Fuel Processing Technology 105 (2013) 113–128

Contents lists available at SciVerse ScienceDirect

Fuel Processing Technology

journal homepage: www.elsevier.com/locate/fuproc

On predicting the ash behaviour using Computational Fluid Dynamics

Roman Weber ⁎, Marco Mancini, Natalia Schaffel-Mancini, Tomasz Kupka
Institute of Energy Process Engineering and Fuel Technology, Clausthal University of Technology, Agricolastr. 4, 38678 Clausthal-Zellerfeld, Germany

a r t i c l e i n f o a b s t r a c t

Article history: The objective of this paper is to examine several approaches for predicting the ash behaviour using Compu-
Received 5 January 2011 tational Fluid Dynamics (CFD). The emphasis is placed on details of the sub-models used. In models that aim
Received in revised form 26 August 2011 at predicting the temperature–time history of fuel particles, from the injection position to the deposit surface,
Accepted 5 September 2011
the information about the char combustion rate is essential. Of particular importance is determination
Available online 26 October 2011
(through measurements) of the rate of char oxidation for the last 20% of burnout. Predictions of the size dis-
Keywords:
tribution of fly-ash particles near the deposition surface seem to be unreliable and measured data are needed
Ash deposition for deriving appropriate corrections. Fragmentation of fuel particles is perhaps one of the unresolved issues
Boilers which require an urgent attention.
CFD The current CFD-predictions of slagging and fouling in industrial boilers are indicative, at their best. This is
certainly the case when numerous parameters, including these describing the particle sticking propensity,
are taken from the literature. It is relatively easy to tune the model predictions to the expected (measured)
results by changing a few key parameters. There is, however, a much more reliable way to proceed. By com-
bining CFD-predictions with an advanced fuel characterization, with the latter being application dependent
and fuel dependent, the deposition problems at hand can be tackled. What is to be determined experimen-
tally and under what conditions, requires a consensus of an experienced fuel engineer/mineralogist and a
CFD-expert.
© 2011 Elsevier B.V. All rights reserved.

1. Introduction and objectives
Ash related problems are for boiler operators of vital importance.
Ash deposition through slagging or fouling is often the main reason
for boiler unscheduled shut downs. Fig. 1 shows the regions of the
ash related problems in typical boilers. The deposits collected in the
boiler areas which are directly exposed to flame radiation, such as
furnace walls and pendant superheaters, are identified as slagging.
Fouling refers to deposits in the boiler areas which are not directly
exposed to flame radiation where convection is the dominant heat
transfer process. Generally, slagging takes place in the hottest parts
whilst fouling occurs as the flue gases and ash particles cool down.
Slagging deposits are chemically active at high temperatures being
partially or completely molten and they are often hard and difficult
to clean by soot blowing. The original chemical and physical struc-
tures of ash particles undergo substantial changes during slagging.
Fouling deposits are formed at lower temperatures and typically the
impacted particles keep their original physical and chemical
structures.
⁎ Corresponding author. Tel.: + 49 5323 722034; fax: + 49 5323 723155.
E-mail address: roman.weber@ievb.tu-clausthal.de (R. Weber).
Photographs of ash deposits from different boiler regions are
shown in Fig. 2. Ash deposition problems manifest themselves in a va-
riety of ways [1–3]:
(a) reduction of heat transfer rate due to deposited ash acting as
an insulation layer. This results in an increase of the furnace
exit temperature and an excessive heat transfer rate to super-
heated steam,
(b) build up of large clinkers on the furnace walls or on the radiant
superheaters tubes. Shedding of such a massive deposit may
damage boiler ash hopper,
(c) partial blockage of heat transfer tube banks which increases
gas velocity and may lead to enhanced corrosion,
(d) impendence to gas flow as a result of excessive fouling in the
convective banks,
(e) burner “eyebrows” deposits distort the burner aerodynamics
which may cause problems with flame stability and ignition.
Nowadays more and more often Computational Fluid Dynamics
(CFD) calculations are used to predict performance of boilers. Above
all, multiphase flow simulations become a common practise in boiler
design. Detailed review of the works on CFD-predictions of industrial
boilers is beyond the scope of this paper, however publications
related to the development of CFD-based ash deposition models
are relevant. There exist not too many publications where CFD-

0378-3820/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.fuproc.2011.09.008
114 R. Weber et al. / Fuel Processing Technology 105 (2013) 113–128
Convective heat exchange surface for
superheating and reheating steam
Convective
7
Combustion surface
chamber lined for water
with tubes for 6
heating
raising steam
8 for superheating
4 chamber lined
5 9
To air preheater
ESP and stack
Burners
3
2 2
Main locations of ash deposition
1 Ash hopper (bridging)
1 2 Ash slope (mechanical damage)
Fouling
3 Burner (eyebrows)
Slagging 4 Wall slag
5 Division wall slag (where approriate)
6 Platen (birdnesting)
Ash 7 Convection bank (bonded deposits)
8 Economiser (bonded deposits)
9 Air heater (gas inlet fouling)
Fig. 1. Regions of ash related problems in boilers: left — conventional pulverised fuel fired boiler configuration; right — tower boiler configuration.
Compiled from Ref. [1,2].
computations have been used to predict slagging and fouling propensi-
ty in industrial boilers, see Table 1. This is not too surprising; although
the vision of possessing CFD-based ash deposition models is in our
opinion realisable, a significant body of the development work is need-
ed to achieve this goal. In short, the currently published papers just ini-
tiate the forthcoming proliferation of scientific articles in this field.
Recently a CFD-based model for straw-fired grate stocker was pre-
sented [5]. The model described combustion on the grate as well as in
the boiler volume above the grate. The emphasis was placed on de-
posit build up on both platen superheaters and tube banks. The sub-
model handling the ash transport accounted for inertial impaction,
turbulent transport and thermophoresis. An ash sticking sub-model
specific to straw ash was developed and the deposit mass flux was
calculated. Although reasonable predictions of the deposition rate
were achieved, the model is used to identify “the areas of potential
deposit formation problems” [5]. It was again observed that the iner-
tial impaction was the dominant mechanism of particle transport and
its rate was around an order of magnitude larger than the turbulent
particle transport rate through the boundary layer. It was postulated
that the initial deposit layer was formed mainly through the alkali va-
pour condensation which was then followed by a much faster ash
particles deposition phase.
In a similar publication, Forstner et al. [6] performed CFD-computa-
tions of a pilot scale 440 kWth biomass-fired grate furnace to predict
the growth rate as well as physical and chemical structures of deposits.
Inertial impaction and condensation of ash forming vapours were con-
sidered as mechanisms of matter transport to boiler surfaces. The con-
centrations of the vapour species were computed using chemical
equilibrium calculations whilst the inertial impaction using the discrete
phase model (see Section 2.2 below) of the CFD Fluent 6.1 code. It was
observed that in the primary combustion zone, deposits were basically
formed by impaction of coarse particles. In the cooled fire tube zone,
the wall temperature was low enough to enable condensation of alkali
vapour, however the inertial impaction was still the dominating mech-
anism of solid matter transport towards the walls.
Convective surface Economiser
for water heating 8
Convective heat
exchange surface
9
and reheating
steam To air preheater
7 ESP and stack
6 Combustion
4 with tubes for
raising steam
Burners
3
1
Ash
Epple et al. [8] used a CFD-based mathematical model to predict
combustion of lignites in a 330 MW boiler. A part of the work was
dedicated to identification of the boiler areas that might be exposed
to excessive slagging. Here also a Lagrangian particle tracking, pre-
sumably similar to the one described in Section 2.2 below, was used
and the assumption was that the CFD code itself took care of the par-
ticle impaction efficiency. In other words, the CFD-computed flow
pattern in the vicinity of the deposition surface and the Lagrangian
particle tracking determined whether a particle hit a surface or not.
Mueller et al. [9] used a combination of CFD and advanced fuel
analysis to predict the ash deposition behaviour in a 105 MW biomass
fired bubbling fluidized bed boiler. The ash particle trajectories were
computed using the Lagrangian model (see Section 2.2 below) and
the impaction efficiency was assumed to be 1 since the worst case
scenario was considered. The model correctly identified the location
of the boiler areas where intensified slagging occurred.
Eddings et al. [11] presented a case study illustrating how knowl-
edge of the transformation of pyrites coupled with CFD models could
be used to explain problems of corrosion that were encountered in
utility boilers as a result of modifications to reduce emissions of nitro-
gen oxides. Schell et al. [12] computed a 450 MWel boiler fired with a
brown coal using AIOLOS CDF-code of TU Stuttgart. Transport of ash
particles towards the deposition surfaces was computed assuming
that particles smaller than 20 μm followed the gas flow without any
slip, whilst for larger particles a Lagrangian approach was used.
Thus, turbulent eddy transport and thermophoresis were the trans-
port mechanisms for particles smaller than 20 μm whilst particles
larger than 20 μm were transported by inertial impaction and tur-
bulent eddy transport. The adhesion behaviour was approximated
by a linear dependence of the sticking probability with particle tem-
perature; from zero at 970 °C to one at 1230 °C. Their considerations
on the effect of deposit formation on heat transfer rate make the pub-
lication of Schell et al. [12] very informative.
Bernstein et al. [13] presented CFD-simulations of 800 MW brown
coal fired boiler (Schwarze Pumpe, Germany) of which the slag
 R. Weber et al. / Fuel Processing Technology 105 (2013) 113–128 115

1 2

5 6

Fig. 2. Examples of ash deposits in different boiler regions [4] 1 — superheater; 2 — burner area and boiler walls; 3 — boiler walls; 4 — burner quartz; 5, 6 — superheater tubes in
waste incinerator.

deposition model was a part of. Perhaps the most interesting part of
the Bernstein et al. [13] work was the usage of a function describing
the sticking probability. The function was a product of three terms ac-
counting for the temperature effect, the particle concentration effect
and the oxygen concentration effect (oxidising versus reducing
environment).
The objective of this paper is to examine several approaches used
to predict slag deposition. The emphasis is placed on details of the
sub-models used and the goal is to determine under which circum-
stances the accuracy required for quality predictions is achievable.
Many models attempt to predict the time–temperature history of
the fuel particles from the injection to the deposition. In our opinion,
in such models, it is not possible to decouple modelling of ash depo-
sition from simulations of pulverised coal combustion. Both are so
strongly coupled that they must be discussed together. Whilst dis-
cussing accuracy and relevance of sub-models we are going to use
the measurement of Kupka et al. [14–16] where in the slagging rig
shown in Fig. 3, blends of a South African bituminous “Middleburg”
coal were co-fired with a municipal sewage sludge, a sawdust and a
Refuse-Derived-Fuel. In this paper we consider the pure coal flame
only.
This paper is not a review of all aspects needed for successful pre-
dictions of slagging and fouling in power station boilers. The issue is
so complex and so interdisciplinary that producing such a review
exceeds our time resources or perhaps even our competence. Instead,
we wish to underpin some critical modelling issues which have often
been overseen even by experienced CFD modellers. Predicting the
particle arriving rate to the deposition surface or predicting the car-
bon content of solid particles hitting the surface, belong to such is-
sues. Furthermore, the issues of particle rebounding or sticking to
the deposition surface are so complex that there are no universal
modelling approaches available. In other words, our paper examines
details of a number of CFD-sub-models and underpins their short-
comings rather than suggests remedies. In order to emphasise the
shortcomings of the sub-models, in the paragraphs that follow, we
quote equations which are known in the CFD-community. We include
them in order to avoid confusions as to which formulation/equation is
being actually considered.
2. Mathematical modelling
In coal-fired boilers the solid fuel is pneumatically supplied to
the burner with solids to transport-air ratio (load) of 1–2 kg/kg,
typically. In the burner zone the ratio rapidly decreases to 0.1–
0.2 kg/kg whilst in the superheater section typical values are an
order of magnitude smaller. Two-phase flows of 1–2 kg/kg load
are regarded as medium loaded flows whilst the flows in the super-
heater zone are weakly loaded. Whilst computing such flows, the
116 R. Weber et al. / Fuel Processing Technology 105 (2013) 113–128

Table 1
CFD-based predictions of deposit formation in industrial boilers.

Work Boiler/fuel Mathematical model

Kær et al. [5] (2006)
Grate-fired, Combined Heat&Power Plant,
Masnedø, Denmark
Fuel: straw
Deposit regions considered: free-board;
platen superheaters and tube banks
Fluent CFD code; Ash formation model for grate combustion of straw,
Ash transport through inertial impaction and thermophoresis, Ash
sticking propensity determined using melting curves for potassium
salts and silica-rich particles

Forstner et al. [6] (2006) Grate-fired 440 kWth boiler Fluent CDF code, Ash release model for grate combustion of wood,
Fuel: waste wood Ash release model for ash forming vapours, Ash transport through
Deposit regions considered: combustion inertial impaction and condensation of vapours, Ash sticking
chamber walls, fire tube walls propensity determined using Urbain [7] viscosity model

Epple et al. [8] (2005) 330 MWel boiler of Meliti Achlada Florina, Greece FLOREAN code of TU Braunschweig, Fly ash properties predicted using
Fuel: lignite a coal combustion model, Ash transport through inertial impaction,
Deposit regions considered: furnace walls Ash sticking propensity based on ash fusion temperature

Mueller et al. [9] (2003)
105 MW fluidised bed boiler of Combined
Heat&Power plant, Idbäcken, Nyköping, Sweden
Fuel: wood
Deposit regions considered: Free-board, secondary
combustion chamber, boiler walls
Fluent CFD code, Ash formation model includes fly ash and bottom ash
fractions which are determined using chemical fractionation analysis,
Ash transport through inertial impaction, Ash sticking propensity
determined using melting curves for the fly ash; sticking temperature
corresponds to 15 wt.% of molten phase in the fly ash

Fan et al. [10] (2001) 300 MWel tangentially-fired boiler In-house CFD code, Fly ash properties predicted using a coal combustion
Fuel: hard coal model, Ash transport by inertial impaction, Ash sticking propensity using
Deposit regions considered: boiler walls a critical viscosity model

Eddings et al. [11] (2001)
900 MWel tangentially-fired boiler
Fuel: bituminous coal
Deposit regions considered: boiler walls
GLACIER code of Reaction Engineering International, Fly ash properties
predicted using a coal char combustion model, Separate treatment of
extraneous and included pyrite, Ash transport by inertial impaction,
turbulent eddy transport and thermophoresis, Ash sticking propensity
based on the critical particle temperature T250 of the fly ash

Schell et al. [12] (2001)
450 MWel German boiler AIOLOS code of IVD of Stuttgart, Fly ash amount and size distribution are
Fuel: brown coal model inputs, Ash transport through inertial impaction, turbulent eddy
Deposit regions considered: boiler walls transport and thermophoresis, Ash sticking propensity determined using
a linear dependence of the sticking probability with temperature, Effects
of the deposit on heat transfer rates are determined

Bernstein et al. [13] (1999) 800 MWel boiler, Schwarze Pumpe, Germany Fluent CFD code, Fly ash properties predicted using a coal char combustion
Fuel: brown coal model, Ash transport by inertial impaction, Ash sticking propensity
Deposit regions considered: boiler walls expressed as a product of three terms: melting temperature function,
particle concentration function and oxygen concentration function

50 kW vertical combustion chamber Burner design

Secondary air Natural gas

Primary air + Pulverised fuel +
Pulverised fuel Primary air

2. Burner Secondary Air Secondary Air

0.55 m Movable block
Port 1, Residence time: 0.3 s swirl generator
0.55 m
1. Fuel feeder Port 2, Residence time: 0.6 s
0.55 m
Radiative section
(electrically heated) Port 3, Residence time: 1.6 s Burner quarl
2.2 m length 300 mm 0.55 m

3. Reactor Port 4, Residence time: 3.8 s

Convective section Port 9, Residence time: 6 s 7. Filter
(water cooled) Pressurised Air
1.8 m length

5. Quench system
4. Drum, gravity/drag
force seperation 6. Cyclone
Fly ash
Atmosphere
Fly ash 8. Flue gas blower

Fig. 3. A schematic diagram of the slagging reactor [14–16].

 R. Weber et al. / Fuel Processing Technology 105 (2013) 113–128 117

assumption is being made that there is no interaction between the The energy balance equation reads
solid particles. Fly ash particles are typically in the diameter range
from few microns to around 100 μm. However, sub-micron particles  
∂ ∂ μeff ∂h
are also present and in some specific applications (for example for ðρUi hÞ ¼ þ Scomb þ Srad þ Spart ð4Þ
∂xi ∂xi Pr ∂xi
brown coal fired power stations) particles substantially larger
than 100 μm have been observed.
where h stands for the physical enthalpy of the gaseous stream and Pr
is the Prandtl number. The right hand side contains source/sink terms
2.1. Fluid-flow from the burner to the deposition surface
due to combustion of volatiles, radiation and energy exchange be-
tween combusting char particles and gas phase.
2.1.1. The gaseous flow-field
It is imperative to aim at quality predictions of the gas phase temper-
Nowadays the Reynolds Averaged Navier–Stokes approach is the
atures however partial pressures of oxygen and carbon dioxide are also
most common method for predicting particle-laden flows in boilers.
important as far as slagging and fouling are concerned. The question
The continuity equation
arises as to what is needed to obtain quality predictions of both chemi-
cal composition and temperature of the gas phase. Fig. 4 shows the
∂
ðρUi Þ ¼ Spart ð1Þ measured and CFD-computed values for a 50 kW experimental rig
∂xi
shown in Fig. 3. In order to obtain such a good agreement between
the measured and computed values the following is required:
and the momentum balance equation
(a) the fluid flow must be accurately predicted. To this end the
! !
∂
 ∂p ∂ ∂Ui ∂Uj inlet conditions to the computational domain must be known
ρUi Uj ¼ − þ μ þ −ρ P
ui uj þ Spart; j ð2Þ (these have been measured) and fine numerical grids together
∂xi ∂xj ∂xi ∂xj ∂xi
with a higher-order numerical solver are needed. The predic-
tions shown in Fig. 4 have been obtained using a numerical
are accompanied by a model of turbulence that relates the Reynolds
P grid of cell size in the range from 1 mm to 5 mm. The grid
stresses ρ ui uj to known or calculable quantities. In the above equa-
must be body-fitted and the step-wise representation of the
tions Ui, ui stand for time-mean and fluctuating velocities, ρ and μ
domain boundary, which was used in the past, should be aban-
for fluid density and viscosity whilst p is the static pressure. Spart
doned. For the predictions shown in Fig. 4 the k–ε turbulence
and Spart,j are source terms representing interactions between the
model has been used, however many other turbulence models
solid and the gas phase. There exists a proliferation of turbulence
would provide predictions of the same quality.
models which can be used including the most commonly applied
(b) the coal fired has to be accurately characterised. First of all, its
k–ε model. Thus, the question which turbulence model should be
particle size distribution must be known which is self-evident.
used is relevant. The answer however cannot be given without a de-
The high temperature volatiles, their rate of release, and com-
tailed analysis of the numerical solvers and the numerical errors asso-
position have been measured for the Middelburg coal in asso-
ciated with them. A combination “numerical solver–numerical
ciated fuel characterisation experiments using a plug flow
grid-turbulence model” is decisive in obtaining high quality predic-
reactor shown in Fig. 5 [24]. Similarly, the rate of char combus-
tions of a turbulent flow with the first two, namely the grid and the
tion and the parameters of the char combustion sub-model
solver, being more important than the turbulence model itself. One
have been determined in these experiments.
should realise that even with the current generation of fast com-
(c) the radiative heat transfer is accurately predicted. Often a
puters, predictions of turbulent flows in boilers are performed using
question is asked whether radiative heat transfer calculations
numerically coarse grids and the primary concern must be placed
should be performed spectrally or as a grey medium. Spectral
on minimising the numerical related errors. To this end, numerical
calculations for coal combustion are rare. The Discrete Ordi-
solvers of a higher order as well as fine grids are needed. The choice
nates Methods with the gas phase absorption coefficient taking
of the turbulence model is then of secondary importance and the
a value of 0.3 m − 1 and the scattering coefficient of 0.4 m − 1 is
k–ε model is just fine for most of the applications. Whenever a swir-
used in our calculations.
ling flow is considered, for example in the burner vicinity, some im-
provements may be observed when a more complex turbulence It is perhaps worth emphasising that the gas-phase temperatures
model (for example the Reynolds Stress Model) is used. in the near burner zone are determined by the rate of combustion

2.1.2. The chemical composition and temperature of the gas phase

The mass balance equation for species J is as follows: Port 1 Port 2 Port 3

! 25 2000
∂
 ∂ μeff ∂mJ
CO (Vol. ppm dry); Temperature (K)

1800
ρUi mJ ¼ þ SJ ð3Þ
∂xi ∂xi Sc ∂xi
O2 and CO2 (Vol.% dry)

20 1600
1400
with mJ being the mass fraction of species J, Sc is the Schmidt number, 15 1200
SJ is a source/sink term due to gaseous combustion and generation O2-measured CO2-measured 1000
of gaseous species (volatiles, CO) from the coal phase. Typically O2-predicted CO2-predicted
10 800
five gaseous species are considered: oxygen, volatile matter, carbon T-measured CO-measured
600
monoxide, carbon dioxide and water vapour although efforts are T-predicted CO-predicted
5 400
being put into the development of more comprehensive schemes
for gas phase combustion [17,18]. The combustion rate in the gas 200

phase is typically calculated either using the eddy break-up model 0 0

0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4 1,6 1,8
[19] or the Eddy Dissipation Concept [20]. The latter model has
Axial distance (m)
become very popular due to its ability to adapt complex chemical
schemes [21–23]. In terms of the turbulent combustion both models Fig. 4. Measured and computed temperatures and gas composition along the slagging
provide comparable combustion rates. reactor shown in Fig. 1 (15 kW flame of Middelburg coal).
118 R. Weber et al. / Fuel Processing Technology 105 (2013) 113–128

N2 , CO2
Air, O2 Fuel Feeder

N2 ,CO2, O2
Ceramic
pipes
Gas heating
elements N2 ,CO2, O2

Injector

Thermocouples
for Control system

Heating elements Sampling ports

(5KW for segment)
(4 heating elements
for segment)

Sampling probe

Main off-gas
Cooling gas
Gas cooler
Particle cyclone
Cooling water Probe off-gas
Gas cooler
Particle cyclone

Fig. 5. Plug-flow reactor for characterisation of solid fuels.

of volatiles. Consequently, the amount and the chemical composition laden flows and this procedure was used in most of the recent publi-
(LCV) of volatiles are important here whereas heat losses due to radi- cations on slagging [29,5,6,9,13]. In simulations of Epple et al. [8],
ation are of secondary importance. When the volatile combustion Schell et al. [12], and Eddings et al. [11] somewhat similar methods
ceases, the rate of the temperature decay along the flame is deter- were applied although details remain unknown to the authors.
mined by radiative heat transfer and it is altered by the char combus- When only the drag and gravity forces act on a small particle in a gas-
tion rate. This general statement is applicable to single flames as well eous environment the equation of motion of the particle can be
as combustion in boilers. expressed in a Lagrangian frame of reference as:
!
dup
 ρp −ρ
2.2. Properties of char/ash particles along their trajectories from the ¼ −α· up −ũ þ gx · ð5Þ
burner to the neighbourhood of the deposition surface dt ρp

2.2.1. Particle trajectories (Lagrangian tracking) 18μ CD Re

The particle tracking procedure used in our work was developed
at the beginning of the eighties at the University of Sheffield
[25–27] and implemented into the first version of the CFD Fluent
code. Nowadays, it has routinely been used [28] to simulate weakly
α¼ · ð6Þ
ρp D2p 24
where up and ũ are the instantaneous components of the particle and
gas velocity respectively; gx is the acceleration due to gravity. By
 R. Weber et al. / Fuel Processing Technology 105 (2013) 113–128
integrating Eq. (5) one obtains the instantaneous position of the par-
ticle. The drag coefficient is expressed as a function of the relative
Reynolds number with the general form:
2
CD ¼ a1 þ a2 =Re þ a3 =Re ð7Þ
where the a's are given [30] for several ranges of Reynolds number.
To account for the effect of turbulence on the trajectories of indi-
vidual particles the instantaneous gas velocity ũ is decomposed into
time-mean and fluctuating components ū and u′. It is assumed that
the fluctuating velocity prevails for a time period equal to the life
time of the fluid eddy that the particle is traversing. This time can
be expressed in terms of the local kinetic energy of turbulence, k,
and its dissipation rate, ε, as:
k
τ≅0:16 : ð8Þ
ε
The values of the fluctuating velocities can be assumed to possess
a Gaussian distribution so that these are related to the root mean
square fluctuations by:
rffiffiffiffiffiffi

P 1=2 2k
u′ ¼ ϕ u 2
¼ϕ ð9Þ
3
where ϕ is a normally distributed random variable. The overbar indi-
cates time averaging. Details of the integration procedure for Eq. (5)
can be found elsewhere [28].
The above described particle tracking procedure has become very
popular over the last two or three decades due to the widely spread
CFD Fluent code. The prerequisite to an accurate particle tracking,
resulting in a proper distribution of solid phase over the furnace/
boiler volume, is the correctness of predictions of the time-mean
flow field. Although in the above equations the fluctuating velocity
components are sampled over eddies of given (calculated) kinetic
energy, a precise determination of its value is less critical. However,
the CFD user should make sure that the time step of the integration pro-
cedure of Eq. (5) is small enough to minimise the integration errors.
2.2.2. Particle temperature, char burnout and diameter
The particle temperature along its trajectory is calculated using
the particle energy balance which considers both the convective
heat transfer and the absorption/emission of radiation at the particle
surface:
dTp


4 4
mp cp ¼ hAp T∞ −Tp þ εp Ap σ θR −Tp ð10Þ
dt
where mp is the mass of the particle, cp the heat capacity of the parti-
cle, Ap the surface area of the particle, Tp the particle temperature,
T∞ the local temperature of the gas phase, h the convective heat
transfer coefficient, εp the particle emissivity (εp = 0.8), σ the Stefan–
Boltzmann constant, and θR is the radiation temperature. This equation
assumes that there is negligible internal resistance to heat conduction
(small Biot number) i.e. the particle is at a uniform temperature through-
dm
out. During the char combustion the additional term fh · dtp ·Hreac has
to be added to Eq. (10). This term accounts for the energy release in
the surface reactions which is directly absorbed by the char particle.
The coefficient fh takes a value of 0.3 following the work of Field et al.
[31].
The initial temperature (typically 60–70 °C) of coal particles
injected into the furnace is known and corresponds to the tempera-
ture of the gaseous (air) stream used for pneumatic transport of pul-
verised fuel from the mill to the boiler furnace. Along its trajectory
from the injection location to the flame front (volatile matter com-
bustion region), the fuel particle is heated up, its moisture is given
off, and the volatiles are released so that the particle temperature
119
remains lower than the gas phase temperature. In the flame front,
the particle temperatures reach the gas phase temperature values
whilst further downstream they are larger than gas phase tempera-
tures, by 100–200 °C. The larger the particle diameter, the larger is
this temperature overshoot which is due to the enthalpy release dur-
ing char combustion. Far downstream, where the char combustion
ceases, the particle temperatures approach the gas phase tempera-
tures. In CFD-simulations shown in Fig. 4, at port 2 (residence time
around 0.6 s, see Fig. 3) particles larger than 50 μm possess tempera-
tures which are around 100 °C higher than the gas-phase tempera-
ture whilst the temperature of particles smaller than 50 μm equals
the gas-phase temperature. At port 3 (residence time around 1.6 s,
see Fig. 3), for particles larger than around 90 μm, the particle tem-
peratures are higher than the gas-phase temperature by not more
than 30 °C.
Predicting char burnout is difficult [32–36]. There exists a prolifer-
ation of char combustion models and their review is beyond the scope
of this paper. However, it is perhaps fair to say that calculating char
combustion rates along a known trajectory (known gas-phase tem-
perature–time history) requires a comprehensive knowledge of char
particle morphology. Furthermore, one observes a substantial de-
crease of carbon oxidation rates for char burnouts larger than 80%
which may be associated with annealing, carbon being trapped in
the mineral matter, or other factors. In our opinion, the only reliable
way to overcome this difficulty is to carry out measurements of the
char combustion rates for the fuel in question. Such measurements
were performed for Middleburg coal using the isothermal plug flow
reactor shown in Fig. 5 and the parameters of the macroscopic pore
model were fitted to the measured rates. The pore model, with the
fitted parameters, was then incorporated into the CFD code used in
our work. Fig. 6 shows the measured and computed degrees of burnout
at traverses two and three. We wish to stress here again that obtaining
such a good agreement between the measured and computed burnout
was possible due to the experimental determination of the char com-
bustion rates for the coal in question. In terms of carbon in ash, the mea-
sured 97% burnout at port 2 corresponds to 20.12% carbon-in-ash
content whilst the predicted 94% corresponds to 33% carbon-in-ash.
At port 3 the predictions agree very well with the measured values
and 99% char burnout corresponds to 7.7% carbon-in-ash.
The third issue, beyond predicting particle temperature and burn-
out, we wish to deal with in this paragraph, is predicting particle di-
ameter. Fig. 6 shows also that the fly-ash flowrate upstream of ports
2 and 3 can be reasonably well predicted by the CFD-model. However,
this is not enough. It is required that the size (diameter) of the fly-ash
particles is also accurately predicted since the particle diameter is
detrimental in determining the impaction efficiency (see below). To
examine this issue, we sampled the fly-ash at ports 1, 2 and 3 and
analysed the sample for the particle size distribution using the
laser-diffraction method. Using solid dots, the measured distribution
of the sampled solids as a function of particle diameter is shown in
Fig. 7 whilst the CFD-predicted distributions are plotted using solid
100
Measured
90
Predicted
Burnout, %
80
70
60
50
Port 1 Port 2 Port 3
Fig. 6. Measured and CFD-predicted char burnout in the slagging reactor shown in
Fig. 3 (15 kW flame of Middelburg coal).
120 R. Weber et al. / Fuel Processing Technology 105 (2013) 113–128
Measured
Port 1
Port 2
Port 3
With
Correction No
Port 1 Correction
Port 2 Port 1
Port 3 Port 2
Port 3
Fig. 7. Measured and CFD-predicted size distribution of fly-ash (15 kW flame of
Middelburg coal). Eq. (11) shows the correction introduced.
lines. A substantial disparity between the measurements and the pre-
dictions can be seen; the CFD-calculated distributions are clearly
shifted towards larger particles. We recall here that the char burnout
model used does not allow for particle shrinking. It is rather easy to
see that such a substantial disparity cannot be explained by particle
shrinkage only and particle fragmentation plays a key role. There ex-
ists a substantial body of literature on particle fragmentation, as ex-
emplified by Ref. [37,38], and its review is beyond the scope of this
paper. It is perhaps worth stressing that the reliability of the fragmen-
tation models is questionable since the issue is very complex. In order
to account for the change of the particle diameter during char com-
bustion either through shrinkage, or fragmentation, or both, we intro-
duce an ad hoc correction function
 0:7
D C
¼ ð11Þ
D0 C0
where D and D0 are particle diameter and its initial value, respectively
whilst C and C0 stand for carbon content of the particle and its initial
value, respectively. The exponent in the above relationship was
obtained to fit the measured particle diameters. After introducing re-
lationship (11) to the char combustion model, the CFD-predicted
fly-ash mass versus diameter distribution became closer to the mea-
sured values, as shown in Fig. 7 using the solid lines.
At the end of this paragraph we wish to reemphasize that predict-
ing carbon-in-ash content along the particle trajectory is challenging
and quality predictions, such as shown in Fig. 6, are obtainable only
when special techniques to characterise the fuel are used with em-
phasis on the last 20% of burnout. Consequently, the calculated
carbon-content of fly ash particles forming the deposit may contain
substantial inaccuracies which then result in error predictions of the
deposit chemistry. Similarly, obtaining quality predictions of the
size distribution of fly ash particles near the deposition surface is
very difficult, as shown in Fig. 7. The above used “ad hoc” correction
(Eq. (11)) is by no means general and it has been derived knowing
the actual (measured) particle size distribution which in general is
very seldom known. CFD-codes incorporate functions for particle
shrinking, which typically take the form (1-burnout) α, however de-
termination of the α-exponent, so that the char burnout and the par-
ticle size distribution are accurately predicted, is seldom carried out.
On top of these inaccuracies, the particle fragmentation has not
been resolved at all. Thus, it is perhaps fair to say that one cannot ex-
pect good quality predictions of the char burnout and the particle size
distribution near the deposition surface of a boiler by performing
CFD-simulations (“from the fuel injection to the neighbourhood
of the deposit surface”) unless the simulations are supported by
both comprehensive fuel characterisation experiments and com-
bustion system dependent measurements of particle shrinking and
fragmentation.
2.3. Predicting transport of matter towards deposition surface
There exist several mechanisms of transport of matter towards de-
position surfaces and these include transport of solids (particles) as
well as transport of vapours and their condensation. The four basic
mechanisms involved in particle transport are: particle motion due
to drag force, gravity, eddy diffusion and thermophoresis. The most
important forces in the particle tracking procedure described above
are the drag force and gravity. Surfaces on which deposit might be
formed are in different orientations to fluid streams depending on
which part of the boiler is considered. For example, ash deposition
on burner eyebrows is driven by complex recirculating flows, deposi-
tion along a tube wall is driven by fluid streams moving along the
wall, whilst deposition on economizer pipes is governed by
cross-flows. Each of these deposition-surface/fluid-flow orientations
should be analysed separately. In this paper we restrict ourselves to
a deposition pipe in a cross flow, see Fig. 8, and the reason for the
choice is simple; there exists an accurate expression for the impaction
efficiency on the front surface of the pipe. This allows for examining
whether CFD-predictions can provide accurate predictions of the im-
paction efficiency and to determine accuracy of such predictions. For
most of other deposition-surface/fluid-flow orientations, the assess-
ment of the inaccuracies in computing the impaction efficiency
using CFD-codes is not so straight forward and the CFD-user is often
confused. Transport of vapours and their condensation is obviously
independent of the deposition-surface/fluid-flow configurations.
2.3.1. Inertial impaction and thermophoresis
A bank of tubes in a cross flow is a typical arrangement for an
economizer of power station boilers. The pipes are typically of 1–2
inch diameter although in specific applications smaller diameters
are also used. The flue gas velocities upstream of the bank are in the
3–10 m/s range. Thus, it is easy to see that the Reynolds number of
the flow over a single pipe is in the 400–2000 range which is below
the critical Reynolds number of around 200,000. There exist several
flow regimes at a circular cylinder, see for example Schlichting and
Gersten [39]. For the Reynolds number range from 400 to around
200,000 the drag coefficient takes a value of 1.2 and is independent
of the Reynolds number. The flow past the cylinder is laminar and
the laminar boundary layer separates at an angle of around 80°.
Until the separation occurs the flow is symmetric and stationary.
The flow downstream of the cylinder is periodic and its frequency is
independent of the Reynolds number. The dimensionless frequency
called the Strouhal number (= f · D/u, where f is the frequency, D is
the pipe diameter and u is the bulk gas velocity) is typically around
Fig. 8. Mechanism of inertial impaction on a pipe.
 R. Weber et al. / Fuel Processing Technology 105 (2013) 113–128
Wessel and Righi,1988
Israel and Rosner,1982
1,0
Impaction efficiency
0,8 disastrous CFD-predictions
0,6 inaccurate CFD-predictions
potential flow
0,4
0,2
accurate CFD-predictions
0,0
0,01 0,1 1 10
Stokes number
Fig. 9. Impaction efficiency for spherical particles approaching a circular cylinder [42].
0.20. It is interesting to note that even if the upstream flow becomes
unsteady, the boundary layer remains laminar and the drag coeffi-
cient increases by not more than 40%.
In many works on modelling of slagging and fouling, the particle
tracking procedure described above (see Eqs. (5)–(9)) or similar
methods are used to determine whether a particle hits the deposition
surface. It is then assumed that the particle tracking procedure prop-
erly (accurately) predicts the so called impaction curve e.g. the de-
pendence of the impaction efficiency on the Stokes number 1
defined as:
ρp ·d2p ·up
St ¼
9·ηg ·D
where ρp, dp, up, ηg stand for the particle density, particle diameter,
bulk particle velocity and gas dynamic viscosity, respectively. D is
the diameter of the deposition pipe. Fig. 9 shows impaction efficien-
cy curves for particles impacting on front face of a cylinder. The
curve marked as “potential flow”, has been obtained on the basis
of theoretical analysis of a potential flow around cylinders [40,41].
Its correctness has been confirmed in many experiments and the
curve is commonly used in filtration and aerosol science. This “po-
tential flow” curve can be perfectly reproduced by the Lagrangian
tracking procedure described above (see Eqs. (5)–(9)), however it
requires a very experienced modeller who understands the physics
of flows around cylinders, the CFD-code and details of the particle
tracking. A number of factors must be correctly handled in order
to obtain the required perfection using the CFD-code. An analysis
of these factors will be given in our forthcoming publication [42].
In Fig. 9 we show curves marked as “accurate CFD-predictions”
and “disastrous CFD predictions”; often CFD-calculations produce
a curve lying somewhere between these two solutions which we
name as “inaccurate CFD-predictions”. It is worth observing that
the error in CFD-predictions of the impaction efficiency may be sub-
stantial, in particular for Stokes number in the 0.1 to 4 range. For
example for a 50 μm ash particle approaching an inch (O.D.) pipe
with a velocity of 5 m/s, (St = 1.43) the CFD-inaccurately predicted
impaction would be by a factor of 1.25 too large. For a particle of
25 μm (St = 0.357) the impaction efficiency may be overestimated
by a factor of 2.85.
It should be however emphasised that in several CFD-based math-
ematical models [5,43–45] and in some experiments [46] where de-
position on a pipe in a cross-flow was considered, the impaction
efficiency was calculated using correlations developed for potential
flow over a circular cylinder [40,41]. Below we quote this useful
1
In the literature also another definition of Stokes number appears in which instead
of number 9, the number 18 is used.
121
correlation (see Eq. (6) in Ref. [40]) which is represented in Fig. 9
by the potential flow curve:
Impaction Pef f iciency
1
¼ :
1 þ 1:25ðSt−0:125Þ−1 −0:014ðSt−0:125Þ−2 þ 0:508·10−4 ðSt−0:125Þ−3
ð13Þ
In other words for each particle approaching the pipe, the Stokes
number is calculated and the corresponding impaction efficiency is
determined. This is indeed a simple, elegant and accurate method of
calculating the impaction efficiency on a single tube.
Thermophoresis takes place when steep temperature gradients
100
occur in the hot gases close to the walls or heat transfer surfaces. It
may cause a motion of ash particles from hot to cold surfaces in direc-
tion opposite to the temperature gradient. Thermophoresis is partic-
ularly significant for particles smaller than dp ≤ 3 μm in diameter
[47]. In Eq. (5) the thermophoretic force has been included as an ad-
ditional acceleration (force/unit mass) term, Fx:
1 ∂T
Fx ¼ −DT; p ð14Þ
mp T ∂x
where mp is mass of the particle, T is the local fluid temperature,
whilst DT,p stands for the thermophoretic coefficient calculated as fol-
lows [48]:
6πdp μ 2 Cc ðK þ Ct Kn Þ
DT; p ¼ − ð15Þ
ρð1 þ 3Cm Kn Þð1 þ 2K þ 2Ct Kn Þ
ð12Þ where: Kn is the Knudsen number (= 2λ/dp) in which λ is the mean
free path of the fluid and dp is the particle diameter. K (= k/kp) is
the ratio of the fluid thermal conductivity (k = 15/4 μ R) based on
translational energy to the particle thermal conductivity (kp). The co-
efficients appearing in Eq. (15) take the following values: Cs = 1.17,
Ct = 2.18, Cm = 1.14. The fluid viscosity is denoted by μ . The above
expression is valid for spherical particles and the fluid is an ideal gas.
2.3.2. Rebounding and sticking
Establishing whether a particle bounces off or sticks to the deposi-
tion surface begins with calculating the kinetic energy at the impac-
tion [49,50]. One calculates then the energy dissipated by the
collision and determines whether this dissipation is sufficient to pre-
vent the particle from rebounding away. If the particle has more ki-
netic energy than the energy dissipated in the collision the particle
is judged to bounce off the surface. If the collision can dissipate all
of the particle's kinetic energy, the particle is judged to stick. Obvious-
ly the difficulty is in developing the relationship between the energy
dissipation and kinematic parameters of the particle of a given size
and mass. Typically the coefficients of restitution are involved and
they can be calculated for each particle velocity component [49,50].
Such calculations, which can be performed for each particle tracked
near the deposition surface,
→  result in the determination of so called crit-
 
ical particle velocity (Vc ). The rebounding condition
→ is usually [49,50]
 
set either on the particle velocity absolute value (V ) or on the normal
 
→
component (Vn ), so that the adhesion occurs when
→  → 
   
Vn ≤Vc  ð16Þ
or
→ → 
   
V ≤Vc : ð17Þ
Care must be exercised when applying condition (16) to impaction
on a pipe (or generally on curved surfaces) since it allows for adherence
of particles within the 90° region, as shown in Fig. 10 whereas
122 R. Weber et al. / Fuel Processing Technology 105 (2013) 113–128

impactions and adhesion. Mechanical collisions and adhesion of

Fig. 10. Rebounding criteria: Left — Eq. (17), Right — Eq. (16).
Adapted from Ref. [53].
experience teaches that front face deposits are formed within a 50–60°
sector. Thus, for pipe deposits, condition (17) is recommended.
The above underlined rebounding procedure is rather popular in
particle technology [49,50] to estimate bouncing probability of solid
particles (not molten) impacting on a clean surface (no sticking
layer). There exist only a limited number of publications where the
procedure is used to determine adhesion of ash-forming deposits
[51–54].
Viscosity-based models assume that sticking propensity is deter-
mined by the effective viscosity of fly-ash. The effective viscosity is in
turn calculated as a function of the particle temperature and its compo-
sition. In the simplest models the assumption is that an impact involv-
ing any viscosity below a critical value results in the particle sticking.
In the commonly used model of Walsh et al. [56], it is assumed that
the sticking probability is inversely proportional to its effective viscosity
(η) for viscosities larger than the critical value (ηcritical):
 
η
Sticking probability ¼ min critical ; 1
η
where the operator takes the smallest of the two terms in the brack-
et. For viscosities lower than the critical viscosity, the sticking prob-
ability equals unity. There is no consistency for choosing the critical
elastic–plastic solid spheres under thermal stress are considered by
Losurdo [53]. The Maxwell model, Voigt model and their combina-
tions are used to represent the variation of the strain–stress relation-
ship of deposits as a function of time and temperature. There seem to
be a high potential of such models for predicting not only sticking
propensity but also deposit shedding, deformation and dripping.
Ash fusion tests are historically the most frequently used methods
of assessing slagging propensities of coals. Fig. 11 (top) shows the
German (DIN) and the American (ASME) standards. According to
the DIN standard three characteristic temperatures are measured:
T(A) — softening temperature, T(B) — hemispherical temperature
and T(C) — fluid temperature. The ASME standard defines: initial de-
formation temperature (IT/ID), the first signs of rounding off of the
apex, hemispherical temperature (HT) and the flow/fluid tempera-
ture (FT). In many works on mathematical modelling of slagging,
the softening temperature is used as a particle sticking criterion, as
exemplified by Ref. [8]. When the particle temperature exceeds the
softening temperature, the particle sticks to the deposition surface
otherwise it is rebounded. In the work of Epple et al. [8] it was pre-
dicted that a decrease of the fusion temperature from 1140 °C to
990 °C resulted in an increase of the deposition rate by a factor of 3.4.
The group of Tran et al. [64–66], working on deposition problems
of kraft recovery boilers firing black liquor, was perhaps the first re-
search team that observed the relationship between the liquid
phase content and the adhesion propensity of fly ash particles.
When the content exceeded 18–20% a strong adhesion occurred
[65]. On this basis another concept for sticking propensity has been
developed which requires producing so called melting curves of the
mineral matter in question [64–66]. Such curves can be produced
IT (or ID) - initial deformation temperature
ASTM D1857 ST - softening temperature
HT - hemispherical temperature
FT - fluid temperature
Before IT ST HT FT
heating
DIN 51730
ð18Þ
Before ST HT FT
heating T(A) T(B) T(C)
1600

viscosity and values as low as 1 Pa-s and higher than 10 9 Pa-s have Sum of basic ash compounds = CaO+MgO+Fe 2O3 +Na 2 O+K20 (%)
been reported. Obviously the value of the critical viscosity has been SiO2+Al2O3+TiO 2+CaO+MgO+Fe 2O3+Na2O+K20 = 100%

a topic of much discussion [55]; Walsh et al. [56] used 8 Pa-s whilst 1500
Srinivasachar et al. [57] suggested that for US bituminous coals the
critical viscosity is between 104 and 108 Pa-s, Yilmaz and Cliffe [58] sug-
Softening temperature, °C

gest a value of 6.7× 109 Pa-s whilst Costen et al. [59] use 105 Pa-s for a
bituminous Hem Heath coal ash. For a German brown coal (Lausitz) 1400

a value of 2 Pa-s has been recommended [60]. For calculating the

ash viscosity as a function of composition and temperature the
model of Urbain at el. [7] is often used with an extension of Senior 1300
and Srinivasachar [61] (a comprehensive review of the subject can be
found in Vergas et al. [62]). The above sticking criterion (18) can be
reformulated using the temperature at which an abrupt change occurs
1200
in the viscosity–temperature relationship. Such a temperature is called
the temperature of critical viscosity. Then, when the particle tempera-
ture exceeds the critical viscosity temperature, the particle sticks to
the surface upon impaction. Obviously the critical viscosity temperature 1100
is a function of mineral matter composition. 0 20 40 60 80 10 0 12 0
A fly ash particle may soften showing a viscous liquid-like behav- Sum of basic ash compounds, %
iour. During such a softening the solid acts in between, being in a liq-
uid and a solid state, and its state is viscoelastic. Consequently, Fig. 11. Top — Ash fusibility tests; Bottom — Softening temperature as a function of
viscoelastic models seem appropriate to reproduce inelastic basic ash elements [1,63].

either by using differential thermal/thermogravimetric analysis
(DTA/TGA) or by performing thermodynamic calculations if the sys-
tem is not too complex. Fig. 12, taken from Zhou et al. [45], is an ex-
ample of a typical melting curve showing the melt fraction of the
particle as a function of particle temperature. The curve can be char-
acterised by several points among them the first melting temperature
(T0), the inflection temperature (Tc) and the complete melting tem-
perature (T100). It has been postulated that a temperature corre-
sponding to either ten percent [45] or fifteen percent [52] melt
fraction may be regarded as the sticky temperature (Fig. 12).
The melting curve concept became attractive in predicting ash de-
position of biomass-fired units in particular [45,52]. In the modelling
work of Forstner et al. [6], it has been assumed that above a melt frac-
tion of 0.15, the sticking probability is calculated with a linear in-
crease from zero to unity at a melt fraction of 0.7. Above 0.7 melt
fraction the deposit is regarded as liquid. Similarly, in the modelling
work of Kær et al. [5] (see also Akbar et al. [67]), an approximated
melting curve is produced by a linear interpolation between the melt-
ing curve of potassium salt and the curve for silicate particles.
2.3.3. Condensation of vapours
Fouling on cool surfaces of a boiler involves condensation of ash
forming vapours. It has been postulated that in boilers fired with bio-
mass, or in units co-firing a biomass with a coal, fouling on cool sur-
faces is initiated by a formation of a layer of alkali vapours. Fly ash
particles impacting on such a surface adhere to the sticky layer and
the initial phase of deposit formation begins. This mechanism can
be of a considerable importance in boilers fired with brown-coals or
lignites. In boilers fired with hard coals the alkali vapour deposition
proceeds typically with a much lower rate than the inertial impaction.
A list of compounds to be considered as ash forming would include
KCl, NaCl, PbCl2, ZnCl2, K2SO4, Na2SO4, K2CO3 and some others. The
vapours transport over the furnace volume can be modelled using
Eulerian balance equations (see Eq. (3)) written either for each com-
ponent or for each alkali element. To speed up the calculations, the
concentrations of the ash forming species (vapours) in computational
cells can be determined using thermodynamic equilibrium proce-
dures in combination with an In-Situ Adaptive Tabulation (ISAT) al-
gorithm [68,69]. In the numerical cells adjacent to surfaces, whose
temperature is lower or equal to the condensation temperature, the
rate of condensation of species i, ṁ i;cond (kg/m 2/s), can be calculated
as:
ṁ i;cond ¼ Mi ·β·ðc∞ −cwall Þ
where c∞ and cwall are the species concentration in the bulk flow, and
the saturation concentration at the wall, respectively. The mass
100
90
80 Probe ash deposit
Melt fractionf, %
70 T c and f c
60
50
40
30
20
T10
10
0 T0
0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500
Temperature, °C
Fig. 12. Typical melting curve for a deposit [45].
R. Weber et al. / Fuel Processing Technology 105 (2013) 113–128 123
transfer coefficient β (m/s) can be calculated using appropriate corre-
lations for the Sherwood number.
The above underlined procedure is applicable to vapour species
which are formed in the gas phase with rates that are much faster
than the turbulent transport rates so that the thermodynamic equilib-
rium assumption is applicable. Formation of sulphates (Na2SO4 and
K2SO4) may be kinetically controlled and should be treated different-
ly, provided that kinetic data are available. The main difficulty in the
above underlined procedure for determination of alkali vapours is in
knowing the speciation of the alkali species during devolatilisation
and char combustion and often measurements are needed, for the
fuel in question. The thermodynamic data base needed for such calcu-
lations is far from being complete and requires upgrading and
maintenance.
2.4. Mineral matter transformation along a particle trajectory
Fig. 13 shows a known cartoon representation of processes gov-
erning the transformation of mineral matter during coal combustion.
The sketch, produced by Sarofim and Helble [70], would have to be
modified here and there on the basis of the research carried out
over the last twenty years or so [71]. The sketch is used to highlight
the complexity which a mathematical modeller faces during the
model development. It is commonly accepted that the most impor-
tant path in the above sketch is the ash particle formation route.
Most of the mathematical models of slag deposition assume that the
mineral matter of the original coal is an inert material whilst other
models attempt to take into account transformation processes
which minerals undergo.
The development of analytic techniques based on Scanning Elec-
tron Microscopy (SEM), namely Computer Controlled-SEM (CCSEM)
and SEM-Electron Probe Microanalysis (SEM/EPMA), provides a new
understanding of interactions among minerals during combustion.
CCSEM quantifies particle size, shape and composition of mineral
matter on a two-dimensional basis in a polished section of the coal
or ash embedded in epoxy. It also shows the amount and nature of
liberated minerals in the coal. The SEM-based techniques provide in-
deed much more information than conventional coal ash analysis and
they reveal that transformations involve tens of reactions proceeding
in different parts of the fly ash particle [71].
2.4.1. Mineral matter as inert material
High temperature ashing and X-ray fluorescence are, due to sim-
ð19Þ plicity and low costs, commonly used methods. A coal sample is
burned under oxidising conditions at temperatures above 800 °C (at
815 °C according to the DIN standard) to generate the ash which is
then analysed using X-ray fluorescence. The ash composition is
expressed as a list of oxides: SiO2 + Al2O3 + TiO2 + P2O3 + Fe2O3 +
CaO + Na2O + K2O + SO3 = 100% (on weight basis). Silica, alumina,
titan and phosphorus oxides rank to acid elements of the ash. Iron,
calcium, sodium, potassium and sulphur oxides are of base nature.
T100 Occasionally the content of MnO, ZnO, BaO and CO2 is also deter-
mined. In most of the ash deposition models the composition of the
mineral matter does not change when particles fly through different
boiler zones. Consequently, chemical composition (expressed as ox-
T70
ides) of the ash deposited, is then identical to composition of the
fly ash coal ash sample. This is for example the case in the simulations pre-
sented in Figs. 4, 6 and 7 as well as in the works of Huang [44], Yilmaz
and Cliffe [29], Hecken et al. [72] (see also Costen [59] where ash min-
erals are treated as inert material).
An interesting CFD-based modelling work has recently been pub-
lished [73] with the aim to develop a model for predicting the size
and the chemical composition of a fly ash particle. The starting
point to the modelling work is the CCSEM data showing the distribu-
tion of twelve elements (Al, Si, P, Na, Mg, Cl, K, Ti, Cr, Fe and Ca) in
grains of minerals of a given (measured) diameter. On the basis of
124 R. Weber et al. / Fuel Processing Technology 105 (2013) 113–128

Nucleation
Coagulation

Inorganic
Vapors Submicron and Conden-
Agglomerated sation
Heterogeneous Ash (0.1-1.0µm)
Ash Condensation
Vaporization Thermo-
Particles phoresis
Mineral
Inclusion Coal
Particles Impaction
and Sticking

Char
Organically Particle
Bound or Char and Mineral
Dispersed Fragmentation
Inorganics Fly Ash
(1-20µm)
Excluded Char
Minerals Burnout
Decreasing Temperature

Fig. 13. Schematic diagram of processes governing the transformation of mineral matter during coal combustion (after Sarofim and Helble [70]).

the CCSEM data, a virtual coal [73] is generated in such a way that the
grains are randomly distributed in coal particles. However, the total
mass of the mineral grains equals the ash mass fraction. Then a
Lagrangian particle tracking (see Section 2.2) is performed for the vir-
tual coal with a char combustion model that allows for coalescence of
mineral grains and fragmentation of coal particles. Both diameters of
fly ash particles as well as their composition can be predicted at any
instant. Although no mineral matter transformations are considered
and the particle fragmentation as well as grain coalescence is over-
simplified, there is no doubt that this novel approach is attractive,
indeed.
2.4.2. Thermodynamic equilibrium calculations
ChemSage [74] is one of the available software packages for cal-
culations of equilibrium composition of multiphase and multi-
component systems. It can be used to estimate the chemical compo-
sition of minerals present in the coal/ash particle provided that the
particle temperature is given as well as composition of the gas
phase is known. The software minimises the Gibbs potential of any
mixture of minerals and provides the liquid phase fraction which
can be then used to determine the sticking probability. Liebetruth et
al. [60] have coupled the software with the CFD-Fluent code to calcu-
late the liquid phase fraction as well as the makeup of minerals along
particle trajectories and at impaction. The ChemSage calculations
were initiated after the volatiles were given off and proceeded until
the particle either left the computational domain or was trapped on
the wall. Obviously the main concern about this approach is the fact
that the fly ash particles, when analysed for their chemical composi-
tion, show a huge departure from equilibrium.
The thermal processes during transformation of minerals can be
divided in four groups [63]:
(a) decomposition of minerals into simple compounds,
(b) devolatilisation (sublimation) of single coal components and
formation of aerosols,
(c) formation of new minerals in the combustion zone,
(d) fusion and liquefaction of not combustible matter.
The most important thermal transformations of main minerals
found in coal are:
(a) clay mineral transformations
(b) quartz transformations,
(c) sulphide and sulphate transformations,
(d) carbonate transformations.
Although Eddings et al. [11] considered the transformations of ex-
traneous and included pyrite, the modellers developing CFD-based
slagging predictors typically stay away from considering chemical re-
actions involving minerals with a notable exemption of the TU Braun-
schweig group [77,78]. Božić [76] attempted to predict the fate of
minerals in a semi-industrial non-swirling flame (flame A1 of the
IFRF long flames trial [79]) by considering minerals listed in Table 2.
In his model, each coal particle contained a mineral-core whose
mass was calculated on the basis of proximate coal analysis. It was as-
sumed that the burning particle remained porous and the transport of
the gaseous components from the mineral-core to the bulk flow
through this porous structure was infinitely fast. Thus, combustion
of the porous combustible matter proceeded in parallel with the

2.4.3. Transformation of minerals (kinetics) Table 2

It has often been emphasised that ash deposition cannot be prop- Coal minerals considered in the calculations of Božić [77]; German Saar coal.
erly understood without considering coal minerals and their transfor-
Mineral Formula Mass fraction — %
mations [75,76,63]. As a product of these transformations ash species
Limonite Fe2O3H2O 28.03
are generated consisting of gases, liquids and solids and they react
Siderite FeCO3 8.99
together whilst the particle flies through the boiler, see Fig. 13. The Quartz/Cristobalite SiO2 8.54
properties of the produced intermediates depend on the association Kaolinite Al2Si2O5(OH)4 18.73
and chemical characteristics of the inorganic components, the physi- Gypsum CaSO4·2H2O 27.94
cal characteristics of the coal particle, the physical characteristics of Magnesite MgCO3 7.77
Total 100.00
the coal minerals, and on the combustion conditions.
 R. Weber et al. / Fuel Processing Technology 105 (2013) 113–128 125

transformations/reactions in the mineral-core. It was stated that
dehydration, thermal dehydroxilation and calcination of the minerals
were completed within 10–500 ms. Melting and transformation to
glass of Kaolinite required typically 10–100 ms. Release of SO2 from
sulphates was around an order of magnitude slower. Crystallisation
of Magnetite (Fe3O4) proceeded rapidly being completed within 20–
60 ms, however for other oxides it took 1 to 100 s. Wüstite (FeO)
was produced either through decomposition of Siderite (FeCO3) or
through the reduction of Magnetite (Fe3O4). The decomposition was
unaffected by the CO2 partial pressure of the gas phase whilst the re-
duction of Magnetite occurred in the 950–1420 °C temperature range
and the oxygen partial pressure ranges from 10 − 14 bar to 10 − 6 bar.
Magnetite (Fe3O4) was produced either by Wüstite oxidation or
Hematite (Fe2O3) reduction and both processes occurred when the par-
ticle travels through the flame. Hematite was produced in the thermal
dehydroxilation of Limonite (Fe2O3H2O) or through oxidation of
Magnetite. It was observed that complete transformation of Magnetite
into Hematite was impossible since the partial pressure of oxygen
was too low (not exceeding 0.21 bar) in the flame. Whilst the original
mix of minerals was as given in Table 2, the predicted slag contained
Lime (CaO), Periclase (MgO), Quartz (SiO2) and Cristobalite (high-
temperature polymorph of Quartz), Mullite (3 SiO2 2 Al2O3) and Alumi-
na (Al2O3). Notwithstanding the limitations of the model of Božić [77],
one cannot refrain from noticing the originality of the work.
In the work of Božić et al. [77,78] the kinetic data required for the
modelling of mineral matter transformation were determined in a
numerous thermogravimetry (TG) measurements. As shown in
Fig. 14 [77] some of these data have been generated under remark-
able high heating rates of 2000 K/s. There exists a substantial body
of literature on kinetics of mineral matter transformation reactions,
as exemplified by Ref. [80–82], however it is perhaps fair to stress
that such kinetic data originate almost exclusively from thermogravi-
metry or differential thermogravimetry (DTG) and they pertain to
heating rates not larger than 20–40 K/min. Thus, applicability of
such data for modelling of industrial boilers, where minerals are
Decomposition
Al2Si2H4O9 ---> Al2Si2O7 + 2H2O
1 800
Kaolinite
0,9 Tp 750
0,8 700
0,7 650
Y, [kg/kg]
heated up with rates up to 10 5 K/s, has to be verified. Recent research
[83,84] on mineral matter transformation in drop tube furnaces
operated under high heating rates, when combined with the most
advanced methods of coal/minerals characterization [85,71], can
establish the applicability of the TG-DTG kinetic data to modelling
of slagging in boilers.
2.5. Deposit growth, shedding and deformation
Once a particle adheres to the surface it contributes to deposit
growth. A number of deposit growth models have been developed
in the literature (see Ref. [86]), however most of the deposit growth
sub-models [5–11] implemented into CFD-codes just sum up the par-
ticles and record their properties. The mass and the shape of the de-
posit change with time and attempts have recently been made to
model the process. Tomeczek and Wacławiak [87] used the Fluent
CFD code to simulate the growth of a deposit on a platen superheater.
Their deposits grew due to condensation of potassium and sodium
salts and inertial impaction. For the latter, the algorithm described
in Section 2.2 was used. The adherence probability for solid particles
was expressed as a function of the condensation rate of alkali salts
[88] and was therefore particle size independent. Fig. 15 shows the
2D predictions [87] of the deposit shape as a function of time for a
bank of 38 mm O.D. pipes with longitudinal and transverse pitches
of 2 and 1.5, respectively. For 33 μm ash particles (particle density
around 2200 kg/m 3) possessing 8.36 m/s upstream velocity, the
Stokes number is around 1, so approximately 40% of such particles
(see Fig. 9) hit the front face of the first pipe and they either adhered
or were rebounded. Although particle rebounding plays a crucial role
in the simulations of Tomeczek and Wacławiak [87], no description of
the rebounding sub-model is given in the publication. Fig. 15 shows
that the model allows for deposition in the whole 90° sector of the
front pipe (see Fig. 9). Furthermore, it was argued that “the particles
are able to deposit also on the down-stream side of the first tubes,
which is the result of particle rebouncing from the neighbouring
rows of tubes” [87]. The authors concluded that the horn-like deposit
formed on the front side of the second tube was due to rebounding of
large ash particles on the neighbouring rows located above and below
the row shown in Fig. 15. For the considered configuration, shown in
Fig. 15, the transverse pitch was (1.5 × 38 mm=) 57 mm. When the
transverse pitch was increased to (5 × 38mm=) 190 mm the second
and the third tubes remained clean from the deposit. There is no

0,6 600 doubt that one has to acknowledge the originality of the analysed
Tp, [K]

0,5 550 publication [87]. However, care must be exercised in using

0,4 500 CFD-codes whilst seeking a mechanistic explanation to such a com-
0,3 450 plex problem since (a) the flows in banks of tubes are difficult to pre-
0,2 400 dict, (b) predicting correct impaction efficiencies is not trivial at all, as
H2O
0,1 350 pointed out in Section 2.2, and (c) an accurate model for inelastic
Meta kaolin
0 300 rebounding is needed in particular for surfaces covered with a layer
0,0155 0,016 0,0165 0,017 0,0175 0,018 of condensed liquid.
t [s]
A very interesting experiment concerning a deposition of calcium
carbonate (40 μm) particles on cold (ambient air temperature) bun-
Transformation: Meta kaolin (Al2Si2O7) ===> Glass
1
Tp
1700 dle of tubes was reported by Abd-Elhady et al. [89]. The bundle of
0,9
1500
nine (3 × 3) 32 mm O.D. tubes was used, with 3.15 longitudinal and
0,8 traverse pitches, and the particle laden air flow approached the
1300
0,7
Meta kaolin bank vertically downwards (along the gravity vector) or vertically
Y [kg/kg]

0,6 Glass: SiO2 - Content upwards (against the gravity). The particle velocity was 1.25 m/s cor-
Tp [K]

1100
0,5 responding to the Stokes number of around 0.6. A digital camera
0,4 900
Glass: Al2O3 - Content recorded the initial stages of fouling since the experiment lasted for
0,3 700 3 h only. It was observed that when the flow was upwards, the foul-
0,2
500 ing of the first tube occurred on its bottom part (front face), due to
0,1
the inertial impaction, and also on its top part (back face), due to
0 300
0 0,1 0,2 0,3 0,4 0,5 0,6 0,7
the flow recirculation. The intensity of fouling of the second row
t [s] tube was lower than that of the first row however the fouling pattern
was similar. When the flow was downwards, the fouling occurred on
Fig. 14. Transformation reactions for Kaolinite — TGA analysis [77]. the top part (front face) of the tubes only, with the fouling intensity
126
1 month
3 months
6 months 2 x 38 = 76 mm
9 months
57 mm
first tube
Fig. 15. Deposit shape on the first two tubes as a function of time for 33 μm fly ash par-
ticles (see Fig. 6 of [87]); upstream gas velocity of 8.36 m/s, 38 mm O.D. pipe; longitu-
dinal pitch of 2; transverse pitch of 1.5.
decreasing for the second and the third row tubes. No significant
effect of particle rebounding on the fouling mechanism was reported.
Thus, the experiment confirmed the well known [89] effect of the
flow direction on the fouling mechanism. Since the experiment of
Abd-Elhady et al. [89] was conducted at ambient air temperature
with a dry material (not molten) of known mechanical properties, it
→
was possible to estimate the critical particle velocity (Vc ) for rebound-
ing (see Eq. (17)) to be 0.02 m/s. Obviously, the calculated 0.02 m/s
value of the critical velocity [89] has to be regarded as an estimate
and substantial inaccuracy is plausible.
Losurdo [53] performed 3-dimensional Large Eddy Simulations of
the Abd-Elhady et al. [89] flows using Eq. (17) as the particle
rebounding criterion with the critical velocity of 0.02 m/s. His simula-
tions reproduced well the fouling mechanism observed in the exper-
iments and there were no indications of the importance of particle
rebounding. Realising that the simulations of Tomeczek and Wacła-
wiak [87] were performed for the Stokes number of around one,
which is not so far away from the 0.5 value applicable to the flows of
Abd-Elhady et al. [89], it is worth noting that the main difference be-
tween these two flow cases, is in the transverse pitch (1.5 against
3.15, respectively) and the particle rebounding sub-models. In our opin-
ion, resolving the issue of particle rebounding in building deposit brid-
ges is important and it should be continued.
Permanently changing shape and thickness of the deposit alter the
fluid dynamics near the deposition surface so that care must be taken
to make sure that particle impaction is properly handled, as shown in
Fig. 16. Typically adaptive meshing techniques are used to vary the
h=5mm h=10mm
1,0
Impaction Efficiency, [-]
0,8
0,6
Pipe: 40 mm
0,4
0,2
0,0
-20 0 20 40 60 80 100 120 140 160 180 200 220
Particle diameter, [µm]
Fig. 16. Alterations to the impaction efficiency due to the deposit shape change;
1200 °C flue gas temperature; 40 mm O.D. pipe; 12.8 m/s flue gas velocity.
R. Weber et al. / Fuel Processing Technology 105 (2013) 113–128
numerical grid along the deposit growth. Melted deposits may under-
go several deformation processes which call for a dynamic descrip-
symmetry plane tion of the deposit growth including melting and solidification
[45,90]. To account for such deformations, the heat conduction
through the deposit must be considered and this is typically done
by solving the heat diffusion equation using the concept of the effec-
tive thermal conductivity (see for example [91]).
The CFD-based modelling of deposit formation has not yet reached
symmetry plane the stage that chemical and physical processes, including sintering,
are accounted for on a regular basis. Sintering has been extensively
second tube investigated in ceramics although models describing sintering in boil-
er deposits are scarce [86]. Another important aspect is deposit clean-
ing [92] by soot blowing or any other techniques.
3. Closing remarks
There are not many publications where CFD-computations have
been used to predict slagging and fouling propensity in industrial
boilers. This is not too surprising, although the vision of possessing
CFD-based ash deposition models is in our opinion realisable, a signif-
icant body of the development work is needed to achieve this goal. In
short, the currently published papers just initiate the forthcoming
proliferation of scientific articles in this field.
Having analysed numerous issues involved in the development of
CFD-based models for slagging predictions, it has became apparent
that a certain degree of accuracy is needed in calculating the time–
temperature/chemistry history of coal/ash particles. To this end, the
information about high temperature devolatilization and char com-
bustion is needed and this is obtainable by performing appropriate
fuel characterization experiments for the fuel in question. Of particu-
lar importance is determination (through measurements) of the rate
of char oxidation for the last 20% of burnout. Fragmentation of fuel
particles is perhaps one of the unresolved issues which require an ur-
gent attention.
Almost every publication on slagging and fouling underlines the
primary importance of inertial impaction if compared to thermophor-
esis, eddy diffusion or vapour condensation. An accurate prediction of
impaction efficiency is a prerequisite to further analysis since a parti-
cle that does not impact does not form a deposit. We have demon-
strated that accurate predictions of the impaction efficiency using
CFD-codes are obtainable only when the flow field in the neighbour-
hood of the deposition surface is accurately resolved. In many of the
recent publications the issue is over-looked and the rate of the parti-
cles arrival to the deposition surface is substantially over-estimated.
D = 40 mm
h The above statement has got important implications for predicting
deposits of industrial boilers. The current generation of CFD-codes
h=30mm
can identify boiler areas prone to deposit formation by showing the
h=60mm
presence of ash particles in the neighbourhood of the areas. However,
computing the particle arrival rate must be done by resolving, to a
great detail, the local flow near the surface in question.
The analytic techniques based on Scanning Electron Microscopy
(SEM) quantify particle size, shape and composition of mineral mat-
ter. They also show the amount and nature of liberated minerals in
the coal; it means much more information is accessible than through
conventional coal ash analysis. This is a good opportunity to advance
0 mm Deposition further the models for mineral matter transformation and in particu-
5 mm Deposition
lar to validate them. Rates of reactions involved in transformation of
10 mm Deposition
minerals are nowadays obtained (mainly) from TGA at low heating
30 mm Deposition
60 mm Deposition
rates. Are these rates directly applicable to high heating rates? It is a
long way but is should be followed. Procedures for determining the
sticking propensity of particles either through melting curves or
through thermodynamic calculations are rather new and a thorough
validation is needed. Reactions and processes in the matter deposited
on the walls should be studied further. Sintering has been extensively
investigated in ceramics although models describing sintering in
 R. Weber et al. / Fuel Processing Technology 105 (2013) 113–128
boiler deposits are scarce. Another important aspect is deposit clean-
ing by soot blowing or any other techniques.
The current CFD-predictions of slagging and fouling in industrial
boilers are indicative, at their best. This is certainly the case when nu-
merous parameters, including these describing the particle sticking
propensity, are taken from the literature. It is relatively easy to tune
the model predictions to the expected (measured) results by chang-
ing a few key parameters. But there is a much more reliable way to
proceed. By combining CFD-predictions with an advanced fuel char-
acterization, with the latter being application dependent and fuel de-
pendent, the deposition problems at hands can be tackled. What is to
be determined experimentally and under what conditions, requires a
consensus of an experienced fuel engineer/mineralogist and a CFD-
expert.
References
[1] G. Couch, Understanding Slagging and Fouling during PF Combustion, IEA Coal
Research, London, 1994.
[2] P.J. Jackson, A.R. Jones, Fireside fouling experience in large CEGB boilers, Proceed-
ings of the 1981 Conference on Fouling Slagging Resulting from Impurities in
Combustion Gases, Henniker, New Hampshire, USA, 12–17 Jul. 1981, Engineering
Foundation, New York, NY, USA, 1983, pp. 1–15.
[3] T.F. Wall, L. Juniper, A. Lowe, State-of-the-Art Review of Ash Behaviour in Coal
Fired Furnaces, Tech. Rep. ACARP Project C9055, University of Newcastle, Newcastle,
2001.
[4] Photographs of the ash deposits originate from: 1,2: Doosan Babcock Energy Limited;
3,4: Clyde Bergemann GmbH; and 5,6: provided by Dipl. -Ing. Jürgen Labuschewski.
[5] S.K. Kær, L.A. Rosendahl, L.L. Baxter, Towards a CFD-based mechanistic deposit
formation model for straw-fired boilers, Fuel 85 (2006) 833–848.
[6] M. Forstner, G. Hofmeister, M. Jöller, J. Dahl, M. Braun, S. Kleditzsch, CFD simula-
tion of ash deposit formation in fixed bed biomass furnaces and boilers, Progress
in Computational Fluid Dynamics 6 (4/5) (2006) 248–261.
[7] G. Urbain, F. Cambier, M. Deletter, M.R. Anseau, Viscosity of silicate melts, Trans-
actions and Journal of the British Ceramic Society 80 (1981) 139–141.
[8] B. Epple, B. Krohmer, A. Hoppe, H. Mueller, R. Leithner, CRFD studies for boilers
fired with high ash containing and slagging lignites, Clean Air 6 (2005) 137–155.
[9] C. Mueller, B.-J. Skrifvars, R. Backman, M. Hupa, Ash deposition prediction in bio-
mass fired fluidised bed boilers — combination of CFD and advanced fuel analysis,
Progress in Computational Fluid Dynamics 3 (2–4) (2003).
[10] J.R. Fan, X.D. Zha, P. Sun, K.F. Cen, Simulation of ash deposit in a pulverized
coal-fired boiler, Fuel 80 (2001) 645–654.
[11] E.G. Eddings, K.A. Davis, M.P. Heap, J.R. Valentine, A.F. Sarofim, Mineral matter
transformation during pulverized coal combustion, Developments in Chemical
Engineering and Mineral Processing 9 (3/4) (2001) 313–327.
[12] U. Schell, S. Richter, K.R.G. Hein, Numerical simulation of slagging and fouling in a
pulverized coal-fired utility boiler, United Engineering Foundation Conference on
Heat Exchanger Fouling, Davos, Switzerland, 8–13 July, 2001.
[13] W. Bernstein, V. Hildebrand, T. Holfeld, Modellierung der Verbrennung und ihre
Validierung am Originalbraunkohledampferzeuger eines 800 MW Blokes, VDI
BERICHTE NR. 1492, 1999, pp. 45–55.
[14] T. Kupka, M. Mancini, M. Irmer, R. Weber, Investigation of ash deposit formation
during co-firing of coal with sewage sludge, saw-dust and refuse derived fuel,
Fuel 87 (2007) 2824–2837.
[15] T. Kupka, K. Zajac, R. Weber, Effect of fuel type and deposition surface tempera-
ture on the growth and structure of an ash deposit collected during co-firing of
coal with sewage sludge and sawdust, Energy & Fuels 23 (2009) 3429–3436.
[16] T. Kupka, Ash deposition problems during co-firing of alternative fuels in
coal-fired power plants. Ph.D. Thesis. Clausthal University of Technology, Faculty
of Energy and Management, December, 2009.
[17] H. Stadler, D. Toporov, M. Förster, R. Kneer, On the influence of the char gasifica-
tion reactions on NO formation in flameless coal combustion, Combustion and
Flame 156 (9) (2009) 1755–1763.
[18] J.P. Kim, U. Schell, G. Scheffknecht, Numerical modeling of mild combustion of
coal, Progress in Computational Fluid Dynamics 7 (2007) 337–346.
[19] B.F. Magnussen, B.H. Hjertager, On mathematical modeling of turbulent combus-
tion with emphasis on soot formation and combustion, Proceedings of the Com-
bustion Institute 16 (1976) 719–729.
[20] B.F. Magnussen, The Eddy Dissipation Concept, Task Leaders Meeting, Lund, 1989.
[21] D. Förtsch, F. Kluger, U. Schnell, H. Spliethoff, K.R.G. Hein, A kinetic model for the
prediction of NO emissions from staged combustion of pulverized coal, Proceed-
ings of the Combustion Institute 27 (1998) 3037–3044.
[22] Ch. Mueller, Numerische Simulation industrieller Verbrennungsprozesse unter
Verwendung von reaktionskinetischen Ansätzen und Wirbelzerfallsmodellen.
Ph.D. Thesis, Faculty of Mechanical Engineering, The Ruhr-University of Bochum,
1998.
[23] M. Mancini, Analysis of Mild Combustion of Natural Gas with Preheated Air, Faculty
of Energy and Economics, Clausthal University of Technology, Faculty of Energy and
Management, 2007.
[24] K. Zajac, T. Kupka, S. Brinker, R. Weber, Advanced solid fuel characterisation, in: G.
Tsatsaronis, A. Boyano (Eds.), Proceedings of the International Conference on
127
Optimization using Exergy-Based Methods and Computational Fluid Dynamics,
Berlin, Germany , October 20–23, 2009, Published in the Series “International
Studies in Science and Engineering”, Papierflieger Verlag, Clausthal-Zellerfeld,
2009, pp. 115–122.
[25] F. Boysan, W.H. Ayers, J. Swithenbank, A fundamental mathematical modelling
approach to cyclone design, Trans. I. Chem. E., 60, 1982, pp. 222–230.
[26] R. Weber, F. Boysan, W.H. Ayers, J. Swithenbank, Simulation of dispersion of heavy
particles in confined turbulent flows, AICHE Journal 30 (3) (1984) 490–492.
[27] F. Boysan, R. Weber, J. Swithenbank, C.J. Lawn, Modeling coal-fired cyclone com-
bustors, Combustion and Flame 63 (1986) 73–86.
[28] Fluent: Fluent 6.0 User's Guide, Fluent, Lebanon, USA, 2001.
[29] S. Yilmaz, K.R. Cliffe, Simulation of coal ash deposition on to superheater tube,
Journal of the Institute of Energy 70 (March 1997) 17–23.
[30] S.A. Morsi, J.A. Alexander, An investigation of particle trajectories in two-phase
flow systems, The Journal of Fluid Mechanics 55 (1972) 193–208.
[31] M.A. Field, D.W. Gill, B.B. Morgan, P.G.W. Hawksley, Combustion of Pulverized
Coal, The British Coal Utilization Research Association, Leatherhead, 1967.
[32] I.W. Smith, The combustion rate of coal chars: a review, Proceedings of the Com-
bustion Institute 19 (1982) 1045–1065.
[33] T.F. Wall, The combustion of coal as pulverized fuel through swirl burners, in: C.J.
Lawn (Ed.), Chapter 3 in Principles of Combustion Engineering for Boilers, Academic
Press, 1987.
[34] N.M. Laurendeau, Heterogenous kinetics of coal char gasification and combustion,
Progress in Energy and Combustion Science 4 (1978) 221–270.
[35] R. Hurt, J.K. Sun, M. Lunden, A kinetic model of carbon burnout in pulverized coal
combustion, Combustion and Flame 113 (1998) 181–197.
[36] J. Haas, Experimentelle Untersuchungen der Reaktivität von Kohlenkoks zur
Modellierung seines Verbrennungsverhaltens in Staubfeuerungen. Ph.D. Thesis,
Faculty of Mining, Iron- and Steel-Making and Mechanical Engineering of the
Clausthal University of Technology, 1999.
[37] L.L. Baxter, Char fragmentation and fly ash formation during pulverized-coal
combustion, Combustion and Flame 90 (1992) 174–184.
[38] N. Syred, K. Kurniawan, T. Griffiths, T. Gralton, R. Ray, Development of fragmenta-
tion models for solid fuel combustion and gasification as subroutines for inclusion
in CFD codes, Fuel 86 (2007) 2221–2231.
[39] H. Schlichting, K. Gersten, Boundary Layer Theory, 8th edition Springer, 2000.
[40] R. Israel, D.E. Rosner, Use of generalized Stokes number to determine the aerody-
namic capture efficiency of non-Stokesian particles from a compressible gas flow,
Aerosol Science and Technology 1 (1983) 45–51.
[41] R.A. Wessel, J. Righi, Generalized correlations for inertial impaction of particles on
a circular cylinder, Aerosol Science and Technology 9 (1988) 29–60.
[42] R. Weber, et al. (in preparation), Ash deposition modelling: requirements for accu-
rate predictions of particle impaction using computational fluid dynamics.
[43] J. Pyykönen, J. Jokiniemi, T. Jakobsen, Development of a prediction scheme for
pulverized coal-fired boiler slagging, in: R.P. Gupta, T.F. Wall, L.L. Baxter (Eds.),
Impact of Mineral Impurities in Solid Fuel Combustion, Proc. Eng. Found. Conf.,
Kona, Hawaii, USA, Nov.2–7, 1997, ISBN: 978-0306461262, 1999.
[44] L.Y. Huang, J.S. Norman, M. Pourkashanian, A. Williams, Prediction of ash deposi-
tion on superheater tubes from pulverized coal combustion, Fuel 75 (3) (1996)
271–279.
[45] H. Zhou, P.A. Jensen, F.J. Frandsen, Dynamic mechanistic model of superheater
deposit growth and shedding in a biomass fired grate boiler, Fuel 86 (2007)
1519–1533.
[46] S.S. Lokare, J.D. Dunaway, D. Moulton, D. Rogers, D.R. Tree, L.L. Baxter, Investiga-
tion of ash deposition rates for a suite of biomass fuels and fuel blends, Energy
& Fuels 20 (2006) 1008–1014.
[47] S.A. Benson, M.L. Jones, J.N. Harb, Ash formation and deposition, in: L.D. Smoot (Ed.),
Fundamentals of Coal Combustion for Clean and Efficient Use, Elsevier, Amsterdam,
1993.
[48] L. Talbot, R.K. Cheng, R.W. Schefer, D.R. Willis, Thermophoresis of particles in
heated boundary layer, Journal of Fluid Mechanics 101 (Dec. 29 1980) 737–758.
[49] C. Thornton, K.K. Yin, Impact of elastic spheres with and without adhesion, Powder
Technology 65 (1991) 153–166.
[50] C. Thornton, Z. Ning, A theoretical model for the stick/bounce of adhesive elastic–
plastic spheres, Powder Technology 99 (1998) 154–162.
[51] T. Mao, D.C.S. Kuhn, H. Tran, Spread and rebound of liquid droplets upon impact
on flat surfaces, AICHE Journal 43 (9) (1997) 2169–2179.
[52] C. Mueller, M. Selenius, M. Theis, B.-J. Skrifvars, R. Backman, M. Hupa, H. Tran, De-
position behaviour of molten alkali-rich fly ashes — development of a submodel
for CFD applications, Proceedings of the Combustion Institute 30 (2005)
2991–2998.
[53] M. Losurdo, Particle tracking and deposition from CFD simulations using a visco-
elastic particle model. Ph.D. Thesis. Delft Technical University, The Netherlands,
2009.
[54] M. Losurdo, C. Bertrand, H. Spliethoff, A Lagrangian particle CFD post-processor
dedicated to particle adhesion/deposition. ECI Symposium Series, Volume RP5,
in: H. Müller-Steinhagen, M. Reza Malayeri, A.P. Watkinson (Eds.), Proceedings
of 7th International Conference on Heat Exchanger Fouling and Cleaning —
Challenges and Opportunities, Engineering Conference International, Tomar,
Portugal, July 1–6, 2007.
[55] A. Rushdi, R. Gupta, A. Sharma, D. Holcombe, Mechanistic prediction of ash depo-
sition in a pilot-scale test facility, Fuel 84 (2005) 1246–1258.
[56] P.M. Walsh, A.N. Sayre, D.O. Loehden, L.S. Monroe, J.M. Beer, A.F. Sarofim, Deposi-
tion of bituminous coal ash on an insulated heat exchanger tube: effects of coal
properties on depth growth, Progress in Energy and Combustion Science 16
(1990) 327–346.
128 R. Weber et al. / Fuel Processing Technology 105 (2013) 113–128
[57] S. Srinivasachar, C.L. Senior, J.J. Helble, J.W. Moore, A fundamental approach to the
prediction of coal ash deposit formation in combustion systems, Proceedings of
the Combustion Institute 24 (1992) 1179–1187.
[58] S. Yilmaz, K.R. Cliffe, Simulation of coal ash deposition on to a superheater tube,
Journal of the Institute of Energy 70 (March 1997) 17–23.
[59] P.G. Costen, F.C. Lockwood, M.M. Siddique, Mathematical modelling of ash depo-
sition in pulverized fuel-fired combustors, Proceedings of the Combustion Insti-
tute 28 (2000) 2243–2250.
[60] Liebetruth M., Görner K., Hildebrandt V. Laufzeitkopplung vom CFD und thermo-
dynamischer Realphasenmodellierung zur Darstellung der Verschlackungsneigung.
VDI-Berichte, Nr. 1664, 2002, pp. 41–52.
[61] C. Senior, S. Srinivasachar, Viscosity of ash particles in combustion systems for
prediction of particle sticking, Energy & Fuels 9 (1995) 277–283.
[62] S. Vergas, F.J. Frandsen, K. Dam-Johansen, Rheological properties of high-
temperature melts of coal ashes and other silicates, Progress in Energy and Com-
bustion Science 27 (2001) 237–429.
[63] R.W. Bryers, Fireside slagging, fouling and high-temperature corrosion of
heat-transfer surface due to impurities in steam-raising fuels, Progress in Energy
and Combustion Science 22 (1996) 29–120.
[64] D.W. Reeve, H.N. Tran, D. Barham, The effluent-free bleached kraft pulp mill —
part XI, Pulp & Paper Canada 82 (9) (September 1981) 1–7 T315–T320.
[65] P. Isaak, H.N. Tran, D. Barham, D.W. Reeve, Stickiness of fireside deposits in kraft
recovery units, Journal of Pulp and Paper Science: 12 (3) (May 1986) J84–J88.
[66] R. Shenassa, H. Tran, D.C.S. Kuhn, Dynamic study of carryover deposition in kraft
recovery boilers using an entrained flow reactor, Pulp & Paper Canada 100 (10)
(1999) 56–62.
[67] S. Akbar, U. Schnell, G. Scheffknecht, Modelling of potassium release and the
effect of potassium chloride on deposition mechanisms for coal and biomass-
fired boilers, Combustion Theory and Modelling 14 (3) (2010) 315–329.
[68] M. Braun, S. Kleditzsch, R. Scharler, I. Obernberger, M. Forstner, M. Jöller, G.
Hofmeister, A method for reduction of computational time of local equilibria
for biomass flue gas compositions in CFD, Progress in Computational Fluid
Dynamics 6 (4/5) (2006).
[69] S.B. Pope, Computationally efficient implementation of combustion chemistry using
in situ adaptive tabulation, Combustion Theory and Modeling 1 (1997) 41–63.
[70] A.F. Sarofim, J.J. Helble, Mechanisms of ash and deposit formation, in: J. Williamson,
F. Wigley (Eds.), The Impact of Ash Deposition on Coal-Fired Power Plants, Proc.
United Engineering Foundation International Conf. Solihull, Birmingham, UK, June
20–25, 1993, 1994, ISBN-1-56032-393-4.
[71] F.J. Frandsen, Ash research from Palm Coast, Florida to Banff, Canada: entry of
biomass in modern power boilers, Energy & Fuels 23 (2009) 3347–3378.
[72] M. Hecken, L. Reichelt, U. Renz, Numerical Simulation of slagging films in the
pressurized coal combustion facility Aachen, The 4th International Symposium
on Coal Combustion, Peking, August 1999.
[73] B. Kreutzkam, C. Wieland, H. Spliethoff, A novel code predicting ash formation
during pulverised coal combustion to improve boiler deposition modelling,
Impact of Fuel Quality on Power Production & Environment, August 29th–
September 3rd, 2010, Lapland, Finland, 2010.
[74] G. Eriksson, K. Hack, Chemsage — a computer program for the calculation of
complex chemical equilibria, Metalurgical Transactions 21B (1990) 1013–1023.
[75] E. Raask, Mineral Impurities in Coal Combustion: Behaviour, Problems, and Reme-
dial Measures, New York, USA, Hemisphere Publication Corporation, 1985.
[76] A.A. Benson, E.A. Sondreal, Ash-related issues during combustion and gasification,
in: R.P. Gupta, T.F. Wall, L. Baxter (Eds.), Impact of Mineral Impurities in Solid Fuel
Combustion, Kluwer Academic Publishers, New York, 1999, pp. 1–21.
[77] Božić O. Numerische Simulation der Mineralumwandlung in Kohlenstaubfeuerun-
gen. Ph.D. Thesis. Faculty of Mechanical and Electrical Engineering, TU Braunschweig,
Germany, 2003.
[78] R. Leithner, O. Božić, A. Hoppe, M. Neuroth, Einfluss vom Verschlakung und
Verschmutzung auf Kohleauswahl, Dampferzeugerauslegung und Betrieb; Erfahrung
und Simulation, Verbrennung und Feuerungen. 21 Deutscher Flammentag. VDI-
Berichte 1750, 2003, pp. 11–22.
[79] J.B. Michel, R. Payne, Detailed measurements of long pulverized coal flame for the
characterisation of pollutant emission, IFRF Doc. No. F 09/a/23, 1981.
[80] V. Bialek, M. Murat, The emanation thermal analysis of kaolinite clay minerals,
Thermochimica Acta No. 282/283, Elsevier Science Publishers B.V, Amsterdam,
1996.
[81] J.V. O'Gorman, P.L. Walker, Thermal behaviour of mineral fractions separated
from selected American coals, Fuel 52 (1973) 71–79.
[82] J. Tomeczek, H. Palugniok, Kinetics of mineral matter transformation during coal
combustion, Fuel 81 (2002) 1251–1258.
[83] L. Yan, R. Gupta, T. Wall, Fragmentation behaviour of pyrite and calcite during
high-temperature processing and mathematical simulation, Energy & Fuels 15
(2001) 389–394.
[84] Y. Liu, R. Gupta, T. Wall, Ash formation from excluded minerals including consid-
eration of mineral–mineral associations, Energy & Fuels 21 (2007) 461–467.
[85] R. Gupta, Advanced coal characterization: a review, Energy & Fuels 21 (2007)
451–460.
[86] R.H. Rezaei, R.P. Gupta, T.F. Wall, S. Miyamae, K. Makino, Modelling the initial
structure of ash deposits and structure changes due to sintering, in: R.P. Gupta,
T.F. Wall, L.L. Baxter (Eds.), Impact of Mineral Impurities in Solid Fuel Combus-
tion, Engineering Foundation Conf., Kona, Hawaii, USA, Nov. 2–7, 1997, ISBN-
978-0306461262.
[87] J. Tomeczek, K. Wacławiak, Two-dimensional modelling of deposits formation on
platen superheaters in pulverized coal boilers, Fuel 88 (2009) 1466–1471.
[88] J. Tomeczek, H. Palugniok, J. Ochman, Modelling of deposits formation on heating
tubes in pulverised coal boilers, Fuel 83 (2004) 213–221.
[89] M.S. Abd-Elhady, C.C.M. Rindt, A.A. van Steenhoven, Optimization of flow direction
to minimize particulate fouling of heat exchangers, in: H. Müller-Steinhagen, M.
Reza Malayeri, A.P. Watkinson (Eds.), ECI Symposium Series, Volume RP5: Proceed-
ings of 7th International Conference on Heat Exchanger Fouling and Cleaning —
Challenges and Opportunities, Engineering Conference International, Tomar,
Portugal, July 1–6, 2007.
[90] S. Kalisz, M. Pronobis, Investigations on fouling rate in convective bundles of
coal-fired boilers in relation to optimization of sootblower operation, Fuel 84
(2005) 927–937.
[91] A. Zbogar, F.J. Frandsen, P.A. Jensen, P. Glarborg, Heat transfer in ash deposits: a
modelling tool-box, Progress in Energy and Combustion Science 31 (2005)
371–421.
[92] F. Bartels, S. Simon, M. Frach, C. Mueller, Intelligent on-load cleaning technology
to increase power boiler performance, in: N.S. Harding, T.F. Wall, F. Wigley, F.
Frandsen, M. Hupa, D. Tillman (Eds.), Impact of Fuel Quality on Power Production,
Snowbird, UT, 2006, Oct. 29 to Nov.3, 2006.