You are on page 1of 10

Twenty-first Symposium (International) on Combustion/The Combustion Institute, 1986/pp.

265-274

THREE-DIMENSIONAL FURNACE COMPUTER MODELLING

R.K. BOYD
Electricity Commission of New South Wales

J.n. KENT
Department of Mechanical Engineering
Univer~i(~' o/Sydney
N.S.W. 2006
Australia

A fully three-dimensional computer model of a pulverized fuel, tangentially fired furnace is


presented. The model predicts gas flows, species concentrations and temperatures, particle
trajectories and combustion and radiation heat fluxes. The gas phase conservation equations of
momentum, enthalpy and mixture fraction are solved utilizing the k-{ turbulence model. A
Lagrangian approach is used to predict particle trajectories. Particle dispersion in the turbulent
flow is modelled by a stochastic method. Coal devolatilization and char burnout are modelled
and radiative heat transfer is handled by a discrete transfer method.
The predictions are compared with data obtained in an operating 500 MW(e) furnace. Input
conditions for the computations are the actual coal feed rate, particle size distribution, specific
energy and chemical analysis. The predictions are compared with furnace measurements of
temperatures, 02 concentrations, wall heat fluxes and carbon burnout. Agreement between the
predictions and data is very satisfactory.

1. Introduction tion a n d heat t r a n s f e r a n d includes particle


dispersion in the t u r b u l e n t gas flow.
Many of the boilers c u r r e n t l y being operated T h e f u r n a c e m o d e l l e d is one of the ECNSW
by the Electricity C o m m i s s i o n of NSW Liddell Power Station 500 MW(e) boilers. T h e
(ECNSW) are tangentially pulverised coal fired boiler is a tower type tangentially fired design
designs a n d hence have flows which are t u r b u - having two furnaces s h a r i n g a c o m m o n central
lent, recirculating a n d fully three-dimensional. water wall. Each f u r n a c e has dimensions 10.69
T e s t i n g of these boilers is extremely expensive; • 8.86 x 31.0 m a n d is fired from four vertical
a n d high costs also make extensive plant modi- banks of 8 b u r n e r s a n g l e d at 28 ~ a n d 40 ~ to the
fications prohibitive. T h u s a n analytical f u r n a c e vertical wall n o r m a l a n d with - 1 6 ~ vertical tilt.
m o d e l allowing the p e r f o r m a n c e of new coal T h e exact f u r n a c e g e o m e t r y is described more
seams to be evaluated a n d e n a b l i n g prospective fully elsewhere a3'14.
f u r n a c e designs to be tested, is invaluable from T h e model uses actual f u r n a c e o p e r a t i n g
design a n d operational viewpoints. conditions a n d coal properties. Predictions are
Over the last two decades the e n o r m o u s c o m p a r e d with m e a s u r e m e n t s o b t a i n e d from
growth in available c o m p u t e r power has fos- the o p e r a t i n g furnace. T h e s e include oxygen
tered the d e v e l o p m e n t of n u m e r i c a l flow, com- concentrations, t e m p e r a t u r e distributions, wall
bustion a n d heat transfer models. Early models heat fluxes a n d coal particle b u r n o u t at the exit.
focussed u p o n gaseous fuels being b u r n e d in
laboratory scale c o m b u s t o r s a n d have b e e n
reviewed previously t-3. Pulverised coal combus- 2. Physical Models
tion models c o n c e n t r a t e d u p o n the chemistry
using assumed flow fields 4. Later work led to 2.1 Gas Flow
o n e - d i m e n s i o n a l or t w o - d i m e n s i o n a l axisym-
metric models of laboratory test furnaces ~'6'7. T h e gas flow is described by th~ time m e a n
Recently, models of industrial-scale, pulverised, conservation equations of global mass, m o m e n -
coal fired furnaces have a p p e a r e d 8-12. T h e tum, enthalpy a n d species mass fractions. T u r -
p r e s e n t work extends previous models by de- b u l e n t diffusivity is h a n d l e d by an effective
scribing fully t h r e e - d i m e n s i o n a l flow, combus- viscosity o b t a i n e d f r o m the well k n o w n k-e

265
266 COAL COMBUSTION

model and inolves two additional equations. All combustion rates require the particle trajecto-
equations are coupled by the gas density which ries are accurately modelled. Particle sizes in
is a function o f the combustion heat release and the range 10-200 Ixm are encountered and
radiation losses. particle-gas phase slippage cannot be neglected
The gas phase conservation equations in particularly in regions of high velocity gradi-
three dimensions can be cast in the following ents and recirculation.
form: The Lagrangian approach, whereby each
particle size class is tracked through the gas
field is used here. During its flight the particle
(1) interacts with the gas flow field j~roducing
Oxi t, ~ Ox, J
source terms (particle source in cell IJ method)
in the m o m e n t u m , enthalpy and mixture frac-
Here, r represents any of the above variables, tion equations as shown in Table 1. Particle
$6 is the source or sink term in the gas phase mass and t e m p e r a t u r e are also changing and
and Sp~ is the term due to the particle phase. these properties are continuously updated.
The source terms and exchange coefficients The particle m o m e n t u m equation is,
used are shown in Table 1.
T h e effective turbulent viscosity is IX: = dut, i 1
Cv,Pk2/~ + P4. W h e r e C~< = 0.09. Gas phase mean mt,---~ = ~C~p(ui--ul,,)2+ml, g , (2)
density is obtained from the perfect gas equa-
tion using t e m p e r a t u r e derived from enthalpy The drag coefficient CD is a function of Reyn-
and molecular weight from the mixture frac- olds number.
tion equation (described later). If the gas velocity used in the particle
m o m e n t u m equation is the mean velocity then
2.2 Particle Dynamics all particles o f a given size and starting location
The heterogeneous nature of pulverised coal will follow identical paths through the furnace.
combustion and the relatively slow particle Particles disperse because of the turbulent gas

TABLE I
Source Terms and Exchange Coefficients for Gas Phase Conservation Equations

6 F, S, Sp,

dmp
1 0 0
dt

dP d
u, t&* - - - - + pg,
dx i dt (mpuip)

k p,/0.9 G~- pe 0

9 Ixj1.22 (1.44Gk--l.92pe)- ~ 0

dmp
f ~x,/O.9 0 dt

d
h t,/0.9 - Qn --~t tmp apl
* Additional stress terms which occur in the momentum equations are for numerical
convenience placed into Su,.

1 /0ui a~j\/0ui a~,'~


THREE DIMENSIONAL FURNACE MODELING 267

flow and hence their trajectories can be more the corresponding sink term in Table I is the
realistically modelled from a distribution of change in particle enthalpy.
instantaneous gas velocities. T h e coal particle is initially composed of raw
T h e model 16 developed here solves the La- coal and ash. Devolatilization and combustion
grangian equation of motion for a statistically of the volatiles takes place in the gas phase
significant sample of individual particles as they through rapid heating of the particle. The
encounter a r a n d o m distribution of turbulent remaining char particle subsequently burns out
eddies. T h e velocity of each eddy is found by to ash at a chemical or diffusion controlled rate.
making a r a n d o m selection from the probability T h e particle devolatilization is modelled by
density function of velocity. This is assumed to considering a pair of parallel, first o r d e r
be an isotropic Gaussian probability density reactions 17.
function having standard deviations (2k/3) ~
about the mean velocity. T h e integration of the I. (RAW COAL) ~1 (1 - Y~) (CHAR)
particle equation of motion is p e r f o r m e d over a + Y1 (VOLATILES) (7a)
time step which is taken to be the lesser of the
eddy lifetime and a characteristic time taken for II. (RAW COAL) ~2 (1 - Y2) (CHAR)
a particle to traverse an eddy. These times are + Y2 (VOLATILES) (7b)
estimated assuming the characteristic eddy size
is the dissipation length scale16; These two reactions are competing, reaction I
being slow and reaction II fast. T h e volatiles
L~ = C 0"75
b~ kl5/e (3) mass yield Y1 for unit mass of raw coal is
assumed to be equal to the proximate volatile
T h e characteristic time taken to traverse an analysis. T h e yield for reaction II, Y2, is
eddy ~ is estimated as assumed Is to be 2Y> T h e rate coefficients '}1 and
k2 are given by A r r h e n i u s expressions of the
tt = L j u p (4) form k = A exp( - E/RTp). T h e constants A l, A2
and activation energies El, E217 are shown in
where up is the particle velocity entering the Table II.
eddy. This model ensures that the actual volatile
The eddy lifetime is 16 yield is a function of the time-temperature
history of the particle. Following devolatiliza-
te = Le/(2k/3) ~ (5) tion, char b u r n o u t is controlled by the rate of
oxygen diffusion to the particle and the rate of
2.4 Particle Combustion chemical reaction at the surface. T h e particle
mass change due to char b u r n o u t is given by4:
Particle devolatization and char b u r n o u t
rates are strongly t e m p e r a t u r e dependent. T h e
particle energy balance is dmp = --~k,P#~Xo2 (8)
dt
d dmp
dt ( m p h p ) = h p - - - ~ + Q p c + Q p ~ + Q t , b (6) where the overall rate is given by the combina-
tion of chemical and diffusion rate terms:
Radiative transfer between the particle and the
gas @, has been neglected here. Convected heat k, = 1/(1/kd + 1/kc) (9)
transfer 0_~ = ha t, (T - Tp) is obtained for a
sphere moving at the slip velocity t h r o u g h the T h e chemical rate is expressed in Arrhenius
gas. form as, kc = Ac exp ( - Ec/RTp) with Ac and Ec
During char b u r n o u t part of the heat release given in Table 24 .
may be directly transferred to the particle
rather than the gas phase. T h e term xq in 0.~b =
- - Xq q~
2.5 Gas Phase M i x i n g and Combustion

dmp The previous section specified the particle


Qt, b = --Xqq~ dt mass loss rate to the gas phase by devolatiliza-
tion and char burnout. These processes pro-
represents the fraction of the particle heat of duce gas phase fuels which provide source
combustion which is transferred to the particle. terms in the gas phase species and enthalpy
Here 4 xq = 0.3. Since all the particle standard- equations.
ized enthalpy change, including combustion, T h e overall complete combustion reaction
takes place by interaction with the gas phase, for the raw coal is,
268 COAL COMBUSTION

TABLE II having the composition of the raw coal. (Water


Operating Conditions is considered separated d u r i n g coal preheating
prior to furnace entry.) The rate of production
ULTIMATE COAL ANALYSIS Mass % of this gaseous fuel is determined by the
Carbon 53.86 relations in 2.4.
Hydrogen 3.64 Chemical reactions in the gaseous phase are
Nitrogen 1.36 considered to occur rapidly and are limited by
Sulphur 0.88 the mixing rate of fuel and oxidant. This allows
Oxygen 6.66 the well known mixture fraction approach to be
HzO 7.40 used. The mixture fraction f is the mass
Ash 26.20 fraction of the gaseous fuel atoms and for
complete combustion all species mass fractions
Specific Energy (dry ash free) 32.83 MJ/kg
and molecular weight can be algebraically re-
COAL COMBUSTION PARAMETERS lated tof. Thus, for pure f u e l f = 1, in air f = 0
Devolatilization: A1 1.3 x 105 s-1 and where f has the stoichiometric value the
A2 1.5 x 1013 s-1 composition is given by the products in Eq. (i0).
Et 74 x 103 KJ/mol The time mean mixture fraction in the furnace
E2 251 x 103 KJ/mol is obtained by the solution of t h e f c o n s e r v a t i o n
Yl 0.27 equation with the source term in Table 2.
Y2 0.54 The molecular weight from f together with
Char Combustion: Ac 0.8596 kg/(s - m2)/ temperature from enthalpy allow the time
(N/m 2) mean gas density to be evaluated. No fluctua-
Ec 1.49 x 105 kJl(kg-mol) tions in f (the "g" equation) have been con-
sidered. The distributed source terms mean
RADIATION PARAMETERS that f gradients are much smaller than for
Ks 0.13m -1 b u r n i n g fuel jets where turbulent fluctuations
% 0.67 can affect the computed time mean variables.
T~ 660 K
BOILER OPERATING CONDITIONS 2.6 Energy Transport
Burner tilt - 16~
Total Coal Flow 30.9 kg/s Standardized enthalpies are used for all
Primary Airflow 56.0 kg/s species including the coal particles. T h e gas
Secondary Airflow 224.0 kg/s phase enthalpy equation contains only two
Primary Air Temperature 353 K source terms; the cell radiation and that due to
Secondary Air Temperature 533 K the enthalpy of fuel transferred from the solid
Excess Air 15.5% to the gas phase. T h e enthalpy together with
the species mass fractions from f allow the
determination of gas temperature.
Radiative heat transfer is handled by the
discrete transfer method of Lockwood and
Shah 19. This method embodies features of the
zone, flux and Monte-Carlo methods and has
m
--' vSO2 + ~ - H 2 0 + nCO 2 been found to yield good results and be
computationally efficient. The furnace is di-
vided into 6 x 6 x 17 cartesian cells, each
3 rn P having u n i f o r m temperature and absorption
+ .76(n+-~-~-+r)N2 (10)
coefficient. A discrete n u m b e r of rays are
tracked to the centre of each wall cell from the
The reaction occurs by gas phase combustion surounding walls. As each ray encounters a cell
of the volatiles and char burnout. The char is at temperature T and absorption coefficient K
considered to be pure carbon and the gas phase the change in intensity of the ray is given by:
fuels thus have differing compositions, heat of
reaction and oxygen requirements. However,
these differences are not large when the com- dI KoT 4
d--~= - K I + --~r (11)
plete conversion of char to CO2 is considered.
(The char surface reaction is generally con-
sidered to produce CO which rapidly oxidises Hence, the change in intensity of any ray
to CO2 in the gas phase.) over a cell is f o u n d by the integration of
A simplified approach is, therefore, to treat equation (11). T h e net absorption or emission
all fuel sources in the gas phase as one gas of energy Qm from a cell is the summation of
THREE DIMENSIONAL FURNACE MODELING 269

the change in intensity o f each ray passing


through that cell.
T h e treatment used here assumes the com-
bined particle and gas absorption coefficient in B
B
each cell is 2~
t / I p
~ I I,
K(m -1) = 0.32 + 0,28 exp( - 7"/1135)(12) I I ~ - . ~, l l 1
and in the b u r n e r region considering soot the
absorption coefficient increases by 0.4 m -1. ~

Isotropic radiative scatter has been included


I ~ ~ ~ . ~ , . o a

with an assumed scatter coefficient Ks = 0.13


m -1"

3, N u m e r i c a l Solution
T h e finite difference T E A C H - 3 E code 21 de-
# / / / / / ! !
veloped at Imperial College, London provides
the basis for the solution o f the quasi-linearized
elliptic conservation equations. The computa- Ill//Ill
tions are carried out on a 12 x 10 x 31 node
cartesian mesh with a fine node spacing in the
l I / / i / / !
b u r n e r regions. Each inlet p o r t is modelled by
one finite difference cell with a p p r o p r i a t e / l / / I l l /
b o u n d a r y conditions. For the results presented ,I i i I I I I /I
here, only the lower 7 out o f 8 b u r n e r levels are

""331~"II1
rl t i i i I I II
in operation with secondary air supplied from
the s u r r o u n d i n g 8 levels.
... .. Cf
T h e Lagrangian particle trajectory equations
are solved for a representative number of
particles. Four particle injection locations
F
A
.f/
77
l't'/
~.

within each of 28 burners are used along with 5


particle size groups. T h e particle size distribu- I'1"/
tion used is 10% by mass 51xm, 20% 15 txm, 40%
50txm, 20% 1001xm and 10% 170 Ixm. Five
identical particles are tracked for turbulent
dispersion. Thus a total o f 2800 particle trajec-
tories are c o m p u t e d after each gas phase
iteration. \....__ __ _ 71 i'.l
T h e computations are started using a par-
tially converged solution for heated inlet
streams but without trajectories or combustion.
T h e Lagrangian particle tracking code is then
r u n after each flow code iteration and a partialy
converged non-reacting two phase solution is
obtained. Finally, the particles are forced to
devolatilize until t e m p e r a t u r e s are high enough
to sustain reaction and the radiation routine is
r u n every 10 iterations until the system is fully scare: 20m/s
converged. This p r o c e d u r e is followed to en-
sure stability. Overall convergence is obtained FIG. 1. Predicted gas flow velocity vectors in the
in approximately 600 iterations, which takes furnace. Top: Plan view Section A-A at 19.1m
about 15 hours CPU time on a VAX 8600 elevation. Bottom: Elevation at section B-B through
computer, burner centreline Overall furnace dimensions are:
8.86m x 10.69m x 31.0m
4. Results
A testing p r o g r a m 22 on one of the Liddell First, some c o m p u t e d general gas flow pat-
Power Station boilers provided data which is terns are shown in Fig. 1. T h e horizontal
used for comparison with the present predic- section is t h r o u g h the level 1.2m above the top
tions. T h e computation input conditions in b u r n e r (19m elevation) and the vertical section
Table 2 are the boiler o p e r a t i n g conditions. is t h r o u g h the b u r n e r centreline. The clockwise
270 COAL COMBUSTION

swirling flow clearly seen in Fig. 1 is induced by with the predicted and measured values coin-
the higher tangential m o m e n t u m c o m p o n e n t ciding at 4%.
of the 40 ~ burners. This swirling flow pattern is Mean particle trajectories (without disper-
still present at the furnace exit, although the sion) are shown in Fig. 3. T h e trajectories are
tangential velocities become much lower. En-
trainment of gas from the corners adjacent to
the 28 ~ burners causes recirculation zones in
these two corners a n d the 40 ~ b u r n e r jets are
forced against the walls.
The vertical section shows several points of
interest. A stagnation region where the two
opposing jets meet can be seen about 3 metres
from the 40 ~ b u r n e r . The diminishing swirl in
the downstream direction can be seen from the
largely horizontal flow from the b u r n e r s be-
coming almost vertical at the furnace exit. \ \ \ \l
Velocity predictions for an isothermal model
with the same characteristics have been found
to be in good agreement with experiment.
Figure 2 shows predicted and measured O2
volume fraction contours at 1.2m above the top
b u r n e r (19m elevation). The predicted and \\ ~% I
measured profiles show considerable similarity '\ \ \\ ';
except for the region near the side wall. T h e
difference between the measured and pre-
dicted O2 concentrations here can be attributed
X~ ~I u
to measurement error. As furnace pressure is 9 ~1 I] I111
lower than ambient, regions near the wall are
subject to air leakage through the m e a s u r e m e n t
ports, resulting in high 02 values. Elsewhere,
the agreement is good as can be seen by the
similarities in the 4%, 6% and 8% contours.
At 26m elevation near the furnace outlet, the
measured and predicted oxygen concentration
profiles are very flat. Agreement is excellent,
,?sZ E
J
'v2dJ,
j 6
~ ; 1 t/
I0 +

(b)
FIG. 3. Predicted mean particle trajectories (with-
out turbulent dispersion) from one 40~burner bank.
Plan and one elevation view through whole furnace.
Fic. 2. Oxygen volume fractions at 1.2m above top warm up and devolatilization; --- char burn-
burner (19m elevation). (a) Predictions (b) ECNSW out. End of trajectory signifies complete burnout of
measurements. + measurements location. char with ash remaining.
THREE DIMENSIONAL FURNACE MODELING 271

only for particles being injected from one set of


40 ~ burners. T h e mean trajectories show that
considerable particle dispersion occurs because
of differing starting locations and varying
particle sizes. A quite small initial displacement
may lead to a markedly different trajectory.
T h e dispersion increases considerably using the
turbulence interaction model described which
was used for all the predictions.
T h e particle warm up, devolatilization and
char b u r n o u t regions are shown. Clearly, the
devolatilization occurs very rapidly and within 2
m of the b u r n e r exit. Figure 3 also shows that
most of the particles b u r n o u t completely before
reaching the furnace exit. T h e predicted over-
all carbon b u r n o u t is 96% o f the initial mass of
the combustible matter which compares with
the experimentally obtained value of 97%.
T h e predicted gas t e m p e r a t u r e s and those
obtained from suction p y r o m e t e r measure-
ments are presented in Fig. 4 for the level 1.2m
above the top b u r n e r (19m elevation). Tem-
peratures are similar except for a predicted hot
region of 2000 K. T h e reason for the high
prediction is not clear at this stage, but may be
due to too rapid devolatilization or inadequate
particle radiation losses.
At the furnace outlet the predicted and
measured t e m p e r a t u r e profiles are both rela-
tively flat. T h e average m e a s u r e d t e m p e r a t u r e
is 1473K and the average predicted tempera-
ture is 1560K, showing that the overall heat Fro. 5. Wall incident heat flux (kW/m2). (a) Predic-
transfer predictions for the furnace are good. tions (b) ECNSW measurements. + measurement
F u r t h e r detail of the heat transfer predic- location.
tions can be seen from the comparison of
measured and predicted incident heat flux to a
side wall in Fig. 5, T h e agreement here is very
satisfactory. T h e only area in which the predic-
tions significantly differ to the measurements is
1800 near the furnace exit, where predictions are
slightly lower than m e a s u r e d fluxes. T h e re-
gion of maximum heat flux which is important
to furnace design is predicted quite well.

Conclusion

~/ ' lz ~ ~`~1I ~176 A fully three-dimensional flow, combustion


and radiation model applied to power station
furnace conditions yields very encouraging
predictions of measured quantities. Predicted
+ 1L,O0
,~_ temperatures, oxygen concentrations, heat
fluxes and carbon b u r n o u t are in generally
,o, Iol good agreement with the available data. The
78"I gas velocities and particle trajectories show
where ash loadings are high and where erosion
problems may occur in the furnace.
Fro. 4. Temperatures (~ at 1.2 m above top T h e r e are many areas o f further investiga-
burner (19m elevation). (a) Predictions (b) ECNSW tion p r o m p t e d by the results obtained. The
measurements. + measurement location. effects of numerical grid size, radiation pa-
272 COAL COMBUSTION

rameters, devolatilization rates, sensitivity to Acknowledgement


particle size distribution and particle dispersion
are a few which may lead to more accurate and The authors are grateful to Dr. F.C. Lockwood of
efficient computations. Sub-models and pa- Imperial College, London for his assistance in em-
rameters such as those used for devolatilization ploying the basic TEACH code and supplying the
and b u r n o u t may be easily changed allowing radiation code. The support and guidance of Dr. A.
the sensitivity o f these factors, or coal types, Lowe of the Electricity Commission of NSW is
upon such problems as b u r n e r flame stability to gratefully acknowledged.
be investigated. Research in these areas is This project was supported by grants from the
currently being undertaken. Electrical Research Board, and has been sponsored
by the Electricity Commission of NSW.

Nomenclature
REFERENCES
Ap particle surface area (m 2)
CD particle d r a g coefficient 1. McDoNALD,H: Prog. Energy Combust. Sci., 5, 97
dp particle d i a m e t e r (m) (1979).
E activation energy (kJ/kmol) 2. KHALIL, E.E.: Modelling of Furnaces and Comb~r
f mixture fraction tors, Abacus (1982).
G, turbulence generation term (J/(m3-s)) 3. GREEN, A.S., WHITELAW,J.H.: J. Fluid Mech.,
gi gravitational acceleration - 9 . 8 1 m / s 2 for 126, 2 (1983).
vertical direction, otherwise zero 4. BAUM, M.M., STREET, P.J.: Comb. Sci. Tech., 3,
h specific enthalpy (J/kg) 231, (1971).
I radiation intensity (W/m ~) 5. SMITH, P.J., FLETCHER,T.H., SMOOT,L.D.: Eigh-
K gas radiation absorption coefficient (m -1) teenth Symposium (International) on Combustion, p.
k turbulence kinetic energy (m2/s 2) 1285, The Combustion Institute, 1980.
Ks particle radiation scattering coefficient 6. LOCKWOOD,F.C., RIzvL S.M.A., LEE, G.K., WHA-
(m -i) LEY,H. : Twentieth Symposium (International) on Com-
mp particle mass (kg) bustion, p. 513, The Combustion Institute, 1984.
P pressure (N/m 2) 7. LOWE, A., WALL, T.F., STEWART, I.M.: Fifteenth
Q4b heat transfer rate to particle due to com- Symposium (International) on Combustion, p. 1261,
bustion (W) The Combustion Institute, 1974.
Qpc convective heat transfer rate to particle 8. LocKwooo, F.C., SALOOJA, A.P., SYED, S.A.:
(w) Comb. and Flame, 38, 1, (1980).
Q4r radiative heat transfer rate to particle 9. LOCKWOOD, F.C., RIZVI, S.M.A., SHAH, N.G.:
from gas (W) Some Predictive Experiences of Coal Firing, Mech.
QR cell net radiative power gain (W/m 3) Eng., Imperial College of Science and Technol-
qc char enthalpy o f combustion (J/kg) ogy Report FS/83/37, (1983).
s distance coordinate (m) 10. BENESCH, W., KREMER, H.: Twentieth Symposium
T T e m p e r a t u r e (gas) (K) (International) on Combustion, p. 549, The Com-
t time (s) bustion Institute, 1984.
ui velocity c o m p o n e n t (m/s) 11. TRUELOVE, J.S.: Twentieth Symposium (Interna-
X mole fraction tional) on Combustion, p. 523, The Combustion
xi length coordinate (m) Institute, 1984.
F effective t u r b u l e n t diffusivity (kg/(m-s)) 12. SMITH, P.J., SMOOT, L.D.: Coal Combustion and
turbulence dissipation (m2/s 3) Gasification, 1985.
r wall radiation emissivity 13. BOYD, R.K., KENT, J.H.: Three Dimensional Gas
viscosity ( k g / ( m - s ) ) Flow Modelling of Boiler Furnace, Eighth Australa-
p density (kg/m 3) sian Fluid Mechanics Conference on Fluid Me-
chanics, 1B14-17, 1983.
14. BOVD, R.K., ROSCAREL,K.J. AND KENT,J.H.: Gas
Subscripts Flow and Mixing in a Tangentially Fired Furnace,
Third Australasian Conference on Heat and
c chemical Mass Transfer, May 1985, Melbourne.
d diffusion 15. MIGDAL, D., AGOSTA, V.D.: ASME, Applied
laminar Mech., p. 860 (1967).
p particle 16. SHUEN,J.S., CHEN, L.D., FAETH,G.M.: AIChEJ.,
t turbulent 29, No. 1, January 1983.
w wall 17. UBHAYAKAR, S.K., STICKLER, D.B, VON
T H R E E D I M E N S I O N A L FURNACE MODELING 273

ROSENBERG JR., C.W., GANXON, R.E.: Sixteenth 21. CARVALHO, M.G.M.S., LOCKWOOD, F.C.: Predic-
Symposium (International) on Combustion p. 427, T h e tion of the Combustion of an End Wall Fired Glass
Combustion Institute, (1977). Furnace, hnperial College, Dept. of" Mech. Eng,
18. JAMALUtmlN, A.S., TRUELOVE, J.S., WALt, T.F.: Internal Report, 1981.
Combustion and Flame 62, 1, 85 (1985). 22. LOWE, A.: The Instrumentation of a Large Utility
19. LOCKWOOD, F.C., SHAH, N.G.: Eighteenth Sympo- Boiler for the Determination of the Effects of Mineral
sium (International) on Combustion, p. 1045, T h e Matter on Heat Transfer, Conference on Slagging
Combustion Institute, 1981. and Fouling Due to Impurities in Combustion
20. GIBB, J., JOYXER, P.L.: LiddeU Boiler Reheat Metal Gases, C o p p e r Mountain, Colorado, Aug. 1984.
Temperature: Gas Side Heat Transfer Analysis, Cen-
tral Electricity Generating Board, Res. Div. R/M/
N1029 (1979).

COMMENTS

F.C. Lockwood, Imperial College, Great Britain. Con- 3. Have you ever used p d f to calculate your
cerning your stochastic particle/turbulence interac- reaction rate in turbulent flow?
tion model, I have three questions:
1) You have stated that you select the gas velocity Author's Replr.
direction randomly, do you also select the velocity 1. T h e physical basis of the assumption of 0.3 o f
magnitude randomly about 2~K1'27 the particle heat release to the solid phase and the
2) All o f this computational effort plus the need to r e m a i n d e r to the gas phase is that the C+(1/2)O2
sample large n u m b e r s o f particles in each size group CO reaction occurs at the particle surtace and the CO
in o r d e r to ensure statistical accuracy suggests ex- + (1/2)O2 ~ CO2 reaction occurs awa} from the
treme c o m p u t e r time will characterize your particle particle. T h e factor o f 0.3 is based upon the relative
dispersion model? Is this indeed the case? heats o f combustion o f the above two reactions.
3) What influence on i m p o r t a n t predictable quan- 2. As described in the text, the mass of volatiles
tities do you observe which are due to this model? released appears as a source in the mixture fraction
equation.
Author's Reply. 3. No concentration fluctuation model is included
1. As stated in the text, the velocity o f each at present, as it is believed that it will make little
turbulent eddy is estimated by making a r a n d o m difference to the results.
selection from the probability density function o f
velocity which is assumed to be isotropic Gaussian
having standard deviations (g~)l/2.
2. T h e stochastic dispersion model is computation- W. Zinser, Univ. of Stuttgart, W. Germany. In your
ally expensive, as you point out. However, only model you are proposing to employ the pyrolysis
between 10 and 20 trajectories for each particle size model o f Kobayaskhi, et. al. in o r d e r to predict the
and starting location are required for the dispersion devolatilization rate o f coal particles in the furnace.
treatment. W h e n more trajectories than this are One o f the aims o f this model is to give a guess about
considered for the stochastic dispersion treatment, the yield o f volatiles d u r i n g fast pyrolysis by consider-
negligible change in the predicted temperature, ing two competing reactions with low and high
species and heat flux occurs. activation energy, respectively. This again requires a
3. With the inclusion o f particle dispersion, the rather accurate prediction o f the temperature history
t e m p e r a t u r e and oxygen concentration fields showed o f the coal particles d u r i n g devolatilization. Now, in
lower gradients as would be expected with more the coarse grid you are required to use for compu-
dispersed fuel sources. tional reasons, devolatilization takes place within very
few gas phase control volumes near each burner,
where you have to assume constant gas temperature
within each volume. Should this not give a rather
Zhou Lixing, Univ. of Beijing, China. erroneous picture o f the volatiles yield in the fur-
1. What is the physical basis for your assumption o f nace? As your fast gas reaction scheme will probably
energy distribution between particle and gas phases? overpredict t e m p e r a t u r e s in this region, I would
2. Have you accounted for the source term due to expect a c o r r e s p o n d i n g strong overprediction of the
particle devolatilization and combustion in the mix- volatiles yield. On the o t h e r hand, you report satisfac-
ture fraction equation? tory predictions o f heat fluxes to the walls and o f char
274 COAL COMBUSTION

burn-out. Should this not indicate that the devolatil- 2) Regarding the modeling of the heating value of
ization model has no strong impact as the prediction the volatile matter (which you have taken as 70% of
of furnace performance? that of the coal)---could you obtain a better estimate
by taking the heating value of char as carbon and the
Author's Reply. Particle heatup is taking place char yield as calculated by your model?
during transit through about two (2) cells at the
burner. During this time, the particle temperature is Author's Reply.
rising asymptotically toward the cell gas temperature. 1) To achieve the correct furnace outlet tempera-
It is worthwhile to consider refining the grid system ture, a wall emissivity of approximately 0.8 is
in this area, but it is not likely to affect the total heat required.
release in the volatiles phase significantly, as heatup 2) The volatile matter and char are assumed to
rates are, in any case, quite rapid. have identical heating values. This assumption will
result in negligible error for the coal considered, as
the dry ash free specific energy of the coal of 32
MJ/Kg is equal to the specific energy of pure carbon.
N. Arai, Nagoya Univ., Japan. I understand that this
kind of modelling is very difficult. What do you think
are the main causes of the discrepancies among the
predictions and experiments, especially for tempera- C. Lawn, Marchwood Eng. Labs., Great Britain. I don't
ture profiles? I think the effect of scattering for energy think you should worry unduly about the discrepancy
treatment is really important in such calculations. between your predicted furnace temperatures and
the measured values. Measurements with suction
Author's Reply. The region of main temperature
pyrometers in furnaces are notoriously difficult, but
discrepancy occurs near the burners, where turbu-
more substantially, I doubt whether the measured
lence is high and gradients are steepest. Fluid
wall heat fluxes are consistent with those tempera-
mechanics, turbulence and finite difference grid
tures, taking any credible value of wall emissivity.
refinement may affect the profiles here. Also, there is
We did some radiation calculations for the very
some uncertainty in the experimental data. The
same furnace some years ago and our results for gas
predicted heat flame agreement with data seems to
temperature were similar to the ones you predict, and
indicate better temperature correspondence than
indeed, to those calculated for a variety of other
shown. Scattering is already included in the radiation
furnaces of this scale.
model and is important with these high ash coals.

N. Shah, Combustion Dept. Rolls Royce, Great Britain. K. GOrner, Univ. of Stuttgart, W. Germany. What is
In answer to a question from a Japanese participant, the velocity inlet condition used in your calculation? I
the scattering in the model is isotropic. Secondly, wonder why you have not found any recirculations
rather than change wall emissivitives, one should region in the funnel shaped bottom and above the
examine the effects of absorption coefficients both upper burner plane. I did some furnace calculations
for gas and scattering. and found some very strong recirculations because of
the normally high momentum input through the
Author's Reply. The absorption coefficients for gas burners.
and particles and the particle scatter coefficients are
presently receiving attention. Preliminary calcula- Author's Reply. The modelled inlet Primary Air and
tions using coefficients calculated from the local gas pulverised fuel velocities are equal to those found in
and particle properties resulted in improved incident the actual furnace--although a flat velocity profile is
heat flux correlations and closer predicted and imposed in the model. Strong recirculation zones are
measured furnace outlet temperatures. present between the 28 ~ burners and the side walls. A
recircu[ation zone with low gas velocities is present in
the region below the burners.

T. Wall, Univ. of Newcastle, Australia.


1) You have fixed the wall emissivity at 0.67, what
should it be to match predicted total heat pick-up?

You might also like