Simulating the combustion of

gaseous fuels
6
th
OpenFoam Workshop Training Session
Dominik Christ
This presentation shows how to use OpenFoam to simulate gas
phase combustion
Overview
Theory
Tutorial case
Solution strategies
Validation
The focus of combustion simulation depends on the application
Furnace operation/retrofit Pollutant formation Burner design
The focus of the present tutorial is simulating a model flame
I
Model flames are a basis to test and evaluate
combustion solvers
I
Tutorial case is a turbulent methane/air flame
(Flame ! from Sandia/T"F wor#shop$
I
Solver applications used are
rhoReactingFoam (%aS& model$
edcSimpleFoam ('( model$
I
Validation with e)perimental data to assess
the solver/model accuracy
Photo: Sandia/TNF
Overview
Theory (combustion* radiation$
Tutorial case
Solution strategies
Validation
(ombustion simulation is characteri+ed by chemical reactions
,lobal reactions summari+e the combustion process-
etailed chemical mechanisms describe events on molecular level-
from ,&./Mech 012 (034 reactions5 40 species$
I Chemical mechanisms need to be used within their specification limits,
eg. GRI-Mech 3.0: methanenatural gas, ! in "000-#$00 %, & in 0."-$
I
'etailed mechanism are more accurate (e.g. )*
+
, ignition dela,-, but
computationall, much more e+pensi.e
/ le.el of detail needs to be chosen b, the user
CH
4
+2O
2
CO
2
+2H
2
O
CH
4
+OOH+CH
3
CH
4
CH
3
+H
CH
4
+HCH
3
+H
2
CH
4
+OHCH
3
+H
2
O

Chemical reactions can be described with e6uilibrium or #inetic

rates (incl1 infinite rate!$
'6uilibrium calculation depends only on thermodynamic data- h75 s75 cp7
0ut concerning combustion, man, things are not in e1uilibrium2
(hemical #inetics determine the reaction rate e1g1 with an 8rrhenius type
formulation-
.nfinite rate! chemistry is a special case5 where reaction rates are assumed
to be infitely fast
#inetically
inhibited
R=AT
b
exp
(

E
T
)
C
CH4
C
O2
0.5
.n turbulent flows5 turbulence/chemistry interaction defines the
reacting flow
I
Turbulence
enhances mi)ing of
species such as
fuel5 o)idi+er and
products
I
Strong turbulence
can suppress
combustion
9 local e)tinction
Turbulent flow (hemical #inetics
I
.n a laminar flow5
combustion is
controlled
e)clusively by
chemical #inetics
I
(ombustion leads to
flow acceleration
9 modification of
flow field
ifferent approaches e)ist to model the turbulence/chemistry
interaction
cf. Poinsot, Veynante Theoretical and Numerical om!ustion"
#
!urbulent mi+ing (e.g. 3'C, 4a5R-
Geometrical anal,sis (flamelet-
*ne-point statistics (4'6-
f
l
a
m
e

F:';
or
F&'S< M.=T:&'
O=..>'&
or
?:&"T M.=T:&'
R
i
= p
y

(1y

)
(Y
i
Y
i

)
The 'ddy/issipation (oncept ('($ assumes reactions in fine
structures
@
i
A
m

@
i
7
Y
i
=

+(1

)Y
o
&elation of mean5 fine structure
and surrounding state-
Fraction of the flow occupied
by fine structures:
y

'( reaction rate depends on turbulent flow properties and

chemical #inetics approach
The fraction of the flow
occupied by fine structures:
Mass transfer rate between the
fine structures and the surroundings:
y

= 9.7
(
vc
k
2
)
3
4

= 2.45
(
c
v
)
1
2
(hemical #inetics approaches for fine structure composition @
i
A
-
Fast
Chemistry
Local Extinction Perfectly Stirred
Reactor (PSR)
accuracy
computational cost
The Fast hemistry approach assumes infinitely fast reactions
I
8ssumes sufficient time to achieve e6uilibrium inside fine structures
I
Bor#s only with irreversible global reactions
(ombustion occurs if fuel5 o)idi+er and products meet simultaneously
/ 4roduct mass fractions must be initiali7ed accordingl,
fuel
o)idi+er
products
ignite
combustion
The %S& approach determines the steady/state of a perfectly
stirred reactor
dp

dt
= 0
dh

dt
=0
dY
i

dt
=u
i
+

m

(Y
o
Y

)
(hemical #inetics5
u
i
=u
i
( p, T , Y
i

)
@
i
A
m

@
i
7
f

=
1

;ocal ')tinction approach employs data from a priori %S&

calculations
f

f
ch
R=0
C
ch
is the minimum residence time which sustains combustion in %S&1
high turbulence
(CAD31e/E$
medium turbulence
(CAD31e/F$
T D 022 G
(C
ch
D H1e/F$
T D I22 G
(C
ch
D 31e/4$
')ample H-
(lose to burner
')ample 3- Free
stream reaction +one
J
K
The %aS& combustion model derives the reation rate in a transient
manner

R
i
=x
C
i , 1
C
i , 0
At
Lt
Mi)ed fraction of cell that can react- M
(
2
(
H
The parameter M is based on two time scales
x=
f
ch
f
m
+f
ch
1
f
ch
=max
(

R
fuel
pY
fuel
,
R
O
2
pY
O
2
)
f
m
=
.
k
c
(
v
c
)
1
2
(hemical time scale (infinite or finite rate$- Turbulent mi)ing time scale-
Mi)ed fraction that reacts-
1
f
ch
=
R
pY
.n OpenFO8M5 mi)ing time scale is implemented slightly different

m
=

(
v

)
1
2

m
=C
mix

eff
p
.n rho&eactingFoam- .n (homia# (HIIE$-
?oth can be transformed into each other5 using-

t
p
=C

, Sc
t
=1, Re
t
=
k
2
v
8s result5 we obtain-
C
mix
=

1
1+C

Re
t
homia$ %&''(): Flame Liftoff in Diesel Sprays *+
th

Symp. ,nt. on om!. pp. *++-.*+(/
The value for (
mi)
needs to be estimated a priori
Typical
tur!ulent flo0
1e
t
&222

mi#
2.&
3aminar flo0
1e
t
4 2

mi#
4 &.2
5#tremely
tur!ulent flo0
1e
t
6

mi#
6 2.2
C
mix
=

1
1+C

Re
t
Nordin %*22&): Complex Chemistry Modelling of
Diesel Spray Comustion! Ph7.Thesis
"ypical #alues for C
mix
$ 2.22& 8 2.9: cf. Nordin %*22&)
&adation heat transfer needs to be considered in combustion
simulation
1adiation Transport 5;uation:
P& . Transport
7iscrete <rdinates %7<=)
>as.?!sorption =odelling:
constant
1?7?3.Polynomials
@S>>= %custom)
Pea% temperature &'( ) higher
*ithout radiation modelling
Overview
Theory
Tutorial case
Solution strategies
Validation
The tutorial case is a non/premi)ed piloted flame (Flame !$
(haracteristics-
Steady/state5 piloted5 methane/air5 diffusion flame5 some local e)tinction
,eometry-
8)i/symmetric5 3
Main Net ((<
F
/air$
%ilot (hot flue gas$
(oflow (air$
The boundary conditions are identical for rhoReactingFoam and
rhoSimpleFoam
A#
Bm/sC
p
BPaC
T
BDC
Ei
B.C
$
BmF/sFC
epsilon
BmF/sGC
=ain Het
/'.( zeroGraient *'/ fi#edValue ,
tur!
42.2/+I 3
tur!
4+.2/e./
Pilot
&&./ zeroGraient &II2 fi!e"alue ,
tur!
42.2(*I 3
tur!
4-.9+e./
oflo0
2.' zeroGraient *'& fi!e"alue ,
tur!
42.2/-& 3
tur!
42.2&'-
<utlet
Jero>radient &22222 Jero>radient Jero>radient Jero>radient Jero>radient
@alls
2.2 zeroGraient Jero>radient Jero>radient 0all function 0all function
Front and ac%side of axi+symmetric domain are specified as ,*edge,-
edcSimpleFoam- Flow field initiali+ed as re6uired by chemical
#inetics approach
I Fast (hemistry and ;ocal ')tinction- Set (O
3
and <
3
O mass fraction to 212H
everywhere1
I
%erfectly Stirred &eactor- .nitiali+e with Fast (hemistry or ;ocal ')tinction
solution1
I
Setup chemistryProperties:
edcFastChemCoeffs
{
oxidiserName O2;
mainFuelName CH;
!
edcPSRCoeffs
{
relati"e#olerance $%e&';
a(solute#olerance $%e&$;
max)terations $%e*;

use+inary#ree off;
(inary#ree#olerance $e&;
(inary#reeSi,e $%e-;
!
edc./Coeffs
{
oxidiserName O2;
mainFuelName CH;
auto)gnition#emperature
*'*;

cur"e$
{
temperature 011;
tauCh2in -%11/&113;
!
4
!
no local e+tinction
abo.e this
temperature
rhoReactingFoam- (hoosing (
mi)
and O' intergrator
I
'stimate turbulent &eynolds number-
1e
t
4+22 6
mi#
4 2.&+
I
Setup chemistryProperties:
odeCoeffs
{
O5/Sol"er S)+S;
eps 3%1e&;
scale $%1;
!
5I05 is stable
enough for sol.ing
detailed chemistr,
Setting/up discreti+ation schemes
I
Con.ecti.e term- ;inear upwind discreti+ation (3
nd
order accurate$
default 6auss linear7p8ind cell.imited 6auss linear $;
For species @
i
(for rho&eactionFoam- @
i
and hs$
di"9phi:;i< 6auss multi"ariateSelection
{
==hs linear7p8ind cell.imited 6auss linear $;
CH linear7p8ind cell.imited 6auss linear $;
O2 linear7p8ind cell.imited 6auss linear $;
4
!
I
!ime discreti7ation: (%seudo$ steady/state
edcSimpleFoam (steady/state solver$
default steadyState;
rho&eactingFoam (transient solver$
default S.#S phi rho 1%-;
default Co/uler phi rho 1%;
global under-rela+ation
factor
ma+. C68 number
Setting/up fvSolution
"umerical solver precision depends on solver type-
Transient solver re6uires each time/step to be accurate
all variables- relTol D 21 *
Steady state solver can reach solution through intermediate results
pressure- relTol 2122H*
other variables- relTol 21H*
intermediate results *ill not e
accurate
(hemical mechanisms can be defined in (hem#in or OpenFO8M
native format
CH > 2O2 ?@ CO2 > 2H2O 0e$0 1%1 -e0
FOR5 = CH 1%- =
FOR5 = O2 1%* =
Chem9in format uses units- mol5 cmO5 s5 G* cal
')ample- 8rrhenius type #inetics
irre"ersi(leArrheniusReaction
CHB1%- > 2O2B1%* ? CO2 > 2H2O 9C%*'*0e>$$ 1 20'31<
*pen6oam nati.e format uses units- #mol5 mO5 s5 G* G
:se chemDin#oFoam to con.ert chem9in files (or to chec9 their consistenc,-
R=A T
b
exp
(

E
T
)
C
CH4
0.7
C
O2
0.8
Overview
Theory
Tutorial case
Solution strategies
Validation
(ombustion simulation often faces stability issues
Many error sources are possible
because numerous models are applied
simultaneously5 for e)ample-
I
(ompressible flow
I
(oupling of transport e6uations
I
"umerically stiff reaction
mechanisms
Iterations
Residuals
Solution strategies include good initiali+ation and under/rela)ation
%ossible well initiali+ed! flow fields are-
I
(old flow
I
Starting solution (steady/state* H/ or 3/Step$
I
%roducts
Strong coupling between transport e6uations may be bro#en with
different under/rela)ation factors1
species- 213
(reaction model$
temperature- 21P
density- 21H pressure/
velocity- 210/21P
.nder+relaxation only applicale to steady+state cases-
.nsteady sol#er ased on "ransient+S/MPLER needed0
Tabulation of %S& reactor results can provide speed/up and
additional stability
(
y

Y
1
:
Y
n
)
(
T

Y
1

:
Y
n

)
Mean flow data %S& integration
results
Tabulation of %S& reactor results can provide speed/up and
additional stability
(
T

Y
1

:
Y
n

)
"ormali+ed mean
flow data
%S& integration
results
H
H
Tabulation of %S& reactor results can provide speed/up and
additional stability
"ormali+ed mean
flow data
%S& integration
results
H
H
H
H
H
H
H
H
"ormali+ed mean
flow data
%S& integration
results
Tabulation of %S& reactor results can provide speed/up and
additional stability
H
H
H
H
"ormali+ed mean
flow data
%S& integration
results
Tabulation of %S& reactor results can provide speed/up and
additional stability
H
H
H
H
"ormali+ed mean
flow data
%S& integration
results
Tabulation of %S& reactor results can provide speed/up and
additional stability
;imiting temperature is possible in steady/state cases
.n a steady state case5 intermediate time steps! are not accurate1
9 Temperature may temporarily increase and needs to be limited
Solver level implementation in edcSimpleFoam-
I "ew enthalpy field calculated from species mass fractions and T
min
I 8nother enthalpy field calculated for T
ma)
I
?oth fields are used to limit enthalpy field
I T
ma)
and T
min
are specified in thermophysicalProperties
&e6uest for integrated limitation filed in Open(FQs bugtrac#er (.ssue R4P$1
Optimal paralleli+ation depends on the comple)ity of the chemical
model
optimi+ed for networ# bandwidth optimi+ed for %S&/integration
/ntegrating complex chemical mechanisms is computationally much more
expensi#e than sol#ing transport e1uations (e#en if there are many)2
Overview
Theory
Tutorial case
Solution strategies
Validation
etailed reaction mechnism predicts temperature profile accurately
Radiation modeling used *ith EDC! not used *ith PaSR-
/nfluence of C
mix
is minimal
e#p. data: Barlo0, 1. S. and Fran$, K. L., Proc.
om!ust. ,nst. *-:&2I-.&2'+ %&''I)
.ntermediate species and pollutants are more difficult to predict
%araFoamQs calculator! can be used to chec# &e
t
assumption
(omparison with measurements may re6uire special post/
processing
(ommon difficulties when comparing simulated mass fractions with measurements-
I
Measured data are often mole fractions or concentrations
If not all (ma;or- species are measured, correct con.ersion to mass fractions
impossible
I
Flue gas or emission monitoring can be measured in dry gas!5 i1e1 after water
vapor has been condensed out
5imulated data comprise a complete set, therefore the, can be accuratel,
con.erted
)ew utilit, massToMoleFraction handles con.ersion together with /dry,as! option

for the '( model-
?1 Magnussen- !he 3dd, 'issipation Concept: < 0ridge between
5cience and !echnolog,, '((OM8S Thematic (onference on
(omputational (ombustion5 ;isbon5 %ortugal5 3224
for the validation of the OpenFoam implementation-
?1 ;illeberg5 1 (hrist5 .1S1 'rtesvSg5 G1'1 &ian5 &1 Gneer5
)umerical simulation with an e+tinction database for use with
the 3dd, 'issipation Concept for turbulent combustion
(submitted$
Final note- Bhen using edcSimpleFoam or edc%isoFoam5 please cite

Than$ you for your attentionM