Upload
voxuyen
View
218
Download
2
Embed Size (px)
Citation preview
PDF Modeling of Turbulent Coal CombustionPDF Modeling of Turbulent Coal Combustion
Michael StöllingerPhD student, Department of Mathematics
International Advanced Coal Technologies Conference June 24, 2010
Outline
Introduction Standard coal combustion model PDF coal combustion model Simulation results Summary and Outlook
Introduction
Goal:
We want to predict the behaviour of coal combustion and gasification devices.
Approach:
Formulate a mathematical model and use numerical methods to solve the equations.
Example: IFRFFurnace for pulverized coal combustion
● Semiindustrial scale: coal feed rate 212 kg/h, 2.1MW
● Simple geometry > ideal as a test case
● Measurements of velocity, temperature, CO2 , O
2
Flame B1
Modeling challenges for coal combustion:
Turbulent reactive flow Large number of chemical reacting compounds Turbulence causes fluctuations of temperature and concentrations Fluctuations drastically change the chem. reaction rates
Turbulent reactive flow + coal particles Heat and momentum exchange between gas and particles Heterogeneous reactions between solid carbon and oxygen Radiative heat transfer becomes very important
Standard Turbulence – chemistry interaction model:
Reduce the number of independent scalars by introducing mixture fractions
f c=mc
mcmair
f c=mc
mvmcmair
Char:(solid carbon) f v=
mv
mvmcm air
Volatiles:
Assume chemical equilibrium
The chemical composition is only a function of the mixture fractions and enthalpy
= f c , f v , h
Mean values and variances of scalars
In a turbulent flow the mixture fraction can be viewed as random variables: to obtain the mean values of the scalars we need the PDF p(f):
=∫0
1
f p f df
Example: one mixture fraction, adiabatic conditions
Mean:
' ' 2=∫
0
1
f −2 p f dfVariance:
A standard Turbulence – chemistry interaction model: continued
Mean and variance of mixture fraction
Assume the shape of the PDF
Example: Gaussian PDF p f =1
2 ' ' 2exp− f −f 2
2 ' ' 2 Parametrize the PDF by the mean and variance of the mixture fraction
Modeled transport equations
are the source terms due to the mass transferred from the coal particles¹Sm f ; ¹Sm ff
Drawbacks of the standard model Shape of the PDF not known
Independence assumptions needed
Gaussian, Beta, doubleDelta .... there is no reason to believe that any one of them is a good approximation
For 2 mixture fractions (char and volatiles) and enthalpy we would need the joint PDF p = p(fc,fv,h)
Instead: p(fc; fv; h) ¼ p(fc)p(fv)±(h¡ ¹h)
¹Á =
Z ZÁ(fc; fv; ¹h)p(fc)p(fv)dfcdfv
Char and volatile mixture fraction are assumed independentEnthalpy fluctuations are assumed independent of the enthalpy level
Can we validate these assumptions?
PDF coal combustion model
Idea: Solve a transport equation for the joint PDF
Approach:
1) Derive a transport equation for gas phase and particle phase PDF~U; ~Á = [fc; fv; h]T are random variables which are governed by
DÁmDt
= £mDUi
Dt=
@Ui
@t+ Uj
@Ui
@xj= Ai
Fine grained PDF: g(~V ; ~Ã; ~x; t) = ±(~U(~x; t)¡ ~V ) ¢ ±(~Á(~x; t)¡ ~Ã)
Using the chain rule:
@g
@t= ¡ @g
@Vi¢ @Ui
@t¡ @g
@Ãm¢ @Ám
@t
@g
@xi= ¡ @g
@Vj¢ @Ui
@xj¡ @g
@Ãm¢ @Ám@xi
½(~Ã)@g
@t+ ½(~Ã)Vi
@g
@xi= ¡ @
@Vj
µ½
µ@Uj
@t+ Ui
@Uj
@xi
¶g
¶¡ @
@Ãm
µ½
µ@Ám@t
+ Ui@Ám@xi
¶g
¶
Using the definition of the joint velocityscalar PDF p(~V ; ~Ã) = hg(~V ; ~Ã)i
½(~Ã)@g
@t+ ½(~Ã)Vi
@g
@xi= ¡ @
@Vj(½Ajg) ¡
@
@Ãm(½£m g)
and the definition for the conditional average for any function
we arrive at the transport equation for the PDF:
½(~Ã)@p
@t+ ½(~Ã)Vi
@p
@xi= ¡ @
@Vj
³½hAjj~V ; ~Ãip
´¡ @
@Ãm
³½h£m j~V ; ~Ãip
´
This equation is exact, but not closed!
2) Introduce closure models
hAjj~V ; ~Ãi
h£m j~V ; ~Ãi
Langevin model
IEM – mixing model
hAjj~V ; ~Ãi ¢ p(~v; ~Ã; ~x; t) = hAjg(~V ; ~Ã;~x; t)i
3) Monte Carlo methodDimension of PDFequation: velocity 3 + scalar 3 + space 3 + time 1 = 10
Conventional methods (like FV) not efficient
Solution through Monte Carlo method: equivalent system of stochastic ODE's
dXi(t) = Ui(t)dt
Fluid particles:
dfm = ¡1
2!¡fm ¡ fm
¢dt + _Mm dt m = c; v
dh = ¡1
2!¡h¡ h
¢dt + ( _Qs + _Qrad)dt
dUi(t) = ¡1½
@hP i@xi
+ Gij¡Uj(t)¡ Uj)dt
¢+ C0"dWi(t) + S1dt
Coal particles:dXp
i (t) = Upi (t)dt
dUpi (t) =
Usi¡ Upi
¿pdt¡ 1
½p
@hP i@xi
+ gidt
dUsi(t) = GsijUsj(t)dt + Cidt + BsijdWj(t)
dmp = ( _mchar + _mvol) dt
mpcppdTp = Nu¼¸g(Ts¡ T p)dt + _Qp
rad
4) Means values
¹T =
PN p
wpT p
PN p
wp
5) Radiation model
The presence of particles increases absorption and emission drastically
We use the discrete transfer method (DTM) to solve the radiative heat transfer equation
dI
ds= ¡(·g + ·p + ¾p)I + ·gIb; g + ·pIbp +
¾p4¼
Z 4pi
0Id
Simulation results: flame B1● Simple first order models for devolatilization and char reaction● 15 size classes to represent coal diameter distribution approx. 150 parcels per cell→● 50 gas particles per cell● 2d axisymmetric domain with 204 axial x 54 radial cells● Simulation time 30h on a single core of a Q6600 Core2 Quad
Mean axial velocity
Summary and Outlook
✔ Development of a transported PDF coal combustion model
✔ Application to semiindustrial scale furnace✔ Initial simulation results are encouraging Need to adjust devolatilization model to improve
results Analysis of the standard model assumptions Application to flameless combustion