CFD8 Computer Fluid Dynamics 2181106

E181107

Combustion,
multiphase flows

Remark: foils with „black background“

could be skipped, they are aimed to the
more advanced courses

Rudolf Žitný, Ústav procesní a

zpracovatelské techniky ČVUT FS 2010
CFD8 COMBUSTION
Use EBU (Eddy
Breakup models)

Laminar
Premixed (only one flame
inlet stream of mixed
Homogeneous fuel and oxidiser) Turbulent
reaction in gases flame
Use mixture
fraction method
(PDF)
Laminar
Non premixed flame
(separate fuel and
Liquid fuels (spray
oxidiser inlets) Turbulent
combustion)
flame
Combustion of
particles (coal)
Lagrangian method-trajectories of a
representative set of droplets/particles
in a continuous media
CFD8 COMBUSTION
Premixed combustion
Fuel and oxidiser are mixed first and even then burn
Bunsen laminar flame uL burning velocity Turbulent premixed flame
(=umsin)

um

flame front flamelets

Flat flame
um Bunsen burner fully
premixed. Colour
corresponds to radicals
porous plate CH (photo Wikipedia)

R.N. Dahms et al. / Combustion and Flame (2011)

Laminar premixed flames are characterised by a narrow flame front (front width depends upon the rate of chemical reaction and
diffusion coefficient). Laminar burning velocity (uL) must be greater than the velocity of flow (um) in case of a flat flame front,
otherwise the flame will be extinguished (in fact this is the way how to determine the burning velocity experimentally). Example:
flat laminar flame in porous ceramic burner. Nonuniform transversal velocity profile is manifested by a conical flame front, and
the flame front cone angle depends upon burning velocity u L=um sin . Example: Bunsen burner.
In case of turbulent flow the flame front is wavy and fluctuating (flamelets, the turbulent flame can be viewed as an ensemble of
premixed laminar flames). Example: Combustion in a cylinder of spark ignition engine, gas turbines in power plants.
Advantage of the premixed combustion: In a gas turbine operating at a fuel-lean condition high temperatures and associated
NOx formation are avoided. Soot formation is also suppressed. Disadvantage: Risk of uncontrolled explosion of premixed
reactants.
CFD8 COMBUSTION
Nonpremixed combustion
Fuel and oxidiser enter combustion chamber as separated streams. Mixing
and burning take place simultaneously (nonpremix is called diffusive burning).
Laminar nonpremixed flame Turbulent nonpremixed flame
Laminar coflow (left) stoichiometric
Flame fuel+flue line
counter flow (right).
front gas
air+flue
Burned gases gas

air

fuel air

Nonpremixed flames unlike premixed flames do not propagate. Flame simply cannot propagate into the fuel side because
there is no oxygen, neither to the oxidising region with lack of fuel. Only burned products diffuse to both sides of the flame
front. Flame front position corresponds to stoichiometric ratio fuel/air – there is also the highest temperature.
In turbulent nonpremixed flames the flamelet concept can be used again. Chemistry of nonpremixed combustion is more
complicated because on one side of the flame front is the fuel-rich burning accompanied by formation of sooth and on the
air side a fuel lean combustion occurs. Existence of glowing sooth is manifested by yellow luminiscence.
Examples: Industrial burners, combustion engines.
CFD8 COMBUSTION
Flame size
The size (length) of a nonpremixed burning jet can be estimated from the velocity
of jet and from the diffusion time necessary for mixing in the transversal direction.
Burke and Schumann (1928), Ind.Eng.Chem 20:998

Velocity of circular jets decreases with the distance from nozzle (u~1/x) however
this approximation assumes a constant velocity at axis and also cylindrical form
of jet, having radius r of fuel nozzle. Mixing length (r) is associated with the
mixing time by theory of penetration depth
r-mixing
length h r  Ddif t
and the height of flame h is estimated as the product of the mixing time and axial
velocity

Vfuel
Vfuel r 2 Vfuel
h  ut  2 t  2  2
r r Ddif  Ddif

This is qualitatively correct conclusion stating that the flame length is proportional
to flowrate and indirectly proportional to diffusion coefficient
(Warnatz 1996 documents the diffusion coefficient effect by comparing the height of flame at hydrogen combustion that is
2.5 shorter than the flame height at carbon monoxide combustion)
CFD8 COMBUSTION aims

Primary purpose of CFD analysis is to evaluate

Temperature field (therefore thermal power, heat fluxes through wall…)
Composition of flue gas (environmental requirements, efficiency of burning)

To do this it is necessary to calculate

Velocities and turbulent characteristics (mixing intensity) – NS equations
Transport of individual components (mass balances of species)
Chemical reactions (reaction rates)
Energy balances (with special emphasis to radiation energy transfer)
CFD8 COMBUSTION balances
mi mass fraction of specie i in mixture [kg of i]/[kg of mixture]
mi mass concentration of specie [kg of i]/[m3]

Mass balance of species (for each specie one transport equation)


(  mi )    (  mi u )    (i mi )  Si
t
Because only micromixed Rate of production
Production of species is controlled by reactants can react of specie i [kg/m3s]

Diffusion of reactants (micromixing) – tdiffusion (diffusion time constant)

Chemistry (rate equation for perfectly mixed reactants) – treaction (reaction constant)
tdiffusion
Damkohler number Da 
treaction
E a
Da<<1 Si   A exp( )mi Reaction controlled by kinetics (Arrhenius)
RT

Da>>1 Si   Cmi Turbulent diffusion controlled combustion
k
CFD8 COMBUSTION problems
Specific problems related to combustion (two examples A and B)

A) It is not correct to use mean concentrations in reaction rate equation

Rotating turbulent eddy Bimolecular reaction A+B→C
Production rate S  kmA mB locally
A B mA mB S  kmA mB globally

mA  0, mB  0 but mAmB  0
Actual reaction rate
is zero even if the
mean concentrations
are positive

m A mB  m A mB
CFD8 COMBUSTION problems
Specific problems related to combustion (example B)

B) It is not correct to use mean temperatures in reaction rate equation

Bimolecular reaction A+B→C
E locally
Production rate S  AmA mB exp( )
RT
SNOx E
Example NOx production S  AmA mB exp( ) globally
RT
Underestimated reaction rate
Tmin Tmean Tmax
follows from nonlinear temperature
dependence of reaction constant
Actual production
rate of NOx
E E
exp( )  exp( )
RT RT
1500 2000
T[K]
CFD8 COMBUSTION enthalpy
Enthalpy balance is written for mixture of all species (result-temperature field)


(  h)    (  hu )    (h)  Sh
t
Sum of all reaction enthalpies of all reactions

Sh   Si hri
i
It holds only for reaction without
phase changes h ~ cpT

Energy transport must be solved together with the fluid flow equations (usually
using turbulent models, k-, RSM,…). Special attention must be paid to
radiative energy transport (not discussed here, see e.g. P1-model, DTRM-
discrete transfer radiation,…). For modeling of chemistry and transport of
species there exist many different methods and only one - mixture fraction
method will be discussed in more details.
CFD8 MIXTURE Fraction method

Bacon
CFD8 MIXTURE Fraction method
Non-premixed combusion, and assumed fast chemical reactions (paraphrased
as “What is mixed is burned or is at equilibrium”)

mfuel Calculation of fuel and oxidiser

consumption is the most
Flue gases difficult part. Mixture fraction
moxidiser method is the way, how to
avoid it

Mass fraction of fuel

Mass balance of fuel (e.g.methane)


(  m fuel )   (  m fuel u )   (m fuel )  S fuel
t
kg of produced fuel
[ ]
Mass balance of oxidant s  m3


(  mox )   (  mox u )   (mox )  Sox
t
Mass fraction of
oxidiser (e.g.air)
CFD8 MIXTURE Fraction method
Stoichiometry
1 kg of fuel + s kg of oxidiser  (1+s) kg of product

Introducing new variable

  sm fuel  mox

(  sm fuel )   (  sm fuel u )   (sm fuel )  sS fuel
and subtracting previous equations t

(  mox )   (  moxu )   (mox )  Sox
t


(  )   ( u )   ( )  sS fuel  Sox
t

This term is ZERO

due to stoichimetry
CFD8 MIXTURE Fraction method
Mixture fraction f is defined as linear function of  normalized in such a way that
f=0 at oxidising stream and f=1 in the fuel stream
mox is the mass fraction of

   0 sm fuel  mox  mox ,0

oxidiser at an arbitrary point
x,y,z, while mox,0 at inlet (at
f   the stream 0)

1   0 sm fuel ,1  mox ,0
Resulting transport equation for the mixture fraction f is without any source term


(  f )   (  fu )   (f )
t
Mixture fraction is property that is CONSERVED, only dispersed and transported
by convection. f can be interpreted as a concentration of a key element (for
example carbon). And because it was assumed that „what is mixed is burned“ the
information about the carbon concentration at a place x,y,z bears information
about all other participating species.
CFD8 MIXTURE Fraction method
Knowing f we can calculate mass fraction of fuel and oxidiser at any place x,y,z
At the point x,y,z where f=fstoichio
mox ,0
f stoichio  are all reactants consumed
(therefore mox=mfuel=0)
sm fuel ,1  mox ,0
For example the mass fraction of fuel is calculated as
f  f stoichio
f stoichio  f  1 (fuel rich region, oxidiser is consumed m ox =0) m fuel  m fuel ,1
1  f stoichio
0  f  f stoichio (fuel lean region) m fuel  0

The concept can be generalized assuming that chemical reactions are at

equlibrium

f → mi mass fraction of species is calculated from equilibrium constants

(evaluated from Gibbs energies)
CFD8 MIXTURE Fraction method
Equilibrium depends upon concentration of the key component (upon f) and
temperature. Mixture fraction f undergoes turbulent fluctuations and these
fluctuations are characterized by probability density function p(f). Mean value of
mass fraction, for example the mass fraction of fuel is to be calculated from this
distribution
Mass fraction corresponding to an
arbitrary value of mixture fraction is
calculated from equlibrium constant
1
m fuel   m fuel ( f ) p ( f )df
p 0 Probability density function, defined in
terms of mean and variance of f
p 1
f (1  f ) q 1
Frequently used  distribution p( f )  1


p 1
f (1  f ) q 1 df
0 fmean 1 0

Variance of f is calculated from another transport equation

 
(  f ' )   (  f ' u )   ( f  f ' )  Cg t (f )  Cd  f '2
2 2 2 2

t k
CFD8 MIXTURE Fraction method
Final remark: In the case, that mfuel is a linear function of f, the mean value of
mass fraction mfuel can be evaluated directly from the mean value of f (and it is
not necessary to identify probability density function p(f), that is to solve the
transport equation for variation of f). Unfortunately the relationship mfuel(f) is
usually highly nonlinear.

1
m fuel   m fuel ( f ) p ( f )df  m fuel ( f )
0
CFD8 COMBUSTION of liquid fuel
Lagrangian method: trajectories, heating and evaporation of droplets
injected from a nozzle are calculated. Sum of all forces acting to liquid
droplet moving in continuous
fluid (fluid velocity v is calculated
du
F
by solution of NS equations)
mfuel m
dt Relative velocity
Drag force (fluid-particle)
1
FD   cD A | u  v | (u  v )
2
Drag coefficient cD Effect of cloud (c volume cD  cD 0 /  c3.7
depends upon Reynolds number fraction of dispersed phase-gas)

24 3 cD
cD  (1  Re) Re  5 Oseen
Re 16
24
cD  (1  0.15 Re0.687 ) Re  800 Schiller Nauman Newton’s region
Re cD=0.44
cD  0.4 1000< Re  3.105 Newton

1 104 105 Re
CFD8 COMBUSTION of liquid fuel

Evaporation of fuel droplet

Sherwood Mass fraction of
Diffusion from droplet surface to gas: number fuel at surface

dm d  D 2
 (  p )   D 2 Sh g Ddif (m fg  ms )
dt dt 6
Ranz Marshall correlation for mass transport Schmidt number =/Ddif

Sh  2  0.6 Re05 Sc 0.33

CFD8 MULTIPHASE flows examples
Fluidised bed Mixer (draft …and others
reactor or tube) Hydrotransport,
combustor cyclones, free
surfaces, breakup
of liquid jets,
expanding foams,
aerated reactors,
cavitation, mold
filling…

Phases
Spray Flow Annular
dryer boiling flow Gas-liquid
Gas-solid
Liquid-liquid
Slug flow
visualisation
THERMOPEDIA
Bubble
flow
CFD8 MULTIPHASE flows methods

Methods
Lagrange (see liquid fuel burners, suitable for low concentration of particles)
Mixture (not significant difference between phases, e.g. sedimentation)
Euler (the most frequently used technique for any combination of phases)
VOF (Volume Of Fluid) (evolution of continuous interface, e.g. shape of free
surface modeling, moving front of melted solid…)
CFD8 MULTIPHASE EULER
For each phase q are separately solved
Continuity equation (mass balance of phase)

( q  q )   ( q  q vq )   m pq
t p
Volumetric Velocity of Mass transfer from
fraction of phase phase q phase p to phase q
q

Momentum balance (each phase is moving with its own velocity,

only pressure is common for all phases)


( q  q vq )   ( q  q vq vq )   q p   ( q q )   R pq
t p

Stresses are calculated in the Interphas

same way like in one phase flows e forces
CFD8 MULTIPHASE EULER
Specific semiemprical correlations describe
interaction terms

Mass transfer
for example Ranz Marschall correlation for Sh=2+…

Momentum exchange Rpq  k pq (vq  v p )

Special models for kpq are
available for liquid-liquid, liquid-
solid, and also for solid-solid
combinations
CFD8 MULTIPHASE MIXTURE method
Mixture model solves in principle one-phase flow with mean
density m , mean velocity vm
Continutity equation for mixture
 m
   (  m vm )  0
t
Momentum balance for mixture (with corrections to drift velocities)

( m vm )    ( m vm vm )  p    ( m (vm  (vm )T )    (  p  p vdr , p vdr , p )
t p

Drift velocities are evaluated from algebraic models vdr , p  v p  vm

(mixture acceleration determines for example centrifugal
forces applied to phases with different density)

Volumetric fraction of secondary phase (p)


( p  p )    ( p  p vm )    ( p  p vdr , p )
t
CFD8 MULTIPHASE VOF
Evolution of clearly discernible interface between
immiscible fluids (examples: jet breakup, motion of large bubbles, free surface flow)

There exist many different methods in this category, Level set method,
Marker and cell, Lagrangian method tracking motion of particles at interface.
Fluent
Donor acceptor

 =0 =0 =0 Geometric reconstruction


=1   =0  u   0
t
=1 =1 =1  Dissadvantage: initially sharp
interface is blurred due to
numerical diffusion
CFD8 Level Set method
The level set method represents a closed curve (in 2D) or a closed surface (in
3D)  by using an auxiliary function , called the level set function.
Dimension of level set function is the dimension of  + 1.

Properties of level 0 at boundary 

set function (x,y,z)
0 inside 

Example: breakup of bubble

CFD8 Level Set method
CFD8 RADIATION

Pollock
CFD8 RADIATION
Heat flow (W) between gas and wall


Q  S  g TS4  Ag Tw4 . 
Absorptivity of gas
corresponding to wall
temperature Tw
Emissivity of gas
corresponding to
temperature of gas Ts

Hottel’s diagram for

emissivity of CO2 and H2O
as a function of temperature and pL
(pressure x length)


 pH 2O  pCO 2 L 8160p 
H 2O  13,810 T
4 
 g T   1  e 10 ,

Kirchhoff’s law (=a)

Emissivity=Absorptivity
at the same wavelength
CFD8 RADIATION Intensity
Spherical coordinate
system and definition  s Total power emitted from the small surface dS
of solid angle d
n EdS [W]
Power of photons moving  inside the solid
angle d in the direction s and emittedby
surface dSs perpendicular to the beam s is
 d   d d sin 
I (s ) d  dSs =I (s )d d sin  cos  dS
radiation solid projected
intensity angle surface dS
d
dS  The radiation emitted by blackbody surface is isotropic (intensity Ib
d is independent of the direction). Integrating the radiation power for
solid angles covering the whole upper hemisphere results to
dSs  cos  dS relationship between the emissive power Eb and the intensity Ib
2  /2  /2


 

0 0
I b sin  cos  d d =2 Ib


0
sin  cos  d   I b  Eb

1
I b (T )  E (T )
  b
intensity emitted power
CFD8 RADIATION
 
participating media
Intensity of radiation I( r , s ) is not a constant in the case that the radiation
penetrates through a semitransparent material (for example flue gas).
Photons moving through space filled by heteropolar gases (H2O, CO2, CO,…) can be absorbed by
molecules of gas (a [1/m] is coefficient of linear absorption) or scattered (s [1/m] is coefficient of linear
scattering). On the other hand new photons are emitted by the gaseous molecules having temperature T
(this contribution to the intensity of radiation is determined by the Stefan Boltzmann
 law) and new
photons are also coming from outside due to scattering into the direction s .
H2O

Intensity I is a scalar H2O

I
Incomming radiation I (s) I (s)  ds outgoing radiation
function of spatial
absorbedH O 2
s
 H2O

coordinate r=(x,y,z) and H2O 

emitted
s
the direction s . I recommend you to H2O
H2O
read again the precise definition of radiation Scattering addition H2O
intensity in previous slides.
ds
Variation of intensity I is described by the integro-differential RTE equation that
follows from the photon balance in the control volume oriented in the selected
direction of beam.
2 
4π
dI (r , s ) (s) 4
T σs
ds
 (a   s ) I (r , s )  an

 
4π 0
I(r,s')Φ(s.s')dΩ'
absorbed and scattered
( sI ) radiation from the control emitted thermal background scattered radiation from environment
volume radiation. n-is refractive index,  is phase function, describing distribution
relative speed of light in vacuum of scattering directions
and in medium
CFD8 RADIATION Fluent models
Radiation models are selected according to optical thickness of media (flue gas)
a.L V
L  3.5
S

aL<1
DTRM (discrete transf.radiation modelling)
dI a T 4
 (a   s ) I 
DO (discrete ordinates) dt 
Absorption and
scatter
1<aL<3
P-1 model (transport equation for radiation temperature)

3<aL
Roseland model (simplified P-1 model)