Turbulent flow and combustion in real-life syngas burner

monkeyresultΜηχανική

22 Φεβ 2014 (πριν από 3 χρόνια και 4 μήνες)

68 εμφανίσεις

Turbulent flow and combustion in real-life syngas burner
Kamil Kwiatkowski
1;2
,Konrad Bajer
1;2
,Karol We¸dołowski
1;2
1
Institute of Geophysics,Faculty of Physics,University of Warsaw
2
Interdisciplinary Centre for Mathematical and Computational Modelling,University of Warsaw
kamil@igf.fuw.edu.pl
Abstract
When multi-species,non-premixed flows combine with chemical reactions,a common situation in industrial devices,the mixing or
diffusive phenomena become the key factors of understanding the whole process.Turbulence is conducive to mixing of species and
consequently may increase the reaction efficiency [1].That motivates us to study in details the mixing of fuel with air,in an industrial
burner of complex geometry (see fig.1a).The fuel is syngas (gas obtained fromthe process of biomass gasification).In order to isolate
the effects of turbulent mixing in a complex-geometry flow,we first neglect ignition and possible re-ignition and employ a relatively
simple parametrisation of chemical reactions.
We create a 3Dmodel of a real industrial syngas burner of complex geometry involving an axial co-flowjet and six additional air inlets.
These extra inlets enhance the swirl and promote the generation of vortices,thus making the flow more turbulent,and consequently
accelerate mixing [1,2].The exploratory simulations of non-reactive gas flows are prepared using the commercial CFD Fluent solver
as well as the open source CFD package OpenFOAM[3].Both simulations are based on the RANS turbulence closure.
We also design and run a series of non-premixed combustion simulations.To reconstruct the flowfield correctly we use Steady Flamelet
[4] approach for measured syngas composition.Despite the fact,that several simplifications are made,the results agree qualitatively
with the observations.
1.Introduction
Gasification appears to be the most economically promising
way of thermal treatment of solid biomass waste [5,6,7].
An in-house industrial installation enables clean waste disposal,
which would otherwise require costly outside contracting,and,
in the simplest version,generates heat used in the plant’s tech-
nological process.The combination of the two benefits makes
even the simplest gasifier-combustor-boiler systems commer-
cially viable provided the process is carefully designed to be
clean enough,so that the composition of the flue gases meets
the environmental norms ([8,9]) and the cost of their additional
purification is limited.Electric power co-generation can con-
siderably improve the economy of the whole process but any
efficient co-generation,whether with a gas turbine or internal
combustion engine,requires much bigger initial investment in a
systemof syngas purification.
All three components of the gasifier-combustor-boiler sys-
tem must be individually designed for different types of solid
biomass waste.Off those three it is the combustor which has
the main effect on the composition of the flue gases.It has to
be individually designed for the target flow rate and for the par-
ticular syngas composition determined by the kind of biomass
fed into the gasifier [10].Some of the pollutants present in the
flue gases are robust.Their content is determined by the chemi-
cal composition of waste biomass and they have to be removed
from the flue gases by specific chemical treatment.This is the
case,for example,with sulphur for which stable and reliable re-
moval processes are known and ready desulphurisation instal-
lations are available.Other pollutants are highly sensitive to
the details of the combustion process and their concentration in
the flue gases may vary by more than one order of magnitude
depending on the apparently small change in the design of the
combustion chamber,burner,oxidant supply configuration,etc.
This is the case with the NO
x
which are extremely sensitive to
the details of the reacting flow inside the combustion chamber
[11,12].Their concentration measured in the flue gases may
vary wildly without any noticeable change in the control pa-
rameters of the industrial process.
To somewhat lesser extent this is also the case with carbon
monoxide (CO).Its chemistry is more robust and predictable
than that of the NO
x
but its emission may be affected by possi-
ble irregular dynamics of the flowcausing puffs of unburned CO
to exit the combustion chamber from time to time.A straight-
forward counter measure would be to raise the temperature of
the combustion process but this is antagonistic to the reduc-
tion of the NO
x
.Another limit on lowering temperature is the
requirement imposed on the combustion of syngas originating
from waste material.Such syngas must be subject to tempera-
ture no less than 850

C for at least 2 seconds ([8,9]).
Since changing temperature in the combustor is restricted
by such opposing environmental constraints it is most important
to optimise the combined CO and NOx emissions by means of
careful design of the reacting flow in the combustion chamber
and that requires numerical simulations which are the subject of
this paper.
In order to have as accurate and reliable numerical simu-
lations as possible we develop two independent codes that are
run in parallel enabling quantitative intercomparison and cross-
checking of the results.One set of simulations is done with
FLUENT and another,under the same conditions,with Open-
Foam [13,3].With these two numerical codes we solve a se-
quence of problems of increasing physical complexity building
up to the ultimate goal of time-dependent simulations of the
combustion chamber of a real industrial installation with true
chemistry of the syngas actually produced froma particular type
of waste biomass.
All simulations are run in the true geometry of the com-
bustion chamber with an axial co-flow jet burner at the top,ad-
ditional air inlets introducing swirl,and asymmetric outlet near
the bottom(see fig.1).The flowis driven by suction,i.e.,by the
pressure drop imposed at the outlet.In practise this is realised
by electric fans at the exit of the flue gases from the boiler far
away fromthe combustion chamber.
All results reported in this paper are for steady-state simu-
lations.First runs for unsteady flow are now under way.With
unsteady simulations we hope to capture the slowmodes of tem-
poral evolution of the flow that we have actually observed in
the combustion chamber.Yet,even the steady state simulations
are quite revealing and indicated hitherto unknown characteris-
tics of the temperature distribution and velocity field inside the
chamber.
The first stage is the simulation of ‘cold’ flow,i.e.,the flow
without chemical reactions,through the combustion chamber.
Somewhat remote from the ultimate goal of pollutant emission
modelling,yet these simulations serve two important purposes.
The first is to compare FLUENTand OpenFoamperformance in
the simulations where the momentum transport dominates and
heat transport does not play a role.The second is to compare
such ‘cold’ simulations with measurements,not yet available,
that will be taken in the industrial installation during the next
maintenance break.We will then test the flow with syngas be-
ing replaced by air.Simple comparison of the relation between
the pressure drop and mass flow rate through this complex ge-
ometry will provide valuable means of validation of the code
and will be a test for different sub-grid scale turbulence models.
The second stage is the simulation of ‘hot’ flow.Then we
include those chemical reactions which strongly affect the en-
ergy balance and therefore influence the flow.These are pri-
marily the combustion of CO and of H
2
.Large energy re-
lease causes qualitative changes in the flowpattern due to buoy-
ancy of the hot gas.Strong counterflow (updraft) appears in
prominent recirculation zones and temperature distribution is
changed.
2.Turbulent cold flow
2.1.Flow conditions
In large-scale industrial syngas devices turbulence occurs natu-
rally.Estimated values of Reynolds number which cover the
range of flows in our simulations of non-reactive case,i.e.
‘cold’ flow,is of the order 10
5
,hence the flow is clearly turbu-
lent and turbulence modelling must be used.The highest value
of Reynolds number,based on the diameter of the fuel inlet and
on the fuel inlet velocity,is equal 1:87  10
5
.
Environmental regulation require that the combusting gases
spend no less than 2 seconds in the region of high temperature
(minimum...).For a given gas volume flux at the inlet,de-
termined by the production rate of the gasifier,this imposes a
lower bound on the volume of the combustion chamber which,
in the installation we are modelling is 36m
3
.A sketch of the
simulated combustion chamber is presented in figure 1.It con-
sists of a vertical cylinder with the system of fuel and air in-
lets installed in the top surface and an outlet located near the
bottom of the cylindrical surface.The position of the outlet in
the wall is a major departure from axial symmetry.The axial
(vertical) cross-section which includes the centre of the outlet
pipe is denoted (x-z) plane.Another asymmetry is due to the
additional swirl-enhancing air inlets in the top surface.Sym-
metrically placed on a circle concentric with the main axial jet
those alone would make only small,possibly negligible,depar-
ture from axial symmetry.However,due to the combination of
discrete auxiliary inlets at the top and the fixed position of the
outlet the systemhas neither planar,nor axial symmetry.
Figure 1:The 3D model of the syngas burner.Six additional
inclined inlet pipes are mounted in the top wall induce swirl.
Details of inlets are presented in the lower picture.
Table 1:Boundary conditions for non-reactive flows.
Boundary
V
axial
[
m
s
]
V
radial
[
m
s
]
V
swirl
[
m
s
]
Fuel inlet
8
0
0
Main air inlet
2.5
0
-2.5
Swirl air inlets
0.5
0
0.5
In our simulations the values of gas and air inlet veloci-
ties correspond to the typical values measured in industrial de-
vices (particularly in the industrial installation for thermal waste
conversion working in Olsztyn).The boundary conditions are-
clearly explained in fig.1.Numerical values of the boundary
conditions are listed in table 1.Since the cold flow simulations
include no chemical reactions,and density of the syngases is
comparable with density of air,we use air as the fluid for both
the oxidiser inlets and for the fuel inlet.
The simulations are run using two alternative Computa-
tional Fluid Dynamics codes,commercial code Fluent and
open-source package OpenFoam.For both CFD packages we
use the same setup,steady-state,incompressible solvers with
RANS k   turbulence modelling.These requirements are
fulfilled by the pre-defined OpenFoam solver called simple-
Foam.Having the experience of two previous long series of
two-dimensional RANS simulations closed by several turbu-
lence closure models,we decided to choose the k   closure
of RANS equations for the present series of simulations.Since
three-dimensional flows are naturally more complex we are go-
ing to make a comparative study of other closures as well.
2.2.Comparison of Fluent and OpenFoamresults
First,we present contours of the axial velocity on the slice in
x-z plane (fig.2).The jet is composed of the streams of co-
flowing fluid issuing from the fuel and oxidiser inlets.Addi-
tional swirls,which supply air mainly for cooling the chamber
walls,does not show in this section.Their influence on the jet
is negligible.Qualitative observations show that in the Open-
Foamsimulations the jest is slightly wider and has weaker core.
Quantitative analysis,the results of which are presented in fig-
ure 3,confirmed systematically lower values of axial velocity in
the jet centre.
Interesting problemwhich we try to answer in this paper is
the influence of the geometrical asymmetry on the direction of
the main jet.In figure 3 the left panel shows the profiles of axial
velocities in the plane x-z plane and the right panel in the plane
y-z.
We can notice that Fluent simulations slightly break axial
symmetry (see fig.3,left panel),this effect does not occur in
the OpenFoam simulations.We repeated the calculations with
OpenFoamfor three different meshes,all of themunstructured,
with 0:5  10
6
(called C-Course),0:75  10
6
(M-Moderate) and
1:0  10
6
(F-Fine) nodes.All results,plotted in figure 4 (plane
x  z in the left panel,plane y  z in the right panel),are al-
most symmetrical.In our later simulations,with combustion
processes enabled,additional turbulence kinetic energy is pro-
duced and departures fromaxial symmetry are prominent all the
time.
3.Turbulent,non-premixed combustion
Table 2:Notation.
C
d
model constant of the order of unity
D molecular diffusion coefficient
~
k Favre-averaged turbulent kinetic energy
N scalar dissipation rate
Sc
t1
,Sc
t1
model constants of the order of unity
u
k
k-component of velocity vector
Y

mass fraction of specie 
~ Favre-averaged energy dissipation

turb
turbulent viscosity
 mixture fraction

 mean value (in sense of Reynolds averaging) of
quantity 
~
 mean value (in sense of density-weight averag-
ing) of quantity 

0
fluctuation of  fromReynolds averaged value

00
fluctuation of  from density-weight averaged
value

st
value of quantity  where mixture fraction is
equal its stoichiometric value
!

production termof specie 
3.1.Characteristics of syngas
Process of biomass gasification gives unique combustible gas
called biomass syngas.This fuel is fully renewable and con-
tains mainly nitrogen (if gasification was made in pure oxygen
and biomass was nitrogen-free,nitrogen would be absent),hy-
drogen and carbon monoxide (detail compositions,measured in
real-life gasifiers,are presented as mass fractions in table 4 and
as volume fractions in table 3).Due to its relatively low mean
calorific value,strongly fluctuating composition and admixture
of soot and other solid components,syngas is most often burnt
in combustion chambers.Due to the fact that produced syngas
contains methane and other hydrocarbons,we have to use the
scheme for chemical reactions that allows for the combustion
of not only hydrogen and CO ([11]) but also of hydrocarbons.
The syngas generally does not contain hydrocarbons higher that
three atoms of carbon (table 3).Hence we can use the GRI3 [14]
detailed mechanism of chemical reactions and thermodynamic
data.We use the GRI3 for laminar flamelet.It is based on
6 elements and 53 species appropriate for syngas and includes
more than 300 reactions.The simulations were performed for
the whole set of 53 species and for a simplified set of 20 species.
3.2.Conserved scalar approach
All species present in the combustion chamber obey the
advection-diffusion (1) with nonlinear source term describing
the chemical conversions of the species.
@(Y

)
@t
+
@(u
k
Y

)
@x
k
=
@
@x
k

D
@Y

@x
k

+!

(1)
Under the assumption of equal diffusivities of all species,a
reasonable assumption for turbulent flow,we can use the con-
served scalar approach to separate the modelling of flow and
combustion [15].For turbulent,non-premixed combustion it is
convenient to introduce the mixture fraction variable, (some-
times denoted also as f),defined over the whole space and time,
to be the mass fraction of the material that originated from the
fuel stream.The origin of the material is constant during chem-
ical reactions or conversions,so the mixture fraction is con-
served (eqn.2).If combustion does not occur,the mixture
fraction simplifies to fuel mass fraction.
@()
@t
+
@(u
k
)
@x
k
=
@
@x
k

D
@
@x
k

(2)
We usually normalise mixture fraction,so that  = 1 in the
fuel stream and  = 0 in the oxidiser stream which simplifies
the boundary conditions for fuel and oxidiser inlets (see table
5).Mass fractions of the particular species of fuel are presented
in table 4.The sum over all species,according to definition of
mixture fraction,should be 1.
For large-scale complex flows with turbulence the set of
conservation equations (eq.2 also) has to be averaged.For
flows with strong variation of density,which occurs naturally
in the combustion processes,the method of choice is density-
weighted average (also called Favre average,[4]) defined as
~
 =


:(3)
Here the averaged quantity is split, =
~
+
00
,where
~
is the
Favre-averaged value and 
00
is the departure (fluctuation) from
the density-weighted mean.The second
0
is added to distinguish

00
fromthe fluctuations in RANS averaging 
0
..
The Favre-averaged version of the equation of mixture frac-
tion conservation takes the following form:
@(
~
)
@t
+
@( ~u
k
~
)
@x
k
=
@
@x
k

D
@
@x
k


@(
^
u
00
k

00
)
@x
k
:(4)
For non-reacting scalar the gradient assumption can be used
for the last term on the right hand side (see [4]),so
]
u
00
k

00
=
Figure 2:Comparison of the OpenFoam(left) and Fluent (right) results,fro the simulations on the fine grid (1:010
6
nodes).Qualitative
observation is that the OpenFoam jet spreads slightly wider and slower than that in Fluent.Quantitative comparisons are presented
below.
Figure 3:Comparison of the simulations with Fluent (FL solid lines) and OpenFoam (OF dashed lines),for the fine grid (1:0  10
6
).
Four radial profiles of axial velocity V
axial
at different z stations (distance measured from the top surface) are plotted:0:5m (red),
1:5m(navy),2:5 (brown) and 3:5m(green).Left panel shows the cross-section along x-axis and right panel - along the y-axis.
Figure 4:Dependence of the OpenFoamresults on the mesh size.Shown are the resulta for three meshes:0:5  10
6
(called C-Course),
0:75  10
6
(M-Moderate) and 1:0  10
6
(F-Fine).Four radial profiles of axial velocity V
axial
at different z stations (distance measured
fromthe top surface) are plotted:0:5m(green),2m(black),3:5 (violet) and 5m(red).Left panel shows the cross-section along x-axis
and right panel - along the y-axis.
Table 3:Composition of syngases (volume fraction) produced in the process of biomass gasification.Syngas composition depends on
the type and parameters of the biomass and on the conditions of the gasification process.
Volume fraction
syngas 0
syngas 1
syngas 2
syngas 3
syngas 4
syngas 5
syngas 6
syngas 7
Nitrogen (N2)
0.486
0.538
0.531
0.493
0.543
0.562
0.518
0.550
Carbon Mono.(CO)
0.240
0.264
0.302
0.310
0.257
0.270
0.285
0.200
Hydrogen (H2)
0.180
0.081
0.074
0.089
0.079
0.065
0.061
0.080
Carbon Di.(CO2)
0.060
0.073
0.051
0.058
0.059
0.044
0.054
0.040
Oxygen (O2)
0.004
0.019
0.021
0.019
0.0500
Water (H2O)
0.024
0.026
0.033
0.021
0.021
0.039
Methane (CH4)
0.03
0.019
0.017
0.019
0.017
0.014
0.018
CH2
0.040
Ethyne (C2H2)
0.040
Ethane (C2H6)
0.001
0.001
0.002
Propane (C3H8)
0.003
0.003
0.003
0:Biomass syngas,source:www.treepower.org.
1-6:Syngases produced fromwood wastes.Composition measured in industrial gasifiers.Courtesy of Modern Technologies and Filtration Sp.z o.o.
7:Syngases produced fromturkey feathers.Courtesy of Modern Technologies and Filtration Sp.z o.o.
Table 4:Composition of the same biomass syngases given as mass fractions.
Mass fraction
syngas 0
syngas 1
syngas 2
syngas 3
syngas 4
syngas 5
syngas 6
syngas 7
Nitrogen (N2)
0.568
0.567
0.562
0.529
0.567
0.583
0.536
0.590
Carbon Mono.(CO)
0.281
0.277
0.319
0.333
0.268
0.280
0.295
0.214
Hydrogen (H2)
0.015
0.006
0.006
0.007
0.006
0.005
0.005
0.006
Carbon Di.(CO2)
0.110
0.121
0.085
0.097
0.096
0.071
0.088
0.067
Oxygen (O2)
0.005
0.023
0.025
0.023
0.061
Water (H2O)
0.016
0.018
0.023
0.014
0.014
0.027
Methane (CH4)
0.020
0.012
0.010
0.012
0.019
0.015
0.020
CH2
0.021
Ethyne (C2H2)
0.040
Ethane (C2H2)
0.001
0.001
0.001
Propane (C3H8)
0.005
0.005
0.005
Table 5:Boundary conditions for turbulent non-premixed combustion.
Boundary
V
axial
[
m
s
]
V
radial
[
m
s
]
V
swirl
[
m
s
]
[]

0
2
[]
T
emp
[K]
Fuel inlet
10
0
0
1
0
800
Main air inlet
2
0
3
0
0
300
Swirl air inlets
6
0
6
0
0
300
Walls
0
0
0
0
0
1000 (300)


So far we have done simulations for hot and cold walls with uniform temperature distribution.Simulations with a more accurate temperature profile
measured in a working industrial combustion chamber are currently under way.
D
t
@
@x
k
.In a turbulent flow the turbulent diffusion coefficient
D
t
is much larger than molecular diffusivity,so the latter is usu-
ally neglected.
As we shall see later,the second moment of mixture frac-
tion,
g

00
2
,is a parameter in the assumed probability density
function (PDF,see eq.8),so we must solve an additional equa-
tion for its temporal evolution (eq.6,ref.[16,17]).
@(
g

00
2
)
@t
+
@( ~u
k
g

00
2
)
@x
k
=
@
@x
k



turb
Sc
t1
@
g

00
2
@x
k
!
(5)
+2

turb
Sc
t2

@
~

@x
k

C
d

~
~
k
g

00
2
;
The mean values of the reacting scalars (mass fractions of
all species) can be calculated fromthe equation 6 using the pre-
sumed shape of the PDF of the mixture fraction,P(),and in-
tegrating over the whole mixture fraction space,
f
Y

=
Z
1
0
Y

()
~
P()d:(6)
The shapes of this PDF is an empirical result of many exper-
iments (cf.[18]).In numerical simulations of turbulent com-
bustion mainly two kinds of PDFs are used:the clipped Gaus-
sian function or a beta function,both parametrised by the mean
value of mixture fraction (obtained by solving equation 4 with
a CFDcode) and by root mean square of the fluctuation of mix-
ture fraction (obtained by solving equation 6).Alternatively,
the whole set of equations for the reacting species should be
solve directly,which is a lot slower and less convenient.The
beta-function PDF of the mixture fraction is defined by the the
formulae 8 and plotted in figure 5 for several values of the mean
mixture fraction and of the variations of mixture fraction.
P() =

1
(1 )
1
R

1
(1 )
1
d
(7)
 =
~


~
(1 
~

g

00
2
1
!
 =

1 
~



~
(1 
~

g

00
2
1
!
Due to large density fluctuations in the combustion process
we use the Farve-averaged version of the beta-function PDF
~
P() =
()

P() (8)
Figure 6:Typical mixture fraction distribution neat to inlets.
3.3.Laminar flamelet
Combustion takes place when two conditions are fulfilled.
Firstly,the value of the mixture fraction should be close to sto-
ichiometric which ensures appropriate proportion of fuel and
oxidiser.Secondly,there should be a large gradient of the mix-
ture fraction,which distinguishes the case when fresh fuel is
mixings with oxidiser fromthe case when fuel and oxidiser are
already burnt [15].Often such zones (flamelets) are consid-
erably thinner than the smallest characteristic length scale of
turbulence (Kolmogorov scale) and can be treated as a wrin-
kled sheet of constant,stoichiometric,value of mixture fraction
[19].Locally it is possible to introduce the coordinate system
with two coordinates parallel to the isosurfaces of the mixture
fraction (see fig.6) and with mixture fraction as transversal co-
ordinate.Peters [4] reviewed the formal introduction of this co-
ordinate transformation and derived the equations for the mass
fractions of all species (eq.9) and for temperature (10).The
main processes in the such ‘mixture fraction space’ are diffu-
sion and chemical reactions.The influence of the fluid flow
on mixing is accounted for by considering the quantity called
scalar dissipation rate,N = 2Dr  r,which may be inter-
preted as the inverse of the characteristic time scale of diffusion
[20].

@Y

@t
=
1
2
N()
@
2
Y

@
2
+!

(9)

@T
@t
=
1
2
N()
@
2
T
@
2
+!
T
:(10)
Figure 5:Beta-functions PDFs plotted for values of mixture fraction from0:1 to 0:5 (function is symmetrical with respect to  = 0:5).
Each plot shows the PDFs for 6 values of the mixture fraction variance.
Figure 7:Scalar dissipation rate in the laminar flamelet model.
Boundary conditions for these equations are such like for
counter-flowing jets of fuel and oxidiser with a stagnation point
(again see fig.6.In such situation the scalar dissipation rate can
be analytically calculated ( see fig.7),
N() = N
max
e
2(erfc
1
(2))
:(11)
This formula is an approximation valid especially in the regions
of stoichiometric value of mixture fraction (
st
) and it gives the
stoichiometric scalar dissipation rate,
N
st
(
st
) = N
max
e
2(erfc
1
(2
st
))
:(12)
In a turbulent flow to obtain the mean values of the mass
fractions,
f
Y

,one has to integrate over mixture fraction and
over all possible values of the stoichiometric scalar dissipation
rate,
f
Y

=
Z
1
0
Z
1
0
Y

(;N
st
)
~
P(;N
st
)ddN
st
:(13)
Above a certain value of the scalar dissipation rate reactions
are always quenche so,in practice,the integration over  in the
formula 13 is limited to a sub-interval of [0;1].
The joint probability density function of
~
P(;N
st
) is gen-
erally unknown.In the literature of the subject ([19,4,21,22])
it is discussed and commonly assumed that  and N
st
are sta-
tistically independent,so
~
P(;N
st
) 
~
P()P(N
st
):(14)
Under the assumption of weak fluctuations of N
st
,the proba-
bility density function P(N
st
) reduces to
P(N
st
) = (N
st

~
N
st
);(15)
where
~
N
st
is typically modelled in the following way ([23])
~
N
st
=
C
N

g

00
2
k
:(16)
Great advantage of the methods based on tabulated chem-
istry,like laminar flamelet method,is the posibility of doing
all calculations of chemical reactions once,before running the
the fluid-mechanical simulations.The Flamelet Library built in
such way consists of solutions of eqs.9 and 10 depending on
the values of the mixture fraction and of the scalar dissipation
rate.Both these quantities,and also the variations of the mix-
ture fraction that are nessesary for the beta-function PDF,are
computed in the CFD code.In figure 8 we show so tabulated
functions (temperature and Y

for selected species) from the
Flamelet Library computed with Fluent Selected dependencies
fromFlamelet Library are plotted in fig.8.
Figure 8:Solutions of the Flamelet equations computed by Fluent.The panels show the dependence of temperature and of the selected
species concentrations on mixture fraction.
3.4.Results
The results are obtained using steady-state,pressure-based Flu-
ent solver.Among simulating processes are turbulent flows,en-
ergy transfer,radiation and non-premixed combustion,details
of Fluent setup are listed at the table 6.
Table 6:Summary of Fluent setup for the simulations of turbu-
lent combustion.
Models
Energy
Turbulence
Realizable k 
Radiation
P1
Non-premixed combustion
Steady Flamelet
Solver
Steady-state
Pressure based
Even in this early stage the simulations correctly reproduce
major features of the real combustion process which we ob-
serve in the working industrial burner.Firstly,the length of
the jets is consistent with observations.Although their extent
varies somewhat fromsimulation to simulation and depends on
the mesh size and on the wall temperature,they reach down,
roughly,to the middle of the combustion chamber (see figures
9).Secondly,simulations in all cases predict the presence of the
updraft near the wall of the chamber.The effect of this updraft is
clearly visible in the pattern formed by solid particles deposited
on the wall.Thirdly,none of the computational results show
symmetry,neither axial nor planar.In fact the flow that we ob-
serve in the working combustion chamber always lacks symme-
try.Large departures from axial symmetry are clearly seen and
is currently being documented by the measurements of external
wall temperature.Visual observations as well as the analysis of
the high-speed camera recording show the combustion process
to be unsteady.The time dependence of the flowand the effects
of unsteadiness cannot be captured by our steady-state (RANS)
simulations.In particular we are not able to predict any time-
dependent phenomena,like oscillations of the jet length or jet
precession.
4.Conclusions
The simulations of non-reacting flows,which we presented in
the first part of the paper,were made using both Fluent and
OpenFoam.In both cases the results are in reasonable agree-
ment.Since the OpenFoam CFD package seems reliable for
non-reactive flows we are going to test its reacting flow solvers
in the near future.Only a few OpenFoam solvers are dedicated
to combustion or reacting flows,i.e.reactingFoam,fireFoam.In
addition there are solvers designed specially for Diesel engine
simulations.
The simulations of combustion are highly sensitive to
changes in the process parameters,mainly to the fluctuations
of temperature.
Qualitative observations made in the real-life device work-
ing in an industrial plant provide strong evidence of the flow
being time-dependent.Through a window located opposite the
outlet we observe four main modes:1) flame not visible at all;
2) clockwise flame vortices appear momentarily and then are
transported by the flowto the outlet;3) entire bottompart of the
chamber is filled with flames;4) anticlockwise flame vortices
flow to the outlet.
From these observations we can infer the following four
distinct possibilities:
i) the dominant mode is oscillatory;
ii) there coexist different steady-state regimes and fluctuations
cause the flow to ‘switch’ between those regimes;
iii) the flow is fully unsteady but on the slow time scale well
separated from the much short time scales of turbulence and
combustion.
High-speed camera recordings,with time resolutions vary-
ing from400 frames per second to 10000 frames per second,al-
lowus to estimate the characteristic times scale of these modes.
The fastest modes are 1) and 3) with time scale of the order
0:01s while modes 2) and 4) are one or even two orders of mag-
nitude slower.
a) b) c)
Figure 9:Colour map of velocity magnitude in the x z plane sections:
a) fine mesh (10
6
nodes),hot wall (1000K);b) coarse mesh (0:5 10
6
nodes),hot wall (1000K);c) coarse mesh (0:5 10
6
nodes),cold
wall (300K).
a) b) c)
Figure 10:Results of the combustion simulations computed with Fluent.Colour maps of temperature the x z plane sections:
a) fine mesh (10
6
nodes),hot wall (1000K);b) coarse mesh (0:5 10
6
nodes),hot wall (1000K);c) coarse mesh (0:5 10
6
nodes),cold
wall (300K).
In order to capture and analyse temporal variations we cur-
rently running the unsteady RANS simulations.We are also
planning to performLES simulations for both reactive and non-
reactive flows.Those require much finer mesh.The LES solvers
are also being developed for OpenFoam and we are going to
compare their results with Fluent.
The main long-termgoal of this work,is to reduce the emis-
sion of nitric oxides.The production of the NO
x
is a highly sen-
sitive process and its modelling requires very accurate simula-
tions of the flowfield as well as the most sophisticated chemical
reaction model.
5.Acknowledgements
This research is sponsored by the Foundation for Polish Science
programme VENTURES (operated within the Innovative Econ-
omy Operational Programme 2007-2013).Numerical computa-
tions were performed in the Interdisciplinary Centre for Com-
putational and Mathematical Modelling (ICM),University of
Warsaw,grant number G34-8.KK’s visitsg to the Univer-
sity of Cambridge are supported by the Young Scientists Pro-
gramme of the British Council and Polish Ministry of Science
and Higher Education.All experimental works were carried in
the gasification system designed by Modern Technologies and
Filtration Sp.z o.o.(MTF).
6.References
[1] K.Bajer,A.P.Bassom,and A.D.Gilbert,“Accelerated
diffusion in the centre of a vortex,” J.Fluid Mech.,vol.
437,pp.395–411,2001.
[2] J.P.Crimaldi,J.R.Cadwell,and J.B.Weiss,“Reac-
tion enhancement of isolated scalars by vortex stirring,”
Physics of Fluids,vol.20,no.7,p.073605,2008.
[3] H.Weller,G.Tabor,H.Jasak,and C.Fureby,“A ten-
sorial approach to computational continuum mechanics
using object-oriented techniques,” Computers in physics,
vol.12,no.6,pp.620–631,1998.
[4] N.Peters,Turbulent Combustion.Cambridge University
Press,Jan 2000.
[5] D.L.Klass,Biomass for Renewable Energy,Fuels,and
Chemicals.Academic Press,1998.
[6] P.Quaak,H.Knoef,and H.Stassen,Energy frombiomass.
A review of combustion and gasification technologies,ser.
World Bank Technical Paper.The World Bank,1999,
vol.422.
[7] C.Higman and M.van der Burgt,Gasification.Second
Edition.Gulf Professional Publishing,2008.
[8] “Council directive 75/442/eec on waste (changed by di-
rective 91/156/ewg and directive 91/689/ewg).”
[9] “Directives 89/429/ewg,94/67/we and 2000/76/we.”
[10] M.Dudyski,“Rights of protection no.(21) 385282:Sposb
i urzdzenie do zgazowania odpadw organicznych,” Mod-
ern Technologies and Filtration Sp.z o.o.,Tech.Rep.,
2008.
[11] A.Frassoldati,T.Faravelli,and E.Ranzi,“The igni-
tion,combustion and flame structure of carbon monox-
ide/hydrogen mixtures.Note 1:Detailed kinetic modeling
of syngas combustion also in presence of nitrogen com-
pounds,” Int.J.Hydrogen Energy,vol.32,no.15,pp.
3471–3485,2007.
[12] A.Cuoci,A.Frassoldati,G.Buzzi Ferraris,T.Far-
avelli,and E.Ranzi,“The ignition,combustion and flame
structure of carbon monoxide/hydrogen mixtures.Note 2:
Fluid dynamics and kinetic aspects of syngas combus-
tion,” Int.J.Hydrogen Energy,vol.32,no.15,pp.3486–
3500,2007.
[13] OpenFoam User Guide,2009.
[14] G.P.Smith,D.M.Golden,M.Frenklach,N.W.Moriarty,
B.Eiteneer,M.Goldenberg,C.T.Bowman,R.K.Han-
son,S.Song,W.C.Gardiner,V.V.Lissianski,and Z.Qin,
http:==www:me:berkeley:edu=gri
mech=.
[15] R.Bilger,S.Pope,K.Bray,and J.Driscoll,“Paradigms in
turbulent combustion research,” Proceedings of the Com-
bustion Institute,vol.30,no.1,pp.21 – 42,2005.
[16] R.Cant and E.Mastorakos,An Introduction to Turbulent
Reacting Flows.Imperial College Press,2008.
[17] T.Poinsot and D.Veynante,Theoretical and Numerical
Combustion.R.T.Edwards,Inc.,2005.
[18] A.Y.Klimenko and R.W.Bilger,“Conditional moment
closure for turbulent combustion,” Progress in Energy and
Combustion Science,vol.25,no.6,pp.595 – 687,1999.
[19] N.Peters,“Laminar diffusion flamelet models in non-
premixed turbulent combustion,” Progress in Energy and
Combustion Science,vol.10,no.3,pp.319 – 339,1984.
[20] R.Bilger,“Some aspects of scalar dissipation,” Flow,tur-
bulence and combustion,vol.72,no.2,pp.93–114,2004.
[21] D.Veynante and L.Vervisch,“Turbulent combustion
modeling,” Progress in Energy and Combustion Science,
vol.28,no.3,pp.193–266,2002.
[22] R.W.Bilger,“Turbulent diffusion flames,” Annual Review
of Fluid Mechanics,vol.21,no.1,pp.101–135,1989.
[23] W.Jones and J.Whitelaw,“Calculation methods for react-
ing turbulent flows:A review,” Combustion and Flame,
vol.48,pp.1 – 26,1982.