Progress in Energy and Combustion Science 31 (2005) 371–421
www.elsevier.com/locate/pecs
Heat transfer in ash deposits: A modelling tool-box
Ana Zbogar, Flemming J. Frandsen *, Peter Arendt Jensen, Peter Glarborg
CHEC Research Group, Department of Chemical Engineering, Technical University of Denmark, Building 229, Lyngby DK-2800, Denmark
Received 8 March 2004; accepted 7 August 2005
Available online 11 November 2005
Abstract
The objective of this paper is to review the present state-of-the-art knowledge on heat transfer to the surface of and inside ash
deposits formed in solid fuel-fired utility boilers, and-based on the review-to propose models for calculation of heat transfer, e.g. in
deposition models. Heat transfer will control the surface temperature of the deposit, thereby influencing the physical conditions at
the deposit surface, e.g. if the surface is molten. The deposit surface conditions will affect the deposit build-up rate as well as the
removal/shedding of deposits: molten deposit may lead to a more efficient particle capturing, but may also flow down the heat
transfer surfaces.
The heat transfer parameters of prime interest are the convective heat transfer coefficient h, the effective thermal conductivity of
the deposit keff, and the surface emissivity 3 of the deposit. The convective heat transfer coefficient is a function of flow
characteristics, and can be calculated using different correlation equations, while the other two parameters depend on the deposit
properties, and can be calculated using different structure-based models.
The thermal conductivity of porous ash deposits can be modelled using different models for packed beds. These models can be
divided into two major groups, depending on the way they treat the radiation heat transfer, i.e. the unit cell models and the pseudo
homogeneous models. Which model will be suitable for a particular application depends primarily on the deposit structure, i.e.
whether deposit is particulate, partly sintered or completely fused.
Simple calculations of heat transfer resistances for deposits have been performed, showing that major resistances are in the heat
transfer to the deposit (by convection), and the heat transfer through the deposit (by conduction). Very few experimental data on the
thermal conductivity of ash deposits, especially at high temperatures where radiation is important, are found in the literature.
Although the structure of the deposit is essential for its thermal conductivity, most of the measurements were done on crushed
samples. The results obtained using different models were compared with the experimental data published in Rezaei et al. [Rezaei,
Gupta, Bryant, Hart, Liu, Bailey, et al. Thermal conductivity of coal ash and slags and models used. Fuel 2000;79:1697–1710.],
measured on crushed coal ash samples. Although errors of the predictions were very high in most cases, two models were proposed
as suitable for heat conductivity calculations, i.e. the Yagi and Kunii model for particulate deposits, and the Hadley model for
sintered and fused deposits.
This literature study showed the need for a wide range of experimental data, which would help in evaluating and improving the
existing thermal conductivity models. Also, it is necessary to formulate a more accurate model for the thermal conductivity of solid
mixtures, in which potentially important sources of errors can be identified.
q 2005 Published by Elsevier Ltd.
Keywords: Heat transfer; Radiation; Conductivity; Convection; Deposits; Combustion; Models
* Corresponding author. Tel.: C45 4525 2883; fax: C45 4588
2258.
E-mail address: ff@kt.dtu.dk (F.J. Frandsen).
0360-1285/$ - see front matter q 2005 Published by Elsevier Ltd.
doi:10.1016/j.pecs.2005.08.002
372
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
Contents
1.
2.
3.
4.
5.
6.
7.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Theoretical background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1. Fundamental description of an ash deposit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2. Heat transfer through ash deposits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.1. Thermal properties of a deposit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3. Heat convection in a boiler . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.1. Heat convection to the tubes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.2. Heat convection to the fluid inside the tubes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Models for the effective thermal conductivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1. Theoretical boundaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2. Unit cell models for packed beds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.1. Additive effective thermal conductivity models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.2. Resistance network models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3. Pseudohomogeneous models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4. Dedicated models developed for ash deposits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Estimation of thermal conductivity of ash deposits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1. Estimation of the general parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.1. Structural parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.2. Radiation parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2. Thermal conductivity of gas and solid mixtures gas mixtures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.1. Solid mixtures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3. Emissivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3.1. Influence of various factors on emissivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3.2. Models for emissivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3.3. Experimental data on emissivities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Experimental validation of models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Thermal conductivity of ash deposits obtained by different models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.1. The thermal conductivity of solid material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.2. Simple models for the thermal conductivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.3. Complex structure models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.4. Radiative heat transfer models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.4.1. The exchange factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.4.2. Additive models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.4.3. Resistance network models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
373
374
374
374
375
377
377
378
379
379
380
381
394
397
398
399
399
399
401
403
403
404
404
405
406
408
411
411
412
412
413
413
414
414
416
420
Nomenclature
cp
d
D
h
k
keff
kr
n
p
R
heat capacity
diameter
diffusivity coefficient
convective heat transfer coefficient
thermal conductivity
effective thermal conductivity
radiative thermal conductivity
refractive index
porosity, i.e. volume fraction of gas phase
heat resistance
r
T
u
ratio between the thermal conductivities of
discrete and continuous phase
temperature
velocity
Greek letters
a
surface absorptivity
3
surface emissivity
k
ratio between the thermal conductivities of
solid and gas phase
m
dynamic viscosity
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
r
s
f
c
density
Stefan–Boltzmann constant (5.67!10K8
W/m2 K4)
volume fraction
radiative exchange factor
Subscripts
c
continuous phase
1. Introduction
In solid fuel combustion, non-combustible material
in the fuel forms an ash fraction, which may be
deposited inside the boiler. Ash deposits, formed on a
furnace walls and a convective pass tubes during fuel
thermal conversion, see Fig. 1, may seriously inhibit
heat transfer to the working fluid and hence reduce the
overall process efficiency [11–13].
Heat is transferred between a hot flue gas and the
steam inside the heat transfer tubes by convection,
conduction, and radiation. Convective heat transfer to
the deposit surface and to the working fluid is not
influenced by the deposit characteristics, but mainly by
the flow characteristics of the fluids and by the
geometry of the system. On the other hand, the
formation of porous, low-conductivity deposit layers
on the heat transfer surfaces affect the conductive and
radiative heat transfer. The physical properties of a
deposit, which are thus the most important for the heat
transfer, are the effective thermal conductivity, keff, and
the surface emissivity, 3. These properties will be
discussed in details throughout this paper, both
qualitatively and quantitatively. Ash deposit acts as a
heat transfer resistance between the flame/hot flue
gases, and the working fluid, e.g. by changing the
radiation characteristics of the heat transfer surfaces,
Fig. 1. Deposit build-up on superheaters after 1 week of co-firing (coal
and straw), Amager Power Station, Denmark.
d
eff
g
m
p
rad
s
373
discontinuous phase
effective
gas phase
mixture
particle
radiation
solid phase
i.e. the colour and the surface smoothness, properties
that directly influence the surface emissivity.
The effective thermal conductivity and the emissivity depend on the deposit structure. As shown in Fig. 2,
an ash deposit consists of several layers, having
different compositions, porosities, and microstructures,
and, thus, different thermal conductivities. As the result
of the thermal resistance, the deposit surface temperature increases. This may lead to deposit softening and
melting, and, conditions favourizing accelerated
deposit formation, may be introduced.
In Section 2, the basic theory on heat transfer within
a utility boiler is presented, i.e. the overall heat transfer
equations, correlations for convective heat transfer for
different geometries, and the various deposit properties
influencing the heat transfer characteristics, are outlined. The thermal conductivity is highly dependent
upon the deposit structure, i.e. whether the deposit is
particulate, partly sintered or completely fused. Thus,
deposit structures are also explained in Section 2.
In Section 3, different models for predicting the
thermal conductivity within a porous structure are
presented. Due to its porous structure, a deposit layer
can be modelled as packed bed. Depending on the
actual theoretical handling of the radiation effects, two
groups of models can be distinguished, i.e. the unit cell
models, which treat radiation as a local effect taking
place between adjacent particle surfaces, and, void
boundary surfaces in the unit cell, and pseudo
homogeneous models, which considers the packed
bed to be a continuum for radiation.
In Section 4, we show how to estimate the different
parameters, needed for the calculation of the effective
thermal conductivity of an ash deposit, i.e. the thermal
conductivities of the gas and the solid phase, and the
emissivity. These parameters are later used to calculate
the thermal conductivity of ash.
In Section 5, experimental data on thermal conductivities of solids, available in the literature, are
presented. Different experiments were done on the
synthetic oxide-mixture powders and crushed coal
374
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
Fig. 2. Layered structure of ash deposits.
ashes, with only one published work dealing with the
in situ measurements of thermal conductivity of ash.
In Section 6, a comparison between model predictions and a set of experimental results from Rezaei et al.
[1], performed on crushed coal ash samples, is
presented. A discussion, considering the applicability
of the mentioned models, is presented.
2. Theoretical background
2.1. Fundamental description of an ash deposit
Ash deposits are porous materials, whose physical
and chemical characteristics change with time, during
the process of initiation, growth and maturation, due to
retention of different ashy material and deposit
consolidation. Deposit consolidation includes sintering,
where the contact between ash particles is increased.
Un-sintered deposits consist of distinct particles that
appear isolated from their neighbours. When the
deposit is sintered, solid phases are interconnected
and forms large plate-like features near the outside
surface of the deposit, where the temperature is highest.
Scanning Electron Microscopy (SEM) analyses
indicate that ash deposits have a layered structure,
with a relatively un-sintered innermost layer. The initial
layer, adjacent to the heat transfer surface, usually has a
particulate structure, consisting of a continuous gas
phase and discrete solid particles. Due to its high
porosity, the thermal conductivity of this layer is very
low, causing decreased heat transfer and increase of the
deposit outer layer temperature. According to Robinson
et al. [2], this layer largely determines the overall
effective thermal conductivity of the deposit. On the
other hand, due to higher surface temperatures, the
outermost layer can be completely fused with voids
embedded in a solid continuous phase. This layer has
high thermal conductivity.
2.2. Heat transfer through ash deposits
In utility boilers, heat needs to be transferred from
hot flue gases (Tg) to the steam (Ts). As shown in Fig. 3,
heat is first transferred from the flame/hot flue gas to the
deposit surface (Td) by radiation and convection.
Consider a composite cylindrical wall, consisting of
ash deposit and a tube wall. Heat is assumed to be
conducted through the deposit, and through the tube
wall (Tt). Finally, heat is transferred by convection to
the steam.
The deposit influences the heat transfer by forming
an insulating layer on the tube surface, and by changing
the radiation properties of the surface. As shown in
Fig. 3, the energy transferred to the deposit surface,
should be equal the energy conducted through the
Steam Tube
Deposit
Flue gas
Tg
Ts
Q = Qconvection
+ Qradiation
Qconduction
Qreflected = Qradiation x ρ
Qemitted = Qradiation x ε
Fig. 3. Heat transfer to and through ash deposits.
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
composite wall:
Qconduction Z Qconvection C Qradiation KQreflected KQemitted
(1)
If heat conduction through the tube wall and steam
film is neglected, and the flame emissivity is equal to
unity (for large boilers), Eq. (1) can be rewritten for a
cylindrical geometry:
Qcond Z
Td KTt
Zhconv ðTg KTd ÞCa$s$Tg4 K3$s$Td4
dd
dd
ln
2$p$kd
dto
375
dependent on the temperature and the chemical
composition of the respective phases. Since the
thermal conductivity of a solid phase is two to three
orders of magnitude greater than the one of a gas
phase, heat conduction through a deposit, will
primarily occur through the solid phase [2].
The emissivity is affected by: (1) the deposit surface
conditions (i.e. particulate, partly fused or completely
molten), (2) the size distribution of particles in the
surface layer, and (3) the chemical composition (i.e.
colour of the deposit).
(2)
where subscripts d and t refers to deposit, respectively,
to the inner tube inner diameter of the tube. A more
detail discussion about heat transfer mechanisms
together with an example of a heat transfer resistance
calculation is provided in Appendix A. A calculation
showed that heat convection from the flues gases to the
deposit, and conduction through the deposit are the
major thermal resistances in the overall heat transfer.
Ash deposits are often characterized by an effective
thermal conductivity, keff, which is used to represent the
combined conductive and radiative heat transfer that
may occur both in parallel and in series. The effective
thermal conductivity is defined such that the following
form of the heat transfer equation for one-dimensional
steady-state conduction is satisfied:
dQ
d
dT
ZK
k ðTÞ
Z0
(3)
dy
dy eff
dy
An overall effective thermal conductivity, representing the entire ash deposit thickness, can be calculated
by integrating the local value with the respect to the
location (y) and dividing it by the deposit thickness.
2.2.1. Thermal properties of a deposit
The deposit thermal properties, i.e. the thermal
conductivity and the surface emissivity, are of great
importance for an accurate heat transfer model. The
thermal properties are strongly influenced by the
deposit physical structure, i.e. the particle size
distribution, the porosity and the degree of sintering.
The radiative properties depend solely on the surface
conditions, while the conductive properties depend on
physical data throughout the deposit.
The effective thermal conductivity of a two-phase
gas–solid system depends on: (1) the thermal
conductivity of solid phase ks, (2) the thermal
conductivity of gas phase kg, (3) the porosity, (4)
the size distribution of pores or particles, and (5) the
deposit sintering state. The properties ks and kg are
2.2.1.1. The influence of the temperature. The thermal
conductivity of deposits generally increases when the
temperature increases. Fig. 4 shows experimental data
on the temperature dependency of the deposits thermal
conductivity, as obtained by Rezaei et al. [1].
It can be seen that thermal conductivity is slightly
influenced by the temperature change up to the
sintering point. In this area, the thermal conductivity
changes primarily due to increase of the gas phase
thermal conductivity. Authors defined the initial
sintering temperature, as temperature where the deposit
thermal conductivity starts to be significantly influenced by the increasing temperature. For the ash sample
shown in Fig. 4, the initial sintering temperature is
identified to be equal 650 8C. According to Anderson
et al. [3], two effects appear at high temperatures: (1)
the radiation becomes very important (for glasses at
temperatures above 500 8C), causing a reversible
thermal conductivity change, and, (2) sintering occurs,
causing an irreversible morphology, and thereby a
change in the thermal conductivity. As stated by Gupta
et al. [4], sintering for coal ashes occurs in the
temperature range 900–1200 8C.
Fig. 4. Effect of temperature on thermal conductivity of ash (the
arrows indicating heating and cooling) [1].
376
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
Fig. 4 shows that the deposit has a higher thermal
conductivity after fusion. This is due to the irreversible
changes in the physical structure and porosity.
Experiments showed that the change of porosity in
the two pellets, before and after heating, was zero for
the pellet with porosity 0.48, while there was less then
5% of porosity reduction for the pellet with porosity
0.3. This, together with the data in Fig. 4, indicate that
the existence of irreversibility in thermal conductivity
measurements for heating and cooling cycles can be
attributed to the better connections between particles,
more than to a reduction in porosity.
The emissivity is affected by the temperature
increase due to surface structural changes, i.e. through
the particle size increase due to sintering. Influence of
particle size on emissivity is explained in Section
2.2.1.3.
2.2.1.2. The influence of the chemical composition. The
chemical composition determines indirectly the structure of the deposit at a given temperature. The silica
ratio1 is known to affect the fusion characteristics of a
coal ash, i.e. the sintering temperature. According to
Raask [5], iron-rich slags have been observed to have a
lower sintering temperature above, which their density
increases rapidly when compared to low-iron slags.
Hence, iron-rich slags may have better conductivity
properties.
The thermal conductivity of solid oxide mixtures is
approximately equal to the weighted mass portions of
the oxides. This correlation is not apparent for ashes,
probably because the ash is not composed of the oxides,
even though an ash analysis obtained by the analysis of
elements, is generally reported in terms of equivalent
oxides [6].
According to Wall et al. [7], the chemical
composition of a deposit affects the emissivity due to
the presence of colouring agents (e.g. iron [8]), which
increase the absorptivity of the deposit, and together
with the particle size distribution, will influence the
temperature of the onset of sintering, and the course of
fusion of the deposit.
approximately the same for all particle diameters of the
same packing material. As the temperature increases,
the difference in effective thermal conductivity
increases. Since it is shown that the effect of particle
size is more significant at high temperatures (in [9],
above 800 K), it indicates that the particle size
distribution influence the radiative properties of the
deposit.
The initial deposit consisting mainly of sub micronsize particles of sulfates and chlorides constitutes a
low-emittance surface. Thus, even a thin initial deposit
layer can significantly reduce the radiative heat transfer
to boiler tubes. As Wall et al. [6] showed, below
sintering temperature the emissivity of coal ash
decreases when the temperature increase, and then
increases sharply at higher temperatures, as sintering
and fusion of the ash occur. As the deposit melts to form
a slag, the absorbance increases to values approaching
0.9, as shown in Fig. 5.
The particle size distribution strongly affects the
emissivity of unsintered samples. With an increase in
the mean particle size (diameter), the emissivity
increases, because small particles reflect a significant
portion of the incident radiation, while the large
particles absorb most of it. Fig. 6 shows the effect of
particle size, on the emissivity. Also, it is show that a
marked increase in the surface emissivity results from a
formation of a sintered or fused matrix.
2.2.1.3. The effect of the particle size. According to
Gupta et al. [4], experiments in [9,10] showed that an
increase in the mean particle size leads to an increase in
the effective thermal conductivity. For the alumina–air
packed beds, used by Nasr et al. [9], the effective
thermal conductivity at low temperatures (350 K) is
1
Silica ratioZ[(SiO2)/(SiO2CFe2O3CCaOCMgO)].
Fig. 5. Expected trends in deposit properties during their growth,
according to Wall et al. [6].
377
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
Re Z
r$u$d
m
(5b)
Pr Z
m$cp
k
(5c)
where d is a characteristic dimension. Gas properties
were calculated at the film temperature, i.e. at an
average between the flue gas temperature, and the
deposit surface temperature. Correlations, which are
used for heat transfer coefficient calculations, are
different in case of heat transfer from the flue gas
(heat transfer onto the cylinder) and from the tube wall
to the steam (heat transfer inside the tube).
Fig. 6. The effect of particle size and temperature on emittance for
particulate ash and the change when a slag layer (sms) is formed on
heating. The transition path from a particulate layer to slag on heating
is also indicated [6].
2.2.1.4. The effect of the porosity. A decrease in
porosity leads to increase in the effective thermal
conductivity of the deposit. Porosity is a function of
sintering conditions, and time. According to Gupta
et al. [4], even a half percent increase in porosity can
decrease the effective thermal conductivity by an order
of magnitude.
The effective thermal conductivity varies significantly with the deposit type. The slag structure thermal
conductivity is considerably higher than the particulate
structure thermal conductivity, especially in the
porosity range 0.2–0.8 [1]. The initial stages of
sintering are accompanied by an increase in the deposit
thermal conductivity. Subsequent sintering continues to
densify the deposit, but has little effect on the deposit
thermal conductivity.
2.3. Heat convection in a boiler
As stated above, in a utility boiler, the heat is
transferred partly by convection from the flame/hot flue
gases to the deposit surface and from the tube wall to
the stream inside the tubes. Correlations, which are
used for the calculation of the heat transfer coefficient h
(W/m2 K), can be as:
h Z f ðNu; Re; PrÞ
2.3.1. Heat convection to the tubes
First, correlations for the convective heat transfer to
a cylindrical tube as shown in Fig. 7 will be presented.
As stated in Incropera and DeWitt [14], calculation
of the Nusselt number can be done using the empirical
correlation of Hilpert, based on experimental values of
the average heat transfer coefficient from a heated
cylinder:
Nu Z C$Rem $Pr 1=3
(6)
Values for parameters C and m in the above equation
are given in Table 1.
Re, Nu and Pr are calculated for the characteristic
dimension d, equal to the cylinder diameter, dc. All the
properties are evaluated at the film temperature. An
approach of Gnielinski is presented in [15], where the
characteristic dimension d equals dcp/2. Correlations,
which are valid over the ranges 10!Re!107 and 0.6!
Pr!1000, are shown
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
(7)
Nul;0 Z 0:3 C Nu2l;lam C Nu2l;turb
where
pffiffiffiffiffiffi
Nul;lam Z 0:664 Re$Pr 1=3
(8a)
Flue gas
Tube
(4)
Steam
where Nu is the Nusselt number, Re is the Reynolds
number and Pr is the Prandtl number of the actual
system. These dimensionless parameters are defined as:
Nu Z
h$d
k
(5a)
Fig. 7. Heat transfer to cylinder tubes.
378
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
Table 1
Constants of Eq. (6) for cylindrical cylinder in cross-flow [14]
Re
C
M
0.4–4
4–40
40–4000
4000–40,000
40,000–400,000
0.989
0.911
0.683
0.193
0.0266
0.330
0.385
0.466
0.618
0.805
fA Z 1 C
0:7 ðb=aK0:3Þ
j1:5 ðb=a C 0:7Þ2
(11)
where
Transverse pitch ratio:
aZ
SQ
D
(12a)
Longitudinal pitch ratio
0:8
Nul;turb Z
0:037$Re $Pr
1 C 2:443$ReK0:1 ðPr 2=3 K1Þ
(8b)
Whitaker [16] recommends the following correlation for the mean Nusselt number:
1=4
1=2
2=3
0:4 mN
Nu Z ð0:4$Re C 0:06$Re ÞPr
(9)
ms
in the range 1!Re!105, 0.67!Pr!300, and 0.25!
mN/m0!5.2. Subscript N stands for the viscosity at the
bulk gas temperature, and s for the viscosity at the
surface temperature. The values of the viscosity and the
thermal conductivity in Re and Pr are estimated at those
at the approaching stream temperature.
In the case of a tube bundle, heat transfer coefficient
can be calculated as follows:
(10)
Nul0;bundle Z fA $Nul;0
where fA is a function of the tube bundle spatial
geometry.
In the case of an in-line arrangement, as shown in
Fig. 8, the parameter fA may be calculated from:
SQ
bZ
SL
D
(12b)
with parameters shown in Fig. 9.
Void ratio:
j Z 1K
p
; bR 1
4a
(12c)
j Z 1K
p
; b! 1
4ab
(12d)
Nul,o is calculated using Eq. (7), with RejZ(r.ud)/
jm.
In the case of a staggered arrangement, as shown in
Fig. 9, the parameter fA may be calculated from:
fA Z 1 C
2
3$b
(13)
According to [14], each correlation is reasonable
over a certain range of conditions, but for the most
engineering calculations one should not expect accuracy to be much better than 25%.
2.3.2. Heat convection to the fluid inside the tubes
If we assume turbulent flow of steam inside the heat
transfer tubes (ReO10,000), an average value of the
heat transfer coefficient can be calculated using
SL
SL
D
SQ
ugas
ugas
Fig. 8. In-line arrangement in a tube bundle.
Fig. 9. Staggered arrangement in a tube bundle.
379
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
Table 2
Values of 3l parameter
Re
4
10
2!104
5!104
105
106
l/d
1
2
5
10
15
20
30
40
50
1.65
1.51
1.34
1.28
1.14
1.50
1.40
1.27
1.22
1.11
1.34
1.27
1.18
1.15
1.08
1.23
1.18
1.13
1.10
1.05
1.17
1.13
1.10
1.08
1.04
1.13
1.10
1.08
1.06
1.03
1.07
1.05
1.04
1.03
1.02
1.03
1.02
1.02
1.02
1.01
1
1
1
1
1
Miheiev correlation [17]
Nu Z 0:021$Re
0:80
$Pr
0:43
3. Models for the effective thermal conductivity
Pr
Prw
0:25
(14)
where Prw is the value of the Prandtl number at the tube
wall temperature. Dimensionless numbers are calculated for the cylinder inner diameter. This equation can
be used for ReO104 and PrZ0.7–2500, and ratio tube
length to tube diameter ratio l/dO50. If l/d!50, then
calculated heat transfer coefficient should be multiplied
by 3l, which values are given in Table 2.
Since the temperature at the tube inner diameter and
steam can be considered very close, the ratio of
viscosities or Prandtl numbers can be neglected (set
equal to 1). One such correlation is given by DittusBoelter [14]:
From a modelling point-of-view, ash deposits are
porous structures that can be approximated as packed
beds. Thus, the thermal conductivity of deposits may be
estimated using well-documented models for the
thermal conductivity of packed beds. Un-sintered
(particulate) deposits should be considered as systems
with continuous gas phase and discontinuous solid
phase. Fused deposits (slags) should be considered as
systems with discontinuous gas phase (voids), and
continuous solid phase. Sintered deposits can be treated
as complex systems, where both phases are continuous.
Thermal conductivity models can be divided into
two main groups depending upon the handling of
radiative effects in the system: unit cell models and
pseudo homogeneous models.
3.1. Theoretical boundaries
4=5
Nu Z 0:023$Re $Pr
0:4
(15)
This equation has been confirmed experimentally
for the range of conditions: 0.7!Pr!160, ReO10000,
l/dO10.
This expression should be used only for small to
moderate temperature differences between fluid and
wall [14], with all the properties evaluated at the bulk
fluid temperature.
Theoretically, the most general approach to calculate the upper and the lower boundaries value of the
thermal conductivity correspond to the cases schematically shown in Fig. 10. Assuming that the gas and the
solid phase conduct heat independently, these boundaries can be formulated as:
(1) The upper boundary value, where heat is transferred in a direction parallel to the layers as shown
gas flow
Maximum K
Minimum K
Rayleigh
CPS
CSP
Fig. 10. Models for thermal conductivity estimations (maximum and minimum k—for theoretical boundaries, Eqs. (16a), (16b), (17a) and (17b)),
according to [4].
380
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
in Fig. 10a. Here, the heat resistance is calculated
as a parallel connection of thermal resistances:
keff Z ks C
(16a)
Since the thermal resistance is equal RZ1/
(thermal conductivity, k), the maximum effective
thermal conductivity can be calculated from:
keff;max
k
Z p C ð1KpÞ s
kg
kg
(16b)
(2) The lower boundary value, where heat is transferred in a direction normal to the layers as shown
in Fig. 10b. Here, the heat transfer resistance is
calculated as a serial connection of thermal
resistances:
Rtot Z p$Rg C ð1KpÞRs
(17a)
As in the previous case, the minimum effective
thermal conductivity is calculated as:
keff;min
1
Z
p
kg
p C ðk1K
s =kg Þ
(17b)
These are not rigorous theoretical bounds, but are
made in intention to provide a rational basis for
comparison. For example, radiation can result in an
effective thermal conductivity greater than the upper
limit provided in Eq. (16a) and (16b) and deposits with
small sized pores could have thermal conductivity less
than the lower limit in Eq. (17a) and(17b). Measurements presented in Robinson et al. [2], showed that the
overall trend in the thermal conductivity of particulate
deposits follows the lower theoretical limit. This is
most probably due to the weak solid–solid contact of
deposited particles, which is a limiting factor, while the
deposit structure is particulate.
According to Bird et al. [16], the first major
contribution to the estimation of the conductivity of
heterogeneous solids was made by Maxwell (1873),
who’s equation represent a material consisting of noninteracting spheres suspended in a fluid matrix. As
stated in Nimick and Leith [18], these equations were
shown to be the most stringent upper and lower
boundaries for homogeneous, isotropic, two-phase
mixtures. The upper-bond equation, describing low-
1
kgKks
p
p
C 1K
3ks
(18)
while, the lower-bond equation, where solid spheres are
suspended in a fluid matrix is:
keff Z kg C
1Kp
C 3kpg
(19)
1
ksKkg
These equations can be derived using the steadystate temperature distribution in a sphere (T1) and in the
surrounding medium (T0). The volume fraction p of
embedded gas spheres (porosity), is taken to be
sufficiently small that the spheres do not ‘interact’
thermally; that is, one needs to consider only the
thermal condition in a large medium containing only
one embedded sphere. Due to the non-interacting
spheres assumption, these equations are limited to
porosities close to 0, i.e. Eq. (18), or 1, i.e. Eq. (19).
This is why the model cannot be used for estimating the
thermal conductivity of a granular porous media, but
only the limiting cases.
These two approaches to the theoretical boundary
limits are shown in Fig. 11, where the effective thermal
conductivity is plotted versus porosity.
3.2. Unit cell models for packed beds
The unit cell models treat radiation as a local effect
taking place between adjacent particle surfaces, and
void boundary surfaces in the unit cell, i.e. long range
effects of radiation are neglected. Vortmeyer [19] lists
five assumptions under which these models are
developed:
Effective thermal conductivity, kef
1
p
1Kp
Z C
Rtot
Rg
Rs
conductivity (fluid) spheres in a high-conductivity
(solid) matrix, is
1
0.8
0.6
0.4
0.2
0
0
0.2
equ. 16b
0.4
0.6
Porosity
equ. 17b
equ. 18
0.8
1
equ. 19
Fig. 11. Theoretical boundary limits for the effective thermal
conductivity.
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
†
†
†
†
†
381
Particle diameter[wavelength
Grey emitting surfaces
Opaque material
Absence of free convection
DT/T/1 across a particle layer
Unit cell models treat effective thermal conductivity
in two ways:
(1) By adding the radiative conductivity to the
conduction component, i.e. additive models:
keff Z kgKs C krad
(20)
(2) By including the radiative conductivity as a
resistance in the network of resistances representing the unit cell, i.e. resistance network models.
3.2.1. Additive effective thermal conductivity models
3.2.1.1. Simple models. The simple models are
restricted to low temperatures, since they neglect
radiative heat transfer through the packed bed.
According to [3,4], radiation will affect the thermal
conductivity in the temperature range: 900–1250 8C.
Neglecting radiative transfer is reasonable if it can be
assumed that the conductive component is independent
of the radiative component. The models have a
simplified approach to the system structure, and thus
can be applied only when one phase is continuous and
the other is discontinuous, i.e. for particulate or fused
deposits. Using these models, a heat conductivity
component, kgKs, can be obtained. The usual assumption is that temperature isotherms are normal to the
direction of heat transfer, i.e. due to an infinite lateral
conductivity assumption the temperature gradient
perpendicular to the heat flow does not exist. The
larger the ratio ks/kg, the greater the errors introduced
by this assumption, which is only totally valid when
ksZkg [20], will be.
Lord Rayleigh [4,10] derived an equation for
estimating the thermal conductivity of a two-phase
system, schematically shown in Fig. 10c. The model
system consists of a cubic array of uniform spheres
(discrete phase), being embedded in a continuous
phase:
keff
ð2 C rÞ=ð1KrÞK2p
Z
ð2 C rÞ=ð1KrÞ C p
kc
Fig. 12. Comparison between experimental data and estimations
based on the Rayleigh model for the thermal conductivity of sintered
ash, at a porosity of 0.3 [1].
spheres, the Rayleigh model is limited to a discrete
phase volume fraction of up to p/6, which is explained
in Appendix B.
Predictions based on the Rayleigh model has been
compared to experimental results presented by Rezaei
et al. [1]. Fig. 12 indicates that the thermal conductivities of unsintered ash samples, during the heating
cycle, are closer to the thermal conductivity of
particulate structure. The thermal conductivity of
samples during the cooling cycle is closer to predictions
for the slag type structure, due to partial sintering.
Fig. 13 shows that experimental results for sintered ash
samples are closer to predictions for fused material.
Although an exact mathematical calculation is being
provided, this model is rigid and artificial, permitting
the spheres to be located only in the center of their unit
cubical species, which radically departs from the
powders encountered in practice.
Russel [4] derived a thermal conductivity model,
schematically shown in Fig. 14, for an array of cubes in
a cubical structure:
(21)
where p refers to the volume fraction of the discrete
phase. Like most of the other models that consider
Fig. 13. Comparison between the experimental data and estimations
based on the Rayleigh model for thermal conductivity of unsintered
ash samples, at a porosity of 0.3 [1].
382
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
Fig. 14. Structure used in the Russel model [4].
keff
rp2=3 C ð1Kp2=3 Þ
Z 2=3
kc
rðp KpÞ C ð1Kp2=3 C pÞ
(22)
where p refers to the volume fraction of the discrete
phase. When r is large, Eq. (22) can be simplified to:
keff
1
Z
kc
1Kp1=3
(23)
The Russell model can be applied over the complete
range of porosity. Since the mathematical treatment is
unrigorous, the model permits an arbitrary, irregular
location of obstacles in their unit cubical species, and
hence is more realistic than the Rayleigh model [10].
On the other hand, pores of cubical shape are not likely
to be found within ash deposits.
Woodside [20] considers a cubical array of noninteracting spherical particles in a continuous medium,
shown in Fig. 15. Snow, with the density range 0.10–
0.48 g/cm3, was used in the experiments.
The thermal resistance of the model cube against the
heat conduction, in the direction shown in Fig. 15,
equals the thermal resistivity of the composite material.
The total thermal resistance in the cube is:
ðR
1 1KR
dx
Z
C
2
k
kg
ks $p$r =4 C kg ð1Kp$r 2 =4Þ
(24)
0
The first term in Eq. (24) represents thermal
resistance of the shaded gas layer, and the second
term is the thermal resistance of the composite gas–
solid layer of thickness dx. Since r2ZR2Kx2, the
resulting equation, derived by integration of Eq. (24), is
1=3
2
kg
6S
a K1
a C1
1K
ln
Z 1K
p
2a
aK1
keff
(25)
where
Fig. 15. Representative unit cell of a material consisting of uniform
solid spheres, distributed in a cubic lattice in a gas, used for estimation
of the thermal conductivity by Woodside [20].
"
4
a Z 1C
pðks =kg K1Þð6S=pÞ2=3
S Z 1Kp Z p$R3 =6 Z
#1=2
rKrg
;
rS Krg
(26a)
(26b)
0% S% p=6 ðsimple cubic cellÞ
If the system density approaches the solid phase
density, then S approaches 1, and k approaches the solid
thermal conductivity. On the other hand, if the system
density approaches the gas density, then S approaches
0, and k approaches the gas thermal conductivity. If the
subscripts s and g are interchanged in Eqs. (25), (26a)
and (26b), Eq. (25) can be rearranged to become the
equation for the conductivity of a material consisting of
uniform spherical pores distributed in a cubic lattice
solid.
Krupiczka [21] developed a model for calculating
the effective thermal conductivity of granular material
of spherical and cylindrical grains embedded in a
continuous fluid phase, schematically shown in Fig. 16.
The simplifying assumption is that the total heat flux
through the unit cell consists of the independent heat
fluxes, Q1 and Q2.
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
B ZK0:057
Q2
φ
α
log k Z 0:48$log dp K1:75
Fig. 16. Cubic element of the model of spheres under consideration
[21].
The quantity of heat conducted only by the gas is:
p
Q1 Z kg $A$DT Z kg 1K
(27)
4
where A is the area normal to heat transfer direction,
when a cubic cell has unity diameter, DT is the
temperature gradient.
The quantity of heat that proceeds from the gas
through surface of the sphere to the solid is
ð
vt2
Q 2 Z kg
dF
(28)
vr rZ1 n;rZ1
F
dFn;rZ1 Z ðr cos 4Þr d4 da Z cos 4$d4$da
(31b)
According to the experiments, Eq. (19) can predict
the effective thermal conductivity of particulate beds
(powders in different gases) with an accuracy of G
30%.
Boow and Goard [5,22] investigated the thermal
conductivity and the total emittance of coal ash deposits
and synthetic slags. Their experiments showed that
these properties are influenced by the particle size and
the chemical composition of the material. Empirical
correlations for the thermal conductivity of ashes, in the
particle size range of 20–300 mm and at TZ700 8C,
were obtained:
for colorless synthetic slag:
Q1
ρ
383
(29)
(32)
for iron-containing slag:
log k Z 0:56$log dp K1:63
(33)
where k [Z] W/m K, and dp [Z] mm. The underlying
experimental results are provided in Fig. 17.
Applying laboratory-prepared ashes, Boow and
Goard obtained exceptionally low values of the thermal
conductivity, from 0.02 to 0.06 W/m K, in the
temperature range of 500–1000 K (less than for air).
In order to reduce the thermal conductivity of a porous
body below that of air or boiler flue gas, the pore size
must be less than 0.5 mm.
Leach [23] developed models for foam like
materials (cellular material, plastic foams, other lightweight materials), where air is considered to be the
discrete phase. Model unit cells are shown in Fig. 10d
and e.
where dFn is the area of the sphere slice shown in
Fig. 16, and t2 is the temperature distribution with
respect to the y-axis, in the external zone. When Eqs.
(27) and (28) are combined, the effective thermal
conductivity of the system can be estimated. Since the
expression for the temperature distribution is quite
complicated, an attempt was made to approximate the
final equation for thermal conductivity by a simple
function. The final model equation is given as:
ACB logðks=kgÞ
keff
k
Z s
(30)
kg
kg
where A and B are parameters derived based on
experiments. For the model based on spheres, the
parameters A and B can be estimated from:
A Z 0:280K0757$log p
(31a)
Fig. 17. The influence of mean particle size on the thermal
conductivity at 975 K. , furnace deposits, ! precipitator and
cyclone dusts, C laboratory ashes, 6 primary superheater, C
furnace floor deposits (pfc) and cyclone dust, $, :, & synthetic slag
[22].
384
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
The Cubic Parallel Series (CPS) model considers a
composite material to consist of vertical side cell walls
and the air inside the cell. This composite material is
put in series with the horizontal solid layers. The model
is analog to the Russell model, with gas as the discrete
phase. It predicts a higher thermal conductivity than
other models. The Cubic Series Parallel (CSP) model
considers horizontal cell walls and the air inside the cell
as a composite material. This composite material is put
in parallel with the vertical cell walls. This model is
equivalent to the model of Diesser and Eian.
The main difference between CSP and CPS models
is the manner in which heat flow at the cell corner is
treated: CSP assumes zero horizontal conductivity and
CPS assumes infinite horizontal conductivity. The
model equations are:
CSP yCSP C 1 Z M
CPS yCPS C 1 Z
f2KM C ðMK1Þ½x C ð1KxÞ1=3 g
1 C ðMK1Þð1KxÞ1=3
(34)
MKðM K1Þð1KxÞ2=3
1Kð1K1=MÞ½ð1KxÞ2=3 Kð1KxÞ
(35)
k
Z y C1
kg
(36)
ks
ZM
kg
(37)
x Z 1Kp
(38)
These models are valid for fused deposits; for
particulate deposits discrete and continuous phase have
to be interchanged.
For low density foams or slags (x/1 and M[1),
CPS, CSP, and spherical models, can be reduced to
1
1
y z ð2MK1Þx C q M$x2
3
9
(39)
where qZ1, 2, 3 (CSP, spherical models, CPS).
Rezaei et al. [1] showed (Fig. 18) that all these
models (CPS, CSP, Rayleigh) lie within the two
extreme limits for a given ratio of the thermal
conductivity of gas to the thermal conductivity of
solid. The Rayleigh model (spheres) gives values that
are between the CPS and CSP models (cubes).
3.2.1.2. Thermal conductivity of complex structures. In
models considering complex structures, both fluid and
gas phase are continuous, as is in reality the case with a
Fig. 18. Three simple models for thermal conductivity of particulate
(P) and slag (S) type structures (radiation neglected) [1].
sintered deposit. Radiation is not discussed here, so the
influence of the particle size will be neglected.
Parameters, which would be required for a complete
analysis, are the fraction of each continuous phase, and
the porosity of each phase. But with the present
knowledge it is not possible to provide such information, so simplified models will be used.
Brailsford and Major [24] derived expressions for
the thermal conductivity of two-phase media for
various types of structures, based on the Maxwell
theory. The unit cell, which is shown in Fig. 19, consists
of particles, i.e. phase 1, surrounded by phase 2, which
is in turn surrounded by a material having an average
conductivity equal to that which we wish to calculate.
The temperature distribution in each region satisfies
the Laplace equation:
V2 T Z 0
(40)
The general solution in each of the regions, in
spherical coordinates, with the origin at the centre of
the suspended particle, is
385
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
equation is:
2
n
0
0
k0 p0 C k1 p1 ð2k3k
C k2 p2 ð2k3k
0Ck1 Þ
0Ck2 Þ
o
kZ n
0
0
p0 C p1 ð2k3k
C p2 ð2k3k
0Ck1 Þ
0Ck2 Þ
1
r
?
Fig. 19. Unit cell for Brailsford and Major model.
vT
vz
d
cos q
r2
(41a)
c
r1 ! r! r2 T Z br C 2 cos q
r
(41b)
r! r1 T Z ar cos q
(41c)
rO r2 T Z
zC
ex
o
(45)
Brailsford and Major also developed a model for
complex systems, which can be interpreted as a random
two phase assembly, containing regions of a gascontinuous-phase, and a solid-continuous-phase
embedded in a random mixture of the same two phases.
They regard both phases as continuous and assume the
porosities in these two phases to be similar. The
material used in the experiments was sandstone. Using
Eq. (45) with k0Zk the following equation was
where zZr cos q, and a, b, c, d are constants.
Conditions of a uniform external temperature gradient
requires that dZ0. This gives the equation
n
o
r3
r3
2 1K r13 C kk12 1K2 r13
k
2
2 o
(42)
Z n
r13
r13
k1
k2
1K 3
2C 3 C
r2
k2
r2
where r1 is the radius of spherical particles of phase 1,
surrounded by phase 2 out to radius r2. If phase 1 (gas)
is the continuous phase, then
n
o
k2
k2
k
k1 3 k1 C 2$p1 1K k1
n
o
Z
(43)
k2
k2
3Kp1 1K kk21
where p1 is the volume fraction of phase 1 ðp1 Z r13 =r23 Þ.
This equation is presented in Fig. 20, as curve 4. In
case phase 2 (solid) is the continuous phase, then:
1Kk1 =k2
1K2p
1 2Ck1 =k2
k
Z
(44)
k2
1 C p 1Kk1 =k2
1 2Ck1 =k2
This equation is equivalent to Maxwell’s result, and
is illustrated in Fig. 20 as curve 3. For three (0
(continuous), 1, 2) or more phase systems, the model
Fig. 20. (1) Minimum conductivity, Eq. (17a) and (17b); (2)
maximum conductivity, Eq. (16a) and (16b); (3) phase 2 continuous,
Eq. (43); (4) phase 1 continuous, Eq. (42); (5) random mixture,
Eq. (45), [24].
386
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
obtained:
k Z fð3p1 K1Þk1 C ð3p2 K1Þk2
C ½fð3p1 K1Þk1 C ð3p2 K1Þk2 g2 C 8k1 k2 1=2 g=4
(46)
where k1, k2 are thermal conductivities of fluidcontinuous phase and gas-continuous phase and p2Z
1Kp1. Eq. (46) is shown in Fig. 20, as curve 5.
If it is assumed that sandstone can be subdivided into
small regions, each of which may be regarded as a fluidcontinuous or solid-continuous region, the appropriate
conductivities within these regions are given in (43) or
(44). The resultant, overall thermal conductivity can be
obtained by using Eq. (46). The volume fraction of the
fluid-continuous phase pf is related to the volume
fraction p of fluid phase through a simple oneparameter equation:
pf Z pfa C ð1KaÞpg
(47)
where a is a factor relating to the structure of the
deposit, Eq. (47) reflects the physically reasonable
assumption that the phase having the major portion of
continuous phase is likely to predominate; thus pf/1
when p/1, and pf/0 when p/0. The result of fitting
experimental data for sandstone, shown in Fig. 21,
gives the value aZ0.8.
Rezaei et al. [1] compared results obtained using the
Brailsford and Major model, with the measurements of
sintered ash samples. According to Figs. 22 and 23, the
agreement is satisfactory. The model describing a
random distribution of mixed phases, gives results,
which are within 20% of the measured values for a
partially sintered sample.
Nozad et al. [25] used the method of volume
averaging for estimating the thermal conductivity of
two- and three-phase systems. The volume averaging
method provides a tool for establishing macroscopic
governing equations from microscopic equations and
boundary conditions. It considers the volume to be
Fig. 22. Comparison between the experiments and prediction results
based on mixed structure (Brailsford and Major) model for thermal
conductivity of sintered ash samples [1].
Fig. 21. Comparison of the experimental thermal conductivities of dry
sandstone at 0 8C with the theoretical predictions assuming either
phase continuous or a random distribution [24].
Fig. 23. Comparison between the experiments and prediction results
based on mixed structure (Brailsford and Major) model for thermal
conductivity of sintered ash samples [1].
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
huge, compared to the small-scale deviations in volume
fraction of each component, but small enough so that
the temperature, vary only slightly over the volume.
The experimental measurements included three
fluids (air, glycerol, and water) and five solids (glass,
stainless steel, bronze, urea-formaldehyde, and aluminium) with particle diameters ranging from 2.5 to
4.0 mm. The measurements were made on a fixed bed
of unconsolidated spherical particles, at average
temperature of 45 8C. In order to illustrate this method,
the governing differential equation and boundary
conditions for transient heat conduction, in the system,
are given in the form
ðrcp Þb
vTb
Z Vðkb VTb Þ; in Vb
vt
(48)
where subscripts b, s represents the continuous,
respectively, discontinuous phases, Tb, Ts are the
local temperatures in the b- and s-phases, respectively,
Abs is the interfacial area between two phases, Abe, Ase
are the bounding surfaces of the entrances and exits for
phases, and nbs is the unit outwardly directed normal
for the b-phase over the area Abs(nbsZKnsb).
B:C:1 Tb Z Ts ; at Abs
(49a)
B:C:2 nbs $kb VTb Z nbs $ks VTs ; at Abs
(49b)
ðrcp Þs
vTs
Z Vðks VTs Þ; at Vs
vt
(49c)
B:C:3a Tb Z g1 ðtÞ; at Abe
(49d)
B:C:3b Ts Z g2 ðtÞ; at Ase
(49e)
resistance analogy for cubical particles and square
contacts
keff
1
(51)
Z rð1Ka2 Þ C
1
ks
C 2 1Ka 2
a$r
r$a Cð1KrÞc
p Z a3 C 3$c2 ð1KaÞ
Abs
ð1KrÞ
rV
ð
1
ðn f
2 bs 1
Abs
C f1 nbs ÞdA
(50)
where keff is the effective thermal conductivity tensor
(kcal/ms K), ps, pb are the volume fraction of phases s
and b, respectively, I is the unit tensor, V is the
averaging volume (m3), f is the vector that relates PhTi
to
Tb
(m)
and
is
represented
like
f Z f0 C p$f1 C p2 f2 C/.
Finally, a model in which both phases are considered
to be continuous is derived using the electrical
(52)
where r is the ratio of the thermal conductivity of the
discontinuous phase to the thermal conductivity of the
continuous phase and a and c are the ratios of particle
size, and neck size to the unit cell size as given in
Fig. 24. Such structure may form when the particulate
phase starts to sinter, establishing contacts between the
particles, thereby resulting in a solid phase continuous.
At the same time, the continuity of the fluid phase is not
destroyed.
Experimental validation of the model is illustrated in
Fig. 25, where the experimental point having the
highest value of r represents the author’s result for the
air-aluminum system.
Hadley [26] developed a theoretical model for
predicting the thermal conductivity of consolidate,
based on the volume averaging method. Consolidate
mixture of two metal powders were used in the
experiments: brass powder (highly angular particles)
and stainless steel powder (approximately spherical
particles). Consolidation of the particles was done by
cold pressing.
A multiphase mixture containing n phases was
considered, and an averaging volume of arbitrary shape
The solution to the model is
ð
keff
ð1KrÞ
1
ðn f
Z ðpb C ps rÞI C
V
2 bs 0
kb
C f0 nbs ÞdA C
387
Fig. 24. Model for particle–particle contact [25].
388
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
constant for a continuous solid, and, kZks/kg. For
granular systems, a is small and the first term in Eq.
(57) is dominant for moderate values of k (for pZ0.4,
the first term will be dominant for kz100). For
consolidated materials, a is moderate (for porosities
less than 0.1) and the second term may be dominant
over most of the range of k.
The values for f o and a are experimentally
determined. If vacuum is used as a pore fluid, then
k/N. If this is used in Eq. (57), the thermal
conductivity at the low gas pressure (close to vacuum)
can be predicted as:
Fig. 25. Comparison between the theoretical predictions for the
effective thermal conductivity, obtained using the Nozad model, and
different experimental measurements [25].
was constructed. Spatial derivates of volume-averaged
quantities may be defined as the change in that quantity
per unit distance movement of the averaging center
(point which represent averaging volume), with the
shape and orientation of the averaging volume fixed.
When volume averaging the temperature, respectively,
the heat flux, we obtain:
hTi Z
n
X
hFi Z
n
X
iZ1
iZ1
hTi i
(53)
n
X
k¼i hVTi i
hFii ZK
k
1Kp
Z 2$a$
ks
2 Cp
(58)
Thus, a measurement of the conductivity, made with
pore pressures low enough to eliminate the influence of
the pore fluid, will provide a value for a. Fig. 26 shows
values of a obtained from Eq. (58) plotted vs. d (percent
theoretical density), i.e. dZ1Kporosity. The figure
shows that a depends primarily upon the material
porosity, and that it is approximately independent of the
other parameters, such as a particle shape or the brass
volume fraction (due to its angular fraction, brass will
(54)
iZ1
After introducing intrinsic averages in the above
equation, the final working equations are:
VhTi Z
n
X
iZ1
¼
kVhTi Z
pi hVTi ii
n
X
iZ1
pi k¼i hVTi ii
(55)
(56)
Combination of the equations for conduction
through a continuous media (the Maxwell upper
formula, Eq. (18)), with the equation for suspension
of particles (obtained from reformulating Eqs. (55) and
(56)), gives a model for the thermal conductivity of a
two-phase system
k
p$f0 C ð1Kp$f0 Þk
Z ð1KaÞ
kf
1Kpð1Kf0 Þ C pð1Kf0 Þk
Ca
2ð1KpÞk2 C ð1 C 2$pÞk
ð2 C pÞk C 1Kp
(57)
where a is a mixing parameter (function of degree of
consolidation), f0 is a parameter which is approximately
Fig. 26. Values of consolidation parameter a determined from
experimental measurements of evacuated samples. The curve
represents an approximate fit [26]. PTD (d): percent theoretical
density (dZ1Kp).
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
389
cause the more efficient packing). This dependence can
be presented as:
p Z 0:3K0:572 a Z 1:5266ð1KpÞ8:7381
(59)
p Z 0:0975K0:3 a Z 0:7079ð1KpÞ6:3051
(60)
In general, a is the most sensitive parameter in this
theory, as it scales the matrix conduction component.
The parameter f0 should be determined from
measurements using a higher conductivity pore fluid
such as water (in that way the second term in Eq. (42) is
reduced, so that maximum sensitivity to f0 is obtained).
Even though it is not understood why the angular brass
particles should require a higher value of fo than the
more spherical steel particles, a linear formula for
determination of fo was obtained:
fo Z 0:8 C 0:1pb
(61)
where pb is a volume fraction of brass. The value of f0 is
between 0.8 for the stainless steel, and, 0.9, for brass
samples.
In order to compare predictions by the Hadley model
with previous two-phase measurements, the author used
the large number of two-phase experiments compiled
and catalogued by Crane and Vachon. The parallel lines
in Fig. 27 represent 20% deviation from unity. The
comparison with the Crane and Vachon list gives an
average error of 22%.
Nozard et al. [25] compared Hadley’s model to their
own numerical calculations, for a two-phase granular
material, which is shown in Fig. 28. Since the
parameter a, defined in the work of Hadley, represents
the fraction of heat conducted through matrix contact
only, the expected relationship between a and the
Fig. 28. Comparison of the Hadley model with the numerical
calculations of Nozard et al., and previous two-phase heat
conductivity data [26].
relative ‘bridge’ area fraction, defined by Nozad, is
evident.
Nimick and Leith [18] developed a semi-empirical
model for estimating the effective thermal conductivity
of a granular porous media, in which convection and
radiation are neglected. The authors assumed that
granular porous medium comprises regions of solidcontinuous material and fluid-continuous material.
A porous medium composed of fused silica powder
and air was used in the experiments. Using Eq. (18),
with the substitutions: pZxfc, ksZksc, kgZkgc, the
authors obtained the following equations
3$xfc ð1KAÞ
keff Z ksc 1K
(62)
2 C A C xfc ð1KAÞ
where
ksc Z ks C
kfc Z kf C
AZ
1
kfKks
psc
psc
C 1K
3$ks
1Kpfc
pfc
C 3$k
f
1
ksKkf
kfc
ksc
(63a)
(63b)
(63c)
psc Z pKpm ðmO 1Þ
pfc Z pn ð0! n! 1Þ
Fig. 27. Thermal conductivities determined from the Hadley model
vs. measured values for a large sampling of previously published twoand three-phase data [26].
(63d, e)
xfc C xsc Z 1
(63f)
p Z psc ð1Kxfc Þ C pfc xfc
(63g)
where p is the bulk porosity in the packed bed, psc is the
solid continuous region (psc(p), pfc is the fluid
continuous region (pfcOp), and xfc is the volume
fraction of the fluid-continuous region. In order to
390
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
estimate the parameters in Eq. (63a,b), i.e. the thermal
conductivity of the solid-continuous phase, and the gascontinuous phase, Eqs. (18) and (19) should be used,
respectively, with the appropriate use of porosity.
Parameters m and n, which are the greatest drawback
to the use of this model, can be determined using leastsquare curve-fitting techniques (in this case: mZ2.127,
nZ0.929, with a correlation coefficient 0.999). No
assumptions have been made about geometric characteristics, so the values of m and n derived for fused
silica should be applicable for any powder that is
approximately similar to the fused silica powder (silicarich glasses, anhydrous silicate minerals). Expected
values for granular media, with a large kf/ks-ratio:
m2[1.9; 2.3], and n2[0.88; 0.98].
Fig. 29 shows the data obtained with fused-silica,
together with the boundaries given by Eqs. (18) and
(19), and the predictions obtained using Eq. (62), with
mZ2.127 and nZ0.929. The semi-empirical model
defined by Eqs. (62) and (63) gives an excellent
representation for the fused silica. Fig. 30 shows that
Hadley’s model underpredicts the Nimick and Leith
least-square-based fit of fused silica-air measurements
by 3–8%.
3.2.1.3. Additive models that include radiation effects.
Models, which consider radiation as a mechanism
contributing to the overall heat transfer, will be
discussed below. The effective thermal conductivity is
affected by radiation in two ways:
1. Radiation across solid transparent medium
2. Radiation across voids (important here)
Fig. 29. Comparison of model calculations based on Eq. (62) with
experimental data for fused silica powder (Nimick, 1990).
Fig. 30. Comparison of the prediction of Hadley [26] with
experimental data, bounding estimates and Eq. (62) [18].
The contribution of the first mechanism (photon
conductivity) is usually implemented into the thermal
conductivity of the solid medium with zero porosity,
and can be determined by solving radiation and
conduction simultaneously. The contribution by the
second mechanism, which is important in this case, is
dependent on the porosity and the pore size distribution
in the case of slags, or particle size distribution in the
case of particulate materials.
Laubitz [10] measured the effective thermal conductivity of several powders (MgO, Al2O3, ZrO2) and
developed the model equivalent to Eq. (20). Powders
with particle sizes dpZ0.011, 0.040, 0.140 cm were
used (several powders of uniform dp, and one of graded
particle size), in the temperature range TZ100–
1000 (8C). The Laubitz model is based on the onedimensional radiant heat transfer through a void,
containing solid cubical obstacles.
In order to include radiation in the models outlined
previously, some structural changes have to be made.
To obtain a finite radiation path, particles must be
randomly positioned in their respective unit cells,
which is possible only in the models by Woodside and
Russell. The Laubitz model for the effective thermal
conductivity of powders, which includes the contributions of a two-phase gas–solid conductivity (first
term) and an equivalent radiation thermal conductivity
(second term), is
d
keff Z 2$kðRÞ C 4$s$T 3 $3 ð1Kp2=3 C p4=3 Þ
p
(64)
391
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
z
t1
t2
d
r
x
–x
Fig. 31. Thermal conductivity for Al2O3 and ZrO2 (porous bubblesceramic shell surrounding air) [10]. CCC experimental points for
Al2O3; !!! experimental points for ZrO2.
where k(R) is the thermal conductivity calculated
according to the Russell model (W/m K). A detailed
derivation of the radiative contribution is provided in
Appendix C. Reasons for doubling the value of the
Russel model for the thermal conductivity are given in
Appendix D. Experimental results obtained by Laubitz
are shown in Figs. 31 and 32.
Laubitz’s experiments showed that this model
cannot be used for calculating the thermal conductivity
of powders with graded particle size. Botterill [27,28]
concluded that the Laubitz model gives arbitrary
predictions, which are good for alumina but not for
silica. Though this model predicts higher values of the
thermal conductivity compared to the other models
investigated in [27,28], it predicts more accurate by the
temperature dependence of the conductive component
of the effective thermal conductivity.
Schotte [29] developed a model in order to
determine the radiation contribution to the thermal
conductivity of packed beds.
When considering the radiation from a plane located
on one side of a particle, to a plane located on the far
x
Fig. 33. Model for derivation of the radiation contribution to the
thermal conductivity of a packed bed [29].
side of the particle, as shown in Fig. 33, two
mechanisms may be included into the analysis. First,
there is radiation from the particle, in series with the
conduction through the particle, i.e. the first term in Eq.
(65), and radiation across the void space past the
particle, i.e. the second term of Eq. (65)
p 2 ks hdp
dt
p 2 p
dt
d
Kh
d
q ZK dp
ks C hdp
4
dx
4 p 1Kp p dx
(65)
where q is the rate of heat transfer. In order to determine
the thermal conductivity correction caused by the
radiation heat transfer, the rate of heat transfer across
the total area can be written as:
p 2 1
dt
dp
(66)
q ZKkr
4 1Kp dx
The radiation contribution is now found by equating
the right-hand sides of Eqs. (65) and (66)
kr Z
1Kp
C p$kr0
C k10
1
ks
(67)
r
where
kr0 Z 4$s$3$dp $T 3
(68)
According to Vortmeyer [19], the final equation of
this model can be written as
keff
ðk Þ
1Kp
Z eff c C 1
C p$Nuw ð2
1
kg
kg
C
X
Nu ð2K0:2643Þ
w
K0:264$3ÞX
Fig. 32. Thermal conductivity for Al2O3 (graded particle size):
CCC experimental points [10].
(69)
where (keff)c is the effective thermal conductivity
without radiation, XZks/kg, and:
392
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
Nuw Z
4$s$dp $T 3 1
2=3K0:264 ks
(70)
Godbee and Ziegler [30] developed a model for the
thermal conductivity of powders, applying MgO,
Al2O3, and ZrO2 powders (dpZ211–1023 mm) in their
experiments. They assumed one dimensional heat
transfer, with a unit cell given in Fig. 34, isotherms
being parallel planes perpendicular to the direction of
heat transfer, and with the pressure sufficiently high to
have no influence on the gas conductivity.
Godbee and Ziegler expressed the effective thermal
conductivity as the sum of three components
(71)
keff Z ksc C kgsc C kr
where ksc is the conductivity by solid only, kgsc is the
conductivity by solid and gas, and kr is the conductivity
by radiation. The influence of gas conductivity on the
thermal conductivity of powders is greater than the
influence of the solid conductivity, so the authors
assumed ksc to be negligible. Thus, the simplified
expression for kgsc is
kgsc Z
kc
Ds =X
ð1KS=X 2 ÞCðkd =kc ÞðS=X 2 Þ
C 1K DXs
(72)
where Ds is the solid length parallel to the heat flow
(considered to be equal to the mean particle size), S is
the solid area perpendicular to heat flow, and X is the
length of representative cell. This equation was used in
the calculations done by Botterill [27,28].
Some investigations, e.g. Woodside [20] have
shown that the thermal conductivity of a powder–gas
system decreases with decreasing pressure, at a much
faster rate than can be explained by the decrease in
conductivity of the pure gas by reduced pressure. In
order to correct for this, the following equation is
derived
kgsc Z h
kg
Ds =X
ðkg =kg Þð1KS=XK2 ÞCðkd =kg ÞðS=X 2 Þ
where
i
C 1K DXs
kg Z kg ð1KS=X 2 Þ C kg0 ðS=X 2 Þ
(73)
(74a)
kg
kg0 Z h
2Ka g T i
1 CZ a
1Cg
Pdf2 N
(74b)
pr
and a is the thermal accommodation coefficient for the
gas–solid surface, g is the ratio of the gas heat capacity
at constant pressure, respectively, at constant volume, d
is the distance between close parallel plates as shown in
Fig. 34, i.e. dZXKDs, T is the absolute temperature, P
is the absolute pressure, f is the molecular diameter of
gas as determined from viscosity (using the kinetic
theory for gases), NprZhCpg/kg is the Prandtl number, h
is the viscosity of gas, Cpg is the gas heat capacity at
constant pressure, and ZZ1.26!10K19 for cm g K
units (1.26!1024 m kg K) (a constant comprised of
fundamental constants and conservation factors).
The parameters describing the bed in this model
were all evaluated from a log probability plot of the
sieve size against the accumulative weight fraction,
according to Woodside [20]
Ds =X Z ðVd =aÞ1=3
(75)
where
a h S=D2s
(76)
and Vd is the volume fraction of the discrete phase, and
a is a shape factor [20]. The quantity a is a property of
particles, and can be related to the heterogeneous body
through a volume balance of two phases. For a, the
authors used the following equation
aZ
D
ðb
Da
2
eKu =2
du
ð2pÞ1=2
(77)
where
u Z ð1=Sln ÞlnðD=D50% Þ
Fig. 34. Unit cell model for model of Godbee and Ziegler [30].
(78)
and Da, Db are the lower and upper points of truncation
(mm) (or smallest and largest particle diameter), D is the
mean particle diameter, D50% is the median particle
diameter, Sln is the logarithmic (to the base e) standard
deviation. Results for some of the powders used in these
experiments are given in Table 3.
A value for S/X2 is obtained with Eq. (75) and the
fact that SDsZVdX3:
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
393
Table 3
Parameters of magnesia, alumina and zirconia powders obtained from
screen analysis [30]
expression, covering a number of radiative models, is
derived
Powder
Median particle size, D50%
(mm)
Shape factor, a
kr Z 4$s$c$d$T 3
MgO
MgO
Al2O3
ZrO2
ZrO2
180
235
166
198
333
0.739
0.735
0.737
0.853
0.851
where c is the radiative exchange factor. Radiative
exchange factors for different models are provided in
Table 4. This parameter is further discussed in
Section 6.
The radiative exchange factor, c, is a function of the
size and the emissivity of the pores or particles.
Botterill’s study showed that:
S
Z ða$Vd2 Þ1=3
X2
(79)
The equation for radiative conductivity is given as
kr Z 4n2r s3ð1=Vd K1ÞDs T 3
(80)
where nr is the refractive index of the media between
surfaces.
According to Botterill [27,28], the Godbee and
Ziegler model gives a conductive component that is
consistent with the experimental results, and a radiative
component that is generally larger than those of the
other unit cell models. Fig. 35 shows how this model
predicts the temperature dependence of thermal
conductivity of silica sand. As the emissivity increases,
the temperature dependence of the model curve
increases but even with the maximum emissivity
value of unity, the temperature dependence never
attains that experimentally observed value.
Botterill [27,28] tested a number of published
models, using experimental data obtained on alumina
and sand, in the temperature range TZ400–950 (C. An
(81)
† Models, which characterize packing using the
voidage, predict similar values of thermal conductivity.
† Models, which take into account the particle size
distribution, give very similar estimates (Godbee
and Ziegler; Bauer, Schlunder; Zehner).
† According to the experiments, most of the models
reported cannot deal with packing substantially
different to the originally tested.
† Good agreement with the model predictions is
obtained, when experimental measurements at a
ambient temperature are used (when radiation is
negligible).
† None of the tested models could predict the
temperature dependence of the effective thermal
conductivity at high temperatures. Possible reason
for this could be that particles are partially
transparent to radiation and not opaque as the unit
cell and pseudohomogeneous models assume (Al, Si
transmit certain frequencies).
† According to Botterill, the models of Godbee and
Ziegler, and Kunii and Smith give reasonable
predictions at low temperatures. The radiative
models of Zehner, and Kunii and Smith give good
estimates for the thermal conductivity for packed
bed of particles at high temperatures with some
modifications (given in Nasr [9]).
Table 4
Radiation exchange factor for various unit cell models [9,28]
Fig. 35. Comparison between the effective thermal conductivity for a
bed of 410 mm silica sand, of different emissivities, and predictions by
the Godbee and Ziegler model; (1) exp. values; (2) 3/0; (3) 3/0.37;
(4) 3/1; (5) 3/0 and silica conductivity perpendicular to c-axis.
[28].
Authors
c
Godbee and Ziegler
Laubitz
Schotte
Zehner; Bauer and
Schlunder
Kunii and Smith
3/(1Kp)
ðð1Kð1KpÞ2=3 C ð1KpÞ4=3 Þ=ð1KpÞÞ3
3
3/(2K3)
p: porosity; 3: emissivity.
1=ð1C ðp=1KpÞð1K3=23ÞÞ
394
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
Mills and Rhine [31] measured the thermal
properties of gasifier coal slags in the temperature
range 298–1800 K, and developed a model for the
radiant component of the heat conduction. The radiant
conductivity increases with the optical thickness,
i.e. a$d (where a is the absorption coefficient and d
is the slag thickness), until a constant value is
achieved. At this point a$d is greater than 3.5, and
the slag is said to be ‘optically thick’. The radiative
thermal conductivity of optically thick slag can be
determined as
16$s$n2 $T 3
kr Z
3$a
2
∆l
3
4
6
1
Fig. 36. A model of heat transfer in a packed bed (adapted from [34]).
(82)
where n is the refractive index. Since it is very
difficult to measure the absorption coefficient of
molten slags, measurements made at 298 K, have
been assumed to give an approximate value for a at
high temperature. Since the absorption coefficient is
highly dependent on the concentration of Fe2C in the
slag, Miles and Rhine suggested that it could be
determined, using the following empirical rule
(particle size range 2–4.5 mm):
a Z 11ðwt% FeOÞ
5
(83)
6. Heat transfer by convection, solid–fluid–solid
7. Heat transfer by lateral mixing
When the Reynolds number is small, the boundary
layers around the solid packing are thick, and therefore
mechanisms 1, 3, 4 and 5 above, are predominant. In
case of gas-filled voids, the model equation is
dp $hrv
keff
bð1KpÞ
C p$b
Z k
g
1
kg
kg
g ks C ð1=4ÞCðdp $hrs =kg Þ
(84)
According to Mills, the absorption coefficient
predicted using this rule, are in reasonable agreement
with the experimental data reported by Finn et al.
b Z lp =dp
(85a)
g Z ls =dp
(85b)
3.2.2. Resistance network models
The following group of models includes the
radiative conductivity as a resistance in the network
of resistances, representing the unit cell. In these
models, it is assumed that there is dependence between
the conductive and the radiative component.
Yagi and Kunii [33] developed a semi-empirical
model for estimating the thermal conductivity of
packed beds. The model is divided into two terms,
one term being independent of the fluid flow, and the
other one dependent on the lateral mixing of the fluid in
the packed beds.
Heat transfer mechanisms, considered are (Fig. 36):
4 Z lv =dp
(85c)
hrs Z 0:1952ð3=ð2K3ÞÞððt C 273Þ=100Þ3
(85d)
1. Thermal conduction through the solid phase
2. Thermal conduction through the contact surface
between neighboring particles
3. Radiant heat transfer between adjacent solid
surfaces
4. Radiant heat transfer between neighboring void
spaces
5. Thermal conduction through the gas film near the
contact surfaces between adjacent particles
2
hrv Z 4
0:1952
p
1K3
1 C 2ð1K
pÞ 3
3
5ððt C 273Þ=100Þ3
(85e)
where hrs is the effective radiation heat transfer
coefficient at the contact surface, hrv is the effective
radiation heat transfer coefficient of the voids, t is the
temperature (8C), lp is the average length between the
centers of two neighboring solids in the direction of
heat flow, ls is the equivalent thickness a layer of solid
should have to represent the same thermal resistance as
the sphere, i.e. lsZ2dp/3, and lv is the equivalent
thickness a layer of fluid should have to represent the
same thermal resistance as the fluid film.
Petersen (as stated in [34]) formulated a following
expression for the calculation of parameter lv
0:3716
kg
p1:7304
(86)
lv Z 0:15912$b
kair
395
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
where kair is the thermal conductivity of air, calculated
using the following correlation [34]
kair ðW=mKÞ Z2:286 !10K11 $T 3 K7:022
K8
2
K4
!10 $T C 1:209 !10 $T
K5:321 !10K3
ð87Þ
where T [Z] K.
In practical packed beds, b takes a limited value
between 0.82 and 1, for a wide range of packing
characteristics. A possible way to estimate the
parameter lp, which is used to determine b, is given in
Appendix E. Parameters g and f are to be determined
experimentally, and according to existing data they can
be approximated as g z1, respectively, f z0.04.
Kunii and Smith [35] developed a model for the
effective thermal conductivity of beds of unconsolidated particles containing a stagnant fluid, which is then
extended to the beds of consolidate particles (materials
like sandstone or porous metals). The authors assumed
1-dimensional heat flow with constant linear heat flux,
through the gas and the solid, and solid–solid contact
conduction to be negligible for atmospheric pressure
conditions. This model is semi-empirical, involving
parameters that were determined using an extensive
comparison with the available experimental data [28].
Similar to the Yagi and Kunii model, heat transfer is
assumed to occur in the vertical direction accounting
the following mechanisms:
1. Heat transfer through the fluid in the void space by
conduction, and radiation between adjacent voids
(when the voids are assumed to contain a nonabsorbing gas)
2. Heat transfer through a solid phase
a. Heat transfer through contact surfaces, between
the solid particles.
b. Conduction through the stagnant fluid near the
contact surface.
c. Radiation between surfaces of solid (when voids
are assumed to contain non-absorbing gas).
d. Conduction through a solid phase.
Mechanisms 1 and 2 are in parallel to each other.
Mechanism d is in series with the combined result of
mechanisms a, b and c. Mechanism a is only important
at low pressures, and when it is neglected, the model is
simplified to
b$hrv $dp
keff
bð1KpÞ
3
Zp 1C
(88)
C2
kg
kg
6 1 g7
C k5
4
1 dp hrs
dC kg
where gZ2/3, and b is in the range 0.9 (for close
packing) to 1.0 (loose packing)
d Z d2 C ðd1 Kd2 Þ
pK0:26
;
0:216
0:26! p! 0:476
d Z d1 ;
pO 0:476
d Z d2 ;
p! 0:260
(89a)
d1 Z
1 2
0:352 kK
k
1 2
ln½kK0:545ðkK1ÞK0:455 kK
K 3k
k
1 2
0:072 kK
k
1 2
d2 Z
ln½kK0:925ðkK1ÞK0:075 kK
K 3k
k
4$s$T 3
i
hrv Z h
p
ð1K3Þ
1 C ð1K
pÞ 23
hrs Z 4$s$T 3
3
2K3
(89b)
(89c)
(89d)
(89e)
The radiant heat transfer coefficients were
calculated from expressions derived by Yagi and
Kunii. The data reduction procedure involved a
temperature-dependent fluid thermal conductivity, a
temperature-independent solid thermal conductivity,
and the solid-surface emissivity. According to the
model, the temperature dependence of the heat
conductivity of solids can be neglected (25%
decrease of ks leads to 5% decrease of effective
thermal conductivity).
The model of Kunii and Smith was evaluated using
the experimental data in Nasr et al. [9]. It can be seen
that the accuracy of model predictions depends on the
particle composition and diameter. Good results were
obtained for alumina–air packed bed (dpz6.6 mm), at
temperatures below 900 K. For smaller particle size
(dpz2.8 mm), better results were obtained for glass–air
beds, at temperatures below 600 K. According to
Botterill [27,28], the model gives reasonable predictions of the conductive component for the silica beds, at
lower temperatures, when radiant transfer is negligible
(up to around 900 8C).
As mentioned, this model was extended to predict
the effective thermal conductivity of consolidate porous
media. The consolidation might occur by partially
clogging with a cementing substance or by sintering.
The idea of clogging of the original bed of the
unconsolidated particles, with a cementing substance
or by sintering, is shown in Fig. 37.
396
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
Heat
flow
z
Dp
Contact area
dr
Fig. 37. Kunii and Smith model of consolidated porous media [35].
α
r
Now, heat is flowing in the vertical direction through
the void space, p 0 and through the solid fraction, 1Kp 0 .
In this latter part, the flow is in series through a solid of
thickness l 0 , and a fluid of thickness lv. The resulting
equation for the effective thermal conductivity is
ð1Kp 0 Þ 1 C p 0 d
0
kg
dp $hrv
p
keff
p
Z p0 C p
1C d C
ðp 0 =pÞ
p
ks
ks
ks
1 C kg dp ðhpChrs Þ
1
d
ks
C
ks
(90)
where p 0 is the void fraction of the bed of consolidated
porous media and hp is the heat transfer coefficient
representing the heat transfer rate through the contact
surface between solid particles in a bed of unconsolidated particles or between clogged particles in a
consolidated bed.
The heat transfer through the contact surface between
consolidated particles, as represented by the dimensionless group hpdp/ks is difficult to quantify. It is regarded as
a consolidation parameter and determined by comparing
experimental data with the form of Eq.(90), when
radiation was neglected. This mechanism is influenced
by the characteristics of solid material and the type of
consolidation, but does not depend upon the fluid in the
pores. The experiments showed that when a value of
0.075 is assumed for the dimensionless group, agreement with the experimental data is satisfactory.
Bauer and Schlunder [36] developed a model for
random packing of uniform spheres, which takes into
account the radiation and pressure effects. The same
data reduction procedure, as in Kunii and Smith [35],
was applied. The model assumes heat transfer by
conduction, through a bed of non-conductive particles
surrounded by a conducting gas. Analogy with the
diffusion in a packed bed was considered. Fig. 38 shows
the unit cell for the model.
Heat flow is divided into three parallel paths:
† Conduction and radiation through
the gas-filled
pffiffiffiffiffiffiffiffiffiffi
voids within the area fraction 1K 1Kp
1-α
1-√(1-p)
√(1-p)
Fig. 38. Unit cell for model of Bauer and Schlunder [36].
† Conduction through the solid and the gas phase,
with radiation between
solid surfaces; within the
pffiffiffiffiffiffiffiffiffiffi
area fraction ð1KaÞ 1Kp
† Solid–solid
conduction, within the area fraction
pffiffiffiffiffiffiffiffiffiffi
a 1Kp
where p is the bed voidage, and a is the area fraction
of the solid-solid contact conduction path. According to
Nasr et al. [9], the final model equations are:
pffiffiffiffiffiffiffiffiffiffi
keff
k
Z ð1K 1KpÞ 1 C p r
kg
kg
pffiffiffiffiffiffiffiffiffiffi ks
kso
C 1Kp a C ð1KaÞ
(91)
kg
kg
1
2
0
kg
ks
kr
ks
kr
C
C
K1
B
kg
kg
kg
kg
ks
kso
2
A K BK1
Z 4
ln@
N
N
B
kg
N2
k
B C1
C r KB
2$B
kg
kg
kr
N Z 1C
KB
kg
ks
1
1 C Biox
ks Z ks
Biox Z
Sox
kox
ks
dp
(92a)
(92b)
(92c)
(92d)
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
3
x $T 3
kr Z 4$s
2K3 R
(92e)
xR Z Rform $dp
(92f)
B Z Cform
aZ
1Kp
p
10=9
23$r2k
1 C 22$rk4=3
(92g)
(92h)
where Cform is the particle shape factor, Rform is the
determinative influence coefficient for the transport
mechanism resulting from thermal radiation (for
further details, see [9]), xR is the radiation effect
factor, r2k is the contact conduction factor, and
(Sox/kox) is the oxidation effect factor ([9]) equal to
the ratio between the oxide layer thickness, and its
thermal conductivity. Parameters Cform, Rform, r2k ,
(Sox/kox) must be determined experimentally. For
monodisperse spherical packing CsphereZ1,25 when
RsphereZ1.
The experiments done by Nasr et al. [9] showed that
this model overpredicts the measured values of the
effective thermal conductivity at low temperatures
(around 20%, for temperatures below 700 K), but
gives reasonably good results at higher temperatures
(where radiation is important).
3.3. Pseudohomogeneous models
This group of models considers packed beds to be
a continuum for radiation. The bed is considered to
be a pseudo homogeneous medium, through which
radiation can penetrate freely, so equations describing radiation through an absorbing, emitting, and
scattering medium are used:
dI
ZKða C bÞI C bK C asT 4
dz
(93)
dK
K ZKða C bÞK C bI C asT 4
dz
(94)
where a is the absorption cross-section per unit
volume of bed (mK1), b is the backscattering crosssection per unit volume of bed (mK1), I is the
forward component of the radiative flux (W/m2), and
K is the backward component of the radiative flux
(W/m2), (both fluxes are parallel to the z-axis).
According to [37], Eq. (93) states that in traversing
397
the distance dz, the radiant flux density in the
forward direction is decreased by bIdz due to back
scattering and aIdz due to absorption; it is increased
by bKdz due to backscattering of the backward
radiant flux density and asT4dz due to radiation. Eq.
(94) is the analogous balance for the backward
radiant flux density. Eqs. (93) and (94) imply that
the spectral distribution of the intercepted and
reemitted radiation is the same, which is a
reasonable assumption except for the situation with
extreme temperature gradients.
The net radiative flux is:
qr Z I KK
(95)
It was assumed that the medium is grey, the
scattering is isotropic, and that the Schuster–Schwarzchild approximation (two flux model) is valid, i.e.
radiation can be scattered only in two directions,
backward and forward.
The framework for pseudo homogeneous models,
requires radiative heat transfer through a packed bed to
be governed by the equation of heat transfer. Thus, the
radiative transport (i.e. the radiative heat flux) may be
accurately evaluated even for a comparatively thin
packed bed, by numerical solution of the heat transfer
equation. This is not possible for the cell models,
because they do not take into account the long-range
effects of radiative transfer through the voids of a
packed bed.
It is required that the radiative properties, which
appear in the equation of heat transfer, are known. It
is essential for the pseudo homogeneous models, to
describe these quantities in terms of the void
fraction, the particle size and the surface reflectivity
of the particles. According to Kamiuto et al. [38],
reflectivity cannot be determined by independent
scattering theory such as Mie theory,2 since it is not
appropriate for this purpose.
Vortmeyer [19] had compared the unit cell and
pseudo homogeneous models in handling radiation in
packed beds. His conclusions are:
(1) In the cell models, the opacity of the packed bed is
large, and the particles are much larger than the
wavelength of radiation.
(2) In the pseudo homogeneous models, the dispersed
phase is considered to be a continuum for radiation.
The effective thermal conductivity is given in terms
2
Only the interactions of a single particle with light of arbitrary
wavelength are considered.
398
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
of the effective scattering and an extinction
coefficient.3
T1f
θ
T1s
3.4. Dedicated models developed for ash deposits
T2s
r
Using three characteristic values of the solid volume
fraction, mean particle size, and tortuosity, Baxter [39]
developed a model connecting the average thermal
conductivity to the structural properties of a boiler ash
deposit. The void space was assumed to be nonconducting.
The equation for 1D, transient heat transfer through
a cylindrical surface, may be written as:
vT
1 v
vT
1 v2 T
Z keff
r
r$Cp
(96)
C 2
vt
r vr
vr
r vq2
The model system for heat transfer through a
cylindrical wall is shown in Fig. 39. When the transient
term is ignored, the model for temperature field through
the cylindrical body is
q$r
r
TðrÞ Z T1 K 1 ln
(97)
r1
k
where r1, r2 are radial directions of the deposit surface.
Eq. (97) reduces to a linear dependence of the
deposit temperature on distance, i.e. plate geometry, in
the limit of a small deposit thickness relative to the
radius of curvature.
The relationship between the effective thermal
conductivity, the thermal conductivity, the porosity,
and the tortuosity, is given by:
keff Z
kð1KpÞ
t
(98)
The tortuosity represents the shortest average path
length between two points, divided by the straight-line
distance between the same points. Thus, as the solid
phase becomes less connected, the tortuosity increases.
The change in tortuosity as a function of the degree of
sintering is illustrated in Fig. 40. Two values of the
tortuosity were used this study, 1 and 2. Deposits with a
solid fraction lower than 0.5, and tortuosities higher than
2, are common in many systems. According to Fig. 41,
the temperature range scales linearly on the tortuosity
and inversely on the porosity, meaning that a change in
3
The fraction of light lost to scattering and absorption per unit
distance in a participating medium. Normally, it is given in standard
units as a fraction per meter. It equals the sum of the absorption
coefficient and the scattering coefficient.
Heat flow
1 – hot fluid
(flue gas)
2 – coldfluid
(steam)
f – fluid
s – surface
Fig. 39. Heat transfer through a cylindrical wall.
either property, changes the difference between the
deposit surface temperature and the tube surface by the
same factor. The extent of curvature is determined by the
system geometry, not by the real deposit physical
properties.
Robinson et al. [2] made in situ measurements of the
thermal conductivity of deposits formed during the
combustion of a mixture of coal and wheat straw
(65/35 wt%). This is the first time, the deposit
conductivity has been measured on original, noncrushed deposit samples.
The model describes the thermal conductivity of a
layered structure, since the analysis of ash deposits
showed that sintering creates a layered deposit
structure with a relatively unsintered innermost
layer. Assuming steady state, two-dimensional heat
transfer through the deposit and a uniform deposit
thermal conductivity, the temperature distribution
within the deposit can be described using Eq.
(130). The average radial temperature gradient at
the inside edge of the deposit is
1
dT Z
2p
2p
ð
0
vT
j ðqÞdq
vr in
(99)
where ((T/(r)(in(q) is the radial temperature gradient
as a function of q, determined from a numerical
solution of Eq. (40). The effective thermal conductivity of the deposit is
keff Z
Q
2$p$rin $L$dT
(100)
where rin is the radius of the inside surface of the
deposit, and L is the length of the tube. If
the deposit is treated as a two-layer material (the
simplest case), its overall conductivity is
keff Z
L
kun
1
LÞ
C ð1K
ksi
(101)
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
399
Fig. 40. Conceptual illustration of the changes in contacting efficiency and tortuosity with sintering/melting [39].
where L is the thickness of the unsintered layer,
normalized by the total deposit thickness (this value
was kept constant, Lz0,4), kun is the thermal
conductivity of the unsintered (inner) layer of the
deposit and ksi is the thermal conductivity of sintered
(outer) layer of the deposit.
2. Estimation of the conduction part of the thermal
conductivity of the zero-density material at the
given temperature.
3. Estimation of the effective thermal conductivity of
the slag or particulate material or partially
sintered deposit.
4. Taking into account the effects of radiation.
4. Estimation of thermal conductivity of ash deposits
According to Gupta et al. [4], practical estimation of
the thermal conductivity of ash deposits should consist
of:
1. Estimation of the thermal conductivity of the zerodensity material at room temperature.
4.1. Estimation of the general parameters
4.1.1. Structural parameters
The mean particle size, dp, which characterizes the
bed packing, can be calculated as
"
#K1
N
X
Dmi
dp Z
(102)
dpi
iZ1
where Dmi is the weight fraction of the ith sieve
fraction, and dpi is the mean sieve spacing for the ith
sieve fraction (m).
The structural parameter c, as defined in Robinson
et al. [2] is:
cðfÞ Z
Fig. 41. The deposit surface temperature as a function of the porosity
and emissivity, assuming no intra-deposit radiative heat transfer and a
non-conducting gas phase. Tortuosity is assumed to be unity [39].
kmeas ðfÞKklow ðfÞ
; 0% cðfÞ% 1
kup ðfÞKklow ðfÞ
(103)
where kmeas is the measured value of deposit thermal
conductivity, kup is the upper limit of the thermal
conductivity of deposits, defined by Eq. (18), klow is the
lower limit of the thermal conductivity of deposits,
defined by Eq. (19), and f is the volume fraction of
solid in the deposit. This parameter provides a useful
measure of the microstructure of the deposit. Using
400
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
Eqs. (18) and (19) as boundary values of the thermal
conductivity, the structural parameter can have values
between 0 and 1. Thus, as shown in Fig. 9, values close
to 0 will indicate a deposit with a more layered
structure, while values close to 1 indicate a more
columnar structure.
4.1.1.1. Porosity change with time. The temperature
profile through a deposit layer increases with time,
which causes densification of the deposit. This affects
the porosity of the deposit, and thereby the thermal
conductivity, which will again affect the rate of heat
transfer by conduction, and thereby the temperature
profile in the deposit.
According to Rezaei et al. [40], during the process of
deposit densification, the porosity will decrease while
the extent of shrinkage DV/V will increase. Two groups
of models, which describe the porosity change during
shrinkage, can be distinguished [40].
The first group is based on geometrical assumptions
that relate the porous system to an idealized system, i.e.
assembles of spherical, viscous particles whose
dimensions and physical properties (surface tension)
and viscosity remain constant during processing.
The second group of models, which are based on a
phenomenological approach to densification process.
Frenkel (in [40]) was first to propose a model for the
sintering of viscous material from geometric assumptions. He modelled the first stage of vitrification, by
equating the energy variation due to a decrease in the
surface area, to the energy dissipated by viscous flow
DV=V ZK9$s$t=ð4$h$ro Þ
(104)
where s is the surface tension (according to Senior [41]
swT, can be neglected), t is time of sintering, h is the ash
viscosity (hwexp T [41]), and r0 is the initial pore radius
(developed on the basis of Scherer’s structural model for
densification of a cubic lattice; the pore radius was set
equal to half the average particle diameter). According to
this model, the volume shrinkage is proportional to the
duration of the thermal treatment.
The shrinkage level or degree of sintering, is defined
as the ratio of the neck radius, x, to the radius of
particles, r, as shown in Fig. 42.
Using the Pythagoras Theorem, we can obtain the
shrinkage level as
x=r Z ððr2=3 Kr0 Þ=ðr2=3 ÞÞ1=2
(105)
where r0 is the density of the initial powder, and r is the
density of the sintered product. The derivation of Eq.
(105) is shown in Appendix F. The powder density at
r0
x
r
Fig. 42. Schematic diagram of the sintering of two spheres [40].
time t, can be expressed as a function of the porosity p
rðtÞ Z rs ð1KpðtÞÞ
(106)
where rs is the density of solid phase. For the initial
density r0Zr (tZ0), i.e. p (tZ0)Zp0. Thus, the
porosity of sintered product at time t can be obtained
from Eqs. (105) and (106) as:
p Z 1K
1Kp0
ð1Kx2 =r 2 Þ2=3
(107)
This model is limited to a shrinkage level of 0.3 [40].
Assuming that during the time period t, the original
radius r0 of the particle does not change much, i.e.
rtzr0, the ratio of the pore volume at time t to the initial
pore volume is:
Vpt =Vpo Z 1K3=2ðx2 =r02 Þ
(108)
Comparing this equation to Eq. (104), x2 equals
3str/2m, thus for the ratio of pore volumes we obtain:
Vpt =Vpo Z 1K9$s$t=ð4$m$ro Þ
(109)
Ivsen (in [40]) connected the pore volume change to
the density change during sintering, using phenomenological assumptions, as:
Vpt
r ðrKrt Þ
Z 0
rt ðrKr0 Þ
Vpo
(110)
According to Rezaei et al. [40], the porosity at time t
can be obtained using Eqs. (106) and (110), as:
9st
1K 4mr
po
p
pZ
(111)
9st
1Kpo C 1K 4mr
p
o
p
Mackenzie and Shuttleworth (1949), suggested the
following band between the actual porosity p at time t,
and the initial porosity po (source: Senior [41]):
p
3$s$t
exp K
(112)
p0
2$rp $m
This model can be used for sintering from app.
50–0% porosity.
401
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
In the above models, an estimate of the surface
tension and the viscosity in each deposit sub-layer is
needed. The surface tension of an amorphous phase can
be estimated from its chemical composition. As stated
in Rezaei et al. [40], Tillotson and Oppen showed that
the surface tension of a material could be expressed as a
linear function of its oxide composition:
N
X
sZ
ðxi $si =100Þ
(113)
iZ1
where si is the surface tension corresponding to each
oxide i, and xi is its molar percentage. The surface
tension of each oxide is calculated from
0:004
si Z Ai C
DT ½N=m
(114)
100
where Ai (N/m) is a component-specific constant,
provided in Table 5 for some oxides present in ash
deposits, DT is the temperature difference to reference
temperature TrefZ1300 8C (negative value, if the actual
temperature is less than Tref).
Concerning the viscosity of a multicomponent
oxide mixture, Vargas et al. [42] provided an outline
of the existing experimental viscosity data, as well
as models for estimation of the viscosity as a
function of the chemical composition and the
temperature. In this work, the models of Hoy et al.
(1965), Watt and Fereday (1969), Urbain et al.
(1981), Kalmanovitch and Frank (1988), and Senior
and Srinivasachar (1995), among others, are
discussed.
Senior [41] experimentally determined a change in
deposit porosity with time, for coal-derived ash deposit,
as shown in Fig. 43.
4.1.2. Radiation parameters
Radiation parameters such as the refractive index
and other optical parameters, can be used in different
models in order to estimate the emissivity of a deposit.
Thus, some theoretical and semi-empirical methods for
estimating these parameters are outlined below.
Table 5
Pure component surface tensions, valid at 1300 8C
Component
Ai (N/m)
Component
Ai (N/m)
K2O
PbO
TiO2
SiO2
Na2O
0.01
0.12
0.25
0.29
0.295
Li2O
CaO
MgO
Al2O3
0.45
0.51
0.52
0.58
The coefficients are to be used in Eq. (107), for estimation of pure
oxides surface tensions [40].
Fig. 43. Densification of Wyodak-Rochelle deposit [41].
When radiation incident onto a particle, it is
absorbed (and then emitted), or scattered. The particle
properties defining these interactions are:
† Particle size parameter [43]
x Z pd=l
(115)
where d is the particle diameter and l is the
wavelength of the incident radiation.
† Complex refractive index
m Z nKiki
(116)
where n is the real refractive index. The parameter ki
is the absorption index, which is a measure of the
attenuation caused by the absorption of energy per
unit of distance that occurs in an electromagnetic
wave of a given wavelength propagating in a
material medium of a given refractive index. Both
indices depend on the wavelength of the light, and
the temperature. Complex refractive index is a
function of material, particle size, and temperature.
Values of m valid in room temperature, for slags
with composition similar to ash are provided by
Goodwin and Mitchner [44].
Due to the lack of reliable and sufficient experimental
data, most studies have assumed that the optical
constants n and ki for ash deposits are independent of
the wavelength, with n taken as 1.5 (because constituent
oxides have a real index of about 1.5 up to the
wavelength of 5 mm), and ki ranging from 0.005 to
0.05 [45]. Based on a limited amount of experimental
data, the complex refractive index of coal ash is known
to have a spectral character, with n varying from 1.5 to 2,
and k varying from very low values of 0.001 to about 1.
Bhattacharya [45] showed that for the coarse size
fraction of the fly ash (big particles, in the range 1.563–
212.65 mm), the effect of the real index is insignificant.
For the fine size fraction, where the effect of the real index
is more pronounced, the difference in total emittance over
402
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
the entire temperature range 800–1800 K, is around 0.05.
Calculations with a real index of 1.5, very closely match
the calculations using the spectral values. These
calculations suggest that the choice of real index is not
critical for the estimation of the hemispherical emittance
of an opaque ash deposit, unless the deposit is composed
of very fine particles. On the other hand, for both size
fractions, a significant effect of absorption index (iron
content) is observed. For a smooth deposit, which is
analogous to molten a slag, no significant variation with
the real as well as the absorption index was observed. It
was also observed that the real index, does not depend on
the chemical composition.
Goodwin and Mitchner [44] investigated the
influence of the chemical composition of coal ash
slags on their optical properties n and k. A mixture rule
was developed to allow calculation of the refractive
index n, as a function of its composition in the
wavelength range 1–8 mm. If the composition is
expressed on a mass basis, the value of n for the slag
may be written as
simplified dispersion equations of the form
X Xm;i n2i K1
n2 K1
Z
r
ri
n2 C 2
n2i C 2
i
Fi ðlÞ Z ðai $l2 Kbi Þ=ðci $l2 Kdi Þ
(117)
where r is the sample density, Xm,i is the mass fraction
of component i in the slag, and ri and ni are the density
and refractive index, respectively, for component i in a
pure reference’ state, for which refractive index data are
available. The slag components are taken to be
the oxides: SiO2, Al2O3, CaO, Fe2O3, TiO2, and
MgO. The refractive index of pure substances that
can be found in combustion systems is given in Table 6.
The reference states of these oxides are assumed to
be fused silica, saphire (Al2O3), crystalline CaO,
hematite (Fe2O3), rutile (TiO2), and crystalline MgO,
respectively. The refractive index values for these
minerals were taken from the literature and were fit to a
Table 6
Refractive index of pure substances that can be found in combustion
systems
Component
Silicon dioxide
Aluminum oxide
Iron (III) oxide
Titanium oxide
Calcium oxide
Magnesium oxide
Potassium oxide
Sodium oxide
SiO2
Al2O3
Fe2O3
TiO2
CaO
MgO
K2O
Na2O
Transmitted
color
Refractive
index
Colorless
White
Red or black
Colorless
Colorless
Colorless
White
White
1487
1768
3042
2615
1873
17,364
–
–
n2i K1 Z Ci C ðBi l2 Þ=ðl2 Kl20;i Þ
(118)
where Bi, Ci, and li are three parameters, the values of
which are provided in Table 7 for each oxide component,
along with their density in the reference state.
Owing to the insufficient experimental data, the
value of n for Fe2O3 is taken to be independent of
wavelength and equal to its value at visible wavelengths. The sample density (g/cm3) is given as
r Z 2:54 C 0:00978ðFe2 O3 Þ
(119)
where (Fe2O3) denotes the weight percent Fe2O3.
But since the above equation slightly underpredicts the
values of n, the authors tried to improve it by using two
empirical adjustments, obtaining a final mixture rule
X
n2 K1
Zr
Xm;i $Fi ðlÞ
2
n C2
i
(120)
where
(121)
Parameters ai, bi, ci and di are given in Table 8.
In order to develop a correlation for the absorption
index, ki, it was assumed that at infrared wavelengths ki
could be written as the sum of two terms describing the
infrared tail of the charge transfer band, and the
absorption due to Fe2C, respectively:
ki ðlÞ Z a1 expðb=lÞ C a2 gðlÞ
(122)
The exponent b and the function g(l) are taken to be
independent of composition. The final equation is
obtained as
ki ðlÞ Z 3:61 !10K7 $r2 $rð1
KrÞðFe2 O3 Þ2 expð1:75=lÞ C r$rðFe2 O3 Þ
!½0:0963 C 0:0011r$rðFe2 O3 ÞgðlÞ
(123)
where r is the sample density (Fe2O3) is the Fe2O3
weight percent, and r is the ferrous ratio:
Table 7
Dispersion equation parameters for pure oxides [44]
Oxide
Density (g/cm3)
Ci
Bi
lo,i (mm)
SiO2
Al2O3
CaO
Fe2O3
TiO2
MgO
2.20
3.97
3.31
5.24
4.86
3.58
1.104
2.082
2.31
8.364
5.031
1.962
0.8975
5.281
11.32
0.0
7.764
2.470
9.896
17.93
33.90
0.0
15.60
15.56
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
Table 8
Parameters defining the function Fi(l) in the mixture rule for n [44]
Oxide
ai
bi
ci
di
SiO2
Al2O3
CaO
Fe2O3
TiO2
MgO
0.9389
1.914
4.250
1.647
2.720
1.278
53.00
174.0
827.7
0.00
260.0
136.9
5.001
10.36
16.63
11.36
15.80
7.433
420.0
1634
6102
0.00
1954
1201
rZ
½Fe2C
½Fe2C C ½Fe3C
(124)
In the case of ash deposits, in the compositional
analysis all the iron is presented as FeC3, thus
parameter r can be taken to be 0.
The values of the function g(l) are shown in Fig. 44.
4.2. Thermal conductivity of gas and solid mixtures
gas mixtures
According to Botterill [27,28], the gas mixture in a
porous structure can be approximated with air. In the
present study, the thermal conductivity of air (W/m K)
has been calculated as a function of temperature, using
data from Touloukian [47]
kg Z 6 !10K12 T 3 K3 !10K8 T 2 C 9 !10K5 T C 0:0007
(125)
where T [Z] K.
403
4.2.1. Solid mixtures
Wall et al. [6] indicated that the thermal conductivity
of a deposit solid-phase cannot be approximated using a
weighted average, based on the mass fraction of the
oxides, although it presumably could be approximated
as the weighted average of the species actually present
in the deposit.
The particulate ash is seen to have a thermal
conductivity below that of the oxides. According to
Wall et al. [48], the thermal conductivity of an actual
furnace deposit is many times higher than that of the
particulates and is closer to that of the sintered or fused
ash, which in turn is close to the data for refractory
bricks. Robinson et al. [2] used a constant value for the
thermal conductivity of solid, ksZ3 W/m K, which is
comparable to the thermal conductivity of silicacontaining materials, at high temperatures.
Rezaei et al. [1] developed an empirical expression
for the thermal conductivity of the solid phase,
assuming it to be dependent only on temperature
ks Z 0:0015 !T 1:1
(126)
where ks is the thermal conductivity of the solid phase
(W/m K), and T is the temperature (K). The equation is
obtained using experimental data for sintered ash
samples.
In the present study, the thermal conductivity of the
solid phase was calculated as the sum of weighted
average of the oxides present in the ash deposit. The
thermal conductivity of the oxides present in the ash,
has been derived using data from Touloukian [47] and
Raznjevic [49].
4.2.1.1. Models for the thermal conductivity of solid
mixtures. Ratcliffe [50] developed an additive formula
based on derived oxide factors, which are claimed to
enable the thermal conductivities of most glasses within
a precision of 5% or better. In the measurements, the
electrically heated disc method was used. An additive
formula, for computing the thermal conductivities at
three temperatures (K100, 0, 100 8C), from the
percentage weight compositions of oxides, were
derived by solving 22 equations for the unknown
oxide factors
105 k Z x1 a C x2 b C x3 c C/C x12 m
Fig. 44. The two functions appearing in the correlation for ki (Eq.
(122)). The function e1.75/l characterize the tail of the Fe2C–Fe3C
charge transfer band, and the function g(l) characterized the Fe2C
absorption spectrum [44].
(127)
where k (cal/cm s C) is the thermal conductivity of
glass at a particular temperature, {x1,x2,.} are the
percentage weight compositions of component oxides,
and {a,b,c,.} are the factors for the particular oxides,
given in Table 9.
404
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
The most important drawback of this approach is
that the experimental points used for the correlation, are
completely out of the range of interest for ash deposits.
As some results showed, extrapolation to higher
temperatures will not give good results. If the oxides
of concern are present in low concentrations, or not
distributed quite uniformly throughout the glass, the
factors could be considerably in error.
4.3. Emissivity
The emittance, 3, is a surface radiative property,
which is defined as the ratio of a body’s emission rate to
that from a black body
3ðTÞ Z
EðTÞ
Eb ðTÞ
(128)
between total and spectral emittance is
3t ðTs Þ Z
1
EB ðTs Þ
N
ð
3v ðTs ÞeB ðl; Ts Þdl
(129)
vZ0
where EB (Ts) is the emissive power, eB (l, Ts) is the
spectral black body flux density, and Ts-surface
temperature. According to Wall [52], if 3l does not
vary with wavelength, meaning that 3t do not vary with
temperature.
According to the energy conservation law (the first
law of thermodynamics), all radiation incident on a
surface must be reflected, absorbed or emitted
a Cr Ct Z 1
(130)
where a is the absorbance, r is the reflecttance, and t is
the transmittance. In the absence of non-linear effects
(i.e. Raman effect, etc.), Eq. (131) can also be applied to
the spectral properties (this can be applied for deposits).
Kirchoff analysis, outlined in Appendix G, showed that
emissivity is equal to absorptivity:
where E is the energy emitted by a body, and Eb is the
energy emitted by a blackbody. Since the spectral
radiation emitted by a real surface differs from the
blackbody distribution, the emissivity can assume
different values according to whether one is interested
in the emission at a given wavelength or in a given
direction, or in integrated averages over wavelength
and direction. Spectral, directional emissivity 3l,q of a
surface at the temperature T is defined as the ratio of the
intensity of the radiation emitted at the wavelength l in
direction of q to the intensity of the radiation emitted by
a blackbody at the same value of T and l. Total,
directional emissivity 3q represents a spectral average
of 3l,q. The spectral hemispherical emissivity, 3l
represents the body’s emission in all directions. Total,
hemispherical emissivity, 3, represents an average over
all possible directions and wavelengths. The connection
3l Z 1Krl
Table 9
Factors for calculating thermal conductivity at three temperatures
from wt% composition, derived from author’s measurements
assuming linear relation [50]
† The physical state (morphology) of the deposit
surface (fused, sintered or packed particles).
† The chemical composition (minimal dependence).
Oxide
The deposit morphology influences the scattering
from surfaces:
SiO2
K2O
Na2O
PbO
Sb2O3
B2O3
Al2O3
ZnO
CaO
BaO
Fe2O3
MgO
Temperature
K100 8C
0 8C
100 8C
2.44
0.54
K1.24
0.60
K5.11
1.09
3.23
1.95
2.82
0.39
1.61
6.37
3.07
0.58
K1.29
0.76
K4.16
1.59
3.72
2.02
3.17
0.46
1.90
5.92
3.44
0.39
K0.67
0.96
1.12
2.49
2.14
1.64
2.39
0.75
1.73
4.53
al Z 3l
(131)
For opaque slabs, transmissivity is equal to zero
(tlZ0), so emissivity is:
(132)
According to Wall et al. [7], deposit must be thick
enough (around 1 mm) in order to be considered as
opaque.
4.3.1. Influence of various factors on emissivity
According to Markham et al. [53], the emittance
depends on:
1. Fused deposits solidify from the molten state to
become glassy in nature, which lead to high
emittance values. According to the experiments
obtained by Markham et al. [53], in the temperature
range TZ300–1630 K, the measured values for the
emittance of ash was 3R0,9. In the molten or liquid
state, the spectral emittance is only slightly higher
than in the fused state.
2. Sintered or powdery deposits consist of individual
ash particles that are weakly held together. The bed
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
of particles of the powdery and sintered deposits
increases the scattering efficiency by increasing the
number of boundary surfaces available for scattering (reflecting) of the penetrating wave. Due to their
particle-like morphology sintered deposits exhibited
a wave-length-dependent emittance [53].
The deposit morphology, and thus its emissivity,
depends on the surface temperature of the deposit.
Curve A in Fig. 45 shows that the emittance of a
particulate ash, is decreasing with temperature increase
up to the sintering point. Also, sintered ash (B) and
fused ash (C) have significantly higher emittance. The
marked increase in emittance of sintered and fused ash
is caused partially by reduction in the reflective surface
of particulate ash and partly by the change from ferric
to ferrous state of iron at high temperatures. A similar
emittance dependence on the temperature and the
degree of sintering have been found by Boow and
Goard [22], and Wall et al. [7].
Bhattacharya [8] showed that the chemical composition of the deposit material affects the emittance and
the thermal conductivity due to the presence of coloring
agents that increase the absorptivity of the deposit, and
together with the particle size influence the temperature
of the onset of sintering and the fusion of the deposit
material. Emittance of smooth deposits, which are
analogous to molten deposit showed very little
dependence on the chemical composition.
Iron in the ferrous valency state (Fe2C), gives a
black color to wustite (FeO), hematit (Fe2O3) and
silicate particles when dissolved in concentrations
above 3% by weight. According to Raask [5], the
visual appearance of ash (for coal) depends largely on
the amount and the mode distribution of iron in ash.
Fig. 46 shows the influence of iron on the emissivity of
coal ash.
Fig. 45. Emittance characteristics of boiler deposit: (A) particulate
ash; (B) sintered deposit; (C) slag. Arrows indicate direction of heat
flow [54].
405
Experiments done by Wall et al. [52] showed that
synthetic mixtures of the oxides Al2O3, Fe2O3 and SiO2
give the same trends of emissivity, as those for ashes of
the same Fe2O3-content. It was showed that the total
emissivity of coal ash with high iron-content, might
increase before the sintering temperature for ashes is
reached.
The trends reported by Boow and Goard [22] were:
† A reduction of the total emittance of particulates
with temperature until initial sintering, and fusion,
occurs, consistent with the spectral effects
† A systematic increase in the total emittance with
increasing particle size, prior to the point of
sintering
† An increase in the total emittance with the ironcontent (Fe2O3) and the carbon-content (due to the
presence of unburnt coal in the ash) prior to the
point of sintering
† The values of 3t for a given Ts varied only slightly
for the laboratory ashes, despite the wide range of
silica ratios examined. On the other hand, samples
that had similar composition but different thermal
history have quite different emissivtiy values.
4.3.2. Models for emissivity
Estimating the emissivity requires certain assumptions: (1) black body distribution, (2) a transmittance
zero for wavelengths where data are available, and (3) a
normal spectral emittance independent of temperature.
Boow and Goard [22] measured the emittance of the
following materials: glass powder, laboratory-prepared
and boiler ashes of different characteristic sizes.
Fig. 46. Total emissivity of opaque particulate layer, for two different
iron contents [8].
406
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
The equations are applicable for the particle diameter
range dpZ7–420 mm at TZ775 K.
For colored glass powder:
curvature of the tubes, so the effective emittance, 3e, is
related to the flat surface emittance, 3r, obtained from
the experiments:
3r Z 0:25 log d C 0:13
3e Z
(133)
For typical silicate ash deposit on boiler tubes before
a sintered matrix develop
3r Z 0:30 log d C 0:16
(134)
where 3r is the total surface emittance (dimensionless
ratio), and d is the particle diameter (mm). Eqs. (134)
and (135) are not applicable for calculating the initial
emittance of a clean metal target. The effective thermal
conductivity of unsintered ash depended on the
emittance 3r, and the relationship at TZ775 K is:
3r log k Z 0:753r K1:55
p3r
1 C ðpK1Þ3r
Wall and Becker [52] showed a method to obtain
total emissivity values from spectral data. A (Planckweighted) total emissivity can be found from the
spectral band emissivities of a finite number of adjacent
bands as
3t ðT0 Þ Z
n
P
jZ1
3lj ðT0 ÞDEb;lj ðT0 Þ
n
P
jZ1
(135)
The correlation coefficient is 0.94, for a variety of
ashes and powder glasses. The results are shown in
Fig. 47.
Wall et al. [7] claimed that the results for emissivity
3Z0.55–0.70 as obtained by Boow and Goard [22],
Mulcahy et al. [55] and Goodridge and Morgan (1971),
may be too low, and that the data should be reexamined
and compared with the results of reflective measurements. They suggested that the data obtained from the
experiments are not necessarily applicable to real
systems, because experiments are usually done on flat
surface deposits while real deposits are rarely flat. A
thin layer of deposit on boiler walls will follow the
(136)
(137)
DEb;lj ðT0 Þ
where T0 is the ash temperature, and;
ljC1
3lj ðT0 Þ Z
Ð
lZlj
3l eb ðl; T0 Þdl
DEb;lj ðT0 Þ
(138a)
where 3lj (T0) is the (Planck-weighted) band emissivity
3l,j, at the ash temperature T0, 3l is the emissivity at a
certain wavelength, eb(l,T0) is the spectral blackbody
flux density (W/m2) (given by Planck’s law), and
DEb;lj ðTÞ Z Eb ðljC1 ; TÞKEb ðlj ; TÞ
(138b)
where DEb,lj(l,T) is the incremental band emissive
power (W/m2) and
Eb ðTÞ Z
N
ð
eb ðl; TÞdl
(138c)
lZ0
where Eb(T) is the total blackbody flux density (W/m2).
This integral has been evaluated as
Eb ðTÞ Z sT 4
(138d)
lj is the nominal wavelength position of the band
given by
lj Z
ll C lu
2
(138e)
where subscripts l and u refer to the lower and upper
wavelength limits of a band, respectively.
Fig. 47. Influence of particle size and heating on total emittance of
particles prepared by crushing a synthetic slag of 5% Fe2O3 [22].
4.3.3. Experimental data on emissivities
Markham et al. [53] and Wall et al. [7] showed that
the chemistry of a deposit actually has a weak influence
on its emissivity. Fig. 46 shows how the emissivity
depends on the iron-content. Comparing the emissivity
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
for particulate layers, with 1 and 5% of iron, showed
that the ratio is almost constant, and has a value of 3(5%
Fe2O3)/3(1%Fe2O3)Z1.12, which will be used to
determine the influence of iron on the emissivity of
particulate deposit in this study. According to Wall et
al. [6], the iron-content does not have any influence on
the emissivity values for slags.
Boow and Goard [22], Bhattacharya [8] and Wall et
al. [6,48] showed that, normal emittance of slags is
approximately constant and that it has values greater
than or equal to 0.9. Wall et al. [48] reported values of
0.955 for normal emittance and 0.875 for hemispherical
emittance, both values measured at 1000 K, for a 5%
Fe2O3 slag. The connection between normal (3n) and
hemispherical (3h) emittance is determined by Wall [6]
to be:
3h Z 0:923n
(139)
Mills and Rhine [31] have done experiments on
gasification slags, and obtained results for the total
normal emittance, in the temperature range TZ1070–
1800 K. A constant value of 3Z0.83 was obtained for
the entire temperature range.
4.3.3.1. Estimation of the emissivity. Now we will see
how empirical equations for emissivities of particles of
different sizes, obtained from the Boow and Goard data
[22], can be used to predict the emissivity of an ash.
These predictions will be compared to total emissivities
experimentally determined by Wall et al. [7], and
Bhattacharya [8]. The iron-content will be included,
knowing the following experimental result: 3(5%
Fe2O3)/3(1%Fe2O3)Z1.12.
Ash deposits in all experiments have particle sizes
less than 44 mm (Bhattacharya [8]: 3.06–11.43 mm;
Wall and Becker [52]: Morwell 22 mm, Tarong 9 mm).
In order to predict the emissivity, correlation equations
were used. Analysis of the results shows:
† Correlation equations for smaller particles underestimate experimental results, except in the case of
the Morwell coal ash data, in which the iron content
is very high (around 10%). Its average error is
around 30%.
† Correlation equations for larger particles are
generally able to predict the emissivity values,
except in the case of the Morwell coal ash data,
when it seriously overpredicts them. The average
error is around 5%.
Since different studies [5,22,53] showed that particle
size has a big influence on the value of the emissivity of
407
a deposit, experimental results obtained by Boow and
Goard [22] given in Fig. 6, can be used for the
correlation. They used synthetic slag samples with the
same chemical composition (5% Fe2O3), but different
particle size.
The procedure, which is used in this study, in order
to estimate the emissivity vs. particle size and ironcontent dependency, is:
† At temperatures below the sintering temperature,
the emissivity is determined using the correlating
equations based on the Boow and Gorad [22] data.
† At temperatures above the fusion temperature, i.e.
for fused and molten deposits, a constant value of
0.955 for the normal emissivity can be used. For
estimating the hemispherical emissivity, Eq. (139)
will be used.
† Between the sintering and the fusion temperatures,
the linear correlation between 3sint and 3fused
developed using the Boow and Goard [22] data,
can be used.
† The concentration of iron will be taken into account
by using the empirical value of 1.12 for ratio 3(5%
Fe2O3)/3(1%Fe2O3).
4.3.3.2. Investigation of the chemistry influence on the
emissivity. An attempt was made to calculate the
emissivity based on the chemical composition. Emissivity data, provided by Wall et al. [6] and Touloukinan
[47], for pure oxides, were used for the predictions, and
the results were compared with the experimental data.
The emissivity of ash deposits is determined as the
emissivity of a mixture (the sum of the weighted
average of the compounds) of pure oxides, present in
the ash. A big drawback of this approach is that it does
not give the dependence of the emissivity on the
particle size. Total emissivities of different coal ashes,
determined by Wall et al. [7], and ashes with different
iron content, determined by Bhattacharya [8], will be
used for comparison. The results are shown in Figs. 48
and 49 and from these it can be concluded:
† The results obtained using the Touloukian [47]
emissivity data, largely overestimate the experimental data.
† The results obtained using the Wall et al. [6]
emissivity data, underestimate the experimental
data.
† In the data obtained from Wall’s results (average
error around 20%), a better agreement is obtained on
the estimate for ash with lower iron content.
408
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
Emissivity
Emissivity, data: Bhattacharya, 1997
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
700
900
1100
1300
T, K
1500
1700
Exp (Bhattacharya, 1%Fe)
Touloukian, 1%Fe
Wall, 1%Fe
Exp(Bhattacharya, 5%Fe)
Touloukian, 5%Fe
Wall, 5%Fe
Fig. 48. Data obtained from Bhattacharya [8] compared to emissivity
obtained by weighted average based on data of Touloukian [47] and
Wall et al. [6].
† This method cannot correctly predict the dependence of the emissivity on the iron-content.
the models of Rayleigh, i.e. Eq. (21), Russel (22),
Woodside (25), and Diesler and Eian.
Figs. 59 and 60 show that the Rayleigh (LR) and
Russel (R) models give similar results. The Woodside
(W) model and the model by Deisler and Eian’s (DC
E), deviate considerably from (LR), particularly for
large p and ko/km.
Anderson et al. [3] have performed experiments on
fly ash, slags and particulate (fouling) deposits (13.5–
1015 mm), in the temperature range: 300–1420 8C. The
equation for estimating the overall thermal conductivity, which represents an average conductivity for the
entire ash sample, is
ð
ð
keo h kðTÞdT= dT ZKqH=DT
(140)
where keo (W/m K) is the overall effective thermal
conductivity, q (W/m2) is the heat flux, H (m) is the ash
sample thickness, and DT (K) is the difference between
extrapolated surface temperatures.
The fly ash samples were prepared by loading the asreceived fly ash into the test cell tray. Samples were not
5. Experimental validation of models
Laubitz [10] measured the effective thermal conductivity of several powders in the temperature range of
100–1000 8C (see Section 4.1.2.). Figs. 50 and 51 show
some of the data on k/km, where km is the conductivity
of the solid phase, and k is the effective thermal
conductivity of the two-phase medium, obtained from
Emissivity: Wall, 1984
1.2
emissivity
1
0.8
0.6
0.4
0.2
0
700
900
1100
T, K
1300
1500
Exp (Wall, Morwell)
Touloukian, Morwell
Wall,Morwell
Exp (Wall, Tarong)
Touloukian, Tarong
Wall,Tarong
Fig. 49. Experimental data reported by Wall et al. [52], compared to
estimated values obtained by weighted average based on data of [47]
and [6].
Fig. 50. Variation of the effective thermal conductivity with the
discrete phase volume fraction, for a fixed value of solid-to-gas
conductivities, ko/kmZ100 [10].
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
409
Fig. 52. Overall effective thermal conductivities vs. T, for fly ash [3].
values:
ksp Asp kp Ap
kh Afc
Z
C
H
H
H
Fig. 51. Variation of the effective thermal conductivity with the ratio
of solid-to-gas conductivity, for fixed discrete phase volume fraction,
pZp/6 [10].
compacted during the loading process. Fig. 52 shows
that the magnitude of keo is quite small, indicating that
fly ash will create a large thermal resistance if it is
allowed to accumulate on heat transfer surfaces.
Results for sample A and B are the same, up to a
mean ash temperature of approximately 650 8C despite
the disparity in particle size, porosity and silica ratio.
For crushed slags the influence of radiation can be
neglected for particles up to 100 mm. From Fig. 53,
it can be concluded that for smaller particles, the
thermal radiation becomes significant at higher
temperatures.
When Figs. 53 and 54 are compared, it can be seen
that the thermal conductivities are approximately twice
as large for the larger particles, which is due to the
effect of radiative transfer. Particle size is not an
important parameter for particles up to 715 mm, but
above this particle size, the thermal conductivity an
increases with increase in the particle size.
Effective thermal conductivities of fouling (superheater) deposits (2/3 solid-porous, 1/3 particulate) were
measured. The thermal conductivity is the sum of two
(141)
where A [m2] is the area, and subscripts represent sp
solid-porous, h hybrid and p particulate sample. The
results are shown in Fig. 55.
From Figs. 60–62 we can conclude that the effective
thermal conductivities of all samples were less than
0.5 W/m K and that they were relatively independent of
temperature.
Results from the crushed slag experiments are
compared with the data of Mulcahy and Singer, and
presented in Fig. 56. This comparison showed good
agreement between experiments and results obtained
from the models at temperatures below 1000 8C. Above
1000 8C, a large discrepancy is noticeable, which can
be explained by a slightly different silica-content
(Anderson: 46%, Mulcahy: 41%) and a difference in
the particle size applied (Anderson: 420,715 mm,
Mulcahy: 200 mm, smaller particles can be expect to
sinter and melt sooner).
Fig. 53. Local thermal conductivty for crushed-wall slag deposits with
Dp%715 mm [3].
410
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
Fig. 54. Local thermal conductivities for slag deposits with DpZ
1015 mm [3].
Robinson et al. [2] performed experiments under
thermal conditions that closely replicate those found in
the convective pass of a commercial boiler (see Section
4.1.1). They described a layered structure of the
deposit. The outer layer actually goes through a change
during sintering, while the inner layer mostly stays
unchanged. For kun, a constant value of 0.15 W/m K
(the thermal conductivity of deposit at the start of the
sintering stage of the experiment) is used, while ksi
varies between kun and 3 W/m K (the thermal conductivity of deposit solid phase). With a temperature
increase, ksi increases (due to sintering in the outer
layer), while kun remains constant. According to the
results, once ksi is roughly 6 times greater than kun, the
overall thermal conductivity of the deposit is insensitive to changes of ksi.
A sensitivity analysis reveals that if L is grater than
0.2, and the ratio of ksi-to-kun is greater than about 3,
and then the critical parameter is kun. The authors use
the constant value of 0.15 W/m K for kun, which is the
Fig. 55. Local thermal conductivities for hybrid and particulate
fouling deposits [3].
Fig. 56. Comparison of present data with those of previous
experiments [3].
measured deposit thermal conductivity at the beginning
of the sintering. Since, in practice, the value of the
overall deposit thermal conductivity is relatively
insensitive to the assumed functional form of ksi, it
may be stated that ksi varies linearly with the solid
fraction of the sintered layer.
Now, the effective thermal conductivity is largely
determined by the thermal conductivity of the
unsintered inner layer. The relatively low temperatures
(!600 8C—the temperature of a heat transfer surface
in a typical boiler) will most likely prevent substantial
sintering of the innermost layer. During this experiment, the average deposit temperature was increased
by approximately 200 8C.
The averaged measured thermal conductivity was
w0.14 W/m K. Since these deposits were unsintered
and loose, it can be expected that this value represent
the lower extreme of the range of deposits that might
form in real boilers.
According to Fig. 57, the success of the model,
supports the conclusions about the importance of a
layered deposit structure in determining the effects of
sintering on the deposit thermal conductivity, and
confirms that an unsintered layer with a low value of
thermal conductivity, will limit the overall deposit
thermal conductivity. Fig. 57e also compares the
measurements to theoretical limits the deposit thermal
conductivity. The thick vertical line indicates the
transition between the growth and the sintering phases
of the experiment. Vertical bars in (e) indicate estimate
of experimental uncertainty, G20%.
Fig. 58 shows a comparison of the measurements of
the deposit thermal conductivity shown in Fig. 57e to
predictions of the two-layered structure model as a
function of time. Fig. 59 shows measurements of the
deposit solid fraction and the deposit effective thermal
conductivity as a function of time.
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
411
Fig. 58. Comparison of the measurements of deposit thermal
conductivity shown in Fig. 57e, to predictions of the two-layered
structure model, as a function of time. The dashed lines show the
changes in the thermal conductivity of the unsintered, kun, and
sintered, ksi, layer of the deposit. The elapsed time corresponds to the
sintering stage of the experiment. Vertical bars indicate an estimate of
the experimental uncertainty on selected data points [2].
in the literature. Most of the experiments were done on
crushed deposits whose original microstructure, which
is very important for both the thermal and the radiative
properties, is destroyed. Robinson et al. [2] developed a
technique for measuring thermal conductivity of ash
deposits in situ on the original microstructure.
Experimental data from the work of Rezaei et al. [1]
is used in order to evaluate the results obtained from the
different models. In this study, coal ash deposits were
used, whose content is given in Table 10.
6.1. The thermal conductivity of solid material
Fig. 57. Measurements of (a) average deposit thickness, (b) average
probe and deposit surface temperature, and (c) deposit solid fraction,
(d) deposit structural parameter, and (e) deposit effective thermal
conductivity as a function of time [2].
6. Thermal conductivity of ash deposits obtained by
different models
Very few experimental data of the (effective) thermal
conductivity of ash deposits, especially at high
temperatures where radiation is important, are available
Three different approaches for determining the
thermal conductivity of the solid phase of an ash deposit
will be examined: (1) the weighted-average of component oxides, (2) the Kobayashi model, and, (3) the
Rezaei empirical model. The model predictions are
compared to experimental data derived on a highly
sintered ash sample A, with porosity pZ0.15, as shown
in Fig. 60.
The analysis shows that a fused ash sample, with a
low porosity will have a thermal conductivity, which is
constantly increasing with increasing temperature, and
that the experimental data are lower than any of the
model predictions. The experimental data is best
described by the Rezaei model.
Even though the weighted-averaged model gave
poor results, it will be used in the predictions instead of
the empirical model.
412
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
All the models underpredict the thermal conductivity of the unsintered deposit. For the sample with pZ
0.30, the difference is around 45% up to the sintering
point, above which difference rapidly increases. The
thermal conductivity of the sintered ashes is over
predicted. The Russel model predicts the highest values
for thermal conductivity, the CSP model predicts the
lowest values, and the Rayleigh model lies in between.
The experimental data show that the thermal conductivity decreases with the porosity, but with much less
sensitivity to this parameter than the models predict.
By comparing the curve of the thermal conductivity
of the solid with the curve of the thermal conductivity
of porous material, we can see the great similarity
between them. This indicates that the models used here
predict great influence of the thermal conductivity of
the solid phase on the effective thermal conductivity of
fused ash deposits.
6.3. Complex structure models
Fig. 59. Measurements of (a) the deposit solid fraction, and (b) the
deposit effective thermal conductivity, as a function of time. These
measurements were conducted under the same conditions, but
terminated after different periods of sintering (0, 1, 4, or 12 h) in
order to generate samples for SEM analysis of the deposit
microstructure. The vertical line at 0 h indicates the start of sintering
portion of the experiment. The theoretical limits are calculated from
the solid fraction of the deposit sintered for 12 h. The elapsed time for
some of the thermal conductivity measurements has been slightly
shifted for visual clarity. Vertical bars indicate experimental
uncertainty for selected data points [1].
6.2. Simple models for the thermal conductivity
Predictions by the Rayleigh, Russell and CSP models
(Eqs. (21), (22) and (34)) will now be compared to
experimental data. Fig. 61 shows the comparison for two
porosities of unsintered deposit B, while Fig. 62 shows
predictions for two porosities of sintered deposit B.
Complex structures, which assume the continuity of
both the solid and the gas phase, would be expected to
give better estimations of the thermal conductivity of
sintered deposits. The Brailsford and Major, the Hadley
and Nimick, and the Leith models (Eqs. (46), (57) and
(62)) will now be compared to the experimental data for
sintered deposits. Fig. 63 shows the comparison for two
porosities of the sintered deposit A.
Based on the comparison shown in Fig. 63, the
following can be concluded:
† The Nimick and Leith model gives the highest
predictions of the thermal conductivity.
† Both the Nimick and Leith, and the Brailsford and
Major models are greatly influenced by the thermal
conductivity of the solid phase, which is evident
from the shape of the resulting curves. Also, the
models are greatly influenced by the porosity, which
is not detected in the experimental data.
† Predictions based on the Hadley model give fair
agreement with experimental results. This model
shows a big sensitivity to porosity, so the agreement
Table 10
Ash content from the study of Rezaei et al. [1]
Sample
name
SiO2
Al2O3
Fe2O3
CaO
MgO
K2O
Na2O
TiO2
dp (mm)
Sintering
temp, 8C
Fusion
temp, 8C
A
B
34.37
48.49
23.75
22.95
4.85
7.07
13.41
4.77
4.32
1.36
0.95
0.66
5.04
0.88
1.22
1.01
8
20
600
670
1200
1440
413
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
Thermal conductivity of solid
Sample A; sintered; p=0.30 and p=0.15
10.0
7
6
8.0
k, W/mK
k, W/mK
5
6.0
4.0
4
3
2
2.0
1
0
0.0
0
300
600
Rezaei
900
T, C
WT
1200
Kobayashi
1500
with the experimental data actually depends on
porosity.
6.4. Radiative heat transfer models
6.4.1. The exchange factor
The exchange factor, as given in Boterill [27,28], is
determined using the Laubitz, the Godbee and Ziegler,
the Zehner, the Bauer and Schlunder, and the Kunii and
Smith models. These predictions are shown in Fig. 64.
We can see that the Goodbee and Ziegler model and
the Laubitz model (additive models) predict an increase
in the exchange factor with increasing porosity.
Resistance network models like the Zehner, and the
Bauer and Schlunder model, predicts that exchange the
factor does not depend on the porosity, and the Kunii
Sample B; unsintered; p=0.30 and p=0.50
1.2
300
900
Rayleigh (1)
Rayleigh (2)
1500
T, C
1800
Russell (1)
Russell (2)
CSP (1)
CSP (2)
1200
Fig. 62. Experimental results (sintered ash) compared to models
predictions.
and Smith model predict that the exchange factor will
decrease with a porosity increase.
Fig. 65 shows the predictions for the radiative
conductivity of a porous material. At lower temperatures, the radiative conductivity slowly increases with
the temperature, and then at temperatures above app.
600 8C, the radiative conductivity starts to increase
more rapidly. The variation of the radiative thermal
conductivity, with porosity, for the different models is
the same as the variation of the exchange factor. Again
the Kunii and Smith model predict that radiative
conductivity will decrease when the porosity increases,
but the reason for that can be that the exchange factor
predicted by their model is more sensitive to the value
of the emissivity. We can also see that absolute values
of the radiative conductivity are very small, so
according to these models, this value will not affect
the effective thermal conductivity much, not even at
high temperatures.
On-set of
sintering
1
600
exp,p=0.3 (1)
exp, p=0.15 (2)
exp
Fig. 60. Thermal conductivity of solid obtained by different models.
Sample A; sintered; p=0.30 and p=0.15
6.0
0.8
5.0
0.6
4.0
k, W/mK
k, W/mK
0
1800
0.4
0.2
3.0
2.0
1.0
0
0
300
600
900
1200
1500
1800
exp, p=0.30 (1)
exp, p=0.50 (2)
Rayleigh (1)
Rayleigh (2)
T, C
0.0
0
T, C
Russell (1)
Russell (2)
CSP (1)
CSP (2)
Fig. 61. Experimental results (unsintered ash) compared to models
predictions.
300
600
900
1200
1500
1800
exp, p=0.30 (1)
B&M (1)
N&L (1)
Hadley (1)
exp, p=0.15 (2)
B&M (2)
N&L (2)
Hadley (2)
Fig. 63. Experimental results (sintered ash) compared to models
predictions.
414
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
Sample A; particulate; p=0.29 and p=0.58
2
1.2
1.6
0.8
1.2
k, W/mK
Exchange factor
Exchange fator; p=0.2 and p=0.5
1.6
0.4
0.4
0.0
0
300
600
G&Z; p=0.2
900
T, C
KS; p=0.2
L; p=0.2
1200
1500
1800
0
L; p=0.5
G&Z; p=0.5
KS; p=0.5
6.4.2. Additive models
The Laubitz, and the Godbee and Ziegler models,
i.e. Eqs. (64) and (71), will be compared to
experimental results, below. Since the Laubitz model
uses twice the value of the Russell’s thermal
conductivity, which overpredicts the values of the
thermal conductivity of sintered ash, the Laubitz model
will be used only for particulate deposits. Fig. 66 shows
the Laubitz predictions for particulate ash sample A.
If the values of the Russell model predictions for the
thermal conductivity are compared to the radiative
conductivity, it can be seen that the value of radiative
conductivity is almost insignificant in comparison to
twice the value of the Russell’s conductivity. Fig. 67
shows the Godbee and Ziegler predictions for sintered
ash sample A.
The Godbee and Ziegler model overpredicts the
experimental results, for both the sintered and the fused
Radiative conductivity, p=0,2 and p=0,5
0.025
0.020
0.015
0.010
0.005
0.000
0
300
600
900
0
300
600
ZBS; p=0.5
Fig. 64. Exchange factor, determined using different models, for
porosities pZ0.2 and 0.5.
kr, W/mK
0.8
1200
1500
1800
T, C
GZ, p=0.2 KS, p=0.2 L, p=0.5 KS, p=0.5
L, p=0.2
GZ, p=0.5 ZBS, p=0.5
Fig. 65. Radiative conductivity determined for two different
porosities, when different model for exchange factor are used.
exp, p=0.29 (1)
900
T, C
L (1)
1200
1500
exp, p=0.58 (2)
1800
L (2)
Fig. 66. Predictions based on Laubitz model compared to
experimental data.
deposits. Furthermore, the model shows great sensitivity to porosity.
6.4.3. Resistance network models
The following resistance network models will be
tested in action: the Yagi and Kunii model (particulate
deposits, Eq. (100)), the Kunii and Smith model
(particulate deposit, Eq. (104) and fused deposits, Eq.
(106)), and the Zehner, Bauer and Schlunder model
(particulate deposit, Eq. (107)). These models use
parameters that should be determined experimentally.
In the present work, constant values for these
parameters will be assumed, which are chosen as
averaged values used in the original studies, or have
already proposed by the authors.
Most of these models are developed for porous
media where the continuous phase is a gas, which may
explain why the influence of the thermal conductivity
of the solid phase is less important.
The results obtained using the Yagi and Kunii model
are shown in Figs. 68 and 69.
The Yagi and Kunii model is due to its semiempirical character applied only to unsintered ash
sample, where gas is the continuous phase. The
following can be concluded:
† The Yagii and Kuni model gives far better
agreement with the experimental results than the
models that were previously used.
† This model gives results that are influenced more by
porosity than the experiments showed to be the real
case, but less than the models previously tested.
415
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
Sample A; sintered; p=0.30 and p=0.15
Sample B; particulate; p=0.30 and p=0.50
1.6
7
6
1.2
k, W/mK
k, W/mK
5
4
3
0.8
0.4
2
1
0
0
0
0
300
600
exp, p=0.30 (1)
900
T, C
GZ (1)
1200
1500
exp, p=0.15 (2)
1800
GZ (2)
Fig. 67. Predictions based on the Godbee and Ziegler model compared
to experimental data, corresponding to two different porosities.
† The difference between the results obtained using
bZ1 and 0.90 are about 10%.
† The results for the higher porosities are in better
agreement with the experiments when bZ1, and for
lower porosities when bZ0.9.
Kunii and Smith developed models, which can be
applied for unsintered (Eq. (104)) as well as sintered
deposits (Eq. (106)). These models are also semiempirical with a parameter b, which is to be determined
experimentally. Since the authors suggested that this
parameter should take values between 0.9 and 1.0, the
influence of this parameter will be investigated further,
below. Fig. 70 shows predictions of this model for
particulate ash deposits, and Fig. 71 for sintered
deposits.
Use of different values of b leads to a difference in
thermal conductivity value of approximately 5%
(between bZ0.9 and 1). This model is again very
sensitive to a change in the porosity, for lower
porosities the thermal conductivity of the ash is
300
600
exp, p=0.30 (1)
exp,p=0.50 (2)
900
B=1 (1)
B=1 (2)
1200
1500
B=0.95 (1)
B=0.95 (2)
T, C
1800
B=0.9 (1)
B=0.9 (2)
Fig. 69. Predictions based on the Yagi and Kunii model, compared to
exp. data.
overpredicted, and for higher porosities, the thermal
conductivity was underpredicted.
The predictions for sintered deposits, based on the
Kunii and Smith model, show a great sensibility to the
thermal conductivity of the solid phase. This model
includes two porosity parameters, f and f 0 (before and
after sintering). Three different cases were shown
considering these porosities: KS, 1: the two
porosities are the same; KS, 0.9: a reduction in porosity
of 10% (f 0 /fZ0.9); KS, 0.8: a reduction in porosity of
20% (f 0 /fZ0,8). The difference between the case with
no reduction in porosity and 20% reduction is about 20%.
Finally the Zehner, Bauer and Schlunder model will
be presented. Instead of the experimentally determined
parameters in this model, their values are approximated
as constants: rkZ0 (contact conduction is neglected),
CsphereZ1.25, RsphereZ1 (Sox/kox)Z0 (an oxidation
factor which is assumed to be zero). Results are shown
in Fig. 72.
The trend of the results obtained using the Zehner,
Bauer and Schlunder model, is quite similar to the trend
Sample A; particulate; p=0.29 and p=0.58
2
Sample A; particulate; p=0.29 and p=0.58
1.6
1.6
k, W/mK
k, W/mK
1.2
0.8
0.4
1.2
0.8
0.4
0
0
300
600
exp, p=0.29 (1)
exp,p=0.58 (2)
900
B=1 (1)
B=1 (2)
1200
1500
B=0.95 (1)
B=0.95 (2)
1800
T, C
B=0.9 (1)
B=0.9 (2)
Fig. 68. Predictions based on the Yagi and Kunii model, compared to
experimental data.
0
0
300
600
exp, p=0.29 (1)
exp, p=0.58 (2)
900
B=0.90 (1)
B=0.90 (2)
1200
1500
B=0.95 (1)
B=0.95 (2)
T, C
1800
B=1 (1)
B=1 (2)
Fig. 70. Predictions based on the Kunii and Smith model, for a gascontinuous porous material, compared to experimental data.
416
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
Sample A; sintered; p=0.15 and p=0.30
3
k, W/mK
2.5
2
1.5
1
0.5
0
300
600
exp, p=0.15 (1)
exp, p=0.30 (2)
900
KS, 1 (1)
KS, 1 (2)
1200
1500
KS, 0.9 (1)
KS, 0.9 (2)
T, C
1800
KS, 0.8 (1)
KS, 0.8 (2)
Fig. 71. Predictions based on the Kunii and Smith model for solidcontinuous porous material, compared to experimental data.
of experimental results. The absolute values of the
thermal conductivity are much closer at higher, than at
lower temperatures. This model also shows great
sensitivity to the porosity, so for low porosities it
overpredicts thermal conductivity of ash, while it
underpredicts the thermal conductivity at higher
porosities.
7. Conclusion
The aim of this work has been to review the present
knowledge on heat transfer in porous media, which can
be applied to estimate the thermal conductivities of ash
deposits. Besides the different approaches to heat
conduction modeling, determination of the thermal
conductivity of solid mixtures and the ash deposit
surface emissivity has been presented. Although the
weighted-average method did not give satisfactory
results, it was anyway used in the discussion, due to the
lack of other available models. Thus, it is highly
Sample A; particulate; p=0.29 and p=0.58
1.6
Appendix A. Calculation of heat transfer resistances
The total heat transfer between a flue gas, and the
steam inside a superheater tube consists of convective
and radiative heat transfer from the hot flue gases (g) to
deposit surface (d), heat conduction through deposit
and tube wall, and convection from the inner tube wall
to the steam (s). The heat transferred (W/m), can be
expressed in terms of heat resistances:
qL Z KL ðTg KTs Þ Z
1.2
k, W/mK
recommended that the thermal conductivity of solid
phase, for the particular ash sample, is measured, and
an empirical correlation is derived. A method for
obtaining surface emissivities was proposed, based on
the particle size and iron content of the deposit.
It was found that the major factor influencing the
thermal conductivity of an ash deposit is the deposit
structure, i.e. if the deposit is particulate, sintered or
fused. Different models were tested using experimental results for two coal ash samples, obtained by
Rezaei et al. [1]. It was found that simple structure
models do not give satisfactory results for both nonsintered and sintered samples. Particulate deposits
can be modeled using the Yagi and Kuni semiempirical model, while sintered deposits should be
modeled using the Hadley complex structure model.
Experimental values for fused deposits were not
available, but the suggestion is that they can also be
modeled as sintered deposits, using the Hadley
model.
Finally, it can be concluded that a wide range of
thermal conductivity models exist, but the need exists
for a wide range of experimental data, which would
help evaluate and improve these models. Also, it is
necessary to formulate a more accurate model for the
thermal conductivity of solid mixtures, which is the
property that can be identified as a potential important
source of errors.
0.8
Z
Tg KTs
rc
Tg KTs
rconvCrad C rcond;dep C rcond;tube C rconv;tubesteam
(A1)
0.4
T, C
0
0
300
exp, p=0.29 (1)
600
900
ZBS (1)
1200
1500
exp, p=0.58 (2)
1800
ZBS (2)
Fig. 72. Predictions based on the Zehner, Bauer and Schlunder model
for thermal conductivity, compared to experimental data.
where rc [m K/W] is the thermal resistance to heat
transfer through the cylindrical wall and KL (W/m K) is
heat transfer coefficient.
Heat transfer resistances originating from different
mechanisms equal one another over normalized heat
transfer coefficients, as will be shown.
417
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
The net heat transfer between a flue gas with
temperature Tg, and a deposit surface, with temperature
Td is
The heat transfer characteristics were calculated:
qL (W/m)
hconvCrad (W/m2 K)
kd (W/m K)
1238.47
126.74
0.6328
qL Z qconv C qrad Z hconvCrad ðTg KTd Þpdd
Tg KTd
Z
1
hconvCrad pdd
Z rconvCrad ðTg KTd Þ
where dd equals sum of the tube diameter and the
deposit thickness. The heat transfer coefficient in this
case, i.e. hconvCrad [14,17], equals the summation of
hconv, i.e. heat transfer coefficient calculated from the
criterial equations and hrad, calculated as:
hrad Z
3d sðTg4 KTd4 Þ
ðTg KTd Þ
(A3)
The heat conducted through the deposit to the
outside tube wall of temperature Tto is
qL Z
Td KTto
Z rcond;dep ðTd KTto Þ
dd
1
2pkd ln dto
(A4)
where kd is the heat conductivity of the deposit layer,
dto is the outer tube diameter, and rcond,depZ1/Kcond,dep
is a heat transfer resistance to conduction through
the deposit. The heat conducted through the tube wall
is
qL Z
Tto KTti
Z rcond;tube ðTto KTti Þ
dto
1
2pkt ln dti
(A5)
where Tti is the temperature of tube inner wall, dti is
the tube inner diameter, kt is the heat conductivity of
the tube, and rcond,tubeZ1/Kcond,tube is the heat
transfer resistance to conduction through the tube.
The heat transfer between tube wall and steam can be
written as
qL Z hs ðTti KTs Þpdti Z
ðTti KTs Þ
1
hs pdti
Z rconv;tubesteam ðTti KTs Þ
The obtained temperature profile is:
(A2)
Tg (C)
Td (C)
Tto (C)
Tti (C)
Ts (C)
900
642.86
570.09
567.69
550
The heat transfer resistances are:
rconvCrad (%)
rcond,dep (%)
rcond,tube (%)
rconv,tube-steam
(%)
41.10
46.16
1.53
11.21
It is obvious that the major resistances to heat
transfer in this system, is the convective heat transfer
from flue gases to deposit surface, and heat conduction
through the deposit.
Appendix B. Maximum discrete phase fraction in
thermal conductivity models based on spheres [55]
Application of thermal conductivity models, which
use spheres to present the discrete phase, is limited to a
volume fraction of the discrete phase of p/6. The
spheres in these models are arranged in simple cubic
cell as shown in Fig. B1.
In this case, the overall structure is shown in Fig. B2.
This lattice consists of layers of square packed
spheres. The layers are stacked so that each sphere is
directly above the one in the layer beneath. Since the
spheres in the lattice touches each other, the edge length
of the cell is two times the sphere radius, r. In this 3D
model the packing efficiency, which is equal to the
discrete volume fraction, is given by:
PE Z ðvolume of spheresÞ=ðvolume of cellÞ
(A6)
where rconv,tube-steamZ1/Kconv,tube-steam is the heat
transfer resistance to convection.
Heat transfer coefficients for convection can be
calculated using correlations from Section 2.
The heat transfer resistances, according to the above
equations, were calculated using an iteration method,
with a convergence criteria for Tto to be 20 8C higher
than the steam temperature. This value is empirical.
The heat transfer coefficient for convection is calculated using Eq. (6). Operational conditions, deposit, and
tube characteristics are listed in Table B1.
For a simple cubic lattice, this is:
1 4 3
8 3r p
PE Z 8
Z p=6 Z 52:35%
ð2rÞ3
(B1)
Appendix C. Calculation of the contribution of
radiation to the effective conductivity of powders
In the Russell model (1935), the obstacles are
assumed to be cubes of length a, located in cubical
volumes of the medium of length d. It is then assumed
that the obstacles are randomly scattered in the cubical
volumes, but in such a manner that the faces of the two
418
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
Table B1
A1 Input data for calculation
Tgas (C)
Tsteam (C)
dt (m)
u (m/s)
kt (W/m K)
dep.type
p
dthick (m)
900
550
0.038
8
25
Fused
0.3
0.005
cubes remain parallel. Such a scattering does not affect
the thermal conductivity of a two-phase medium. If the
probability that radiation will pass a cube without
hitting an obstacle is r, then:
LZ
a
p
(C4)
The rate of the heat transfer by radiation is
a2
r Z 1K 2
d
(C1)
If L is the average length that radiation travels
without hitting an obstacle in the first cube, then:
L Z d C rd C r 2 d C/Z
d
d3
Z 2
1Kr
a
(C2)
If p is the particle volume fraction, then
pZ
and
a3
d3
(C3)
Q Z 4sT 3 3ðd 2 Ka2 ÞDT
a2 DTd2
Z 4sT 3L 1K 2
L
d
3
(C5)
where T is the absolute temperature, 3 is the emissivity
of the obstacles, and DT/L is the gradient in the twophase medium, assumed small compared to T. Therefore, the contribution of that radiation effective
conductivity is:
a2
a
3
4sT 3L 1K 2 Z 4sT 3 3 ð1Kp2=3 Þ
(C6)
p
d
If it is assumed that the radiation that does hit an
obstacle in the first cube travels on the average a
distance of (dKa), then the contribution by the
radiation that does hit an obstacle in the first cube of
the medium can be calculated: The rate of heat transfer
by this process is
Fig. B1. Simple cubic unit cell.
QZ
a2 dKr K0 DT
aKr C LK0 d
(C7)
where
Kr Z 4sT 3 3L
(C8)
and its contribution to the effective conductivity is
a2 Kr K0
p1=3 Kr K0
Z
dðaKr C LK0 Þ Kr C K0 =p
(C9)
Since K0/p[Kr, then Eq. (C9) becomes
p4=3 Kr Z p4=3 4sT 3 3
a
p
(C10)
and the total effective conductivity
a
K Z KðRÞ C 4sT 3 3 ð1Kp2=3 C p4=3 Þ
p
Fig. B2. Overall structure in simple cubic cell.
(C11)
where K(R) is the two-phase conductivity calculated by
Russell.
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
Appendix D. Qualitative explanation for using
doubled value of k(R) in the Laubitz model
In actual powders, there are densely packed regions
where particles touch each other (contribute to heat
transfer chiefly by conduction), and empty air pockets
(contribute by radiation). The air pockets have roughly
a length of a/p0 (shown in Appendix C), where a is the
linear dimension of the particles, and p0 the average
particle volume fraction from the aggregate. But when
the p-value for the densely packed region is not p0 but
p01=3 , then the equivalent conductivity of the powder
becomes KðR; p01=3 Þp02=3 , where KðR; p01=3 Þ is the
Russell’s conductivity calculated for pZ p01=3 . For the
powders used in this experiments
KðR; p1=3 Þp2=3 z2KðR; pÞ
which gives a rough reason for the factor of two used
in Eq. (64).
419
The distance between the centres of the particles can
be related to the volume as follows:
3
lp
V
Z
(E3)
lpo
V0
This gives
1Kp 1=3 lp0
Z
1Kp0
lp
(E4)
or
lp Z lp0
1Kp0
1Kp
1=3
(E5)
The last equation relates the distance between
particles when the porosity is varied. The initial value
of p0 is set equal to the particle diameter. The minimum
value of the porosity is calculated for bed of cylinders.
Appendix F. Determination of the shrinkage level,
based on geometry approach
Appendix E. Estimation of the parameter lp in the
Yagi and Kunii Model for the effective thermal
conductivity
The following method of parameter estimation is
developed by Fjellerup et al. [34]. Fig. E1 indicates the
expansion of a volume partly filled with particles from
V0 to V. The amount of particles is constant and it is
assumed that the particles will spread homogeneously
to volume V. The porosity and the distance between
particles will change, and a relation between these two
parameters will be derived.
The porosity is the porosity between particles
(external porosity), and po and p is the porosity before
and after expansion, respectively. Parameters lpo and lp
are the distance between particle centres before and
after expansion, respectively. The initial solid volume
equals the final solid volume and the following
equation can be derived
Vð1KpÞ Z V0 ð1Kp0 Þ
(E1)
which gives
x2 C r 2 Z r02
x 2
r
C1 Z
(F1)
r 2
0
(F2)
r
V0 Z ð4=3Þr03 p Z m=r0
(F3)
V Z ð4=3Þr 3 p Z m=r
(F4)
x 2
r
0
3m
4r0 p
B
Z@
3m
4rp
2=3 1
(F5)
C
2=3 A K1
2=3
1=2
x
r Kr2=3
0
Z
r
r02=3
(F6)
A powder density can be expressed as a function of
porosity p and density of solid:
r0 Z rs ð1Kp0 Þ
1Kp
V
Z
1Kp0
V0
(F7)
(E2)
ro
x
x – radius of the neck
r0 – initial sphere radius
r – sphere radius at time t
r
Fig. E1. Expansion of a porous bed.
Fig. F1. Part of Fig. 46. x: radius of the neck; r0: initial sphere radius;
r: sphere radius at time t.
420
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
r Z rs ð1KpÞ
(F8)
Using Eqs. (F6)–(F8) porosity at time t can be
calculated as:
x 2 3=2
p Z 1Kð1Kp0 Þ 1 C
(F9)
r
Appendix G. Kirchhoff’s Law
Kirchhoff’s law can be written as
EðTs Þ
Z Eb ðTs Þ
a
(G1)
where Ts is the surface temperature, E(T) is the
emission from a body at Ts, a is the absorptivity and
Eb(T) is the emission from a blackbody, at the
temperature Ts. According to Eq. (G1), ratio of emitted
energy by a body at the surface temperature and its
absorption ability does not depend on physical
characteristics of a body; it is the same for all gray
bodies and equals to the emission of the blackbody at
the surface temperature.
Equation for the total, hemispherical emissivity is:
3ðTÞ Z
EðTÞ
Eb ðTÞ
(G2)
Combining Eqs. (G1) and (G2), alternative form of
Kirchhoff’s law is obtained:
a Z3
(G3)
This means that the total, hemispherical emissivity
of the surface is equal to its total hemispherical
absorptivity. Eq. (G2) is derived under the assumption
that the surface irradiation corresponds to emission
from a blackbody at the same temperature as the
surface.
The previous equation can be applied for spectral
properties, when less restrictive conditions are needed
(it is applicable if the radiation is diffuse or if the
surface is diffuse):
al Z 3l
(G4)
A form of Kirchhoff’s law for which there are
no restrictions involves the spectral, directional
properties (they are inherent surface properties,
thus are independent of the spectral and directional distributions of the emitted and incident
radiation).
References
[1] Rezaei HR, Gupta RP, Bryant GW, Hart JT, Liu GS, Bailey CW,
et al. Thermal conductivity of coal ash and slags and models
used. Fuel 2000;79:1697–710.
[2] Robinson AL, Buckley SG, Baxter LL. Experimental measurements of the thermal conductivity of ash deposits: part 2. Effects
of sintering and deposit microstructure. Energy Fuels 2001;
15(1):75–84.
[3] Anderson DW, Viskanta R, Incropera FP. Effective thermal
conductivity of coal ash deposits at moderate to high
temperatures. J Eng Gas Turbines 1987;109:215–21.
[4] Gupta RP, Wall TF, Baxter LL. The thermal conductivity of ash
deposits. In: Gupta RP, Wall TF, Baxter LL, editors. Impact of
mineral impurities in solid fuel combustion. New York: Plenum
Publishers; 1999.
[5] Raask E. Mineral impurities in coal combustion. Washington,
DC: Hemisphere Publishing Corporation; 1985.
[6] Wall TF, Bhattacharya SP, Zhang DK, Gupta RP, He X. The
properties and thermal effects of ash deposits in coal-fired
furnaces. Prog Energy Combust Sci 1993;19:487–504.
[7] Wall TF, Mai-Viet T, Becker HB, Gupta RP. Mineral matter in
coal and the thermal performance of large boilers. Prog Energy
Combust Sci 1979;5:1–29.
[8] Bhattacharya SP. A study on the importance of dependent
radiative effects in determine the spectral and total emittance of
particulate ash deposits in pulverized fuel fired furnaces. Chem
Eng Process 1997;36:423–32.
[9] Nasr K, Viskanta R, Ramadhyani S. An experimental evaluation
of the effective thermal conductivities of packed beds at high
temperatures. J Heat Transfer 1994;116:829–37.
[10] Laubitz MJ. Thermal conductivity of powders. Can J Phys 1959;
37:798–808.
[11] Frandsen FJ. Personal communication. Denmark: Technical
University of Denmark; 2002.
[12] Andersen KH. Deposit formation during coal–straw cocombustion in a utility pf-boiler. PhD Thesis. Technical
University of Denmark; 1998.
[13] Salmenoya K. Field and laboratory studies on chlorine-induced
superheater corrosion in boiler fired with biofuels. PhD Thesis.
Finland: Abo Akademi; 2000.
[14] Incropera FP, De Witt DP. Fundamentals of heat and mass
transfer. New York: Wiley; 1985.
[15] www.cheresources.com/convection.pdf.
[16] Bird RB, Stewart WE, Lightfoot EN. Transport phenomena.
New York: Wiley; 2002.
[17] Djordjevic B, Valent V, Serbanovic S. Thermodynamics and
thermotechniques. Belgrade: Faculty of Technology and
Metallurgy; 1997.
[18] Nimick FB, Leith JR. A model for thermal conductivity of
granular porous media. J Heat Transfer 1992;114:505–8.
[19] Vortmeyer D. Radiation in packed solids. In: Sixth international
heat transfer conference. Toronto: proceedings, vol. 6; 1978. p.
525–39.
[20] Woodside W. Calculation of the thermal conductivity of porous
media. Can J Phys 1958;36:815–23.
[21] Krupiczka R. Analysis of thermal conductivity in granular
materials. Int Chem Eng 1967;7(1):122–45.
[22] Boow J, Goard PRC. Fireside deposits and their effect on
heat transfer in a pulverized fuel-fired boiler: part III. The
A. Zbogar et al. / Progress in Energy and Combustion Science 31 (2005) 371–421
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[33]
[34]
[35]
[36]
[37]
[38]
[39]
influence of the physical characteristics of the deposit on its
radiant emittance and effective thermal conductance. J Inst
Fuel 1969;42:412–9.
Leach AG. The thermal conductivity of foams. I: models for heat
conduction. J Phys D Appl Phys 1993;26:733–9.
Brailsford AD, Major KG. The thermal conductivity of
aggregates of several phases, including porous materials. Br
J Apply Phys 1964;15:313–9.
Nozad I, Carbonell RG, Whitaker S. Heat conduction in
multiphase systems. Chem Eng Sci 1985;40(5):843–55.
Hadley GR. Thermal conductivty of packed metal powders. Int
J Heat Mass Transfer 1985;29(6):909–20.
Botterill JSM, Salway AG, Teoman Y. The effective thermal
conductivity of high temperature particulate beds—I. Experimental determination. Int J Heat Mass Transfer 1989;32(3):585–93.
Botterill JSM, Salway AG, Teoman Y. The effective thermal
conductivity of high temperature particulate beds—II. Model
predictions and the implication of the experimental value. Int
J Heat Mass Transfer 1989;32(3):585–93.
Schote W. Thermal conductivity of packed beds. AIChE J 1960;
6(1):63–7.
Godbee HW, Ziegler WT. Thermal conductivities of MgO, Al2O3,
and ZrO2 powders to 850C. J Appl Phys 1966;37(1):40–65.
Mills CK, Rhine JM. The measurement and estimation of the
physical properties of slags formed during coal gasification. Fuel
1989;68:904–10.
Yagi S, Kunii D. Studies on effective thermal conductivities in
packed beds. AIChE J 1957;3(3):373–81.
Fjellerup J, Henriksen U, Glarborg P, Jensen A, Jensen PA. Heat
transfer in a fixed biomass char bed. DTU, internal report; 2002.
Kunii D, Smith JM. Heat transfer characteristics of porous rocks.
AIChE J 1960;6(1):71–8.
Bauer R, Schlunder EU. Effective radial thermal conductivity of
packings in gas flow. Part II. Thermal conductivity of the packing
fraction without gas flow. Int Chem Eng 1978;18(2):189–204.
Larkin BK, Churchill SW. Heat transfer by radiation through
porous insulations. AIChE J 1959;5(4):467–74.
Kamiuto K, Iwamoto M, Naguro Y. Combined conduction and
correlated-radiation heat transfer in packed beds. J Thermophys
Heat Transfer 1993;7(3):496–501.
Baxter LL. Influence of ash deposit chemistry and structure on
physical and transport properties. Fuel Process Technol 1998;
56:81–8.
421
[40] Rezaei HR, Gupta RG, Wall TF, Miyamae S, Makino K.
Modeling the initial structure of ash deposits and structure
changes due to sintering. In: Gupta RP, Wall TF, Baxter LL,
editors. Impact of mineral impurities in solid fuel combustion.
New York: Plenum Publishers; 1999.
[41] Senior CL. Predicting removal of coal ash deposits in convective
heat exchangers. Energy Fuels 1997;11:416–20.
[42] Vargas S, Frandsen F, Dam-Johansen K. Rheological properties
of high-temperature melts of coal ashes and other silicates. Prog
Energy Combust Sci 2001;27(3):237–429.
[43] Drolen BL, Tien CL. Independent and dependent scattering in
packed-sphere systems. J Thermophys 1987;1(1).
[44] Goodwin DG, Mitchner M. Infrared optical constants of coal
slags: dependence on chemical composition. J Thermophys
1989;3(1):53–60.
[45] Bhattacharya SP. A study on the importance of dependent
radiative effects in determine the spectral and total emittance of
particulate ash deposits in pulverized fuel fired furnaces. Chem
Eng Process 1997;36:423–32.
[47] Touloukian YS. Thermophysical properties of matter. New
York: IFI/Plenum; 1979.
[48] Wall TF, Bhattacharya SP, Baxter LL, Richards G, Harb JN. The
character of ash deposits and the thermal performance of
furnaces. Fuel Process Technol 1995;44:143–53.
[49] Raznjevic K. Handbook of thermodynamic tables and charts.
Washington, DC: Hemisphere; 1976.
[50] Ratcliffe EH. A survey of most probable values for the thermal
conductivities of glasses between about K150 and 100 8C,
including new data on twenty-two glasses and a working
formula for the calculation of conductivity from composition.
Glass Technol 1963;4(4):116.
[52] Wall TF, Becker HB. Total absorptivities and emissivities of
particulate coal ash from spectral band emissivity measurements. J Eng Gas Turbines Power 1984;106:771–6.
[53] Markham JR, Best PE, Solomon PR, Yu ZZ. Measurement
of radiative properties of ash and slag by FT-IR emission
and reflection spectroscopy. J Heat Transfer 1992;114:
458–64.
[54] Mulachy MFR, Boow J, Goard PRC. Fireside deposits and their
effect on hetat ransfer in a pulverized-fuel-fired boiler. Part 1:
the radiant emittance and effective thermal conductance of the
deposits. J Inst Fuel 1966;39:385–94.
[55] http://www.kings.edu/wchemlab/vrml/scubic.html.