Sie sind auf Seite 1von 12

09575820/00/$10.00+0.

00
q Institution of Chemical Engineers
Trans IChemE, Vol 78, Part B, January 2000

MIXING, MODELLING AND MEASUREMENTS OF


INCINERATOR BED COMBUSTION
Y. R. GOH (ASSOCIATE MEMBER), C. N. LIM (ASSOCIATE MEMBER), R. ZAKARIA, K. H. CHAN (ASSOCIATE MEMBER),
G. REYNOLDS (ASSOCIATE MEMBER), Y. B. YANG, R. G. SIDDALL, V. NASSERZADEH (ASSOCIATE MEMBER) and
J. SWITHENBANK (FELLOW)
Shefeld University Waste Incineration Centre (SUWIC), Department of Chemical and Process Engineering, University of Shefeld, Shefeld, UK

he safe disposal of municipal solid waste has now become an urgent environmental
problem. The traditional method of landlling waste has created so many
environmental problems that countries including Denmark, Holland and Germany
have imposed severe restrictions on landlling burnable waste. With up to 1 tonne of municipal
waste being generated by every individual annually in the UK, incineration is now at the
forefront of combustion research, as developed countries recognize the environmentally
friendly advantages of this technology. An efcient incinerator is not only assessed by the
amount of heat recovery but also by the levels of emissions and quality of the ash it produces.
Incinerator designs must therefore be fully optimized so that they can control emissions by
reducing the production of harmful pollutants such as dioxins, furans, NOx and SOx. Hence,
incinerator bed combustion is a vital area that urgently needs further investigation. At
SUWIC, the present work concentrates on the development of a comprehensive and reliable
model for the incinerator bed combustion process. The results from the incinerator burning
bed model can then provide the much needed boundary conditions for Computational
Fluid Dynamics (CFD) modelling of the gas phase reacting turbulent ow in the freeboard
region of an incinerator. In addition to the development of the computational model,
the work involves several parallel activities, including experimental investigations into
waste combustion, solid mixing and prevention of slag formation and instrumentation
development.
Keywords: waste incineration; slag formation; instrumentation; mixing; mathematical
modelling; pollution.

INTRODUCTION

for assessing proposed designs of incinerator, or troubleshooting existing plant designs, without building the plant
or small-scale accurate physical models.
The evolution of better and more efcient incinerators
in the UK is currently hindered by the lack of reliable data,
and fundamentally based design procedures to assist their
design/manufacture, operation and control. The combustion of waste in incinerators is complex, and the design
and control of such processes poses many problems. For
example, many existing grate systems do not agitate the
burning refuse efciently. Consequently, a large volume of
excess air is required to achieve a good burn-out. This leads
to problems with the gas phase combustion, overloading
of the electrostatic precipitators and/or scrubbers, and high
particulate emissions from the bed. SUWIC studies at some
UK municipal incinerators with moving grate systems
also showed high levels of unburned carbon in the residual
and y ash, which increases the cost of ash disposal. In
contrast to the progress in gas phase combustion modelling
with CFD, the open literature contains no satisfactory model
for the burning of municipal solid waste on a travelling
grate. The prediction of the ow rate and composition of
gases emerging from the bed is particularly important as
it provides the upstream boundary conditions for the ow
calculation in the gas phase region.

The amount of waste generated by mankind and the growing


concern over health and environmental problems related
to landll has risen to a point where incineration is perceived to be the best option to dispose of most wastes,
especially when some of the energy content of the waste can
be recovered. Incineration is an environmentally friendly
process for the disposal of municipal solid waste, provided
that it is properly carried out. The main objective of
incineration is to convert a large volume of chemically
and biologically active waste into a small volume of inert
matter.
In any technology, the design process depends on a clear
understanding of the fundamental scientic principles
on which the design is to be based. These fundamental
principles are usually expressed in terms of the governing
differential equations of the process under consideration.
The required design is then obtained by solving these
equations subject to the appropriate boundary conditions,
physical constants, input and output conditions. In incinerators, the main combustion-related design items are the
burning bed of solid waste on the grate and the gas phase
path, including the reactions which take place. For the gas
phase, advances in CFD have proved to be invaluable
21

GOH et al.

22
MODELLING OF SOLID MIXING ON A
MOVING GRATE

A crucial feature of incinerator combustion is the mixing


of the solid material on the grate. To achieve complete combustion of solid waste in a municipal incinerator, producing as few toxic contaminants as possible, it is essential to
acquire experimental information about the combustion
processes on the travelling grate. Existing grate systems
do not produce a uniform gas concentration distribution
above the burning refuse, and a large volume of excess air
is usually required to achieve good ue gas burn-out (typically using 100% excess air). These non-uniformities lead
to the aforementioned problems with the gas phase combustion, overloading of the ue gas scrubbing system, high
particulate carryover from the bed and high levels of
unburned carbon in both the residual ash and the y ash,
which makes it expensive to dispose of the ash.
In this part of the study, the mixing process on scale
models of three industrial grates has been quantitatively
determined by a series of systematic particle movement
experiments (random lateral and vertical motion). For each
conguration, sufcient data have been collected to permit a
new mathematical model for the mixing process statistics
to be developed. To characterize and quantify the mixing
process, experiments were conducted in the following
small-scale industrial grates (1:15 size ratio): (i) Deutsche
Babcock grate (as shown in Figure 1); (ii) Martin grate
(Figure 2); and (iii) ABB double motion overthrust
horizontal grate (Figure 3).
Mixing is essentially the dispersion of material from an
initial position, so to quantify this, it is necessary to see
where material can progress to, once it has left the initial
position. In the experiments, a xed volume of solids was
introduced onto the grate contained in a perspex box, as
shown in Figure 1. Marked or coloured solid tracers were
positioned on the original bed of solids. A digital image
video camera was then used to record the movements of the
tracers. Digital frames of the test were captured at ve
second intervals. The position of each tracer on each frame
was evaluated in terms of (x,y) and time, where x is the
distance measured from the point of refuse input, and y is
the distance measured from the nearest wall. This can then
be used to produce the distribution of the tracers based on

the proportion of tracer at a particular distance and the


number of tests performed.
Figure 4 shows the distribution curves of the solid tracers
in a Deutsche Babcock grate experiment, normalized to the
total number of tests performed. The standard deviations1 for
these curves were calculated, and the mean of the calculated
standard deviations was found to have a value close to 2.
A mathematical model has been developed to simulate
the movement of the solid material on the grate based on
the probability for a solid tracer to swap position with the
adjacent tracer. This model can be applied to the movement
of the solid material in three directions, i.e. axial (x),
transverse ( y) and normal to the grate (z).
As an example, consider the solid movement in the
transverse ( y) direction only. The distance travelled by a
solid tracer in the transverse direction is the result of several
local swaps. In this model, the solid bed is divided into a
number of cells nt, as shown in Figure 5. Initially, cell n is
occupied by a tracer n. In Figure 5, C(t,n) is the number
of the tracer in cell n at time t. Each cell n is also given a
variable P(t,n) which is a random number between 0 and 1,
generated during the numerical simulation.
A sequence of `decision-making processes is carried
out across the bed, over several time steps, to determine

Figure 1. Deutsche Babcock rotating drum grate.

Figure 3. ABB horizontal grate.

Figure 2. Martin grate.

Trans IChemE, Vol 78, Part B, January 2000

INCINERATOR BED COMBUSTION

23

Figure 4. Tracer distribution from Deutsche Babcock grate experiment.

whether a tracer in cell n would swap position with the


tracer in cell n + 1. This sequence is as illustrated in
Figure 6. In this `decision-making process, a global value
Pl is rst set for the probability for a local swap. At the
rst time step Dt, if probability of cell 1 at t = 0 or P(0,1)
is larger than Pl, then tracer m = 1 at time t = 0 or C(0,1)
will stay at cell 1 and no swapping will occur. The process
then proceeds to cell 2 where P(0,2) is compared with Pl
to decide whether C(0,2) will swap position with C(0,3). If
P(0,2) is again larger than Pl, C(0,2) will not swap position
with C(0,3). The process will then proceed to cell 3.
If in the rst instance, P(0,1) is smaller than Pl, then
C(0,1) will swap position with C(0,2). The process then
proceeds to cell 3 where P(0,3) is compared with Pl to
decide whether C(0,3) will swap position with C(0,4). If
P(0,3) is again smaller than Pl, C(0,3) will swap position
with C(0,4). The process will continue at cell 5. If at the
end of the sequence, there is only one cell remaining,
the probability P(0,nt) will not be compared with the
limiting value, Pl. Since the tracer is already in the last
cell adjacent to the wall, no swapping will occur. This
`decision-making process can start at either side of the bed.
In summary, at any time t, C(t,1) can either be C(t Dt,1)

or C(t Dt,2), depending on the values of P(t Dt,1)


and

Pl. Similarly, C(t,2) can either be C(t Dt,1), C(t Dt,2)

or C(t Dt,3), depending on the values


of P(t Dt,1),

P(t Dt,2) and Pl. For cell n = nt , C(t,nt) can either


be

C(t Dt,nt 1) or C(t Dt,nt). In general, C(t,n) can


be C(t

C(t Dt,n) or C(t Dt,n + 1),


either
Dt,n 1),

for 1 < n < nt . This process is essentially


a `random
walk
procedure.
A spreadsheet program based on the aforementioned
procedure has been set up. Each cell of the spreadsheet
represents a cell of the solid bed. The probability value
for each cell, P(t,n), is produced by the spreadsheet
random number generation tool. It returns an evenly distributed random number greater than or equal to 0 and less
than 1. A new random number is returned every time the
worksheet is calculated. The frequency distributions of
the solid tracers across the bed can be obtained after
performing the calculation for several time intervals, Dt
and after a certain number of tests, N. This calculation is
then repeated several times for a range of Pl. The distribution curves, except for the two tracers which started in
the cells immediately adjacent to the side walls (tracers
m = 1 and m = nt = 10), resemble normal distribution
Trans IChemE, Vol 78, Part B, January 2000

curves where the majority of the tracers reside at their


original position.
The spread of the computed distribution is represented
by its standard deviation, s. Figure 7 shows the standard
deviation s computed for the tracer particles over a range
of values of Pl. If the probability to swap is higher, the
solids are more likely to spread as they travel to other cells,
as indicated by the increasing horizontal scatter of the data
at higher values of Pl. Since tracers at the centre of a grate
tend to spread more in comparison to those near to the
wall, more than one standard deviation can be obtained
from each assumed probability of swapping, Pl. A relationship between the probability of swapping, Pl, and standard
deviation, s, can be expressed in mathematical form by
tting a curve through the data.
Multiple regression analysis1 was carried out to determine the best tting line of the simulated data. This line
must pass through the origin since there is no chance for
the tracers to disperse if the probability of swapping is equal
to 0. The least squares curve of points (s1,Pl1), (s2,Pl2),,
(sN,PlN) is found to be of the form:
Pl = 0.04638s 2

(1)

The 95% condence limits for the coefcient in equation


(1) were found to be 0.04638 6 0.0108. The computation
of the standard deviation is somewhat in error due to
grouping of data into classes and this error is adjusted
using Sheppards correction1.

Figure 5. Division of solid bed in the transverse (y) direction.

GOH et al.

24

Figure 6. Local swap procedure in the transverse standard deviation.

By inserting the mean value of experimentally obtained


s into equation (1), the probability for the solid tracers to
swap positions in the experiment, Pl, was obtained. This
value of Pl is then fed into the mixing model to predict the
standard deviation for each tracer distribution.
Figure 8 compares the experimentally obtained and the
computed standard deviations of the tracers. It also shows
the variation of s with respect to distance from one of the
side walls; which was used as a reference wall in this
analysis. It is evident from Figure 8 that there is a
reasonably good agreement between the experimental and
the computed data. The mean of the computed standard
deviations was found to be 2.1; a value which lies within
the 95% condence bands of the best tting curve as
shown in Figure 9. This suggests that the movement of the
tracer particles in the y-direction can be accurately predicted
using this technique.
The same technique can be applied to the x and z
directions and be validated against experimental data
obtained from the particle tracking experiments in the
respective directions. The integration of the mixing model
in these directions will provide a three-dimensional mixing
model which can predict the mixing process on the grate
and be incorporated to a mathematical model for incinerator waste bed combustion simulation.

considered to consist of four components: moisture, volatile,


xed carbon and ash. During incineration, moisture, volatiles
and xed carbon are removed from the solid matter by the
drying, pyrolysis and gasication processes, respectively. A
physical representation of the change of bed volume which is
mathematically linked to these three processes has been
established in the step change model2.
The conservation equations of the solid phase components are expressed as:
s Yi
+ = (s ns Yi ) = = (Ds = (s Yi ) + Si
(2)
t
where = u = u x /x + u z /z for a two-dimensional simulation. The general expression for the energy equation of
the solid phase may be written as:
s Hs
+ = (s ns Hs ) = = (ls = Ts ) + = qr + Qs
t

(3)

MODELLING OF INCINERATOR
BED COMBUSTION
The physical volume of the municipal waste reduces
by 90% and the mass by 70% during combustion. A suitable
bed model must therefore permit simulation of the reduction in the bed volume. In general, the solid waste can be

Figure 7. Probability of swapping versus computed standard deviation.

Trans IChemE, Vol 78, Part B, January 2000

INCINERATOR BED COMBUSTION

25

Figure 8. Variation of standard deviation (s) with distance from one of the side walls ( y).

Figure 9. Comparison of the experimental result with the calculated value from the mixing model.

where qr denotes the radiative heat ux, and the source term
Qs represents the overall effects of heat transfer between
gas and solid and heats of reaction of the various processes
occurring during solid incineration.
The solid velocity in the vertical direction for any control
volume j in the bed (ns)z( j ), depends on the rate of volume
reduction of the bed and can be written as:
1 V
(ns )z ( j) = (ns )z ( j 1) +
(4)

Az ( j ) t j
where
V VB VC VD VA
+
+
+
(5)
=
t
t
t
t
t
and B is the initial waste material, C is the dried solids, D is
the dried and pyrolysed solids and A is the dried, pyrolysed
and gasied solid (ash). The time differential equations
representing the changes in volume of each material are
given as follows2:
VB
(RP)2
(6)
=
2 v 2B (1 B )
t

VC (RP )2 [1 (1 F2 )v 2B ]
(RP )3

(7)
=
3 v3C (1 C )
t
2 v2B (1 C )

VD (RP )3 [1 (1 F3 )v 3C ]
(RP )5

(8)
=
5 v5D (1 D )
t
3 v 3C (1 D )

VA
(RP )5 v 4D
(9)
=
t
5 v 5D (1 A )

Trans IChemE, Vol 78, Part B, January 2000

The steady-state travelling grate combustion problem may


be simplied by using an unsteady-state static bed if it is
assumed that the refuse is supplied to the moving grate at a
constant rate, and there is no movement of bed material
relative to the grate. Predicted variations with respect to
time, t, using the unsteady-state static bed model can be
used to predict variations with distance, x, from the point
of refuse input in the steady-state moving bed model by
using x = nx t, where nx is the local steady velocity of the
grate movement. Any variations of the conditions above
the bed and the rate of underfeed air input with respect to x
for the locally steady-state moving bed can be modelled

Figure 10. Typical gas temperature prole from unsteady state xed bed
model.

26

GOH et al.

Figure 11. Contour plot of predicted solid temperature.

as variations with respect to time in the unsteady-state static


bed using t = x/nx . An example of the typical temperature
prole of the gas phase predicted by the mathematical
model is shown in Figure 10. This prole was obtained
assuming a constant freeboard temperature and that combustion of gaseous volatiles liberated during pyrolysis
occurs above the bed. Similar temperature proles were
obtained during experiments with a batch reactor (see
Figure 14). The slopes of the temperature rise in Figure 14
are not as steep as those seen in Figure 10. This indicates
that a fraction of the volatiles liberated during pyrolysis
burn within the bed causing a more gradual temperature
rise.
Figure 11 shows the temperature contours of the solid
material in the bed, modelled on a sloping grate with the
refuse being fed from left to right, predicted using
the steady-state travelling bed model. The temperature of
the solid material in the bed gives an indication of the
process that is occurring within the bed; drying (between
the wet bulb temperature and the vaporization temperature), pyrolysis (approximately 540 K) and gasication
(above 600 K). The locations of the processes as indicated

in the gure show close similarity to the combustion model


proposed by Kuo et al.3. Typically, the solid waste on the
travelling grate is initially heated and loses most of its
free moisture by the time its temperature reaches 373 K. The
drying zone in the bed as indicated in Figure 11 corresponds
to the region where the mass fraction of the moisture content in the bed is decreasing in Figure 12. In Figure 11, as
the bed temperature increases further, the waste pyrolyses
at about 540 K and then ignites at approximately 600 K.
The waste eventually burns vigorously until either the
oxygen surrounding the solid is consumed, or the solid is
fully devolatilized leaving a carbonaceous char. The residual charred or partially charred element may undergo
further pyrolysis, be gasied by CO2 or H2O to yield CO
and H2, or be oxidized by O2 to form CO2.
MEASUREMENTS OF WASTE BED COMBUSTION
A good dynamic mathematical model of the incineration
process requires a combined knowledge of the physical
characteristics of the grate movements and chemical characteristics of the waste combustion on the grate. Combustion

Figure 12. Contour plot of predicted moisture concentration.

Trans IChemE, Vol 78, Part B, January 2000

INCINERATOR BED COMBUSTION

27

Figure 13. Schematic diagram of the xed bed reactor.

and kinetics data that are required for the bed modelling
programme have been obtained from an experimental
laboratory xed bed reactor built to study the combustion
of municipal solid waste, as illustrated in Figure 13. The rig
consists of a cylindrical chamber of 1.5 m length and 20 mm
internal diameter. The combustion pot was designed to hold
samples of simulated waste up to 0.5 m in height from the
grate (about 2 to 3 kg). Preheated primary air is introduced
through a distributor plate (i.e., the grate) at the bottom of
the chamber. The bed is ignited at the surface and the
process fronts advance from the top of the bed to the grate.
The chamber is mounted on a weighing scale to permit the
burning rates to be determined. Temperatures and gas
composition measurements at xed positions above the
grate are also made. An example of the temperature
measurements obtained during a typical preliminary
combustion experiment are shown in Figure 14. As the
thermocouples are located at xed positions within the fuel
bed, the approach of a reaction front to each position is
shown by the sharp increase in the temperature measured.
The corresponding change in the solid mass in the bed is
as illustrated in Figure 15.
The gas composition in the reactor at 430 mm above the
grate is as illustrated in Figure 16. The time for the increase
in CO (to above the range of the gas analyser) and CO2
concentrations and the reduction in O2 concentration in this
position correspond to the time for the bed temperature
rise as depicted in Figure 9. This notable change is the
ignition front and is due to the ignition of the gaseous
volatiles that are liberated during the pyrolysis process.
Figure 16 also shows that concentration of NOx in this
position is also seen to gradually increase to a maximum
value of approximately 150 ppm before reducing to about
100 ppm. The NOx concentration increases because the
gases coming up to the sampling point contain NOx that
is formed during devolatization of the waste. As the layer
of char close to the surface of the bed passes below the
sampling point, the concentration of NOx decreases slightly.
Trans IChemE, Vol 78, Part B, January 2000

This decrease in NOx concentration can be attributed to


the reduction of NOx in the char layer where CO and NOx
can react to form CO2 and N2 using the carbon in char
as catalyst for the reaction4.
Experimental investigations on the xed bed reactor
must be complemented by studies on a full-scale incinerator
plant. In addition, development work on a suitable mathematical model for incinerator bed combustion is far from
ideal without comparisons with experimental data and
measurements from a full-scale plant to conrm the validity
of the approach used in the model. Conventional measuring
techniques only allow full scale plant data to be obtained
from xed positions in the plant, mainly in the freeboard
region and not inside the burning bed. In-situ instantaneous
measurements of temperature, radiation and gas composition inside a burning bed in a full-scale incinerator plant
are therefore difcult and often restricted for many
reasonssize and integrity of the affordable measuring
instrument, safe access, etc.
The requirement for accurate solid bed combustion
data for the bed model validation, particularly within the
solid waste in large incinerator plants, has initiated the

Figure 14. Gas temperatures above the grate.

GOH et al.

28

Figure 15. Change in bed mass with time.

as it can be introduced into the incinerator with the waste


feed and tumbles along with the burning waste material.
The unit thus experiences the same high temperature conditions as the waste material, whilst measuring and recording
the temperatures, gas composition and heat uxes in the
burning bed. At the end of the process, the capsule can be
recovered at the incinerator exit, and the data stored in
its memory unit can be downloaded to a computer. The
recorded data or process variables can then be saved to
a common delimited le, or copied to another existing
commercial package such as a spreadsheet, statistical
package or a word processor. The electronic instrument
within the capsule can be reused by refurbishing its
protective casing and the whole measuring procedure can
be repeated several times so that the mean and standard
deviation of the measured process variables can be determined. Thus, data on the statistical variability can be
obtained through random motion of the unit with the waste
and not through standard measurements from a xed
point in the plant, e.g. a boiler ue.
The size of the instrument must be optimized to exploit
fully the thermal properties of the insulating material and
geometrical design. To reduce the heating rate to the core
of the instrument, the ratio of the exposed surface area
per unit volume must be minimized. For a simple thermal
insulation system, the heating rate by conduction to the
unit, q, assuming quasi-steady state, is given by:
q = 2 lp(To

Ti )

Figure 16. Gas concentration at 430 mm above the grate.

development and construction of a prototype measuring


instrument. This prototype instrument consists of a family
of built-in electronic sensors and a recording memory unit
or a mini chip installed in a refractory bre or ablating
resin capsule about 12 to 15 cm in diameter and length. It
can record a variety of measurementstemperature (using
thermocouples), radiative ux (using a thin sapphire rod
as a bre optic element), and gas composition (e.g. CO2,
CO, O2) (using a phosphoric acid based electrochemical
cell), for the duration of the test, e.g. one hour, depending
on the thermal properties of the capsule and unit design.
The instrument as shown in Figure 17 is uniquely different from the instruments currently commercially available

(ro

(hi ro ho ri )
2ri ro

+
ri ) ln(hi ro /ho ri ) (ho h1 )

(10)

where, ho, hi, ro and ri are the dimensions of the capsule


as depicted in Figure 17.
The operating time of the instrument in each test depends
on the rate at which heat is transferred to the core and the
maximum allowable operating temperature of the electronics within the unit. To increase the operating time while
simultaneously maintaining a low internal temperature, a
heat sink can be incorporated into the design. Preliminary
testing of the system in a furnace at more than 10008 C with
plaster material and water as the heat sink demonstrated
a working life of more than one hour as shown in Figure 18.
This is sufcient for measurements in an incinerator since
most incinerators operate with a residence time of about
one hour.

Figure 17. Schematic of the prototype heat resistant capsule.

Trans IChemE, Vol 78, Part B, January 2000

INCINERATOR BED COMBUSTION

Figure 18. Change in core temperature with time.

Figure 19. Onset of slag formation.

MEASUREMENTS AND MODELLING OF


SECONDARY JETS IN AN INCINERATOR
In present practice in incinerators, secondary air jets are
located above the bed to burn out hydrocarbons and carbon
monoxide. It is observed that slag tends to build up round
these jets, and in time the jets are often completely closed
off. The build-up of slag around the jets is a major problem
in the operation of incinerators. This slag can accumulate
to a thickness of about one metre with a weight of several
tonnes. The accumulation can reduce the availability of the
plant and it also gives a safety problem since injury can be
caused when the slag is removed by pneumatic drill during

maintenance operations. More importantly, the jets are


often located in a critical mixing region in the incinerator,
i.e. in the throat region between the furnace chamber and
the radiation shaft, and the thick slag layer alters the ow
dynamics in this critical region where secondary air or
recycled ue gases are introduced.
Slag is formed when molten y ash particles which
are entrained into the jet ow impinge near the mouth of
the nozzle. The angle of the turn of the entrained ow in
the vicinity of the jet mouth is particularly sharp. Some
of the particles will have sufcient inertia not to be dragged
along with the jet. The inertia of these particles will lead
to them impinging on the wall as shown in Figure 19.
Since some of the lower melting point particles can be
molten at this stage, they will stick rather than bounce
off, thus initiating the slag layer. Once the slag layer
begins accumulating, even particles that are not molten
or sticky are more likely to be captured5. The solid slag
will then gradually build up until the thermal resistance
of the layer is such that the surface is molten. Particles
with a higher melting point can then stick and the rate
of build-up can be expected to increase. Eventually, the
inertia of the particles causes the layer to build up until
the jet is closed off.
The ow eld around a high velocity jet penetrating a
low velocity cross ow has been experimentally investigated in a wind tunnel specially designed to minimize
the turbulent intensity at the main ow inlet. The ink dot
ow visualization technique6 was used to obtain visible
streak lines on acetate lms that reveal the surface ow
patterns.
Figure 20 shows the surface ow pattern of a 7 mm diameter nozzle with a jet momentum ux ratio of 65 where
the jet momentum ux ratio J, is dened as7:
J=

(u 2 )jet
(u 2 )flow

(11)

It is evident in Figure 20 that the uid from the main


ow around the jet is entrained towards the mouth of the jet.
To prevent the process of slag build-up, the ow eld
around the jet region must be reversed. This can be achieved
by tting a deector near the mouth of the jet to redirect
some of the ow parallel to the wall as depicted in Figure 21.
The purpose of the deector is to skim off the outer layer
of the jet and redirect it tangentially to the surface of the
wall.

Figure 20. Surface ow prole around a jet.

Trans IChemE, Vol 78, Part B, January 2000

29

GOH et al.

30

Figure 21. Annular deector to prevent slag formation.

Figure 22 shows a surface tracing of the same nozzle


and ow conditions, but with a deector attached. The
deector had an orice diameter of 0.9 jet diameters, and
a height of 0.6 jet diameters above the at plate. It can be
seen that the addition of the deector clearly alters the ow
eld. Under these ow conditions, the deector moves
the area of separation in front of the nozzle 1.9 jet diameters
away from the leading edge of the nozzle. There is now
an area around the nozzle swept out by redirected ow
due to the deector. The darkened area seen near the forward facing mouth of the nozzle is due to recirculation
of ow in this area, which prevents the ink from being
blown away from the mouth of the nozzle.
Computational Fluid Dynamics simulation under the
same conditions revealed the recirculation zone under
the deector as shown in Figure 23. The uid from the
cross ow containing molten particles is no longer being
entrained into the jet but is forced away by the ow from
the main jet, hence reducing the probability of particle
impingement near the mouth of the jet. This recirculation
is likely to depend on the height of the deector and hence
the deector distance from the nozzle must be optimized to
minimize the recirculation. Further design to improve and
optimize the deector design, and hot combustion experiments with simulated y ash, are currently under way.
CONCLUSIONS
Incineration is now accepted as the most environmentally friendly means of disposing of municipal solid

waste, and the design of incinerators must be optimized.


In the past, the design of incinerators has not been based
on fundamental understanding and modelling of the process, and empirical rules have had to be used. The topic
is now receiving increased attention from the engineering
research community, and the gap between fundamental
scientic principles and plant construction and operation
is now being bridged. Municipal waste incineration
research carried out at SUWIC is focused on improving
the design of municipal solid waste incinerators so that
future plants can be designed and operated in the most
effective manner possible.
Mixing is the key factor to ensure complete combustion
of solid waste in an incinerator burning bed. Existing grate
systems do not produce a uniform waste distribution,
resulting in channelling of ow through the bed and a
high excess air requirement. A mathematical mixing model,
based on the probability for a tracer particle to swap position with an adjacent tracer on a travelling grate, has been
developed. In parallel with the model development, the
grate mixing process has been quantitatively determined
by a series of systematic particle movement experiments
on scaled models of three of the most common industrial
grates. Comparison between experimental and computed
data has shown good agreement. The grate mixing process
can be quantied and accurately modelled in a systematic
numerical procedure based on `swap probability. The
importance of the probability parameter is that it uniquely
characterizes the `random walk mixing and is essentially
independent of factors such as particle position and grate
length.
A comprehensive model of the burning bed of municipal
solid waste based on the key processes in a waste bed has
been developed in order to provide more accurate data than
presently available on the boundary conditions at the bed
surface. This information is required for use in the CFD
modelling of the gas phase. In the next stage, the mixing
mathematical model will be incorporated into the present
bed model to improve the prediction of the incinerator bed
combustion. The model developed for dynamic simulation
of a xed bed has been expanded for modelling the steady
state combustion process on a travelling grate incinerator.
The present analysis is limited in certain respects and

Figure 22. Flow prole with annular deector.

Trans IChemE, Vol 78, Part B, January 2000

INCINERATOR BED COMBUSTION

Figure 23. Velocity vector plot for ow eld prediction.

improvement to the bed combustion model requires


combustion data acquired from a laboratory-scale xed
bed reactor as well as full-scale incinerator plant. To acquire
high quality plant data, a prototype measuring instrument
is currently being developed. This instrument can be used
to collect the plant data within the burning solid waste bed
in the incinerator.
Another key area to improve the incinerator plant efciency is to prevent the formation of slag. At present, slag
formation around the secondary air jets in the incinerator
poses a major operation problem. SUWICs research also
suggests that the formation of slag around secondary air
jets can be prevented by installing an annular deector
around the jets.
NOMENCLATURE

qo
Qs
ri
ro
(Rp)i
s
Si
Ti
Ts
Tg
To
t
u
ns
nx
V
VL
x
y
Yi
z
l
ls

s
i

Az
C(t,n)
Ds
F2
F3
hi
ho
Hs
j
J
m
M
n
nt
Pl
P(t,n)
qr
q

control volume cross-sectional area normal to z direction


tracer particle in cell n at time t
solid diffusion coefcient
fraction of the waste volume occupied by water which is replaced
by pores during drying
fraction of the waste volume occupied by volatile matter which
is replaced by pores during pyrolysis
capsule core height
capsule external height
solid sensible enthalpy
control volume
jet momentum ux ratio
tracer number
solid mass
cell number
total number of cells
swapping probability
random number generated for cell n at time t
radiative heat ux
conductive heat ux to capsule core

Trans IChemE, Vol 78, Part B, January 2000

viL

Dt

convective heat ux to ablating surface


source term for enthalpy balance equation
capsule core radius
capsule outer radius
rate of removal of component i, where i = 2, 3, 4, 5 or 6
standard deviation
source term for component mass balance equation
temperature of the internal surface
solid temperature
gas temperature
temperature of the external surface
time
uid velocity
solid velocity
moving grate velocity
total volume of solid
volume of solid material L, where L = A, B, C or D
axial direction
transverse direction
mass fraction of component i
direction normal to grate
thermal conductivity of the capsule material
solid effective thermal conductivity
uid density
solid density
density of component i
void fraction of material L
volume fraction of component i in material L
time increment

Subscripts
2
moisture
3
volatile matter
4
bound ash
5
xed carbon
6
free ash
A
dried, pyrolysed and gasied material
B
raw waste material
C
dried material
D
dried and pyrolysed material

31

GOH et al.

32
REFERENCES

1. Spiegel, M. R., 1972, Theory and Problems of Statistics, Shaums


Outline Series in Mathematics (McGraw-Hill Publishing Company)
Chapters 4, 13 and 14.
2. Goh, Y. R., Siddall, R. G., Nasserzadeh, V., Zakaria, R.,
Swithenbank, J., Lawrence, D., Garrod, N. and Jones, B., 1998,
Mathematical modelling of the waste incinerator burning bed, Journal
of the Institute of Energy, 71: 110118.
3. Kuo, T. J. and Essenhigh, R. H., 1970, Combustion and emission
phenomena in incinerators: development of physical and mathematical models of incinerators, Proc 1970 Nat Incinerator Conf
(ASME, New York), 261271.
4. Goel, S., Sarom, A. F. and Lu, J., 1996, A new approach to studying
pore diffusivity during char combustion at FBC conditions, 26th
Symposium (International) on Combustion (The Combustion Institute), 31273155.
5. Glen, N. F. and Howarth, J. H., 1988, Modelling refuse incineration
fouling, 2nd. UK National Conference on Heat Transfer (Glasgow,
IMechE).
6. Langston, L. S. and Boyle, M. T., 1982, New surface-streamline ow
visualisation technique, Journal of Fluid Mechanics, 125: 5357.
7. Ryu, C. K. and Choi, S., 1995, Design Consideration for the Cross Jet

Air Mixing in the Municipal Solid Waste Incinerators HTD 317-2,


American Society of Mechanical Engineers, Heat Transfer Division,
205212.

ACKNOWLEDGEMENT
The authors would like to thank the Engineering and Physical Sciences
Research Council (EPSRC), Kingseld Electronics Ltd., Shefeld Heat
and Power Ltd., Shefeld Municipal Solid Waste Incinerator Plant UK and
ABB Corporate Research (Switzerland) for their nancial and technical
contributions to this project.

ADDRESS
Correspondence concerning this paper should be addressed to
Dr Y. R. Goh, Department of Chemical and Process Engineering,
University of Shefeld, Mappin Street, Shefeld, S1 3JD, UK.
This paper was presented at the 2nd International Symposium on
Incineration and Flue Gas Treatment Technologies, organized by IChemE
and held at the University of Shefeld, UK, 46 July 1999.

Trans IChemE, Vol 78, Part B, January 2000