IMA Journal of Mathematics Applied in Business & Industry (1996) 7, 109-116

An accurate program for radiation modelling in the design of

high-temperature furnaces
D. A. LAWSON
Coventry University

C . D . ZlESLER
British Gas pic

Downloaded from http://imaman.oxfordjournals.org/ at University of Victoria on June 4, 2015

The zone method is well established for determining radiative heat transfer. It has
been used successfully in simple geometries. However the task of determining the
required exchange areas has limited its use in more geometrically complex enclosures.
In this paper, a computer program, RADEX, is described. This program will calculate
direct and total exchange areas using a Monte Carlo method for enclosures with
complex geometries. Some results from a validation study of RADEX are given. Finally,
as an illustration of its potency as a design tool, the results of a comparative study
of nine different designs of radiant tube furnace are given.

1. Introduction
There is a pressing need for practical simulations of heat transfer in high-temperature
industrial processes which can successfully model the mechanisms controlling thermal
performance. Among the benefits of such simulations is that they enable more effective
and efficient original plant designs to be produced in shorter time. They also have
the advantage that they allow the identification of improvements to existing plant
to be made more readily, helping the assessment and implementation of new
technologies, such as regenerative burners, or novel control methods, in suitable
applications.
Zone models have been used successfully in the field of metal reheating furnaces
for some time, since these furnaces are geometrically simple; by contrast, they have
not been applied with any great success in situations like potteries or kilns where
the radiative heat transfer is difficult to evaluate. This restriction has been due entirely
to the limitations of the available radiation submodels. The submodel is responsible
for calculating the exchange areas required by the zone model from the geometric
and thermophysical information. Provided that the radiation is accurately character-
ized, heat-transfer analyses can, in most cases, be performed with the use of relatively
simple models of the other physical processes.
RADEX is a new program developed for evaluating exchange areas in radiative heat
transfer. It was designed to be compatible with previously developed zone models.
This paper describes this radiation program, the principles upon which it is based,
and its application in a design study.
109
£ Oxford Univenily Preii 19%
110 D. A. LAWSON AND C. D. ZIESLER

2. The zone method and Monte Carlo integration

In the zone method, the enclosure walls and the enclosed space are sub-divided into
a number of 'zones' which are assumed to be isothermal and to have uniform
radiative properties. The radiative properties are assumed to be spectrally and
directionally independent. In order to represent the spectral nature of the gas when
combustion products are present, the so-called 'sum-of-grey-gases' model may be
used (Hottel & Sarofim 1967).
The direct-exchange area (DEA) between two 'elements' (an element may be a
surface area or a gas volume) is an important concept in the zone method. It may
be regarded as an effective area for direct radiative interchange between two elements.
Since some of the radiation incident upon an element is reflected, it is necessary to

Downloaded from http://imaman.oxfordjournals.org/ at University of Victoria on June 4, 2015

compute the total-exchange areas (TEAs) in order to evaluate the radiative inter-
change. TEAs are similar to DEAs but give a measure of the radiation emitted by
one element that is absorbed by another (possibly after being reflected at a number
of surfaces). The TEAs are related to the DEAs and can be found either by matrix
inversion (Noble 1975) or using a sole-emitter strategy (Chan 1987). Once the TEAs
are known, evaluating the radiative heat loss from an element is clearly the difference
between emission and absorption.
The bulk of the computation required by the zone method lies in the calculation
of the DEAs. These are evaluated from multidimensional integral formulae (Hottel
& Sarofim 1967). Analytical evaluation of these integrals is rarely possible, even when
the absorption coefficient of the gas is constant throughout the enclosure. When the
absorption coefficient varies (for example, if there is a flame in the enclosure),
analytical evaluation is impossible even for the simplest cases.
An appropriate numerical method for finding DEAs is Monte Carlo integration
which is based on randomly sampling the integrand. This method is particularly
useful for multidimensional integrals since, unlike standard methods, its accuracy
does not decrease as the dimensionality of the integrals increases (Gould &
Tobochnik 1988). A further advantage of Monte Carlo integration is that irregular
geometries require little extra computational effort.
DEAs may be calculated by using an adaptation of the basic Monte Carlo method.
A large number of bundles, or 'rays', are fired in randomly selected directions from
randomly selected points on or in the sending element. Each bundle is monitored
until it reaches a surface. The amount of energy absorbed by each gas volume
traversed is recorded, along with the amount of energy reaching the surface. Summing
over all the bundles and normalizing appropriately gives the DEAs.

3. RADEX: a program to calculate exchange areas

A radiation submodel, RADEX, has been written to enhance the applicability of
existing zone-model programs. One of its major strengths is the range of complexity
of geometries it can represent to any required level of detail. The boundaries of the
enclosure need not lie along cartesian grid lines (as is the case in Omori et al. (1990)).
The enclosure may contain any number of objects, such as radiant heating tubes and
(in the case of furnace applications) the load being heated.
 A PROGRAM FOR RADIATION MODELLING IN FURNACE DESIGN 111

ZONE 2

ZONE 1

BACK VWkLL

HEARTH

Downloaded from http://imaman.oxfordjournals.org/ at University of Victoria on June 4, 2015

FRONT WALL

FIG. 1. Domed-roof furnace. Geometric grid - -; heat-transfer grid •

A key concept in allowing geometrical accuracy while maintaining reasonable

computational times is the use of two grids: a fine (geometric) grid and a coarse
(heat-transfer) grid. The geometric grid permits the detailed representation of the
enclosure geometry. This may, however, produce a prohibitively large number of
elements for the heat-transfer calculations. The coarse grid is a subset of the fine grid.
It groups together fine grid elements to produce a smaller, manageable, number of
heat-transfer elements. The resulting surface elements may not be planar (which is
of no consequence to Hottel's method) but they maintain computational efficiency
without loss of accuracy. This can be illustrated by the example in Fig. 1. As the
shape of the domed roof may need to be accurately represented, RADEX allows the
curved edge of the cross-section to be approximated by a number of sloping line
segments (four are shown but more may be used if greater accuracy is required). It
is possible that, for the heat-transfer calculations, it is desired to divide the roof along
its length, but not across its width. The line segments necessary to capture the
geometric detail produce a fine grid with four divisions across the width of the
chamber. By restricting the coarse grid to just the first and last vertical fine grid
lines a single element across the width is produced. The geometric detail of the fine
grid is retained for use within the Monte Carlo integration.
RADEX allows complete generality of surface emissivity and gas absorptivity for
each element within the enclosure, which is particularly useful for combustion
chambers containing flames. Furthermore, it is possible to specify a number of
different gaseous atmospheres within the enclosure and to calculate exchange areas
for all the gases simultaneously. This is the most computationally efficient way of
using the weighted sum-of-grey-gases model to represent non-grey atmospheres. It
also allows the modelling of atmospheres that vary in time.

4. Verification of RADEX
A detailed verification study has been performed which can be illustrated by the
following examples. In the first study, the view factors produced by RADEX for a
112 D. A. LAWSON AND C. D. ZFESLER

TABLE 1
Comparison of RADEX with analytic values

a(8F)
Ray density Mean error Max. error
(m- 2 ) (%) (%) rms (<5F)

2000 2.13 4.72 0.14

10000 0.93 1.83 0.13
40000 0.39 1.02 0.15
80000 0.36 0.68 0.14

Downloaded from http://imaman.oxfordjournals.org/ at University of Victoria on June 4, 2015

simple cuboid were compared with analytic values. In the second, a comparison was
made between DEAs calculated by RADEX and by RAD3D—a commercially available
Monte Carlo program written for use with computational fluid dynamics models
(Guilbert 1989)—for a geometry consisting of a cylindrical tube running through a
cuboid. In these two studies, the ray density was varied over a wide range in order
to assess its effect on accuracy. In practice, the ray density needed to provide the
required level of accuracy would have to be determined from the details of the exact
nature of the problem.
The expected (rms) error in any Monte Carlo integration is inversely proportional
to the square root of the number of rays (Emery et al. 1991). The last column of
Table 1 shows the normalized standard deviation of the error. This should be constant
and independent of both the number of rays and the value of the integral being
computed. It can be seen that this is the case. Comparisons with values of published
results show that RADEX conforms to the general standard of Monte Carlo
integrations. For example, the test case R-40 in Emery et al. (1991) shows a value
of approximately 0.14. It is therefore concluded that the accuracy of the answers
produced by RADEX behaves in the predicted way for simple analytic cases. The
comparison with RAD3D, presented in Table 2, shows that the agreement between
the programs improves at the expected rate as the ray density is increased.

5. The application of RADEX to a radiant tube furnace

The use of radiant tubes for heat treatment and reheating is becoming an increasingly
attractive option for a range of processes where stock quality is the primary concern.

TABLE 2
RADEX compared with RAD3D

Ray density Mean difference Max. difference

(m" 2 ) (%) <%)

2000 4.56 18.41

5000 2.50 9.66
20000 1.75 7.93
 A PROGRAM FOR RADIATION MODELLING IN FURNACE DESIGN 113

[Tube Height] Zone 1 Zone 2 Zone 3 Zone 4

\
o o
\
Low

FIG. 2. Scheme of furnace. Zone boundaries ; sloped roof •

Downloaded from http://imaman.oxfordjournals.org/ at University of Victoria on June 4, 2015

Such processes include firing of ceramics, metal heat treatment, sintering, and forging.
In some manufacturing processes—for example, the production of turbine blades—
the stock temperature history must be monitored, controlled and documented in
order to guarantee product quality.
Computer modelling can help to demonstrate the potential benefits of this
technology by predicting important process parameters such as the time required for
an item of stock to achieve a specified evenness of temperature through its body (the
soak condition). In a real process, the degree of uniformity is usually judged by the
variation in temperature along the top surface of the stock (lateral uniformity) and
by the difference in temperature between the top and bottom of the stock. If such
applications are to be modelled successfully, then the radiation within the enclosure—
specifically the radiative interchange between the tubes, the load, and the furnace
structure—must be rendered as accurately as possible.
The following design study was carried out for a general radiant tube batch furnace,
whose shape is depicted in Fig. 2. The design features that were being investigated were
• the number of tubes
• the size of the tubes
• the positioning of the tubes
• the effect of the shape of the furnace.
The performance of the furnace was judged by examining the efficiency of the heat
transfer to the stock, and the time-averaged temperature uniformity across the load
surface. The zone furnace model used had the same structure and observed the same
basic principles as the package of programs ZONE (Palmer 1989).
The furnace was zoned in the way indicated in Fig. 2. The enclosed volume was
divided into four equal zones (marked Zones 1-4 in Fig. 2). Each of these volume
zones has two side-wall surface zones and one roof surface zone next to it. In addition
the two end zones each have an end-wall surface zone and a small floor surface zone.
The surfaces of the furnace are therefore represented by 16 surface zones. Each radiant
heating tube gives another surface zone, and in each of the volume zones the load
surfaces make up a further surface zone. There are therefore 24 surface zones in total.
RADEX was used to calculate the exchange areas between each pair of these zones.
(Note that, in'radiant tube furnaces, the enclosure gas does not participate in the
radiative heat exchange since it is free of combustion products).
114 D. A. LAWSON AND C. D. ZIESLER

TABLE 3
Input conditions and results from radiant tube model

Number Soak Lateral

Design of Tube Tube Efficiency time uniformity
number tubes diameter Shape positions % (K)

1 4 D rect see Fig. 2 33.9 100.0 6.4

2 2 2D rect central 34.4 97.9 22.7
3 2 2D sloped central 37.4 89.9 23.4
4 2 2D rect outside 33.7 101.7 12.2
S 2 2D rect left 34.3 96.1 69.1
i 1 4D rect left 34.2 96.9 88.9
7 4 D rect low 33.7 96.1 6.7

Downloaded from http://imaman.oxfordjournals.org/ at University of Victoria on June 4, 2015

8 4 D rect high 35.1 100.3 7.0
% 4 D rect high & low 34.3 99.6 15.4

Efficiency /% Normalized Time

1.25

- 1.15

- 1.05

D.1 D_2 D_3 D_4 D-5 D_6 D_7 D_8 D.9

Configuration
I Furnace Efficiency I I Time to Soak
FIG. 3. Furnace performance.

A total of nine different designs were evaluated, shown in Table 3, to investigate

the effect of the parameters outlined above. Design 1 was taken to be the reference
for evaluation of furnace performance. In all cases, the thermal input was the same
and the total emitting area of the tubes was constant.
Figure 3 shows the efficiencies and the time required to bring the load to the soak
condition, the latter being referenced to design 1. It is clear from Table 3 that the
height of the tubes (designs 1, 7, 8, and 9) makes only a small difference to the
efficiency, while staggering the tubes at two different heights (design 9) affects only
the lateral temperature uniformity. Both the size of the tubes (designs 1, 2, and 6)
 A PROGRAM FOR RADIATION MODELLING IN FURNACE DESIGN 115

Mean Temperature / deg C Vertical Differential / K

1100 16

900

700

500

Downloaded from http://imaman.oxfordjournals.org/ at University of Victoria on June 4, 2015

300

100
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Normalized Time

FIG. 4. Heating curves. Design 1 •; design 3 *; difference 1 x; difference 3 <>.

and their lateral positioning (designs 1, 2, 3, and 6) have little influence on the
efficiency, but cause a marked deterioration in the uniformity. The parameter that
demonstrates an improved efficiency is the enclosure shape which increases it by 3.5
percentage points in this study. The lack of a marked effect on efficiency of the
variations made in this study is, in part, due to the constraint of constant enclosure
size imposed at the start in order to make the results directly comparable. Figure 4
shows the heating curve for the load, along with the temperature difference between
the top and bottom of the load for designs 1 and 3.
The main conclusion which can be drawn from this study is that only the alteration
of the enclosure shape leads to an improved efficiency without undue deterioration
in the uniformity of the temperature across the load.

6. Conclusions
A program, RADEX, has been implemented and tested. This program can calculate the
exchange areas pertinent to a wide variety of zone-model and heat-transfer applica-
tions. It use, and its ability to enhance the value of existing programs, has been
demonstrated in a parametric study of a radiant tube batch furnace. It has been
designed not only for computational efficiency, but also for ease of use and to display
the high level of input flexibility demanded of contemporary applications.

REFERENCES
CHAN, S. H., 1987. Numerical methods for multidimensional radiative transfer analysis in
participating media. Annual Review of Numerical Fluid Mechanics & Heat Transfer 1,
305-50.
116 D. A. LAWSON AND C. D. ZIESLER

EMERY, A. P., JOHANSSON, O., LOBO, M., & ABROUS, A., 1991. A comparative study of methods
for computing the diffuse radiation viewfactors for complex structures. J. Heat Transfer
113,413-22.
GOULD, H. & TOBOCHNIK, H., 1988. Computer simulation methods (part 2). Addison-Wesley,
Reading, Mass.
GUILBERT, P. W., 1989. Computer program RAD3D for modelling thermal radiation.
Technical report AERE-R 13534, Harwell.
HOTTEL, H. C , & SAROFIM, A. F., 1967. Radiative transfer. McGraw-Hill, New York.
NOBLE, J. J., 1975. The zone method: explicit matrix relations for total exchange areas. Int. J.
Heat & Mass. Transfer 18, 261-9.
OMORI, T., TANIGUCHI, H., & KUDO, K.., 1990. Monte Carlo simulation of indoor environment.
Int. J. Numer. Meths Eng 30, 615-27.
PALMER, M. R., 1989. A practical computer package for the thermal design of high temperature
industrial plant. Proceedings of the 1989 International Gas Research Conference, Tokyo.

Downloaded from http://imaman.oxfordjournals.org/ at University of Victoria on June 4, 2015