PERGAMON Progress in Energy and Combustion Science 25 (1999) 117–132

Modelling ash deposition in pulverized coal-fired applications

F.C.C. Lee a,1, F.C. Lockwood b,*
a
Computational Dynamics Ltd, 317 Latimer Road, Olympic House, London W10 6RA, UK
b
Imperial College of Science, Technology and Medicine, Department of Mechanical Engineering, Exhibition Road, London SW7 2AZ, UK

Abstract
A predictive scheme based on CCSEM flyash data and computational fluid dynamics (CFD) was developed to study the
slagging propensity of coals. The model was applied to predict the deposition potential of three UK coals; Bentinck, Daw Mill
and Silverdale, in a pilot scale single burner ash deposition test facility. The project is part of a collaborative research
programme sponsored by the UK Department of Trade and Industry and involved various industrial organizations and univer-
sities. The objective is to understand the fundamental aspects of slagging in pulverized coal-fired combustion systems. The
present model predicts the relative slagging propensity of the three coals successfully, and the predicted deposition patterns are
also consistent with the observations. The results from the model indicate a preferential deposition of iron during the initial
stage of ash deposition. The average compositions of the deposits become closer to that of the bulk ash when the accumulation
of ash deposits is taken into account. 䉷 1999 Published by Elsevier Science Ltd. All rights reserved.

Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
2. Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
2.1. Flow dynamics and combustion modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
2.2. Particle trajectories and flyash properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
2.3. Arrival rate of flyash particles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
2.4. Sticking efficiency of flyash particles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
2.5. Deposits accumulation model and effects on furnace performance . . . . . . . . . . . . . . . . . . . . . . 124
3. Validation case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
3.1. Results and discussions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
4. Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131

1. Introduction Slagging not only reduces the thermal efficiency of a

furnace, but also affects its integrity as a result of corrosion,
The accumulation of furnace deposits is one of the many erosion or impact on the bottom tubes. Slagging when
difficulties faced by utility operators all over the world. severe can lead to substantial financial losses to an operator.
This may be through loss of availability, increased main-
tenance and operating cost. At its worst, deposits can lead to
* Corresponding author. ash bridges or slag flags requiring generator units to be
1
F.C.C. Lee was employed by National Power PLC, Windmill dropped out of service for cleaning or repair. Although
Hill Business Park, Whitehall Way, Swindon SN5 6PB, UK, at the methods (e.g. high pressure jets of air, steam or water)
time of conducting this study. were developed to remove ash deposits from furnace
0360-1285/99/$ - see front matter 䉷 1999 Published by Elsevier Science Ltd. All rights reserved.
PII: S0360-128 5(98)00008-2
118 F.C.C. Lee, F.C. Lockwood / Progress in Energy and Combustion Science 25 (1999) 117–132

Fig. 1. Overview of the presenting modelling problems.

walls, slagging can still occur at locations which the clean- operating conditions, also play an important role in the
ing devices cannot reach, or when the bonding strength deposition process, a ‘‘chemistry only’’ approach is gener-
between the walls and the deposits is too strong for the ally not sufficient to assess such a complex phenomenon.
cleaning equipment to be effective. Recent advances in computer control scanning electron
The slagging propensity of coal is usually measured by microscopy (CCSEM) and other analytical techniques
indices derived from chemical properties of coal and/or its have provided invaluable information, allowing more
ashes. Since external factors, such as furnace design and sophisticated models to be developed. During the last few
 F.C.C. Lee, F.C. Lockwood / Progress in Energy and Combustion Science 25 (1999) 117–132 119

years, prediction schemes [1, 2] were constructed which a number of mono-sized particles having the same mass,
make use of CCSEM flyash data together with some simple momentum and temperature. Dispersion of pc particles by
flow models to determine the deposition rate of flyash par- the action of fluid turbulence is not currently accounted for,
ticles. However, owing to the lack of generality of the flow but will be incorporated in the next development stage using
models employed by these schemes, the external factors that existing theory. The mean pc trajectories and temperature–
control slagging, such as burner arrangement or particle– time histories are determined by solving the corresponding
turbulence interaction, are usually over-looked. Prediction equations of motion and energy balance for the representa-
schemes which emphasize flyash chemistry or particles’ tive particle from each group. Coal pyrolysis is modelled by
dynamics are available, [3, 4] but they fail to integrate all a single reaction rate model [8]. Heterogeneous char
the relevant factors to provide a complete description of coal combustion is approximated by a first-order differential
ash deposition. equation, with a combined rate coefficient that accounts
This article presents an alternative approach to model ash for the global diffusion of oxygen to the char surface and
deposition in pulverized coal fired systems. It accounts for the chemical reaction rate of char oxidation [9]. Char par-
the effects of firing rate, burner arrangement, furnace ticles are considered to burn at constant size, producing
geometry, wall conditions, as well as coal ash properties. carbon monoxide as the main initial product of surface
The model can also be used iteratively with the main furnace oxidation. The CO thus formed is further assumed to be
calculation to simulate the transient nature of ash deposition completely oxidized to carbon dioxide inside the particle
and its effect on furnace performance. The model was boundary layer. The heat liberated in the formation of CO
applied to a single burner pilot scale ash deposition rig is transferred to the particle to raise its temperature, and the
and the results are discussed in Section 3. heat generated by the subsequent oxidation to CO2 is
allowed to enter into the bulk gas [10].
Two combustion schemes were implemented to model
2. Methodology the global combustion rate of volatiles and char in the
CINAR code. The first model involves the solution of a
Since the pulverized fuel ash (pfa) phase is relatively single mixing parameter which defines the concentration
dilute in furnace applications, it is acceptable to ignore its of volatiles and a fictitious carbon source derived from the
contribution to momentum and energy to the gaseous phase. heterogeneous char oxidation [5]. The oxygen and fuels
Although this simplification will slightly affect the accuracy concentration are re-calculated according to a fast chemistry
of the gaseous phase solutions, it will drastically reduce the reaction scheme. The results are then fed into the next
run time of the deposition calculations. This decoupling, iterative loop until convergence is achieved.
however, is only limited to the flyash trajectory simulation. The second model involves the solution of two mixing
The pulverized coal (pc) particles are fully coupled to the parameters representing the volatiles from coal pyrolysis
flow field and temperature calculations. The resulting and the carbon atom in the carbon dioxide consequent on
conservation equations are solved by a computational fluid char oxidation, respectively (N. Kandamby, private com-
dynamics (CFD) method. Further, the effect of wall deposits munication). As in the first scheme, the oxygen and fuels
on furnace performance can still be accounted for by concentrations are re-calculated according to a fast
running the ash-deposition post-processor and the main chemistry reaction model.
CFD calculations in an iteratively manner. Fig. 1 shows The presence of pf particles is made known to the gaseous
the flow-chart of the procedures employed in the present phase by the particle source in cell (PSIC) method [11]. This
study. method couples the two phases by adding source terms, i.e.
mass, momentum and energy, from the pardculate phase to
2.1. Flow dynamics and combustion modelling the corresponding Eulerian gas phase equations [10, 12]. An
iterative procedure is used to obtained the final converged
The CFD combustion code, CINAR [5], developed by solution of the two phases.
CINAR Ltd is used to predict the flow field, temperature
profile and other field variables. The code models the 2.2. Particle trajectories and flyash properties
gaseous phase by solving the corresponding Eulerian
conservation equations of mass, momentum and energy on To calculate the trajectories of the flyash (pfa) particles,
a structured three-dimensional (3D) staggered finite volume the equation of motion for a single particle in a viscous
cartesian grid. Turbulence is modelled by using a Boussi- environment is solved.
nesq effective viscosity approach in partnership with a two- dUp 3 V
equation k–1 turbulence model [6] for closure. Radiation is rV ˆ⫺ r C U ⫺ Uf †兩Up ⫺ Uf 兩 ⫹ V r ⫺ rf †g;
dt 4 d f D p
modelled by the discrete transfer method developed by
1†
Lockwood and Shah [7].
The size distribution of pulverized coal (pc) is divided where r, article density; rf, density of the carrier phase; V,
into a finite number of size groups. Each group represents particle volume,Up, instantaneous velocity vector of a
120 F.C.C. Lee, F.C. Lockwood / Progress in Energy and Combustion Science 25 (1999) 117–132

Fig. 2. Graphical representation of particle/turbulent eddy interaction.

particle, Uf, instantaneous velocity vector of the carrier of the courant timescale, tc; the eddy lifefime, te; and the
phase; d, particle diameter; g, gravitational acceleration; particle transit time, tr. RL(6) is considered to be negligible
and CD, particle drag coefficient. when 6 is greater than the minimum of te and tr. In this case,
The Basset term, the added mass term and the temporal a different correlation domain is computed by using a locally
derivatives of the fluid fluctuating velocities which appear in homogeneous and isotropic turbulence assumption as
the full equation [13] are omitted. The removal of these described by Gosman and Ioannides [16]. On the other
terms was justified previously by other researchers [14], hand, when 6 is smaller than the minimum of te and tr, the
and the results from a similar study carried out by one of contribution from the temporal correlation term u(t)u(t ⫹ 6)
the authors [15]. The solution of Eq. (1) requires the instan- is considerable, and the correlation is incorporated as
taneous velocity of the carrier phase. This may be decom- follow,
posed into a mean, Um, and a fluctuating component, u. The
ui t ⫹ 6† ˆ wi 6†ui t† ⫹ ai ; 3†
mean velocity field is obtained from CFD calculations,
whereas the fluctuating velocity is derived from a modified where wi(6) is a coefficient to be determined, and ai is a
stochastic scheme as described later. function characterizing the randomness at the corresponding
The turbulent flow field is represented by a series of char- fluid location. The derivations of wi and ai and the final
acteristic eddies as shown in Fig. 2. As the particle traverses forms of Eq. (3) are described elsewhere [15].
the flow field, it interacts with the eddies successively. Two The continuous flyash size distribution is divided into a
factors were identified to characterize these interactions: (1) number of groups according to their relaxation times. The
the instantaneous velocity of an eddy; and (2) the interaction aerodynamic properties of each group are represented by the
time. group’s mass averaged diameter and density. A number of
According to the defmition of the Lagrangian auto-corre- flight paths for each representative particle from each group
lation function, RL(6), of a fluid particle, are simulated by solving Eq. (1). For example, if the pfa size
distribution is divided into eight groups and the representa-
u t†u t ⫹ 6† tive particle from each group is tracked for a thousand times,
RL 6† ˆ ; 2†
u t†2 the total number of particle tracks from each injection point
will be equal to 8000. Consequently, a total of 32 000 par-
the correlation becomes zero only when the time lag, 6, ticle tracks from each burner will be simulated when four
tends to infinity. As a result, in order to determine the instan- injection points per burner are defined.
taneous velocities of the fluid phase, the temporal correla-
tion of the fluctuating velocities must also be taken into 2.3. Arrival rate of flyash particles
account.
In the present model, the integration time interval, 6, is The net rate of deposition depends on both the rate of
calculated dynamically at each iteration interval as the lesser transport of the particles to a surface and their propensity
 F.C.C. Lee, F.C. Lockwood / Progress in Energy and Combustion Science 25 (1999) 117–132 121

Table 1 to penetrate the wall boundary layer completely. As a result,

Chemical classification scheme the arrival rate increases with particle size under this type of
flow condition.
Chemical groups Definition
Since the instantaneous velocity of a particle is the true
(1) Alumino-silicates Al2O3 ⫹ SiO2 ⬎ 80 wt.% measure of it’s motion in all the flow regitnes mentioned
and SiO2 ⬍ 80 wt.% earlier, the particle trajectory model discussed in the
(2) Iron–calcium CaO ⬎ 20 wt.% and previous section can, therefore, be extended to unify all
Fe2O3 ⬎ 20 wt.% the transport mechanisms up to the edge of a boundary
(3) Calcium-alumino-silicates Al2O3 ⫹ SiO2 ⬍ 80 wt.% layer. It remains then to model the penetration efficiency
and CaO ⬍ 80 wt.% and of particles through this layer for the determination of the
Fe2O3 ⬍ 20 wt.% and particles arrival rate.
CaO ⬎ Fe2O3
The arrival rate of flyash particles are characterized by a
(4) Calcium rich CaO ⬎ 80 wt.%
(5) Iron-alumino-silicates Al2O3 ⫹ SiO2 ⬍ 80 wt.%
‘‘critical velocity’’ and a predefined displacement thickness,
and CaO ⬍ 20 wt.% and y⫹. The ‘‘critical velocity’’ is defused as the minimum
Fe2O3 ⬍ 80 wt.% and initial velocity required by a particle to traverse a distance
Fe2O3 ⬎ CaO y⫹ in a stagnant and viscous environment within a time
(6) Iron rich Fe2O3 ⬎ 80 wt.% period corresponding to the particle’s relaxation time. In
(7) Non-CFAS (CaO ⫹ Fe2O3 ⫹ Al2O3 ⫹ the present study, the representative flyash particles are
SiO2) ⬍ 80 wt.% tracked until they reach the edge of the viscous sublayer
(8) Silica rich SiO2 ⬎ 80 wt.% extending to a value of y⫹ ˆ 5 [15, 19]. The particles are
then divided into eight chemical classes according to their
composition shown in Table 1. The mass average diameter
to stick once they reach a surface. This section describes the and density from each chemical class are then used to deter-
transport rate of flyash particles to a substrate surface. mine the penetration depth of the particles through the
The mechanism governing the transport of particles onto sublayer. Particles that possess the required ‘‘critical
a surface depends on their sizes. For very fine particles velocity’’ at the edge of the sublayer are assumed to be
(⬍ 1 mm), molecular diffusion and thermophoresis are able to traverse the sublayer and arrive on the surface.
dominant. For larger particles, the inertia of the particles Those that do not have the required minimum velocity are
normal to a surface becomes more significant, and controls considered to be trapped and carried by the boundary layer
their arrival rate on the substrate surface. In regimes where to the outlet.
the mean flow is parallel to a surface, no deposition will
occur if the particles were to follow the mean flow stream-
lines perfectly. In reality, deposition does occur. Previous 2.4. Sticking efficiency of flyash particles
studies suggested that this is caused by turbulence in the free
As mentioned in the previous section, the net rate of
stream, which cause particles to deviate from the mean flow
deposition depends on both the arrival rate of particles and
streamlines.
their ability to stick to a surface. Amongst all the possible
In general, particles obtain energies from the turbulent
qualities, particle viscosity seems to be the most appropriate
eddies when they are caught up by them. If the energies
and quantifiable [2, 20]. Since the objective of the present
given to the particles are not fully dissipated in the boundary
study is to design a scheme which can be apply to a wide
layer, they will traverse the layer and reach the substrate
range of coals, the general Urbain viscosity model [21] is
surface. This arrival rate mechanism was first observed by
chosen to determine the viscosity of flyash particles.
Friedlander and Johnstone [17] and is known as the eddy
Detailed viscosity models for non-silicate systems (e.g.
diffusion–impaction regime. Since the particles cannot
iron-rich pyrite particles) can be incorporated at a later
traverse the boundary layer faster than the turbulent eddies,
stage if necessary.
the maximum deposition velocity of particles via this
The inputs to the Urbain viscosity model are the tempera-
mechanism is therefore roughly equal to the magnitude of
ture and chemical compositions of flyash particles. The
the eddy velocity. When the particle size is bigger than the
flyash chemistry is obtained from CCSEM analysis. The
characteristic size of the eddies, the influence of the eddies
energy balance Eq. (4) of a flyash particle is solved along
on the particle’s motion diminishes. When this happens, the
its trajectory to obtain the corresponding temperature at the
particle will follow the mean flow streamlines again,
point of impact.
causing a decline in arrival rate [18].
In impinging flows, the particles are propelled by the d mp Cp Tp †
mean flow towards the substrate surface. The particles’ ˆ mp Qc ⫹ Qr †; 4†
dt
momentum is, therefore, the driving force for them to devi-
ate from the mean flow streamlines. Particles will reach the where mp, mass of a flyash particle; Cp, particle specific
substrate surface when their inertia are high enough for them heat; Tp, particle temperature; Qc, rate of heat gain or loss
122 F.C.C. Lee, F.C. Lockwood / Progress in Energy and Combustion Science 25 (1999) 117–132

Fig. 3. CRE ash deposition rig.

 F.C.C. Lee, F.C. Lockwood / Progress in Energy and Combustion Science 25 (1999) 117–132 123

Fig. 4. CRE burner arrangement.

by the particle per unit mass by convection; and Qr, rate of detetmined by the ratio between the critical value and their
heat gain or loss by the particle per unit mass by radiation. actual viscosities. The overall scheme may, therefore, be
To assess the sticking probability of flyash particles, a expressed as follow,
‘‘critical viscosity’’ of 10 5 Pa s is chosen [1]. The sticking
probability of particles having viscosities less than this criti- Viscosity ⬍ critical value; Sticking probability ˆ 1
cal value is considered to be unity. The sticking probability :
of particles with viscosities greater than the ciitical values is Viscosity ⬎ critical value; Sticking probability ˆ mcrit =mactual
124 F.C.C. Lee, F.C. Lockwood / Progress in Energy and Combustion Science 25 (1999) 117–132

Table 2 by simultaneously comparing the surface viscosity of the

Operating conditions of the test rig as specified in the present deposits and the particles according to the criteria
simulation mentioned above.
If the effect of wall deposits on furnace performance is
Mass flow rate of pf 0.0059 kg/s
Mass flow rate of primary 0.0161 kg/s required, one can simply re-run the main CFD calculations
air @ 30⬚C with the new furnace wall temperatures obtained from the
Mass flow rate of 0.0463 kg/s above calculations.
secondary air @ 404⬚C

2.5. Deposits accumulation model and effects on furnace 3. Validation case

performance
The model was applied to predict the slagging propensi-
Since deposits accumulation rate is affected by many ties of three UK coals, Bentinck, Daw Mill and Silverdale,
factors (e.g. erosion by impacting flyash particles, viscous in a single-burner pilot scale ash deposition rig [22–24]. A
flow, external removing mechanisms or load cycling), 3D cutaway view of the rig is shown in Fig. 3. The rig is
development of a detailed ash accumulation model is gener- designed to study the ash deposition characteristics of power
ally not feasible or pointless if not all these mechanisms are station grade coals and to simulate the temperatures history
taken into account. In order to estimate the effect of wall of pf particles in a front wall fired furnace. The coal/primary
deposits on subsequent deposition rates, a simple but effec- air stream is conveyed to the burner by a flexible hose. The
tive scheme has been proposed [15] to account for the burner is similar to a gas burner in design with a closed end
changes in thickness and surface temperatures of the wall and 8 nozzles spaced equally on the circumference (Fig. 4).
deposits. Secondary air, preheated to temperatures of up to 450⬚C is
The amount of flyash deposits over a specific time interval introduced at a swirl angle of 45⬚ to the coal/primary air
is calculated by multiplying the initial deposition rate (based stream. The burner fires into a refractory lined chamber
on the initial assumption of a clean wall) by the correspond- where an air cooled mild steel plate serves as the ‘‘slag
ing time interval. Based on the properties of the wall depos- panel’’, simulating a section of a boiler wall. The mass
its and the corresponding wall and adjacent gas flow rate of pf and the corresponding inlet combustion air
temperatures, the thickness of the deposit layer and the are presented in Table 2. The ultimate and proximate
surface temperatures of the layer are calculated. In the analyses of the three coals are given in Table 3.
subsequent time interval, if the calculated surface tempera- The rig is represented by a computational grid consisting
tures of the wall deposits are higher than the fusion tempera- of 58 × 17 × 17 cells (Fig. 5). In order to simplify the mesh
ture of the bulk ash, the local average composition of the structure, all the circular sections of the rig (i.e. burner,
deposits and the new surface temperatures will then be used burner quart and quart extension) are represented by square
to re-calculate the surface viscosity of the wall deposits. The equivalents of the same cross-sectional area. The burner is
sticking probability of flyash particles is then re-examined represented by nine rectangular cells as shown in Fig. 6. The
secondary air is entered into the computational domain
Table 3
through the cell faces representing the secondary inlet at
Proximate and ultimate of the three coals under investigation constant velocity and at the corresponding swirl angle.
The primary jets are modelled by the introduction of sources
Properties Bentinck Daw Mill Silverdale (i.e. momentum, mass, enthalpy, turbulence kinetic energy
and its dissipation rate) into the eight scalar inlet cells as
Ash (%ad) 17.1 17.4 17.6 indicated by the thick arrows in Fig. 6. The pf sample are
Moisture (%ad) 3.6 5.2 2.2
divided equally into four groups with mass averaged repre-
Volatile matter (%ad) 29.2 31.4 32.1
Volatile matter (%daf) 36.8 40.5 40.0
sentative diameters of 16, 32, 51 and 64 mm, respectively.
Calorific value (kJ/kg daf) 33 660 32 800 33 860 These representative particles are introduced into the
Carbon 67.4 65.18 67.8 computational domain at the same velocities as the inlet
Hydrogen 4.6 3.6 4.7 primary air.
Nitrogen 1.4 1.13 1.6 The thin arrows shown in Fig. 6 represent their injection
Oxygen N/D 7.52 5.8 points. Referring to the figure, a total of 32 pulverized coal
Sulphur particles tracks are simulated. This, however, should not be
Pyritic N/D 0.33 2.54 confused with the number of trajectories used for flyash
Organic N/D 1.18 0.83 deposition calculations. In this study, a total of 64 000 pfa
Sulphatic N/D ⬍ 0.1 ⬍ 0.1
trajectories are used (i.e. 8 nozzles × 8 size groups per
Total 1.7 1.51 3.37
Chlorine 0.3 0.16 0.23
nozzle × 1000 trajectories per size group). The temperature
profile of the slag panel is set to the measured values. A
F.C.C. Lee, F.C. Lockwood / Progress in Energy and Combustion Science 25 (1999) 117–132 125

Fig. 5. CFD mesh for the CRE rig.

126 F.C.C. Lee, F.C. Lockwood / Progress in Energy and Combustion Science 25 (1999) 117–132

Fig. 6. CFD mesh representing the burner.

F.C.C. Lee, F.C. Lockwood / Progress in Energy and Combustion Science 25 (1999) 117–132 127

Fig. 7. Predicted flow field.

128 F.C.C. Lee, F.C. Lockwood / Progress in Energy and Combustion Science 25 (1999) 117–132

Fig. 8. Typical ash depositions predictions.

 F.C.C. Lee, F.C. Lockwood / Progress in Energy and Combustion Science 25 (1999) 117–132 129

Table 4 communication). The predicted results were also assessed

Summary of the deposition results for the test rig by the rig operators qualitatively. It has been concluded that
the accuracy of the predictions are adequate for the subse-
Percentage of the Bentinck Daw Mill Silverdale
quent ash deposition calculations.
incoming ash by weight
In general, the predicted deposition patterns for the three
Total amount of deposits 36.8 38.3 41.8 coals, i.e. Bentinck, Daw Mill and Silverdale, are in quali-
Amount of pfa that goes 30.0 28.4 25.2 tative agreement with observations. The predictions for the
straight to outlet Bentinck coal show that the two side wars and the roof are
Amount of pfa that are 25.8 28.2 29.1 entirely covered by a thin layer of ash deposits with the floor
trapped in a boundary exhibiting a thicker deposit layer than for the other three
layer walls. The predicted deposition rate on the roof at the
Amount of non-sticky 7.4 5.1 3.9
converging section near the outlet is also consistent with
flyash particles
the observations, which suggest a relatively low deposition
rate in this region. Both the predictions and the observations
suggest localized deposition spots in the quart extension.
constant wall temperature profile of 1250⬚C is specified for This is believed to be caused by the direct impaction of
the rest of the refractory lined chamber walls. flyash particles on the quarl as they were blown from the
burner. The deposition patterns predicted for the Daw Mill
3.1. Results and discussions coal are similar to those for the Bentinck coal. The two side
walls are covered by a thin layer of ash with little deposit
The predicted flow pattem on a horizontal plane across being formed and downstream from the slag panel. Local-
the centre of the ash deposition rig (ADR) is shown in Fig. 7. ized deposition spots are predicted to form on the quarl
Also shown in the figure are the outlines of the internal extension and on the burner quarl. For the Silverdale coal
(IRZ) and external (ERZ) recirculation zones formed in the model correctly predicts it’s deposition rate relative to
the rig enclosure. The formation of an IRZ in the near burner that from the Bentinck and Daw Mill coal. The overall
region (NBR) is the direct result of the negative pressure deposition pattern is similar to the other two runs. Never-
gradient developed by the swirling secondary air and the theless, a thin layer of ash deposits is predicted to cover the
high velocity primary air jets. The IRZ is enclosed by the entire roof of the rig for the Silverdale coal. This prediction
larger ERZ which extends downstream to about 7 times the agrees very well with the observations. Fig. 8 shows a typi-
secondary air inlet diameter. This flow pattern results in a cal contour plot of the ash deposition pattern for the one of
stable flame and rapid heat release. Coal devolatilization the three coals under investigation. Detailed experimental
takes place at the oxygen-rich IRZ boundaries and burnout results and photographic records of the test rig after each test
is achieved by most of the pf particles before they leave the burn are presented elsewhere [22–24].
quart extension (i.e. about three burner diameters down- Table 4 summarizes the predicted results for the three
stream from the inlet). coals during the initial deposition stage (i.e. starting from
The predicted heat flux to the slag panel resembles the a clean wall condition). The amount of non-sticky flyash
measured values reasonably well (within 8%). The flow shown in the table corresponds to the percentage of ash
pattern and temperature distributions are also consistent particles from the inlet that are able to penetrate the bound-
with those measured previously (M.W. Lewitt, private ary layer completely, but are unable stick to the walls as a

Table 5
Predicted chemical properties of the deposits on the slag panel in the ADR

Bentinck Daw Mill Silverdale

P B I S M P B S M P B SM

SiO2 55.6 57.0 47.6 53.1 53.7 55.1 46.8 53.0 50.0 48.4 48.4 45.7
Al2O3 26.8 25.6 21.7 25.7 25.2 22.8 10.9 25.0 26.7 25.0 25.0 25.0
Fe2O3 9.1 8.3 18.3 12.0 8.7 7.9 30.3 9.9 17.4 18.9 18.9 25.0
CaO 1.1 1.7 3.3 1.9 4.8 7.6 7.5 5.4 1.5 3.3 3.3 1.1
MgO 0.9 1.0 1.3 1.1 1.8 1.4 1.0 1.9 0.6 0.6 0.6 0.6
Na2O 1.2 1.2 0.9 1.3 0.7 0.5 0.4 0.6 0.4 0.4 0.4 0.2
K2O 3.4 3.4 3.3 3.7 3.2 2.9 1.4 3.2 2.1 2.0 2.0 1.8
TiO2 1.2 1.1 1.4 1.2 1.1 1.1 0.5 1.1 0.8 0.9 0.9 0.6
MN3O4 0.2 0.3 0.8 — 0.4 0.4 0.7 — 0.2 0.3 0.3 —
P2O5 0.4 0.6 0.7 — 0.5 0.3 0.4 — 0.3 0.2 0.2 —

P, probe ash; B, backend ash; S, slag panel deposits; and M, predicted compositions of the slag panel deposits.
130 F.C.C. Lee, F.C. Lockwood / Progress in Energy and Combustion Science 25 (1999) 117–132

Table 6
Predicted deposition rate on the slag panel in the ADR

Bentinck Daw Mill Silverdale

Predicted deposition rate (g/h) 1.28 1.93 2.90

Viscosity of the slag panel deposits (Pa s) 9.00 × 10 5 6.11 × 10 5 1.26 × 10 4

result of their high viscosities. Also shown in the table is the a result, a further examination of the chemical properties of
percentage of flyash particles that are trapped in a boundary the slagging deposits was performed.
layer. This refers to the particles that do not have sufficient The predicted results in terms of the chemical cormpo-
momentum to traverse the layer. As a result, they are consid- sitions of the initial deposits formed on the slag panel are
ered to be trapped in the layer, and are carried by the layer to shown in Table 5. The ‘‘furnace ash’’ referred to in the
the furnace outlet. Finally, the percentage amount of flyash Table represents the flyash sample collected in the furnace.
that goes straight to the outlet corresponds to the amount of ‘‘Back end’’ ash, on the other hand, refers to the sample
particles that have not collided with any interior surfaces of collected from the cyclone at the back end of the rig. In
the rig before leaving. order to be more consistent with the present predictions,
Referring to Table 4, the predicted deposition potential the compositions of the ‘‘furnace ash’’ and the ‘‘back end
are generally consistent with the observations. Since the ash’’ were both obtained from CCSEM analysis.
Bentinck flyash sample has a higher proportion of large By comparing the model predictions with the ‘‘furnace
particles (⬎ 50 mm), and the density variations are relatively ash’’ data, it has be shown that all three deposits have an
small amongst the three samples, the average inertia as initial enrichment in Fe and a corresponding reduction in
possessed by the Bentinck flyash particles are, therefore, Al2O3 and SiO2. The enrichment in Fe relative to the flyash
higher than the other two coals. Consequently, the Bentinck is expected and confirms the known role of low-melting
flyash particles are more able to penetrate the boundary point pyrite derived particles in the initial stage of depo-
layer, resulting in the least quantity of trapped particles. sition. Examination of the predicted viscosity of the slag
Because of their high inertia, the Bentinck flyash particles panel deposits shown in Table 6 reveals that the ash deposits
are also less affected by the flow field. This allows more of on the slag panel from the Silverdale coal are generally more
them to traverse the IRZ, and reach the outlet before hitting fluid than those of the other two coals. This can be explained
the wall boundaries. by the relatively large proportion of iron found in the Silver-
Referring to the results shown in Table 4, approximately dale deposits, which substantially lowers their viscosities.
42% of the input flyash particles from Silverdale are depos- Since the present sticking model is solely dependent on
ited during the initial stage of ash deposition. When the viscosity, the lower the viscosity, the higher the predicted
effect of the wall deposits was taken into account by the deposition rate will be. As a result, the predicted deposition
accumulation model nnentioned in Section 2.5, this rate on the slag panel is higher for the Silverdale coal than
increased to nearly 55% after 1 h of prolonged firing. Exam- the other two coals.
ination of the results suggested that this increase is princi- On the whole, the predicted compositions of the deposits
pally caused by the rise in particles arrival rate to the on the slag panel are in reasonably good agreement with the
substrate surfaces and the reduction of non-sticky flyash measurements. However, the dramatic increase in the
particles. The increase in particles arrival rate is thought measured CaO level in the three samples is not immediately
to be the direct result of the increasing thickness of the anticipated. Further examinations of the furnace ash and the
wall deposits as computed by the accumulation model. back end flyash indicate that all the flyash samples collected
Since the viscosity of flyash particles should remain constant at the back end are enriched in CaO. The reasons for this
before and after the accumulation model is used, the reduc- enrichment are not fully understood at present. However, the
tion in non-sticky flyash particles is, therefore, mainly depletion of CaO in the samples taken from the main
caused by the formation of a sticky layer on the surface of furnace, the composition of the collected ash samples is
the wall deposits which retains the otherwise non-sticky undoubtedly affected by the aerodynamic properties of the
particles. flyash particles. Apart from the CaO level the measured
Coal ash deposition is a very complex process, and it is Fe2O3 levels on the slag panel, as obtained from CCSEM
not sufficient to judge the slagging propensity of coal from analysis for the Daw Mill trial, are also exceptionally high
its deposition rate alone. In order to better understand the when compared with the parent flyash sample (almost 250%
slagging performance of a coal, the properties of its deposits increase). Examination of the slag panel deposits data as
must also be taken into account. For example, coals which obtained from atomic absorption iduced couple plasma tech-
produce a large amount of runny slag prove less problematic nique shows a different trend and suggests only a 30%
than those which generate a moderate amount of highly increase in Fe content, which is more consistent with the
fused and viscous slag that is very difficult to remove. As current predictions. Finally, the deposit composition is also
 F.C.C. Lee, F.C. Lockwood / Progress in Energy and Combustion Science 25 (1999) 117–132 131

found to be closer to the bulk ash when the effects of wall of flyash data and other related information for the two test
deposits have been taken into account. cases.

4. Concluding remarks
References
A predictive scheme based on CCSEM flyash data and a
CFD technique has been developed to model ash deposition [1] Srinivasachar S, Senior CL, Helble JJ, Moore JW. A funda-
in a pf-coal field furnace. The model has been applied to a mental approach to the prediction of coal ash deposit forma-
single-burner pilot scale ash deposition rig to predict the tion in combustion systems. 24th Symposium on Combustion,
deposition behaviour of three UK coals (Bentinck, Daw July 1992, Sydney, Australia.
Mill and Silverdale) exhibiting different degrees of slagging [2] Walsh PM, Sayre AN, Loehden DO, Monroe LS, Beer JM,
tendencies. Sarofim AF. Prog Energy Combust Sci 1990;16.
The predicted slagging potential and deposition patterns [3] Wall TF. Mineral matter transformation and ash deposition in
of the three coals are in reasonable agreement with observa- pulverized coal combustion. 24th Symposium on Combustion,
tions. The model also correctly predicts the chemical par- July 1992, Sydney, Australia.
[4] Kalmonovitch DP. Predicting ash deposition from flyash char-
titioning in the initial deposits layer of the three coals, that
ateristics. Proceedings of the 1991 Engineering Foundation
is the enrichment in iron and a depletion in silicates and Conference, March 1991, Palm Coast, FL, USA.
aluminium. The deposit compositions are found to resemble [5] Lazouplous G. Ph.D. thesis, Imperial College of Science
the bulk ash chemistry when the effects of wall deposits are Technology and Medicine, Department of Mechanical
taken into account. This finding is consistent with general Engineering, 1995.
observations, and agrees with the measurements from the [6] Launder BE, Spalding DB. The numerical computation of
full scale plant trails. Further, the results from this study turbulent flows. Comput Meth Appl Mech Engng 1974;3.
indicate that large flyash particles are more likely to deposit [7] Lockwwod FC, Shah NG. A new radiation solution method
on the furnace walls than smaller ones. The sticking for incorporation in general combustion prediction
efficiency of particles are found to be mainly controlled by procedures. 18th Symposium on Combustion, August 1981,
Waterloo, Canada.
their thermal (e.g. temperature) and chemical properties.
[8] Badzioch S, Hawkesley PGW. Kinetics of the decomposition
The quantity of flyash particles trapped in the boundary of pulverized coal particle. Ind Engng Chem Process Des
layer and the particles arrival rate is very sensitive to their Develop 1970;9.
size and density, but it is more sensitive to the particle [9] Field MA, Gill DW, Morgan BB, Hawksley PGW. Combus-
chemistry. In general, the model predictions are found to tion of pulverized coal, The British Coal Utilization Research
be quite sensitive with respect to the physical and chemical Association, 1967.
properties of the input flyash data. As a result, the accuracy [10] Lockwood FC, Rizvi SMA, Lee GK, Whaley H. Coal combus-
of the predictions are dependent on a proper knowledge of tion model validation using cylindrical furnace data. 20th
the flyash properties. Symposium on Combustion, August 1994, MI.
As for future work, there are some features of the model [11] Migdal D, Agosta VD. A source flow model for continuous
gas-particle flow. J Appl Mech 1967;34.
that could benefit from further development. In this respect,
[12] Lockwood FC, Salooja AP, Syed SA. Sonic exploratory
an improved sticking mechanism based on the more sophis- calculations of the flow and combustion in a cement kiln,
ticated viscosity and liquidus temperature of the depositing Report to the associated Portland Cement manufacturers
ashes is currently being investigated. The incorporation of a Ltd, 1978.
flyash generation model will also reduce the uncertainties [13] Belemont A et al. Particle Lagrangian simulation in turbulent
inherent in the measured flyash data. The inclusion of depos- flows. Int J Multiphase Flow 1990;16.
its removal mechanisms, such as soot blowing and viscous [14] Faeth GM. Evaporation and combustion of sprays. Prog
flow, are desirable as well. Finally, the present study has Energy Combust Sci 1983;9.
revealed a serious lack of good quantitative data essential to [15] Lee CCF. M.Phil to Ph.D. Transfer Report, Imperial College
code development and validation for furnace aerodynamics of Technology and Medicine, Department of Mechanical
Engineering, 1995.
and ash deposition.
[16] Gosman AD, Ioannides E. Aspects of computer simulation of
a liquid-fuelled combustors. AIAA 19th Aerospace Science
Meeting, 1980.
Acknowledgements [17] Friedlander SK, Johnstone HF. Deposition of suspended
particles from turbulent gas streams. Ind Engng Chem
1957;49.
Most of this work was conducted at National Power Plc [18] Liu BY, Agarwal JK. Experimental observation of aerosol
Research and Engineering Division, and is published with deposition in turbulent flow. J Aerosol Sci 1974;5.
the permission of National Power Plc. The authors would [19] Reeks M, Skyrme G. The dependenc of particle deposition
like to thank Mr. F. Wigley (Imperial College), Dr. W. Gibb velocity on particle inertia in turbulent pipe flow. J Aerosol
(PowerGen Ltd) and Mr. M. Lewitt (CRE) for the provision Sci 1976;7.
132 F.C.C. Lee, F.C. Lockwood / Progress in Energy and Combustion Science 25 (1999) 117–132

[20] Srinivasachar S, Helble JJ, Boni AA. An experimental study of research programme—tests on Bentinck Coal in the CRE
the inertial deposition of ash under coal combustion conditions. Ash Deposition Rig, CRE report No. BC 21, 1993a.
23rd Symposium on Combustion, July 1990, Orleans, France. [23] Lewitt MW, Keeling J, Hamlin CJ. Collaborative slagging
[21] Frank M, Kalmonovitch DP. An effective model of viscosity research programme—tests on Daw Mill Coal in the CRE
for ash deposition phenomena. In: Bryers RW, Vorres KS, Ash Deposition Rig, CRE report No. BC 20, 1993b.
editors. Mineral matter and ash deposition from coal. United [24] Lewitt MW, Keeling J, Hamlin CJ. Collaborative slagging
Engineering Trustees Inc. pp. 85–101. research programme—tests on Silverdale Coal in the CRE
[22] Lewitt MW, Keeling J, Hamlin CJ. Collaborative slagging Ash Deposition Rig, CRE report No. BC 22, 1993c.