in the calculations of the three scattering cases
Run conditions Gray-band SNB(variable split ting) SNBCK
Isothermal, n D 2 109 m3 8,019.6 61,242.7 75,481.8
Nonisothermal,n D 2 109 m3 9,480.3 61,442.4 92,619.0
Nonisothermal,n D 2 108 m3 4,768.5 54,938.3 48,170.2
Fig. 6 Comparisons of narrowband integrated radiation intensitiesat (1 m, 1 m, 7.9 m) and along the direction (0.0990147, 0.0990147,
0.990147)for the nonisothermal case with scattering. Particle numberdensity is 2 108 m 3.
related to scattering. To be specic, the source terms of Eq. (9)
(terms containingInm)are almost always negative.To examine the performance of the variable splitting method at
lower particle number densities, numerical calculations were con-ducted for the nonisothermalscatteringcase with a particle number
density of 2.0 108 m3 and the results are compared in Fig. 6 fornarrowband intensities along the same line of sight as just men-
tioned. The level of agreement between the results of the SNB andthe SNBCK methods in Fig. 6 is seen to be similar to that observed
in the nonscatteringcase shown in Fig. 3, indicating that the resultsof the variable splitting method for this lower particle loading case
are quite accurate.The relativeimportance of the nonscatteredpart and the scattered
part of radiation intensity for the nonisothermal and lower particleloading case is shown in Fig. 7. The magnitude of scattered part
is reduced by about a factor of four compared to the high particleloading case, Fig. 5.
All of the numerical calculations of the present study were per-formed on a SGI Octane at 175 MHz. The CPU times of the three
scatteringcases(isothermaland higherparticleloading,nonisother-mal and higher particle loading, and nonisothermal and lower par-
ticle loading)are compared in Table 1.The CPU time of thegray-bandmodel is about an order of magni-
tude lower than those of the other two methods. For the two higherparticle loading cases the variable splitting method requires less
CPU time than the SNBCK method. For the lower particle loadingcase, however, the variable splitting method requires slightly more
CPU time.In addition,theCPU time ofthe variablesplittingmethodis less affected by the level of particle loading because most of the
CPU time of the variablesplitting method is consumed in the calcu-lation of the nonscattered part using the ray-tracing algorithm.The
CPU times given in Table 1 were for a relativelylarge relative error
of 1.0 102 used as the convergencecriterion for the iterationanda single radiating gas calculations.Under these conditions the vari-
able splitting method does not offer bettercomputational efciencythan the SNBCK method. If a more stringent criterion is used in
the iteration or the problem involves two or three radiating gases,which is often the situation for practical applications, the variable
Fig. 7 Nonscattered and scattered portions of the narrowband inte-gratedintensities at (1 m , 1 m , 7 . 9 m) andalongthe direction (0.0990147,
0.0990147,0.990147) for the nonisothermalcasewith scattering.Particlenumber density is 2 108 m 3 .
splittingmethod will be more efcientthan the SNBCK method.Atan overlapping band the CPU time of the variable splitting method
remainsalmostthe same as thatat a nonoverlappingband.However,
the CPU time of the SNBCK method increases by about a factor ofseven for an overlapping band calculation.
IV. Conclusions
A variable splitting method was developed to incorporate the
statistical narrowband model into the radiative transfer equation tocalculate narrowband radiation intensities in three-dimensionalab-
sorbing, emitting, and scattering media. The method offers similaraccuracy as the SNBCK method in the isothermal case. For non-
isothermal gas-particle mixtures, the method offers good accuracyfor low to intermediate particle loadings. However, results of the
variable splittingmethod are inaccuratefor nonisothermaland highparticleloadingmixtures.For problemscontaininga single radiating
gas the variable splitting method offers similar computational ef-
ciency to the SNBCK approach. The variable splitting method willbecome more efcient than the SNBCK method for problems in-volving two or more radiating gases. The method developed in this
work alleviates to some extent the band model-scattering incom-patibility difculty of the statistical narrowband model. This study
demonstrates that the statistical narrowband has a good potentialin the predictionof nongray radiative transfer in three-dimensionalscattering media provided that the particle loading is not too high.
Acknowledgments
This work was supported in part by the Canadian Departmentof
National Defence, Task DREV 36-1/96. This is National ResearchCouncil Paper No. 41985. The authors acknowledge the helpful
suggestionsof the reviewers to clarify the presentationof this study.
References1 Denison, M. K., and Webb, B. W., A Spectral Line-Based Weighted-
Sum-of-Grey-Gases Model for Arbitrary RTE Solvers, Journal of Heat
Transfer, Vol. 115, No. 4, 1993, pp. 10041012.2 Denison, M. K., and Webb, B. W., An Absorption-LineBlackbody Dis-
tribution Function for Efcient Calculation of Total Gas Radiative Transfer,
7/23/2019 Application of Statistical Narrowband Model to Three-Dimensional Absorbing-Emitting-Scattered Media
7/7
LIU, SMALLWOOD, AND GULDER 291
Journal of Quantitative Spectroscopy and Radiative Transfer, Vol.50, No. 5,1993, pp. 499510.
3Denison, M. K., and Webb, B. W., The Spectral Line-Based Weighted-Sum-of-Grey-Gases Model in Non-Isothermal Non-Homogeneous Media,Journal of Heat Transfer, Vol. 117, No. 2, 1995, pp. 339365.
4Edwards, D. K., Molecular Gas Band Radiation, Advances in HeatTransfer, edited by T.F. Irvine Jr. and J. P. Hartnett, Vol. 12, Academic, NewYork, 1976, pp. 115193.
5Kim, T-K., Menart, J. A., and Lee, H. S., Nongray Radiative Gas Anal-
ysis Using the S-N Discrete Ordinates Method,Journal of Heat Transfer,Vol. 113, No. 4, 1991, pp. 946952.
6Zhang, L., Souani, A., and Taine, J., Spectral Correlated and Non-Correlated Radiative Transfer in a F inite Axisymmetric System Containing
an Absorbing and Emitting Real Gas-Particle Mixture,International Jour-nal of Heat Mass Transfer, Vol. 31, No. 11, 1985, pp. 22612272.
7Liu, F., Numerical S olutions of Three-Dimensional Non-Grey Gas Ra-
diative Transfer Using the Statistical Narrow-Band Model,Journal of HeatTransfer, Vol. 121, No. 1, 1999, pp. 200203.
8Riviere, P. H., Souani, A., and Taine, J., Correlated-kand Fictitious
Gas Methods for H2O Near 2.7lm,Journal of Quantitative Spectroscopyand Radiative Transfer, Vol. 48, No. 2, 1992, pp. 187203.
9Tang, K. C., and Brewster, M. Q., K-Distribution Analysis of Gas Ra-
diation with Nongrey, Emitting,Absorbing,and AnisotropicScattering Par-ticles,Journal of Heat Transfer, Vol. 116, No. 4, 1994, pp. 980985.
10 Marin, O., and Buckius, R. O., Wideband Correlated-kMethod Ap-
plied to Absorbing, Emitting, and Scattering Media, Journal of Thermo-physics and Heat Transfer, Vol. 10, No. 2, 1996, pp. 364371.
11 Lee, P. Y. C., Hollands, K. G. T., and Raithby, G. D., Reordering theAbsorptionCoefcient within theWideBand forPredictingGaseous RadiantExchange,Journal of Heat Transfer, Vol. 118, No. 2, 1996, pp. 394400.
12 Lacis, A. A., and Oinas, V., A Description of the CorrelatedkDistribu-tionMethodfor ModelingNongrayGaseous Absorption,Thermal Emission,and Multiple Scattering in Vertically InhomogeneousAtmosphere,Journal
of Geophysical Research, Vol. 96, No. D5, 1991, pp. 90279063.13 Souani, A., Gas Radiation Spectral Correlated Approaches for In-
dustrial Applications,Heat Transferin Radiating and Combusting Systems,
edited by M. G. Carvalho, F. C. Lockwood, and J. Taine, SpringerVerlag,Berlin, 1991.
14 Reed, R. A., Brown, D. G., Hiers, R. S., III, Cromwell, B. K., and
Zaccardi, V. A., Compatibility of Infrared Band Mo dels with Scattering,Journal of Th ermophysics Heat Transfer, Vol. 8, No. 2, 1994, pp. 208215.
15 Kim, O. J., and Song, T-H., Implementation of the Weighted Sum ofGray Gases Mod el to a Narrow Band: Application and Validity,NumericalHeat Transfer, Part B, Vol. 30, 1996, pp. 453468.
16 Yang, S-S, and Song, T-H., An Improved WSGGM-Based NarrowBand Model for CO2 4.3 lm Band,Revue Generale de Thermique, Vol.38, No. 3, 1998, pp. 228238.
17 Goody,R. M.,and Yung,Y.L.,Atmospheric Radiation, 2nded.,Oxford
Univ. Press, Oxford, 1989.18 Liu, F., Gulder, O. L., Smallwood, G. J., and Ju, Y., Non-Grey Gas
Radiative Transfer Analyses Using the Statistical Narrow-Band Model,International Journal of Heat Mass Transfer, Vol. 41, No. 14, 1998, pp.
22272236.19 Liu,F., Gulder, O. L., and Smallwood, G. J., Three-Dimensional Non-
Grey Gas Radiative Transfer Analyses Using the Statistical Narrow-Band
Model,Revue Generale de Thermique, Vol. 37, No. 7, 1998, pp. 759768.20 Liu,F., Gulder, O. L.,and Smallwood, G. J., Evaluation of an Approx-
imate Narrow-Band Formulation for Non-Grey Radiative Transfer Calcu-
lation in Three-Dimensional Absorbing-Emitting-Scattering Media, Pro-ceedings of the 7th AIAA/ASME Joint Thermophysics and Heat TransferConference, Vol. 1, American Society of Mechanical Engineers, New York,
1998, pp. 3339.21 Modest, M. F., Radiative Heat Transfer, McGrawHill, New York,
1993,p. 305.22 Hottel, H. C., and Sarom, A. F., Radiative Transfer, McGrawHill,
New York, 1967, p. 278.23 Ludwig, D. B., Malkmus, W., Reardon, J. E., and Thomson, J. A. L.,
Handbook of Infrared Radiation from Combustion Gases, NASA SP3080,1973.
24 Souani, A., and Taine, J., High Temperature Gas Radiative PropertyParameters of Statistical Narrow-Band Model for H2O, CO2 and CO, andCorrelated-K Model for H2 O and CO2 ,International Journal of Heat Mass
Transfer, Vol. 40, No. 4, 1997, pp. 987991.25 Bohren, C. F., and Huffman, D. R.,Absorption and Scattering of Light
by Small Particles, Wiley, New York, 1983, pp. 479482.26 Viskanta, R., and Menguc, M. P., Radiation Heat Transfer in Combus-
tion Systems,Progress in Energy and Combustion Science, Vol. 13, No. 2,
1987, pp. 97160.
Top Related