See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/245363171

Numerical Determination of Radiative View Factors Using Ray Tracing

Article in Journal of Heat Transfer · July 2010

DOI: 10.1115/1.4000974

CITATIONS READS

16 428

3 authors, including:

Shicheng Xue
The University of Sydney
41 PUBLICATIONS 649 CITATIONS

SEE PROFILE

All content following this page was uploaded by Shicheng Xue on 02 August 2015.

The user has requested enhancement of the downloaded file.


T. Walker
S.-C. Xue
G. W. Barton
e-mail: gbarton@usyd.edu.au
School of Chemical and Biomolecular
Engineering,
University of Sydney,
New South Wales 2006, Australia
Numerical Determination of
Radiative View Factors Using Ray
Tracing
A ray-tracing method is presented for numerically determining radiative view factors in
complex three-dimensional geometries. This method uses a set of “primitive” shapes to
approximate the required geometry together with a Monte Carlo simulation to track the
fate of randomized rays leaving each surface. View factors were calculated for an op-
erational fiber drawing furnace using both numerical integration and ray-tracing meth-
ods. Calculated view factor profiles were essentially identical above a ray density of 105
per unit area. Run times for the ray-tracing method were considerable longer, although
the setup time to describe a new geometry is very short and essentially independent of
system complexity. 关DOI: 10.1115/1.4000974兴

Keywords: radiative heat transfer, view factor, ray tracing, Monte Carlo method, poly-
mer optical fiber

1 Introduction 共i.e., nondeforming兲 preform and used in a heat transfer model of

Figure 1 shows an arbitrarily orientated surface Ai having a
constant total hemispherical emissivity ␧i and temperature Ti. Of
the total radiant energy leaving this surface in all directions
共␧i␴0T4i Ai, where ␴0 is the Stefan–Boltzmann constant兲, only a
fraction will reach a second arbitrarily orientated surface A j. This
fraction 共Fi-j兲 is defined as the view factor, a term that is consis-
tent with the fact that its value is dependent only on the geometri-
cal orientation of the two participating surfaces. A view factor can
thus be expressed 共and calculated兲 as follows:
the drawing furnace with close agreement being obtained between
simulated and experimental 共measured at the preform surface and
center兲 temperature profiles. These view factors are presented later
to confirm the accuracy of results obtained using an alternative
ray-tracing method.
The remainder of this paper is thus structured as follows. Sec-
tion 2 presents the concept of ray-tracing and how “primitives”
can be used to calculate view factors. The tapered cylinder primi-
tive is presented in detail, as this is used in calculating view fac-
tors for the case where a cylindrical preform is drawn to fiber.

冕 冉冕 冊
1 cos ␪i cos ␪ j
Fi-j = dA j dAi 共1兲
Ai ␲Si-j
2 numerical integration and ray tracing are used, while in the latter
where dAi and dA j are differential elements within the two sur-
faces with Si-j being the length of the connecting line between the
two, while ␪i and ␪ j are the respective angles between the con-
necting line and the normal vectors 共ni and n j兲 to each surface.
Evaluation of view factors is straightforward provided that Si-j,
cos ␪i, and cos ␪ j can be expressed in terms of the geometrical
parameters that define the two participating surfaces and the nec-
essary integration performed. For radiative exchange between an
infinitesimal and a finite surface, only a single area integration is
needed. However, if two finite surfaces are involved, then a
double area integral is required. Because of the importance of
radiative heat transfer in a wide variety of applications, compila-
tions of analytical or tabulated results 共often in terms of dimen-
sionless geometrical parameters兲 are available in the literature 关1兴.
In some cases, an unknown view factor can be generated from
known factors by making use of view factor algebra 关2兴.
The present work was motivated by the need to calculate the
radiative heat transfer within an operational furnace used to heat a
polymer preform sufficiently to be drawn to fiber. Here the avail-
able view factor compilations were of limited use due to the geo-
metric complexity of our drawing furnace, and the fact that once
the glass transition temperature is reached then the shape of the
deforming preform becomes a strong function of the thermal con-
ditions within the furnace. Initially all view factors were deter-
mined using numerical integration for the case of a cylindrical
Contributed by the Heat Transfer Division of ASME for publication in the JOUR-
HEAT TRANSFER. Manuscript received May 31, 2009; final manuscript received
NAL OF
November 26, 2009; published online April 28, 2010. Assoc. Editor: He-Ping Tan.
Section 3 presents view factor results for both the nondeforming
preform and the tapering preform cases. In the former case, both
only ray tracing is employed. This section also compares the rela-
tive merits of the two methods.
2 Ray Tracing
Ray tracing is a widely used technique in optics whereby the
path of a photon or “ray” is followed through a three-dimensional
共3D兲 environment from its point of origin to its final destination.
This technique has been employed to solve diverse problems from
the generation of images with a high degree of photorealism 关3,4兴
to complex radiative heat transfer modeling 关5–10兴. In the present
context, ray tracing may be adapted to calculate the view factor
between objects through its implementation within a Monte Carlo
simulation 关11兴. Essentially a large number of rays are “fired”
from random points on a given object at random angles into the
3D environment with the first object that each ray intersects being
recorded. The view factor between any two objects may be de-
fined as the fraction of rays leaving one object that reaches the
other.
Accurate view factor estimation here requires that careful atten-
tion be paid to a number of computational issues, the two most
important being the distribution of 共random兲 points on the emit-
ting surface and their 共random兲 emission into the 3D environment.
The major strength of ray tracing as a tool for numerically
estimating view factors is that its application is essentially inde-
pendent of the complexity of the 3D geometry that defines the
system. By comparison, alternatives such as numerical integration
become increasingly challenging with geometrical complexity.

Journal of Heat Transfer Copyright © 2010 by ASME JULY 2010, Vol. 132 / 072702-1

Downloaded 29 Apr 2010 to 129.78.64.102. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 1 Configuration for radiative heat exchange between two
finite surfaces
In passing, it should be noted that although the focus of this
paper is on the use of ray tracing in view factor calculation, the
technique can be conceptually expanded to calculate net radiative
heat transfer within any complex 3D enclosure 关12兴. Here each
ray emitted from a surface is taken as representing a “packet” of
thermal energy that is either absorbed 共and possibly re-emitted兲 or
reflected when impinging on a second surface. However, it must
be said that our 共unpublished兲 case studies have shown this to be
a computationally demanding approach.
2.1 Use of Primitives in Ray Tracing. Ray tracing is not a
new approach for calculating view factors with a literature dating
back some 40 years 关13,14兴. However, most approaches advocate
that any object within a system be described via a mesh which
maps to its surface 关11,15,16兴. While this approach reduces the
complexity of calculating suitable ray launch positions and launch
directions, it can introduce representational errors in terms of ad-
equately modeling an object’s curvature and requires a robust dis-
cretization algorithm.
A more elegant approach is to approximate any 3D geometry as
a composite created from a modest set of simpler generic shapes
共e.g., cylinder, cone, sphere, and plane兲. Each such primitive is
described not in terms of a meshing arrangement but rather by a
set of surface equations. Any primitive may be manipulated via
affine transformations that allow it to be scaled, rotated, and trans-
posed in 3D space to provide an infinite set of options while
requiring a computational method for just a single shape. Thus
any complex 3D geometry may be constructed as a collection of
affine transformed primitives with such an approach being widely
used in the computer-based multimedia industry as previously
noted 关3,4兴.
Intuitively, it might be expected that rays are created and traced
in “world space,” that is, the space where an actual object is
approximated by a composite of 共affine transformed兲 3D primi-
tives. However, such an approach increases the complexity of the
ray-tracing implementation as it necessitates calculation of each
ray’s starting point and directional vector within the transformed
space. It is computationally far more efficient to calculate a ray’s
starting point and launch direction in the primitive space, and
subsequently transform the ray to perform the appropriate calcu-
lations. This ray transformation procedure is shown schematically
in Fig. 2 for the case where the objective is to determine whether
a ray from one object 共A兲 intersects with a second object 共B兲.
In this way, it is possible to describe a wide variety of 3D
geometries in terms of a small number of primitive shapes and to
use these compact descriptions to calculate the required view fac-
tors. For example, a polymer preform that is being drawn to fiber
072702-2 / Vol. 132, JULY 2010
Downloaded 29 Apr 2010 to 129.78.64.102. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
Fig. 2 Procedure for transforming a ray to find ray-object
intersection
can be described in terms of a stack of tapered cylinders. The
description of this primitive is thus considered next.
2.2 Tapered Cylinder Primitive. The primitive here is a cyl-
inder with a base radius of 1 and a variable upper radius. As such,
it can be used to model a wide range of shapes from a cylinder,
through a conical frustum, to a cone. The “surface equation” for
this primitive when its base is centered on the origin in the x-y
plane and it extends along the z axis to a height of 1 is as follows:
x2 + y 2 − 共1 + 共s − 1兲z兲2 = 0 共2兲
where s is the radius of the upper surface such that 0 ⱕ s ⱕ 1.
When ray tracing is used to calculate view factors, it is essential
to ensure an even distribution of ray starting points from the sur-
face. For a simple cylinder 共for which s = 1兲, ray starting points
can be defined in terms of the azimuthal angle 共␾兲 as follows:
z=␹ 共3兲
x = rz cos共␾兲 共4兲
y = rz sin共␾兲 共5兲
where ␹ is a random number between 0 and 1, rz is the radius of
the cylinder at z 共here rz = 1兲, and ␾ is a random angle between 0
and 2p.
However, for a tapered cylinder, it is no longer possible to
simply use a random number to represent the z coordinate but
rather a random surface area fraction 共0 ⱕ A f ⱕ 1兲 is selected. As
the base radius and height for this primitive are both unity, A f can
be expressed as the following function of z:
z共1 + rz兲
Af = 共6兲
共1 + s兲
where rz = 1 + 共s − 1兲z is the radius of the tapered cylinder at the
point z.
Substituting for rz in Eq. 共6兲 gives a quadratic equation whose
positive solution gives a random z coordinate distribution which
reflects the changing surface area of the tapered cylinder:
冑1 − A f + A2f − 1
z= 共7兲
s−1
The coordinates defining the starting point of any ray on this
tapered cylinder are thus
rz = 1 + 共s − 1兲z 共8兲
x = rz cos共␾兲 共9兲
Transactions of the ASME

Fig. 3 Schematic diagram of fiber drawing furnace
y = rz sin共␾兲 共10兲
The second issue is calculation of the ray directional vector, which
is achieved by assuming azimuthal symmetry and calculating a
hemispherical distribution as suggested by Farmer and Howell
关12兴. This distribution is rotated in such a way that it lies parallel
to the surface of the tapered cylinder. The angle 共␪兲 by which the
directional distribution needs to be rotated around the x-axis is
simply given by
␪ = tan−1共1 − s兲 共11兲
Applying this transformation and rotating around the z-axis by the
random angle ␾, a ray’s directional vector may be determined as
follows:
冤 冥冤 冥
xdir x cos共␾兲 − sin共␾兲 ⫻ 共y cos共␪兲 − z sin共␪兲兲
y dir = x sin共␾兲 + cos共␾兲 ⫻ 共y cos共␪兲 − z sin共␪兲兲 共12兲
zdir y sin共␪兲 − z cos共␪兲
It should be noted that this method generates a uniform directional
distribution and thus is unsuitable for cases involving materials
whose emissive characteristics deviate significantly from the un-
derlying gray body assumption.
Using a ray’s starting point and directional vector, as defined by
Eqs. 共9兲, 共10兲, and 共12兲, ensures that the calculated view factors
are unbiased by any shape change, which in the present case
would occur as the preform tapers and is drawn to fiber.
In applying the ray-tracing approach, a key computational pa-
rameter is the ray density ␳ 共i.e., the rays launched per unit area兲
to be used. Once this is set, the total number of rays used is simply
the product of the surface area of the transformed 共into “world
space”兲 tapered cylinder and the chosen ray density. Noting that
rotation and translation do not affect surface area, the following
formula can be used to give the surface area of a tapered cylinder
that has been transformed by scaling factors a and b in the x and
y directions, respectively:
冑
␲共1 + s兲 s共a2 + b2兲
A= 共13兲
s 2
3 View Factor Results for Fiber-Drawing Furnace
In order to evaluate the performance of ray tracing, it was used
to determine the view factors needed in modeling the heat transfer
within a fiber drawing furnace. Figure 3 is a schematic represen-
tation of an operational furnace used to fabricate microstructured
polymer optical fibers 共mPOFs兲. Such fibers owe their light-
guiding properties to a pattern of holes that runs the length of the
fiber. Here the required pattern is created by drilling holes into a
monolithic block referred to as the preform. This is then drawn to
Journal of Heat Transfer
Downloaded 29 Apr 2010 to 129.78.64.102. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
fiber in a relatively low temperature furnace where heat transfer to
the preform occurs by both convection and radiation. A critical
issue in mPOF fabrication is the extent of deformation undergone
by the hole structure as the preform is heated 共above the poly-
mer’s glass transition temperature兲 and drawn to fiber. Details on
mPOF fabrication are provided in the recent literature 关17兴.
In Fig. 3, the preform enters at the top of the furnace through an
adjustable iris while drawn fiber leaves at the bottom through a
second iris. Six external halogen lamps 共under on-off control兲 pro-
vide radiant heating through a central quartz “hot-zone” window
with the rest of the furnace being well insulated. The preform
surface is heated by thermal radiation 共both through the quartz
window and by reradiation from the furnace walls兲 and by con-
vection induced within the furnace. In the present view factor
determination context, two cases were considered. In the first, the
furnace was operated such that the preform never exceeded its
glass transition temperature and thus never underwent any neck-
down in diameter. Here all relevant view factors were determined
by numerical integration of Eq. 共1兲. As noted, very good agree-
ment was obtained between experimental temperature measure-
ments and results from a fully conjugate heat transfer model
where convective heat transfer was estimated using commercial
computational fluid dynamics software.
In the second case considered 共for which experimental mea-
surements were not available兲, it was assumed that the preform
was raised above its glass transition temperature and that the re-
sultant preform neckdown could be described by the “slender
body” approximation 关18兴. Here the preform radius 共r兲 as a func-
tion of distance from the furnace entrance 共z兲 is given by
r = R exp 冉 z ln共Dr兲
2H
冊 共14兲
where Dr is the draw ratio 共set by furnace operating conditions兲, H
is the height of the furnace 共0.18 m兲, and R is the initial preform
radius 共0.006 m兲.
In the latter case, determining all relevant view factors by nu-
merical integration would be a time-consuming exercise. How-
ever, calculating these view factors using ray tracing is scarcely
more involved than for the case where the preform diameter re-
mains unchanged. The two Dr values chosen 共9 and 900兲 gave exit
to inlet radius ratios 共i.e., r / R兲 that ranged from a modest to a
substantial preform neckdown 共33.33% and 3.33%, respectively兲.
3.1 Constant Diameter Preform. Although not required by
this implementation of ray tracing, the preform and furnace walls
were each discretized using 180 “slices” to allow a direct com-
parison of the view factors with those obtained via numerical
integration 共which used this level of discretization to ensure that
the results were grid independent兲. Additionally, ray densities 共␳兲
of 104 and 105 per unit area were used to assess the impact of this
key parameter on the ray-tracing results. As expected, increasing
the ray density improved the “accuracy” of the ray-tracing results
with excellent agreement between the two methods being
achieved at the higher ␳ value.
Results are given in Figs. 4–6 where in each case the view
factors calculated by numerical integration and ray tracing are
shown as solid lines and discrete symbols, respectively. Although
all possible view factors for this furnace enclosure were deter-
mined, values are only given here between the preform surface
and the furnace wall 共observed from three different axial positions
z = 0.0095, 0.0895, and 0.1695 of one surface to the entire other
surface or to itself兲 to provide a direct comparison between the
two methods. Here F p-f refers to the view factor from a position
on the preform surface to the furnace wall, F f-f refers to the view
factor from a position on the furnace wall to itself, and so on. The
z values chosen show the essential geometric symmetry of a sys-
tem that comprises a constant diameter cylinder located along
the axis of a cylindrical enclosure. Under conditions where the
JULY 2010, Vol. 132 / 072702-3

Fig. 4 View factors from three positions „zp = 0.0095,
0.0895, 0.1695… on the preform to the furnace wall „␳ = 104 per
unit area; cylindrical preform…
preform undergoes neckdown, however, this symmetry is broken,
as discussed in Sec. 3.2.
The accuracy of the view factor calculations is implied by both
the close agreement between all Fi-j profiles obtained from the
numerical integration and ray-tracing methods and the “sum rule”
where the total of the view factors for each surface within the
enclosure is essentially unity. Using these view factors, it is pos-
sible to determine the radiative heat transfer, using the net radia-
tion method 关2,19兴, between a stationary, finite length, solid pre-
form, and the furnace surfaces where the heating wall is at a
uniform temperature, and the top and bottom irises are either in-
sulated or treated as black surfaces. For this case, if the aspect
ratio of the cylindrical preform is large enough, the net radiative
heat flux profile over the central section of the preform surface
should approximate that for the limiting case of two infinitely long
concentric cylinders. Figure 7 shows the results when a stationary
polymethylmethacrylate 共PMMA兲 preform 共R = 6 mm with a con-
stant hemispherical total emissivity e p of 0.90; initially at a uni-
form temperature of 293 K兲 is heated inside a furnace 共with a
Fig. 5 View factors from three positions „zp = 0.0095,
0.0895, 0.1695… on the preform to the furnace wall „␳ = 105 per
unit area; cylindrical preform…
072702-4 / Vol. 132, JULY 2010
Downloaded 29 Apr 2010 to 129.78.64.102. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
Fig. 6 View factors from three positions „zf = 0.0095,
0.0895, 0.1695… on the furnace wall to the preform surface „left…
and itself „right… „␳ = 105 per unit area; cylindrical preform…
constant emissivity ew of 0.885, a height H of 180 mm, and a
radius Rw of 32 mm; uniform wall temperature Tw of 373 K兲. Note
that here the net heat flux for the preform has been scaled by the
emitted flux qb from a black surface with the same uniform wall
temperature, that is, by qb = ␴0T4w, while the axial length has been
scaled by H. As expected with an aspect ratio for the cylindrical
preform of 30, when both irises are treated as insulated surfaces,
the net radiative heat flux over the middle portion of the preform
surface is consistent with the analytical result for two infinitely
long concentric cylinders. However, if both irises are treated as
black surfaces, then, due to heat lose thought the irises, the overall
net radiative heat flux to the preform surface is reduced along the
length of the preform, with the reduction becoming more marked
when approaching the ends of the preform.
3.2 Reducing Diameter Preform. A constant diameter pre-
form may be thought of as having a draw ration 共Dr兲 of unity. As
Dr is increased, so the amount of preform neckdown increases.
View factors within the drawing furnace were calculated for two
Dr values 共9 and 900兲 using a ray density of 105 per unit area. The
Fig. 7 Net radiative heat flux profiles along a stationary cylin-
drical preform within a furnace with a uniform heating wall tem-
perature; two different thermal boundary conditions are used
for the irises
Transactions of the ASME

Fig. 8 View factors from three positions „zp = 0.0095,
0.0895, 0.1695… on the preform to the furnace wall „␳ = 105 per
unit area; Dr = 9…
profiles presented in Figs. 8–11 correspond to those given in Figs.
5 and 6 for the constant diameter case.
Figures 5, 8, and 9 show the view factor profiles from the same
positions on the preform surface to the furnace wall for three
different Dr values 共1, 9, and 900兲. The amount of preform thin-
ning increases with Dr and the form of the view factor profile
changes as the system geometry becomes less longitudinally sym-
metrical. Similar changes with preform thinning are apparent in
Figs. 6, 10, and 11, where view factor profiles from three different
positions on the furnace wall to both the preform surface 共F f-p兲
and itself 共F f-f 兲 are given.
3.3 Computational Efficiency. It took ⬃5 s 共using an AMD
Althon 64 processor, 2.21 GHz, 2.0 Gbytes of RAM兲 to perform a
complete set of view factor calculations for the draw furnace with
a constant diameter preform using numerical integration at a level
of discretization necessary to ensure grid independence. By com-
parison, the view factor profiles for a slender body preform with a
draw ratio of 900 took 44.2 h when run on a comparable computer
Fig. 9 View factors from three positions „zp = 0.0095,
0.0895, 0.1695… on the preform to the furnace wall „␳ = 105 per
unit area; Dr = 900….
Journal of Heat Transfer
Downloaded 29 Apr 2010 to 129.78.64.102. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
Fig. 10 View factors from three positions „zf = 0.0095,
0.0895, 0.1695… on the furnace wall to the preform and to itself
„␳ = 105 per unit area; Dr = 9…
and a ray density 共105 per unit area兲 that provided acceptable
accuracy. This difference must, however, be weighted against the
considerable time 共several man months兲 taken in carrying out the
necessary analysis and coding for the numerical integration case
关20兴, the fact that it was not felt to be cost effective to extend this
approach to the general deforming preform case, and the very
modest time taken to set up a new case using the ray-tracing
approach. No attempt was made in this study to speed up the
ray-tracing calculations using either faster hardware or more so-
phisticated numerical techniques. In the latter category, con-
strained maximum likelihood smoothing appears to have promise
关21兴, although its implementation would result in a significant
nonlinear programming problem, somewhat at odds with the rela-
tive simplicity of the coding required for ray tracing built around
the use of primitives.
Finally, it should be noted that if numerical integration is used,
then the extent of the geometric discretization is primarily set by
the need to ensure that the computed view factors are grid inde-
pendent. By comparison, the discretization used in ray tracing is
set simply by the resolution required in any given view factor
profile. Thus in the present calculations, 180 cylindrical slices
Fig. 11 View factors from three positions „zf = 0.0095,
0.0895, 0.1695… on the furnace wall to the preform and to itself
„␳ = 105 per unit area; Dr = 900…
JULY 2010, Vol. 132 / 072702-5
 共either constant or tapered兲 were used to describe the preform
simply for ease of comparison with the numerical integration re-
sults.
4 Conclusions
The determination of view factors for complex 3D geometries
can be a challenge. In some cases, especially those that exhibit a
measure of geometric symmetry, it is possible to employ numeri-
cal integration techniques, although considerable analysis is gen-
erally required and care needs to be taken in the discretization
employed. Ray tracing is an alternative numerical approach that
does not rely on any detailed geometric analysis. In the version
employed in this study, 3D geometries are constructed using a set
of primitive shapes. A subsequent Monte Carlo simulation 共in
which the “fate” of randomized rays leaving a surface are col-
lated兲 is used to determine the required view factors. Although run
times for ray-tracing calculations are significantly longer than for
numerical integration, the “setup” time for a problem using ray-
tracing is essentially independent of geometric complexity, mak-
ing this a highly attractive approach for generic view factor soft-
ware.
References
关1兴 Howell, J. R., 2008, “A Catalog of Radiation Heat Transfer Configuration
Factors,” 2nd ed., http://www.me.utexas.edu/~howell/index.html
关2兴 Siegel, R., and Howell, J. R., 2001, Thermal Radiation Heat Transfer, 4th ed.,
Taylor & Francis, New York.
关3兴 Hill, F. S., 2001, Computer Graphics Using Open GL, 2nd ed., P. Hall, ed.,
Prentice-Hall, Englewood Cliffs, NJ.
关4兴 Shirley, P., Sung, K., Brunvand, E., Davis, A., Parker, S., and Boulos, S., 2008,
“Fast Ray-Tracing and the Potential on Graphics and Gaming Courses,” Com-
put. Graph., 32, pp. 260–267.
关5兴 Farmer, J., 1995, “Improved Algorithms for Monte Carlo Analysis of Radiative
Heat Transfer in Complex Participating Medium,” Ph.D. thesis, University of
Texas, Austin, TX.
关6兴 Wang, A., and Modest, M. F., 2006, “Photon Monte Carlo Simulation for
Radiative Transfer in Gaseous Media Represented by Discrete Particle Fields,”
ASME J. Heat Transfer, 128, pp. 1041–1049.
072702-6 / Vol. 132, JULY 2010
Downloaded 29 Apr 2010 to 129.78.64.102. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
View publication stats
关7兴 Steward, F. R., and Cannon, P., 1971, “The Calculation of Radiative Flux in a
Cylindrical Furnace Using the Monte Carlo Method,” Int. J. Heat Mass Trans-
fer, 14, pp. 245–262.
关8兴 Fan, T., and Fedorov, A. G., 2002, “Apparent Radiative Properties and Radia-
tion Scattering by a Semi-Transparent Hemispherical Shell,” ASME J. Heat
Transfer, 124, pp. 1088–1094.
关9兴 Hu, L., Schmidt, A., Narayanaswamy, A., and Chen, G., 2004, “Effect of
Periodic Structures on Coherence Properties of Blackbody Radiation,” ASME
J. Heat Transfer, 126, pp. 786–792.
关10兴 Xia, X. L., Shuai, Y., and Tan, H. P., 2005, “Calculation Techniques With the
Monte Carlo Method in Stray Radiation Evaluation,” J. Quant. Spectrosc. Ra-
diat. Transf., 95, pp. 101–111.
关11兴 Vujičić, M. R., Lavery, N. P., and Brown, S. G. R., 2006, “View Factor
Method Using the Monte Carlo Method and View Factor Sensitivity,” Com-
mun. Numer. Methods Eng., 22, pp. 197–203.
关12兴 Farmer, J. T., and Howell, J. R., 1998, “Comparison of Monte Carlo Strategies
for Radiative Transfer in Participating Media,” Adv. Heat Transfer, 31, pp.
333–429.
关13兴 Howell, J. R., 1968, “Application of Monte Carlo to Heat Transfer Problems,”
Adv. Heat Transfer, 5, pp. 1–52.
关14兴 Campbell, P. M., 1967, “Monte Carlo Method for Radiative Transfer,” Int. J.
Heat Mass Transfer, 10, pp. 519–527.
关15兴 Vueghs, P., de Koning, H. P., Pin, O., and Beckers, P., 2009, “Use of Geometry
in Finite Element Thermal Radiation Combined With Ray Tracing,” J. Com-
put. Appl. Math. 共in press兲.
关16兴 Sakai, S., and Maruyama, S., 2003, “A Fast Approximated Method for Radia-
tive Exchange for Combined Heat Transfer Simulation,” Numer. Heat Trans-
fer, Part B, 44, pp. 473–487.
关17兴 Large, M. C. J., Poladian, L., Barton, G. W., and van Eijkelenborg, M. A.,
2008, Microstructured Polymer Optical Fiber, Springer, New York.
关18兴 Xue, S.-C., Tanner, R. I., Barton, G. W., Lwin, R., Large, M. C. J., and
Poladian, L., 2005, “Fabrication of Microstructured Optical Fibers, Part I:
Problem Formulation and Numerical Modeling of Transient Draw Process,” J.
Lightwave Technol., 23共7兲, pp. 2245–2254.
关19兴 Xue, S.-C., Poladian, L., Barton, G. W., and Large, M. C. J., 2006, “Radiative
Heat Transfer in Preforms for Microstructured Optical Fibres,” Int. J. Heat
Mass Transfer, 50共7–8兲, pp. 1569–1579.
关20兴 Lee, S. H.-K., and Jaluria, Y., 1995, “The Effects of Geometry and Tempera-
ture Variations on the Radiative Transport During Optical Fiber Drawing,” J.
Mater. Process. Manuf. Sci., 3, pp. 317–331.
关21兴 Daun, K. J., Morton, D. P., and Howell, J. R., 2005, “Smoothing Monte Carlo
Exchange Factors Through Constrained Maximum Likelihood Estimation,”
ASME J. Heat Transfer, 127, pp. 1124–1128.
Transactions of the ASME