Chinese Journal of Aeronautics, (2017), 30(2): 651662

Chinese Society of Aeronautics and Astronautics

& Beihang University
Chinese Journal of Aeronautics
cja@buaa.edu.cn
www.sciencedirect.com

A new model to simulate infrared radiation from an

aircraft exhaust system
Zhou Yue a,*, Wang Qiang a,b, Li Ting a,b

a
School of Energy and Power Engineering, Beihang University, Beijing 100083, China
b
Collaborative Innovation Center for Advanced Aero-Engine, Beijing 100083, China

Received 20 November 2015; revised 10 February 2016; accepted 24 March 2016

Available online 17 February 2017

KEYWORDS
Computational efficiency;
Infrared radiation;
k-distribution;
Multi-scale method;
Radiative contribution
Abstract A multi-scale narrow band correlated-k distribution (MSNBCK) model is developed to
simulate infrared radiation (IR) from an exhaust system of a typical aircraft engine. In this model,
an approximate approach instead of statistically uncorrelated assumption is used to treat overlap-
ping bands in gas mixture. It significantly reduces the requirement for computing power through
converting the exponential increase of computing power consumption with the increase of partici-
pating gas species to linear increase. Besides, MSNBCK model has a great advantage compared
with conventional methods which can estimate each species contribution to the total gas mixture
radiation intensity. Line by line (LBL) results, experimental data and other results in the references
are used to evaluate this new model, which demonstrates its advantage in terms of accuracy and
computing efficiency. By coupling this model and finite volume method (FVM) into radiative trans-
fer equation (RTE), a comparative study is conducted to simulate IR signature from the exhaust
system. The results indicate that walls IR emission should be considered in both 35 lm and
814 lm bands while gases IR emission plays an important role only in 35 lm band. For plume
IR radiation, carbon dioxides emission is much more significant than that of water vapor in both
35 lm and 814 lm bands. Especially in 35 lm band, the water vapors IR signal can even be
neglected compared with that of carbon dioxide.
2017 Chinese Society of Aeronautics and Astronautics. Production and hosting by Elsevier Ltd. This is
an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction

Infrared stealth is one of important abilities for aircraft to pre-

vail in modern air battle. IR emission from the exhaust system
* Corresponding author. takes a great proportion in the total IR emission of an aircraft.
E-mail address: zhouyuebuaa@163.com (Y. Zhou). Therefore, accurate prediction of IR signature from an exhaust
Peer review under responsibility of Editorial Committee of CJA. system is of great significance for IR detection and aircraft
propulsion system design.
IR emission from a specific exhaust system can be attribu-
ted to two parts: hot plume IR emission and high temperature
Production and hosting by Elsevier
engine parts IR emission, such as that of nozzle, and turbine
http://dx.doi.org/10.1016/j.cja.2017.02.014
1000-9361 2017 Chinese Society of Aeronautics and Astronautics. Production and hosting by Elsevier Ltd.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
652
blades. Compared with high temperature engine parts, radia-
tion from the hot plume is much more complicated, since car-
bon dioxide and water vapors IR emission plays a major role,
which shows non-gray characteristics. With respect to non-
gray gas radiative calculation, LBL method is conducted using
detailed information of every needed spectral line from a high
resolution spectroscopic database; therefore it is recognized as
the most accurate method.1 However, the disadvantage of this
method is also very obvious, and it requires huge amount of
computing power which makes it impractical to be applied
to the real world. For this reason, different models have been
developed to provide relatively accurate yet efficient calcula-
tions. From the functional standpoint, all these models can
be divided into two kinds: one is for calculating thermal radi-
ation such as Weighted-Sum-of-Gray-Gases (WSGG) model,2
Spectral Line Based Weighted Sum of Gray Gases (SLW)
model,36 absorption distribution function (ADF) model7
and full-spectrum k-distribution (FSK) model8,9 and the other
is for IR signal calculation such as statistical narrow band
(SNB) model10,11 and narrow band k-distribution (NBK)
model.12 The output parameter of SNB model is transmissiv-
ity, which is suitable for ray tracing RTE solver, but it is diffi-
cult to couple with differential methods such as finite volume
method and discrete coordinate method. Besides, the SNB
model is also incompatible with scattering from particles or
soot in the plume even though different strategies have been
tried to solve this issue.13,14 In comparison, NBK model pro-
vides absorption coefficient as model output parameter which
can be coupled with arbitrary RTE solution method and deal
with scattering simultaneously. The principle of NBK model is
that provided the medium is homogeneous and for a narrow
band which is sufficiently narrow to assume a constant Planck
function, the absorption coefficient can be reordered into a
smooth and monotonically increasing function so that integra-
tion over the narrow band can be replaced by an integral
scheme with several points. Compared with spectral evalua-
tions for LBL, NBK model provides an efficient way to reduce
the computational costs, but it is limited to calculate single
homogeneous medium. Two methods have been used to deal
with non-homogeneous gas, i.e. correlated-k (CK) method
and scaling approximation, and the CK method assumes that
the maximum absorption coefficients across the spectrum
under investigation always occur at the same wavenumber,
regardless of the temperature, pressure and mole fraction,15
and are suitable for all intermediate values. The scaling
approximation is inherently similar to CK method but more
precise in mathematical terms. Because of less restrictiveness
and acceptable accuracy, the CK method is widely used in
both full spectrum and narrow band, forming FSCK model
and NBCK or SNBCK model.16,17 For all band models, treat-
ment of overlapping bands in gas mixture is always somewhat
problematic. One conventional approach is statistically uncor-
related assumption, i.e. transmissivity of a gas mixture can be
obtained by multiplying transmissivity of each species. This
turned to be true for narrow band but the computational costs
become expensive by a factor of NM where N is the number of
integral points in solving RTE and M the number of
participating gases. Depending on different assumptions, some
approximate mixing models such as convolution, multiplica-
tion, superposition and hybrid approaches have also been
proposed by Solovjov and Webb to deal with gas mixtures
radiation in full spectrum.18 Besides, Modest and Riazzi
Y. Zhou et al.
calculated the mixtures cumulative k-distribution function
based on the multiplication of transmissivities and property
of Laplace transformation.19 Instead of treating the mixture
as a single gas, it is noted that a multi-scale approach is applied
in full spectrum to calculate radiation emitted by a certain gas
but absorbed by all the gases. This is an approximate way but
which finally achieved both reasonable accuracy and computa-
tional efficiency.20 Above all, it increases computational costs
by a factor of MN, but to date, the multi-scale concept is
mainly used in full spectrum with several different models hav-
ing been derived.2123
In this paper, we present a narrow band model based on the
CK method and multi-scale method, and to the best of the
authors knowledge, such a combination (MSNBCK) has not
been carried out in the past. Experimental and LBL results will
serve as a benchmark to validate the accuracy of k-distribution
method. When dealing with overlapping bands in gas mixture,
the MSNBCK model is evaluated by comparison with exact
LBL results and SNB model combined with statistical uncorre-
lated assumption for both isothermal and non-isothermal
cases. Using the Finite Volume Method (FVM) to solve
RTE and MSNBCK model and describe the property of
non-gray gas mixture, simulation of the infrared signature
from a realistic three-dimensional aircraft engine is conducted.
Absorption coefficients of participating species are calculated
from the spectroscopic database HITEMP 2010 in advance
and all the IR calculation process in this paper is accomplished
using C program.
2. Numerical modeling and validation
The radiative transfer equation for an absorbing, emitting and
scattering medium is written as1
Z
dIg rsg
kg Ibg  Ig  rsg Ig Ig s0 Ug s; s0 dX0 1
ds 4p 4p
where subscript g is wavenumber, Ig spectral intensity varying
along a path s, Ibg Planck function, kg spectral absorption coef-
ficient, rsg scattering coefficient, Ug s; s0 scattering phase func-
tion, and X0 the solid angle. In the case of a gray surface which
emits and reflects diffusely, boundary conditions subject to
Eq. (1) can be expressed as
Z
1e
Ig eIbgw Ig jn  sjdX 2
p ns<0
in which e is the walls emissivity, Ibgw radiative intensity of the
black wall and n the corresponding surface normal vector.
2.1. k-distribution and spectral reordering
In a narrow spectral interval (generally 450 cm1), the Planck
function Ibg , scattering coefficient rsg and scattering phase
function Ug s; s0 can be treated as constants,8 and the spectral
intensity would obtain the same value repeatedly since the
absorption coefficient attains the same value (Fig. 1)
(1 atm = 101,325 Pa). Therefore, the key point of k-
distribution method is to reorder the absorption coefficient
into a smooth and monotonically increasing function so that
RTE solution with the same absorption coefficient is carried
out only once.
A new model to simulate infrared radiation 653

behavior and a cumulative k-distribution function is further

defined as
Z k
gk fkdk 6
0

which progressively represents fraction of the spectrum whose

absorption coefficient lies below the value of k.
The reordered absorption coefficient corresponding to
Fig. 1 is shown in Fig. 2, and horizontal axis is cumulative
k-distribution function gk, which can be regarded as a
pseudo wavenumber inherently.
Since gk is a monotonically increasing function with
0 6 gk 6 1, the primitive integration turns to be very effi-
cient using numerical method. Accordingly, the narrow band
Fig. 1 Absorption coefficient of CO2 in 4.3 lm band.
averaged intensity could be written as
Z Z kmax Z 1
For a given k value, the k-distribution function fk is 1
Ig Ig dg Ik fkdk Ig dg 7
defined as Dg Dg 0 0
 
1 X dg  in which Ig is the reordered spectral intensity corresponding to
fk dkg  3
Dg i i gk. Similarly, other narrow band averaged property such as
transmissivity or emissivity can be expressed as
in which i denotes spectral locations where kg k mathemat- Z Z 1 Z
ically, and Eq. (3) can be put into a more elegant form as 1 1
Z sg sg dg ekgL dg or eg eg dg
1 Dg Dg 0 Dg Dg
fk dk  kg dg 4 Z 1
Dg Dg
1  ekgL dg 8
where dk  kg is the Dirac-delta function 0


0 jxj > de in which L is the length of gas layer. Eqs. (7) and (8) can be
dx lim 1 5 easily calculated through numerical integration method. For
de!0
2de
jxj < de
this paper, all the numerical integration is accomplished via
Physically, Eq. (3) or Eq. (4) denotes that fk could be con- 7-point Gauss integral scheme. The experimental data mea-
sidered as a weighted sum of the fraction for spectrum where sured by Bharadwaj et al.2427 Exponential Wide Band
kg k. In practical, the function fk always show erratic (EWB) results28 and LBL results are used here for the valida-
tion of k-distribution and spectral reordering. Table 1 shows
the conditions under which the measurements and calculations
are taken.
For all conditions, both measured and calculated data are
shown in Fig. 3, among which the NBK model represents
results using k-distribution method. Generally, the predicted
band transmissivity calculated by NBK model and LBL agreed
with the measured values very well while the NBK model takes
much less computational costs than LBL. For the present case,
the bandwidth of NBK model is 4 cm1 and the wavenumber
resolution of LBL is 0.01 cm1, which means transmissivity
needs to be calculated 7 times for NBK model and 400 times
for LBL method to obtain an average transmissivity in Eq.
(8). In contrast, the EWB model treated transmissivity as a
constant and calculated it only once across a certain band,
Fig. 2 Cumulative k-distribution function corresponding to but loss of detailed spectral information and larger errors are
Fig. 1. also inevitable (Fig. 3).

Table 1 Description of calculated and measured conditions.

Case Specy Measured band T (K) L (cm) Concentration Pressure (atm) Figure
1 CO2 2.7 lm-band 1300 50 1.0 1.0 Fig. 3(a)
2 CO2 4.3 lm-band 1550 50 1.0 1.0 Fig. 3(b)
3 H2O 2.7 lm-band 1000 40 1.0 1.0 Fig. 3(c)
4 H2O 6.3 lm-band 1000 20 1.0 1.0 Fig. 3(d)
654
Fig. 3 Comparison of band transmissivity for cases in Table 1.
Fig. 4 Absorption coefficients at two different temperatures.
2.2. Correlated-k (CK) method for non-isothermal gas
In Section 2.1, the k-distribution method is used for isothermal
gas and is not straightforward to be applied to non-isothermal
medium. To overcome this limitation, the CK method assumes
that the k-distribution functions are correlated regardless of
temperatures.15 It implies that the ups and downs of the
absorption coefficient curve have the same pattern even if tem-
perature changes. This method has been proven very successful
in the atmosphere29; however, the appearance of hot lines at
high temperature makes the pattern of absorption coefficient
curve not strictly match that at low temperature (Fig. 4).
Y. Zhou et al.
Lack of correlation fails to provide accurate results when
CK method is combined with full spectrum (FSCK) to
calculate cases with large temperature gradients, and thus
the multi-group method is subsequently put forward to solve
this problem, which has obtained good accuracy.22,30 In the
multi-group method, the spectrum is broken up into different
groups according to the temperature dependence of absorption
coefficients, such grouping ensures that the absorption
coefficients in every group closely follow the rule of correla-
tion, and CK method is then used in each group. To examine
the accuracy of CK method combined with narrow band
k-distribution model (written as NBCK) as well as the necessity
of multi-group method for narrow band, the NBCK model and
multi-group NBCK model (written as MGNBCK) are used to
calculate narrow band intensities in the following cases.
Considering a slab of CO2-N2 confined between cold black
walls, the left layer is hot gas at 1500 K with a fixed width of
5 cm; the right layer is cold at 300 K and its width is variable
in the calculation; both layers have the same total pressure
1.0 atm and partial pressure of CO2 is 0.1 atm. Non-
dimensional integrated intensity across g = 22002400 cm1
leaving from the cold layer is shown in Fig. 5(a) and (b) corre-
sponds to the computational error which is defined as
jIsymbol  ILBL j=ILBL . By increasing the temperature in left layer
to 2000 K, the present case is recalculated and the results are
shown in Fig. 6.
The results of Figs. 5 and 6 indicate that both NBCK model
and MGNBCK model predict the integrated intensity in non-
isothermal gas very well. Although the accuracy of NBCK
model is not as good as MGNBCK model, its maximum error
A new model to simulate infrared radiation 655

Fig. 5 Computational results of hot layer at 1500 K.

Fig. 6 Computational results of hot layer at 2000 K.

is still quite small as shown in Figs. 5(b) and 6(b). Results of 2.3.1. Statistical uncorrelated assumption
NBCK model also prove that the CK method in a narrow The statistical uncorrelated assumption believes that the total
band is still acceptable although it is not strictly rigid, which transmissivity of an overlapping band obeys the multiplication
is consistent with the conclusion in Ref. 16 to some extent. property,
For MGNBCK model, the present grouping number is 4 as
used in Ref. 30 and integral scheme is 7-point Gauss integra- Y
I
smix si 9
tion, which means that the RTE needs to be solved 4  7 times i1
in every narrow band. In comparison, the NBCK model treats
every narrow band as a whole group and the RTE needs to be where I denotes the total number of individual gas species.
solved just 7 times. Considering the accuracy and computa- Applying Eq. (9) and NBK model to a gas mixture of CO2
tional efficiency together, the NBCK model is a good and H2O, the mixtures transmissivity can be written as
! !
approach and is used to calculate radiation of non- X7 X7

isothermal gases in the following part. smix L xi expkci L xj expkhj L

i1 j1

2.3. Treatments of gas mixture 7 X

X 7
xi xj expkci khj L 10
i1 j1
When radiation calculations in gas mixture are performed, the
NBK model is not straightforward to be used without any pre- where x is the quadrature weight of a gauss integral point, kci
mise. It is mainly because spectral properties of each participat- the absorption coefficient of carbon dioxide at Gauss intagra-
ing gas are often quite different and overlapping in certain tion point i, khj the absorption coefficient of water vapor at
bands, which leads to unknown cumulative k-distribution func- Gauss integration point j, and this implies that the RTE
tion g(k) for gas mixture. To solve this problem, the conven- becomes
tional method: statistical uncorrelated assumption and the dIij
new method: multi-scale method will be discussed as follows. kci khj Iij kci khj Ibg 11
ds
656 Y. Zhou et al.

where i; j represents the Gauss integral point. Based on Eq. In Eq. (19), the interaction between kmg and kg is lumped
(11), the final intensity of gas mixture can be obtained using into km . For Eqs. (14) and (19), the analytical solutions will
Eq. (12), which means that the RTE needs to be solved be obtained if the participating medium is homogeneously sur-
7  7 times. rounded by cold black walls. The exiting averaged intensities
7 X
X 7 of a narrow band at L s for Eqs. (14) and (19) respectively
Imix xi xj Iij 12 are
i1 j1 Z 1 Z
km
Im Img dg Ibg 1  expkm sfm km dkm 23
0 Dg km
2.3.2. Multi-scale method
and
For gas mixture, scattering effects can be neglected in Eq. (1),
Z Z
the absorption coefficient kg and radiative intensity Ig are 1
km
Im Img dg Ibg 1  expksdk 24
summation from M single gas, and accordingly, the total 0 Dg k
RTE can then be separated into M individuals as Eq. (13).
The integrated intensity Im must be equal to Im and this
X
M X
M leads to
dImg
kmg Ibg  kg Img ; kg kmg ; Ig Img 13 km k
ds m1 m1 25
km fm km dkm km kdk
In Eq. (13), it could be explained physically that the inten- or
sity Img is due to emission from the m-th gas species but subject
to absorption by all groups of other gases plus its own group. km fm km dkm km km dkm 26
To reorder the absorption coefficient of each species, Eq. (13) Eq. (26) provides the relationship between km and km . To
is first multiplied by Diracs delta function dkm  kmg , and determine km , Eq. (26) is integrated on both sides.
the RTE is then integrated over and divided by the narrow Z km Z k0 km
band spectral intervalDg, leading to    
k0m fm k0m dk0m km k0m dk0m 27
dImg 0 0
km Ibg T  km Img 14
ds Although Eq. (27) is an implicit equation, km can be deter-
for m 1; . . . M, where mined through numerical integration and linear interpolation.
Z In practical application of multi-scale method, radiative
1 intensity, which is emitted by every single gas (water vapor
Img Img dkm  kmg dg=fm km 15
Dg Dg or carbon dioxide) and absorbed by the total gas mixture, is
Z firstly calculated, and then the single intensity is added
1 together to obtain the total intensity as shown in Eq. (28).
fm km dkm  kmg dg 16
Dg Dg
X
7

Z Imix xi Ihi Ici 28

km
i1
gm fm kdk 17
0 Compared with the statistical uncorrelated assumption in
Z Eq. (12), the multi-scale method turns the exponential calcula-
1
km Img kg /Img dkm  kmg dg=fm km 18 tion into a linear one, which tremendously saves the computa-
Dg Dg tion time. To calculate absorption coefficient km and overlap
In Eq. (18), km is a reordered absorption coefficient of the m- coefficient km for an arbitrary state in Eq. (14), two databases
th scale gas taking into account the overlap with all other spe- for carbon dioxide and water vapor are constructed from
cies. Physically, kmg is reordered during the above deduction HITEMP 2010 in advance, and then the needed coefficients
and the interaction between kmg and kg is lumped into km . In can be obtained through an interpolation between pre-
order to determine km , Eq. (13) can also be reordered in terms calculated states in the database. When non-isothermal gas
of kg ; substitute dk  kg for dkm  kmg , and the final equation mixture is dealt with in this paper, the CK method15 is incor-
will be porated with multi-scale method, forming the final multi-scale
narrow band correlated-k (MSNBCK) model.
dImg k
m Ibg  kImg 19 2.3.3. Validation of MSNBCK model
ds fk
In order to examine the accuracy of both MSNBCK model
where
Z and corresponding databases, the following two cases are sim-
1 ulated under the same computational conditions in Ref. 31 In
Img Img dk  kg dg=fk 20
Dg Dg Ref. 31, the k-distribution function fk was determined
through the inverse Laplace transformation of the gas trans-
Z
1 missivity which was calculated by SNB model. When radiation
fk dk  kg dg 21
Dg Dg in gas mixture was dealt with, the statistical uncorrelated
assumption, i.e., Eqs. (9)(12), was used to treat overlapping
Z
1 bands. For the purpose of comparison, results of NBCK
km kmg dk  kg dg 22
Dg Dg
model using statistically uncorrelated assumption (written as
A new model to simulate infrared radiation
NBCK-uncorrelated) are provided in the following part, and
the bandwidth of both MSNBCK model and NBCK-
uncorrelated model is set as 25 cm1 to keep consistent with
Ref. 31.
Case 1: The gas mixture contains 30% CO2, 30% H2O and
40% N2 (mole fraction), temperature of the mixture is con-
stant at 1800 K, the total pressure is 1.0 atm, and both bound-
aries of the gas column (at x = 0 m and x = 8 m) are black
wall at 300 K. Narrow-band intensities at x = 8 m along neg-
ative normal vector of the wall as well as computational error
defined as jImodel  ILBL j=ILBL are illustrated in Fig. 7, and it
can be seen that the NBCK-uncorrelated model is in good
agreement with exact LBL approach while MSNBCK model
and results in Ref. 31 show some discrepancies in certain
wavenumber. Due to the approximation and interpolation in
multi-scale method, its discrepancies are inevitable but not
serious. Inaccuracy of the results in Ref. 31 is mainly due to
the outdated model parameters apart from the statistical
uncorrelated assumptions involved in the model.
Case 2: the gas mixture contains 10% CO2, 20% H2O and
70% N2 (mole fraction), and temperature of the gas varied
obeying the rule
8 x
>
< 400 2000  x < x
x Considering the compatibility of k-distribution model, finite
Tx
: 800 1600 L  x x P x
> volume method is used to solve RTE and this method is briefly
L  x
*
where x = 1.5 m, L = 8 m. Pressure of the mixture is also
1.0 atm, and boundaries at x = 0 m and x = 8 m are
black wall at 300 K. Narrow-band intensities as well as
computational error of different models are compared in
Fig. 7 Computational results for isothermal gas mixture.
657
Fig. 8. For the non-isothermal path, MSNBCK model and
NBCK-uncorrelated model are consistent with benchmark
solution LBL results while Ref. 31 showed the least accuracy.
No matter the transfer path is isothermal or non-
isothermal, both MSNBCK model and NBCK-uncorrelated
model show satisfactory results. Compared with NBCK-
uncorrelated results in Case 1 and Case 2, the MSNBCK
model missed the intensity slightly for certain wavenumber
even if no essential error is produced. In every narrow band,
the RTE needs to be solved 2500 times for LBL, 72 times for
NBCK-uncorrelated model and 7  2 times for MSNBCK
model. Table 2 shows the CPU time that different models
spend in solving RTE for the cases in Figs. 7 and 8, and all
the calculations were run on a 3.2 GHz Intel Core i7 processor.
Compared with LBL and NBCK-uncorrelated model, the
MSNBCK model reduces the computational time significantly.
Since this paper is focusing on investigating the IR signature
from a 3D aircraft exhaust system, MSNBCK model is used
(Section 3) to save computational costs and acquire results
with good accuracy.
2.4. RTE solver
summarized in this section. Eq. (1) is integrated over volume V
and solid angle X:
Z Z Z
tA Ig s  ndAdX kg Ibg  Ig dVdX 29
X X V
658 Y. Zhou et al.

Fig. 8 Computational results for non-isothermal gas mixture.

Table 2 Computational time of different models. ms.

Item Computational time (ms)
Fig. 7(a) Fig. 7(c) Fig. 8(a) Fig. 8(c)
LBL 578 576 576 577
NBCK-uncorrelated 11 11 11 11
MSNBCK 3 3 3 3

X  Z 
where A is the cells surface and n the corresponding outward 6

face normal vector. Assume that the intensity and radiative aP kg;P VX max Ai s  ni dX; 0 33
i1 X
properties within the control volume and solid angle are con-
stant, which can be characterized by the value at cell center  Z 
P, and Eq. (29) turns to be aAi max Ai s  ni dX; 0 34
X
X 6 Z
Ig;Ai Ai s  ni dX kg;P Ibg;P  Ig;P VX 30 bg kg;P Ibg;P VX 35
i1 X
The above process is applied to general k-distribution
where Ai is the area of cells i-th surface, Ig;Ai the intensity at model, and for MSNBCK model specifically, kg;P in Eq. (33)
the surface center. First-order step scheme shown as Eq. (31) should be replaced by kP .
is chosen to link the relationship between cell-surface intensi-
ties and cell-center intensities.
Z 3. Application and discussion
Ig;Ai Ig;P for s  ni dX > 0 31
X 3.1. Three-dimensional CFD simulations
Then the discretization form of RTE is obtained as follows:
The improved MSNBCK model and finite volume method are
X
6
aP Ig;P aAi Ig;Ai bg 32 used here to calculate IR signature from a real aircraft engine
i1 exhaust system shown in Fig. 9, where h is polar angle and u
azimuthal angle.
where
A new model to simulate infrared radiation
Fig. 9 Exhaust system 3D model and radiative direction
definition.
For a typical condition, the fuel is C12H23; fuelair ratio is
0.0202; free-stream values used are temperature T = 288 K,
pressure P = 101325 Pa and velocity Ma = 0.8; gas tempera-
ture at nozzle inlet is 800 K with a velocity 110 m/s; the wall
is adiabatic. Three-dimensional CFD simulations are con-
ducted using Fluent. Fig. 10 demonstrates the distribution of
static temperature, static pressure, mole fraction of carbon
dioxide and mole fraction of water vapor. Considering symme-
try of the jet flow, only a half cross-section is shown in Fig. 10.
3.2. Radiative characteristics of gas mixture
Cases with different numbers of grid and control volumes are
tested to validate the independence of both spatial and
directional discretization. Finally, the computational domain
Fig. 10 Distribution of nozzles thermodynamic parameters.
Fig. 11 Intensity distribution of 35 lm band (u = 90, h = 25.7).
659
is discretized into 27250 structured grids and total solid angle
range of 4p is divided into Nh  Nu 7  14 non-overlapping
entities. It is necessary to note that the grids used for present
radiation calculation are much sparser than the grids used
for CFD simulation. Corresponding thermodynamic data for
radiation calculation grids is obtained through interpolation
from the CFD results in Section 3.1. Based on MSNBCK
model and the 3D CFD results, we can calculate the intensity
emitted by one specific gas but absorbed by all gases, which
can be a quantitative rule to estimate the role that each partic-
ipating gas plays in the final total intensity. In the present case,
assume the walls emissivity e 0 to ensure that only gases
radiation is considered, and intensity calculated in this part
is emission from one single gas which is then attenuated by
the gas mixture. For infrared detection, the interested spectral
range is usually within 35 lm band and 814 lm band, so
integrated intensities corresponding to g = 20003300 cm1
and g = 7001200 cm1 are calculated respectively. For 3
5 lm band, attenuated intensity emitted by water vapor in
the direction of u 90 , h 25:7 is shown in Fig. 11(a) while
Fig. 11(b) corresponds to that of carbon dioxide. It could be
seen that shock diamonds of the plume is even noticeable in
intensity distribution of carbon dioxide. Compared with car-
bon dioxide, intensity emitted by water vapor can be neglected
for this band.
Accordingly, intensity distribution corresponding to
k = 814 lm is shown in Fig. 12. Comparing scales of the
two graphs, we can obviously see that within 814 lm band,
660
Fig. 12 Intensity distribution of 814 lm band (u = 90, h = 25.7).
Fig. 13 Validation of spectral characteristics of plume.
carbon dioxides emission is still more important than that of
water vapor although the difference is not as huge as that of 3
5 lm band. Results of both bands indicate that less carbon in
the fuel will lead to less IR radiation.
To validate the present code in predicting spectral charac-
teristic of the plume, experimental data in Ref. 32 and calcu-
lated results (at direction u 90 ; h 90 ) are compared in
a non-dimensional form (Fig. 13). In Ref. 32, experiments were
performed to measure intensity from a plume and this could be
referred for the case with e 0. From Fig. 13, we could see
that the spectral intensity of gas mixture is obviously in two
special bands:4.3474.545 lm (22002300 cm1) band and
4.1674.237 lm (23602400 cm1) band. With different
exhaust systems and different operating conditions, the two
results cannot be identical. But they agree qualitatively in the
corresponding emissive bands, which demonstrates the right-
fulness and reliability of the code.
In order to investigate the effects of atmospheric transmis-
sivity in absorbing IR radiation, Fig. 14 shows the local atmo-
spheric transmissivity and non-dimensionalised plume
intensity on the same axis. It can be seen that there are over-
laps between the plume emission spectra and atmosphere
absorption window. This is mainly because both ambient air
and hot plume have the same participating species, i.e. carbon
dioxide and water vapor. But due to higher temperature, pres-
sure and species mole fraction in the plume, emissive bands of
the participating species are broadened and atmosphere is rel-
atively transparent to the emission in extra broadened band.
The emissive band is partially attenuated (Fig. 14), the right
band changes to 4.404.55 lm (2197.02272.5 cm1), and the
left one is 4.144.18 lm (2392.52412.5 cm1). With longer
Y. Zhou et al.
Fig. 14 Plume spectral intensity attenuated by atmosphere.
absorption length or higher participating species concentration
in atmosphere, the current change trend would continue until
the signature peak is totally attenuated.
3.3. Role of gas mixture and solid walls in radiation process
Under practical working conditions, radiation emitted by
gases and walls would interact with each other, and solid walls
absorb and reflect gas radiation while its emissive intensity
provides the initial energy for gas radiation transfer. It is
assumed that the walls are gray and diffuse with e 0:8, the
coupling radiation of solid walls and gas mixture is calculated.
Taking the same direction of observation in Section 3.2,
Fig. 15 demonstrates the intensity distribution of 35 lm band
and 814 lm band. We add the intensity of carbon dioxide and
water vapor in Figs. 11 and 12, and then intensity distribution
emitted by gas mixture e 0 is shown in Fig. 16.
In comparison, emission from solid walls enhanced both
bands radiation, including maximum value as well as high
intensity area. However, compared with 35 lm band, increase
brought by solid walls in 814 lm band is much more signifi-
cant: radiative intensity with and without walls emission even
have different orders of magnitude. Taking another line of
sight u 90 ; h 0 , in which direction high temperature
components are sheltered by the nozzle external surface, the
intensity distribution with and without walls emission is
shown in Figs. 17 and 18. No obvious difference is shown
for outer plume emission in Figs. 17(a) and 18(a), which
demonstrates the effects of participating gases for IR signature
in 35 lm band. By contrast, 814 lm bands intensity from
the nozzle external surface is much stronger than emission
A new model to simulate infrared radiation 661

Fig. 15 Intensity distribution emitted by gas mixture and solid walls (u = 90, h = 25.7).

Fig. 16 Intensity distribution emitted by gas mixture (u = 90, h = 25.7).

Fig. 17 Intensity distribution emitted by gas mixture and solid walls with walls emission (u = 90, h = 0).

Fig. 18 Intensity distribution emitted by gas mixture without walls emission (u = 90, h = 0).

from plume which has a much higher temperature, as shown in 4. Conclusions

Figs. 17(b) and 18(b). This can be attributed to two reasons:
firstly, according to Wiens displacement law, the peak emis-
sion wavenumber of solid walls with environment temperature
(300 K) is located around 10.0 lm, which is located in band of
814 lm; secondly, even though the plumes temperature is
much higher, absorption effects of carbon dioxide and
water vapor in 814 lm band is still not as strong as that in
35 lm band. Through comprehensive comparison from
Figs. 1518, it can be concluded that solid walls played a dom-
inant role in both 35 lm and 814 lm bands while the
plumes radiation is mainly in 35 lm band.
A numerical study is carried out to predict the IR signature
from an engine exhaust system. Combining the k-distribution
and multi-scale method in narrow band, we develop a
MSNBCK model to calculate radiative properties of gas mix-
ture. According to the relevant analysis of numerical results,
several conclusions are listed as follows:
(1) Compared with statistically uncorrelated assumption to
treat radiation in overlapping bands, the MSNBCK
model is constructed based on the physical interaction
662
effects between different participating species. Accord-
ing to LBL and experimental data in reference, the
new model shows satisfactory accuracy, saving tremen-
dous computation time through converting the exponen-
tial increase of computation time with the number of
participating species to linear increase. In addition, the
new model provides a solution to investigate the individ-
ual species contribution to the total intensity of gas
mixture.
(2) Compared with emission in 35 lm band, the plumes
radiative effects in 814 lm band is too weak to make
an influence. No matter the interested band is 35 lm
or 814 lm, carbon dioxide always plays a more impor-
tant role than water vapor, which provides a theoretical
conclusion that less carbon content in aviation kerosene
helps to reduce infrared signature of aircraft.
(3) IR emission from solid walls always greatly enhances the
exhaust systems radiation for both 35 lm and 814 lm
bands since its spectral intensity is more intense and con-
tinuous. Especially, the walls IR makes a significant dif-
ference for 814 lm band because both nozzle external
surface and hot parts of engine show effective emission
in this band.
References
1. Modest MF. Radiative heat transfer. San Diego: Academic Press;
2013. p. 3256.
2. Modest MF, Haworth DC. Radiative heat transfer in turbulent
combustion systems: theory and applications. New York: Springer
Press; 2016.
3. Denison MK, Webb BW. An absorption-line blackbody distribu-
tion function for efficient calculation of total gas radiative
transfer. J Quant Spectrosc Radiat Transf 1993;50(5):499510.
4. Chu H, Liu F, Consalvi JL. Relationship between the spectral line
based weighted-sum-of-gray-gases model and the full spectrum k-
distribution model. J Quant Spectrosc Radiat Transf 2014;143
(8):11120.
5. Doner N, Selcuk N. An application of spectral line-based weighted
sum of grey gases (SLW) model with geometric optics approxi-
mation for radiative heat transfer in 3-D participating media. Appl
Therm Eng 2013;50(1):8993.
6. Denison MK, Webb BW. The spectral-line-based weighted-sum-
of-gray-gases model in non-isothermal non-homogeneous media. J
Heat Transf 1995;117(2):35965.
7. Pierrot L, Soufiani A, Taine J. Accuracy of narrow-band and
global models for radiative transfer in H2O, CO2, and H2O-CO2-
mixture at high temperature. J Quant Spectrosc Radiat Transf
1999;62(5):52348.
8. Modest MF, Zhang H. The full-spectrum correlate-d-k distribu-
tion for thermal radiation from molecular gas-particulate mixture.
J Heat Transf 2002;124(1):308.
9. Pal G, Modest MF, Wang L. Hybrid full-spectrum correlated k-
distribution method for radiative transfer in strongly non-homo-
geneous gas mixture. J Quant Spectrosc Radiat Transf 2008;130
(8):27018.
10. Levy Y, Lev M, Ovcharenko V. Study of infrared CO2 radiation
from liquid-fueled combustor. J Heat Transf 2005;128(5):47883.
11. Hu HY, Wang Q. A numerical simulation on the infrared
radiation of hot exhausting nozzles with a coupled flow and heat
transfer model. Sci China (Technol Sci) 2010;53(10):2699707.
12. Wang AQ, Modest MF. High accuracy, compact database of
narrow-band k-distributions for water vapor and carbon dioxide.
J Quant Spectrosc Radiat Transf 2005;93(1):24561.
Y. Zhou et al.
13. Hiers RS, Cromwell III BK, Zaccardi VA. Compatibility of
infrared band models with scattering. J Thermophys Heat Transf
1994;8(2):20815.
14. Liu F, Smallwood GJ, Oslash, Guacute M, lder. Application of
statistical narrowband model to three-dimensional absorbing-
emitting-scattering media. J Thermophys Heat Transf 1999;13
(3):28591.
15. Lacis AA, Oinas VA. Description of the correlated-k distribution
method for modeling non-gray gaseous absorption, thermal
emission, and multiple scattering in vertically inhomogeneous
atmospheres. J Geophys Res Atmos 1991;96(5):902763.
16. Modest MF. Narrow-band and full-spectrum k-distributions for
radiative heat transfer correlated-k vs. scaling approximation. J
Quant Spectrosc Radiat Transf 2002;76(1):6983.
17. Faycal BN, Akram M, Kamel C. Application of the statistical
narrow-band correlated-k model to entropy generation through
non-grey gas radiation inside a spherical enclosure. Int J Exergy
2011;8(2):12847.
18. Solovjov VP, Webb BW. SLW modeling of radiative transfer in
multicomponent gas mixtures. J Quant Spectrosc Radiat Transf
2000;65(4):65572.
19. Modest MF, Riazz RJ. Assembly full spectrum k-distribution
from a narrow band database: effects of mixing gases, gases and
non-gray absorbing particles and non-gray scatters in non-gray
enclosures. J Quant Spectrosc Radiat Transf 2005;90(2):16989.
20. Zhang HM, Modest MF. A multi-scale full-spectrum correlated-k
distribution for radiative heat transfer in inhomogeneous gas
mixture. J Quant Spectrosc Radiat Transf 2002;73(25):34960.
21. Wang LY, Modest MF. Narrow-band based multiscale full-
spectrum k-distribution method for radiative transfer in inhomo-
geneous gas mixture. J Heat Transf 2005;127(7):7408.
22. Gopalendu P, Modest MF. A narrow band-based multiscale
multigroup full-spectrum k-distribution method for radiative
transfer in non-homogeneous gas-soot mixture. J Heat Transf
2010;132(2):291303.
23. Roger M, Crouseilles N. A dynamic multi-scale model for
transient radiative transfer calculations. J Quant Spectrosc Radiat
Transf 2013;116(1):11021.
24. Modest MF, Bharadwaj SP. Medium resolution transmission
measurements of CO2 at high temperature. J Quant Spectrosc
Radiat Transf 2002;73(25):32938.
25. Modest MF, Bharadwaj SP. Medium resolution transmission
measurements of CO2 at high temperature-an update. J Quant
Spectrosc Radiat Transf 2007;103(1):14655.
26. Bharadwaj SP, Modest MF, Riazzi RJ. Medium resolution
transmission measurements of water vapor at high temperature.
J Heat Transf 2006;128(4):37481.
27. Bharadwaj SP. Medium resolution transmission measurements of
CO2 and H2O at high temperature and a multiscale Malkmus
model for treatment of inhomogeneous gas paths [dissertation].
Philadelphia: Pennsylvania State University; 2005.
28. Edwards DK, Balakrishnan A. Thermal radiation by combustion
gases. Int J Heat Mass Transf 1973;16(1):2540.
29. Fu Q, Liou KN. On the correlated k-distribution method for
radiative transfer in nonhomogeneous atmosphere. J Atmos Sci
1992;49(22):213956.
30. Zhang H, Modest MF. Scalable multi-group full-spectrum corre-
lated-k distributions for radiative transfer calculations. J Heat
Transf 2003;125(3):45461.
31. Liu F, Smallwood GJ, Omer LG. Application of the statistical
narrow-band correlated- k method to non-grey gas radiation in
CO2H2O mixtures: approximate treatments of overlapping
bands. J Quant Spectrosc Radiat Transf 2001;68(4):40117.
32. Shan Y, Zhang JZ. Numerical investigation of flow mixture
enhancement and infrared radiation shield by lobed forced mixer.
Appl Therm Eng 2009;29(17):368795.