See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/282633527

The usability and limits of the steady flamelet

approach in oxy-fuel combustions

Article in Energy July 2015

DOI: 10.1016/j.energy.2015.06.103

CITATIONS READS

3 57

4 authors, including:

Bernhard Mayr Rene Prieler

Graz University of Technology Graz University of Technology
11 PUBLICATIONS 22 CITATIONS 15 PUBLICATIONS 51 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Coupled CFD-FEM simulations of combustion processes View project

All content following this page was uploaded by Rene Prieler on 06 January 2016.

The user has requested enhancement of the downloaded file.

 Energy 90 (2015) 1478e1489

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

The usability and limits of the steady amelet approach in oxy-fuel

combustions
Bernhard Mayr a, *, Rene Prieler a, Martin Demuth b, Christoph Hochenauer a
a
Institute of Thermal Engineering, Graz University of Technology, Inffeldgasse 25/B, A-8010 Graz, Austria
b
Messer Austria GmbH e Kompetenzzentrum Metallurgie, Industriestrae 5, A-2352 Gumpoldskirchen, Austria

a r t i c l e i n f o a b s t r a c t

Article history: This paper investigates two furnaces which work under oxy-fuel condition with natural gas. One is a
Received 20 October 2014 0.8 MW furnace where detailed iname measurements are available. The other furnace is an 11.5 kW lab-
Received in revised form scale furnace with temperature measurements. The furnaces were investigated by CFD (Computational
17 March 2015
uid dynamics) analysis. The main focus was on using combustion models that are not computationally
Accepted 27 June 2015
Available online 21 July 2015
demanding. Therefore the SFM (steady amelet) approach was used with two detailed mechanisms. The
advantage of the SFM is that the calculation time can be reduced from 4 weeks to 4 days on 8 CPU-cores.
The applicability of two detailed mechanisms under oxy-fuel condition is pointed out in this paper. The
Keywords:
Oxy-NG combustion
investigation showed that the skeletal25 mechanism and the SFM are in very good accordance with
Computational uid dynamics measurements. If the strain rate between CH4 and O2 stream is too low, the SFM fails to predict the ame
Combustion kinetic shape correctly. The inuence of three different turbulence models was also investigated. Furthermore
Steady amelet simulations with the eddy dissipation model and numerically expensive eddy dissipation concept model
Limits of steady amelet were conducted. Different WSGGM (weighted sum of grey gases model) were applied. The comparison of
the WSGGMs showed that the difference between them is insignicant for small furnaces.
2015 Elsevier Ltd. All rights reserved.

1. Introduction enhances the energy efciency, e.g., Oliveira et al. [7] showed that
in metal reheating furnaces with a temperature level of 1200
C the
In times of rising fuel prices and global warming, oxy-fuel fuel consumption could be reduced by 46% if pure oxygen was used
combustion can be seen as a vast opportunity to optimize the instead of air for the combustion. Experiments carried out by
processes of many industrial applications. Oxy-fuel combustion is Blohradsk et al. [8] on a test facility with a natural gas burner
also a promising technology to reduce global CO2 emissions. CO2 is showed a similar trend. With increasing oxygen concentration in
mainly produced by fossil fuels, which are accountable for 85% of the oxidizer the energy efciency increased from 60% at 21% O2 to
global energy production [1]. CO2 can also be emitted by chemical 78% at 38% O2. Numerical calculations and experiments carried out
reaction in industrial processes i.e. cement clinker formation, 68% on a lab-scale furnace with a thermal load by Prieler et al. [9] also
CO2 is emitted from carbonates and only 32% by combustion [2]. showed an increase of the efciency from 44% at 25% O2 to 67%
Flue gas mainly consists of CO2 and H2O when using pure oxygen as at 100% O2. In conventional power generating processes, the
an oxidant. This can be used to reduce CO2 emission into the at- maximum temperature is limited through high stresses, i.e. for gas
mosphere by CCS (carbon capture and storage) technologies. Some turbines 1600 K [10]. To control the ame temperature and to
technical and economic studies show that oxy-fuel and CCS could substitute missing N2, the oxidant in power plants is a mixture of
be an upcoming technology in the near future [3e5]. In industries oxygen and recycled ue gas (wet or dry ue gas). Through recy-
with high temperature application, like glass, steel or cement in- cling ue gas, the heat transfer in the convection zone of the boiler
dustries, oxy-fuel or oxygen-enriched air combustion technologies should be approximately the same as under air red condition. The
are already used [6]. The absence of N2 leads to a higher combustion total heat transfer characteristic is considerably different compared
temperature in comparison to conventional air-ring, and this also to air combustion under oxy-fuel condition, with or without recy-
cling of ue gas, the radiative heat transfer is strongly promoted by
the much higher level of CO2 and H2O [11e13]. Despite the above
* Corresponding author. Tel.: 43 316 873 7804. mentioned reviews about CCS and the use of oxy-fuel in power
E-mail address: bernhard.mayr@tugraz.at (B. Mayr). plants, the main focus of this work was to investigate high

http://dx.doi.org/10.1016/j.energy.2015.06.103
0360-5442/ 2015 Elsevier Ltd. All rights reserved.
 B. Mayr et al. / Energy 90 (2015) 1478e1489 1479

temperature processes under oxy-fuel conditions. In high temper- concentration of radiative active gases like CO2 and H2O is low due
ature processes, as in melting processes, NG (natural gas) is mainly to the high concentration of N2. This makes it possible to consider
used as fuel. Due to the higher concentration of H2O therefore, ue the ue gases as a grey gas without spectral dependence. It also has
gas is considerably different than in CCS and in power plants under to be mentioned that this assumption can lead to a temperature
oxy-fuel condition. The ratio of H2O and CO2 is 0.125 for CCS (i.e. dry under prediction of 100 up to 150 K even in air-fuel condition
ue gas recycling) and for oxy-fuel combustion of NG the ratio [20,21]. The concentration of CO2 and H2O are far higher under oxy-
equals 2 and is therefore considerably different. fuel condition due to the lack of N2 which leads to a strong spectral
CFD (Computational uid dynamics) are mainly used to inves- dependence. Therefore, most models for radiative properties used
tigate heat transfer, uid ow and combustion behaviour. With the in commercial CFD codes are out of their validation range. There are
aid of computer power that has strongly increased in recent years, mainly three different types of models to describe the radiative
it is now possible to perform numerical calculation of industrial properties of gases. These are the line-by-line method, band models
processes, which also provides a good opportunity to get a better and global methods. For CFD calculation, global methods are widely
understanding of combustion behaviour and heat transfer under used due to the low computational effort. One of these global
oxy-fuel conditions. Commercial CFD codes have many models for methods is the WSGGM (weighted sum of grey gases model) rst
transport and heat transfer problems in furnaces. The main goal is proposed by Hottel and Sarom [22]. In this model, the non-grey
to nd the right models to simulate combustion accurately. In gas is assumed to be a mixture of a number of grey gases,
combustion modelling, chemistry and radiative heat transfer were weighted by factors. These factors are derived from more accurate
identied to have the greatest impact on the solution. The aim is to models like the line-by-line methods or the banded models, which
nd radiation and combustion models that are in good accordance are computationally highly demanding. Becher et al. [12] compared
with measurements and are not computationally demanding. different LBL (line-by-line) models (HITEMP2010 [23], HITEMP1995
Global mechanisms have been successfully used in the past for [24], HITRAN2008 [25], HITRAN2004 [26]) and band models (EM2C
simulations of air-red furnaces. Very popular are the two global [27], RADCAL [28], EWB [29]) with gas cell experiments. The
mechanisms by Westbrook and Dryer (WD) [14], which is a 2-step comparison showed that the HITEMP2010 is the most accurate one
mechanism, and the mechanism postulated by Jones and Lindstedt with regard to concentration and temperature. The band models
(JL) [15] which contains 4-steps. In oxy-fuel combustion, the for- are less accurate but not so computationally demanding. In CFD
mation of radicals is mainly promoted by the high temperature. In codes, the WSGGM proposed by Smith et al. [30] is widely used.
the WD and JL mechanisms these radicals are only considered in This WSGGM was derived for a H2O/CO2 ratio of 1 and 2, path
the burning rates and not explicitly in the reactions. That means lengths up to 10 m and for temperatures between 600 and 2400 K
that for the radicals no conservation equations are solved explicitly. from a banded model (EWB). Yin et al. [31] investigated the
Yin et al. [16] performed chemical calculation with WD and JL WSGGM from Smith for different H2O/CO2 ratios, path lengths and
mechanisms under oxy-fuel and air-fuel condition and compared it temperature. The WSGGM showed high deviation for small H2O/
with a detailed mechanism. The results of this calculation showed CO2 ratios and large beam lengths. Similar investigations were done
that the adiabatic ame temperature for the WD and JL mechanism by Becher et al. [32]; they compared the WSGGM with the very
are in a good agreement with the detailed mechanism for air-fuel accurate LBL model (HITEMP2010). The WSGGM showed a
condition. In the case of oxy-fuel combustion, the adiabatic ame maximum deviation of 59% under oxy-fuel conditions. This is the
temperature was massively overestimated (300e500 K). It was reason why many new WSGGM have been published for oxy-fuel
concluded that neglected radical formation in the reactions was the condition recently [13,31,33e37]. These new chemical approaches
reason for the overestimation of the adiabatic ame temperature. and the new radiation models were tested for different furnaces.
Frassoldati et al. [17] did a similar investigation and came to the Yin et al. [16] simulates a 0.8 MW furnace (which will be investi-
same conclusion as Yin et al. They rened the JL mechanism to gated in this study as well) and a 609 MW boiler with different
consider the radicals H, OH and O in the reactions. The temperature combustion mechanisms and two WSGGMs. The simulation
prole calculated with the rened mechanism ts well with the showed that the different WSGGMs make a negligible difference for
detailed mechanism. Glarborg and Bentzen [18] conducted further small-scale furnaces because of the small beam length. For bigger
investigations on the inuence of radicals under oxy-fuel condition. furnaces, the WSGGMs showed a higher deviation from each other.
They studied the dissociation of CO2 with the radicals H, OH, O as Another work published by Galletti et al. [38] focused on oxy-fuel
well as hydrocarbon radicals. The study showed that CO also combustion. The authors investigated a 3 MW semi-industrial
occurred at higher temperatures due to the thermal dissociation of furnace equipped with a low NOx burner. The aim of this study
CO2. It can be concluded that the presence of radicals and high was to validate the different sub models (e.g. combustion/kinetics
temperature in oxy-fuel combustion has a great inuence on the and radiation/spectral) under oxy-natural-gas red conditions by
correct prediction of CO concentration in the furnace but it must comparing of prediction and measurements. Global combustion
not be overlooked that most of the CO production occurs bevor CO2 mechanisms were applied just like WD and JL as well as the rened
production (see equations (1) and (2)). Andersen et al. [19] per- mechanism proposed by Anderson et al. [19]. They used two WSGG
formed simulation in a plug-ow reactor under oxy-fuel condition models with coefcients from Refs. [13,39] to calculate the radiative
with the WD and JL mechanism. As a reference they also calculated properties. Results showed a better agreement with the rened
the plug-ow reactor with the detailed mechanism proposed by global mechanism and with WSGG model optimized for oxy-fuel
Glarborg and Bentzen [18]. The global mechanisms were rened to conditions.
improve the prediction of the species concentration in the plug- It can be seen that lots of investigations have been done so far
ow simulation and as well as in the CFD simulation of a propane on oxy-fuel combustion and radiation under oxy-fuel condition.
oxy-fuel ame. The adapted mechanisms showed a good prediction Most researchers used global mechanisms with a maximum of 6-
of the CO concentration with a lower computational time compared steps to reduce the calculation time. However, most of these
to detail mechanism. global mechanisms were derived for air-fuel combustion and were
As mentioned before, radiation heat transfer is a very crucial therefore insufcient for oxy-fuel combustion. New rened global
part in modelling furnaces. The models for the radiative properties mechanisms were derived for oxy-fuel combustion, but these
of gases, which are implemented in the most commercial CFD mechanisms are not as good as detailed mechanisms. The aim of
codes, are developed for air-fuel condition. In air-fuel condition the this paper is to use detailed chemical kinetic by using the SFM
1480 B. Mayr et al. / Energy 90 (2015) 1478e1489

(steady amelet) and different WSGGM with little computational 3. CFD modelling
effort. Using the steady SFM, calculation time can be reduced from
4 weeks with EDC (eddy dissipation concept) (which is mainly 3.1. Flow modelling
used by other authors) to 4 days on 8 CPU-cores. In this research
the limits of the SFM are also investigated. The furnaces examined The uid ow in the furnaces was calculated by a steady-state
in this paper are a 0.8 MW furnace and an 11.5 kW lab scale CFD simulation using a pressure-based solver. Numerical simula-
furnace. The results of the simulations were compared with tions were performed with the commercial code ANSYS FLUENT 14.
measurements. The ow is calculated by solving the RANS (Reynolds-averaged
NaviereStokes) equations. To solve the RANS equations it is
necessary to have turbulence models to close the RANS equations.
2. The furnaces under simulations There are many turbulence models available in commercial CFD
codes. Two equation models are widely used for simulation of in-
2.1. IFRF 0.8 MW oxy-NG ame furnace dustrial application. In this simulation the standard k-, the real-
izable k- and the Reynolds stress model were used. Pressure-
The furnace has a cross-section of 1.05  1.05 m and a length of velocity coupling was computed with the SIMPLE scheme. For
3.44 m (internal dimension), as shown in Fig. 1. The NG (natural spatial-discretization, second order upwind scheme was applied.
gas) was supplied by a 16 mm tube. Oxygen was fed through an The PRESTO! scheme was used for pressure. Also simulations with
annulus tube into the combustion chamber with an inner diameter the body force weighted scheme were carried out. The solution
of 28 mm and an outer diameter of 36 mm. The diameter of the calculated with these two schemes were compared and showed
chimney is 500 mm. The thermal input is about 0.8 MW and the negligible differences. With the PRESTO! scheme a faster conver-
furnace operates at slightly over pressure of 3 mm H2O to prevent gence of the simulation was observed and therefore it was used in
air leaking in. The properties of the fuel red in furnace, operating the simulations. Gradients were calculated by Green-Gau cell
condition and the measured temperatures are listed in Table 1. based methods. The convergence of the solution was determined
These are the main input data for the CFD simulation. Additional by constant monitors with a maximum allowed uctuation of the
information about the furnace and ame can be found in temperature by 5 K at the outlet and as well on different points
Refs. [40e44]. inside the furnace and a maximum allowed change of the species
concentrations of 0.001 mol fraction at the outlet and different
points inside the furnace. All residuals were below 103 and the
2.2. VO2 furnace
overall heat imbalance was lower than 0.1%.
This lab-scale furnace is equipped with a Messer Oxipyr-Flex
3.2. Boundary condition
burner with a thermal input of 11.5 kW. Fig. 2 shows a 3D model of
the furnace. The natural gas was supplied to the combustion
For both furnaces, pure methane was used as a fuel for the
chamber through a 3.5 mm tube. Oxygen is supplied through an
simulation. In the experiment with the IFRF-furnace they did not use
annulus tube with an inner diameter of 10.5 mm and outer diam-
pure methane as fuel, but Yin et al. [16] showed that the inuence of
eter of 15 mm. The internal dimension of the combustion chamber
the used fuel made a negligible difference. The fuel and the oxidizer
is 0.77  0.75  0.99 m, which is surrounded by an approximately
streams were modelled as mass-ow inlets. The turbulent intensity
0.2 m thick ceramic insulation. The outlet of the furnace is a semi-
was set to 10% for the VO2-furnace and 20% for IFRF-furnace. The
circle with a diameter of 0.33 m. The temperature in the furnace
high inlet turbulent intensity of 20% for the IFRF furnace was used
was measured by ve type B thermocouples. The position of these
because it was also used by the researches of the IFRF [41] [44].
thermocouples can be seen in Fig. 3. The furnace operated under a
Additionally both furnaces have been calculated with a turbulent
positive pressure of 0.4 mbar. Further information about the
intensity of 5%. The calculation showed that the inlet turbulent in-
furnace can be found in Ref. [45].
tensity has negligible inuence on the solution. At the inlet of the
fuel and the oxidizer a turbulent velocity prole (1/7th power ve-
locity prole law) was dened via user dened function to ensure
that the ow is fully developed until it reaches the chamber. The inlet
temperature for the fuel and the oxidizer was set to 25
C for both
furnaces. The simulation for the IFRF-furnace was carried out with an
equivalence ratio of 0.9871 and the simulation of the VO2-furnace
with an equivalence ratio of 0.977. That means that both furnaces
work under slightly fuel-lean conditions. The wall temperature of
the IFRF-furnace was given by function, which can be seen in Table 1.
Table 2 shows the remaining boundary condition of the furnaces.

3.3. Combustion modelling

In combustion modelling, reaction kinetic and turbulent inter-

action is crucial. The correct selection of turbulence/chemistry
interaction affects the solution and computing time. In this study,
three different turbulence/chemistry models were tested with
different combustion mechanisms. The rst combustion model
used is the EDM (eddy dissipation model) proposed by Magnussen
and Hjertager [46]. It is widely used for simulation with air but also
Fig. 1. 3D model of the IFRF 0.8 MW Oxy-NG ame furnace used in the simulation for simulation with oxygen. The model provides a chemical reac-
(units in mm). tion rate, dominated by the large-eddy mixing time scale (mixed is
 B. Mayr et al. / Energy 90 (2015) 1478e1489 1481

Table 1
Fuel composition and operating conditions.

Species CH4 C2H6 C3H8 C4H10 C5H12 CO2 N2 O2

Natural gas composition in vol-% 86 5.40 1.87 0.58 0.14 1.79 4.01 0.21

Operating conditions

Natural gas Oxygen Wall temperatures (K)

Mass ow (kg/h) Temperature (K) Mass ow (kg/h) Temperature (K) T(x) 1700,6 212.59x-46,669x2
63 298.15 224.5 298.15 (x: axial distance from the burner wall, (m))

burnt approach) [47]. Another version of the EDM also takes nite- suggested by Magnussen [50]. The disadvantage of this model is the
rate chemistry into account. This model calculates the Arrhenius long calculation time for the chemistry integration.
reaction rate and the eddy dissipation rate. The nite rate model Another opportunity is to simulate combustion with the
was not used in this investigation due to the dominating eddy amelet model. In this model the thermochemical state of the uid
dissipation rate in the furnaces. In non-premixed combustion the is described by only one parameter called mixture fraction. The
oxidizer and the fuel are slowly mixed by the turbulence in the mixture fraction stands for the mass fraction of all fuel (C, H, ) and
reaction zone where it burns rapidly due to the fact that most fuels oxidizer (O, N, ) elements in all species (H2O, CO2, CO, O, OH, ).
are fast burning. In this case it can be said that the combustion is The advantage of this approach is that the mass balance on the
mixing-limited and the chemical kinetic rates can be neglected. The atomic level avoids source terms from the chemical reactions in the
biggest drawback of this model is that only two reactions can be transport equation of the mean mixture fraction. Through the
considered due to the different Arrhenius rates of multi-step introduction of the mixture fraction the chemistry gets reduced to a
mechanisms [48]. For the simulations carried out with the EDM, simple mixing problem [48]. In this work the SFM (steady laminar
the following 2-step mechanism was used [49]: amelet) model was used. In the SFM approach, a turbulent ame
can be represented by a number of small, one dimensional ame-
lets [51]. These amelets are generated by calculating counter ow
CH4 1:5O2 / CO 2H2 O (1)
diffusion ames. The species mass fraction and the temperature can
be easily calculated with complex chemical kinetic and compared
with measurements in such counter diffusion ames. The amelet
CO 0:5O2 / CO2 (2)
is dened by the mixture fraction and the scalar dissipation rate. A
For air-fuel condition, the restriction of the EDM for 2-step big advantage of this approach is that the calculation of the
mechanism might be sufcient. Due to the high temperature un- chemistry can be pre-processed and stored in look-up tables and
der oxy-fuel condition, dissociation effects and formation of radi- only two additional transport equations for the mean mixture
cals have to be taken into account by the chemical mechanism. To
use a multi-step mechanism, many researchers used the EDC (eddy
dissipation concept) model, which is an enhancement of the EDM.
It assumes that reaction takes place in ne scales and was

Fig. 2. 3D model of the VO2 lab-scale furnace used in CFD. Fig. 3. Position of the temperature measurement points.
1482 B. Mayr et al. / Energy 90 (2015) 1478e1489

Table 2 Table 4
Boundary conditions for different furnaces in the CFD calculations. Settings for amelet calculation.

VO2-furnace IFRF-furnace IFRF/VO2-furnace

Mass ow-rate CH4 (kg/h) 0.828 55.422 Maximum number of amelets 20

Mass ow-rate O2 (kg/h) 3.08 224.49 Number of grid points in amelet 32
Initial scalar dissipation (1/s) 0.01
Convective boundary condition at outer wall
Scalar dissipation step 5
Free stream temperature (K) 298.15 e
Heat transfer coefcient (W/m2K) 30 e

Radiation
Internal emissivity 0.9 0.7 The coefcients for the WSGGM used in this simulation are from
Smith et al. [30], Kangwanpongpan et al. [36] and Yin et al. [31]. The
coefcients from Kangwanpongpan were derived for furnaces with
fraction and mixture fraction variance have to be solved. This ue gas recycling where CO2 and H2O are present in oxidant. This is
means that the mass fraction, density and temperature are only not the case in the calculated furnaces due to the use of natural gas
functions of the mixture fraction and enthalpy (for non-adiabatic as fuel and pure oxygen as oxidant. The ratio of H2O and CO2 in oxy-
conditions). In turbulent ows, only averaged values (Reynolds fuel combustion is between 0.125 and 1 for dry and wet ue gas
and Favre) are used. Therefore a b-PDF (probability density func- recycling respectively. For oxy-fuel combustion without recycling,
tion) is used to predict the mean values of these scalars. The PDF the ratio of H2O and CO2 is 2. Therefore the coefcients from Yin
can be seen as the probability that the instantaneous value is equal et al. were used for negligible H2O and CO2 concentration to
to the averaged value at the same place in the ow eld. The set- calculate the radiative properties of the gas near the burner where
tings used for the PDF generation and amelet generation are given the concentration of CO2 and H2O is low.
in Table 3, and Table 4. For the simulation, two detailed chemical
mechanisms are used. The mechanisms skeletal25 [52] and 4. Results and discussion
smooke46 [53] were tested. The applicability of these mechanisms
for oxy-fuel condition should be pointed out. Furthermore, the 4.1. Effect of grid
limits of the SFM under oxy-fuel condition should be investigated.
Two simulations were carried out with two different meshes to
3.4. Radiation model examine the grid independency of the solution for the IFRF furnace.
The detailed mechanism (skeletal25) with the SFM approach was
The DO (discrete ordination) model is applied to solve the RTE used for each of these simulations. The radiative properties of the
(radiative transfer equation) in all cases. The DO model discretizes ue gas were calculated with the WSGGM from Smith et al. The
each octant of the angular space into azimuth and polar angle. For coarse mesh consists of 1 million hexahedron cells with a
each of these angles the RTE is solved. The number of discretization maximum skewness of 0.56 and a minimal orthogonal quality of
strongly affects the calculation time. Porter et al. [54] investigated 0.81. The mesh is ne near the burner due to high velocity, species
the inuence of the number angle discretization. The investigation and temperature gradients. For the ne mesh, 2.5 million hexahe-
showed that an azimuth and a polar discretization of 3 for each dron cells for a quarter of the furnace were used with a maximum
angle and a discretization of 8 make little difference to the solution. skewness of 0.56 and a minimum orthogonal quality of 0.62. The
Furthermore, they compared the P1 model and the DO model and two different meshes can be seen in Fig. 4 and Fig. 5. The calculated
concluded that the DO model is more accurate than the P1 model. volume fractions (dry) of CO2, H2, CO and O2 are plotted in Fig. 6 and
This is the reason why the DO model is used in this simulation. are confronted with measurements. Fig. 6 shows that the size of the
Under oxy-fuel condition the atmosphere inside the furnace mesh has only a small inuence on the solution. Therefore the
consists mainly of CO2 and H2O and due to the strong absorption mesh with 2.5 million cells is used in all simulations of the IFRF
bands of these gases the gas should be treated as a non-grey gas. 0.8 MW furnace. The grid independency test for the VO2-furnace
The big drawback of a non-grey assumption is that the computa- can be seen in Ref. [45].
tional time dramatically increases. Therefore a grey approach is
used to calculate the radiative properties. The WSGGM is widely 4.2. Effect of combustion modelling
used for this. In this model, the radiative gas properties of the ue
gas are considered as a mixture of grey gases. The emissivity of the Three different models were used to calculate the chemistry
gas mixture is calculated by the sum of the emissivity of all ctive turbulence interaction in the IFRF furnace. These models are the
grey gases combined with a weighting factor as displayed in EDM, EDC and the SFM models. For the EDM model a simple 2-
equation (3). Where a;i stands for the emissivity weighting factor, step mechanism (see equations (1) and (2)) was used. The EDC
ki is the absorption coefcient of i-th ctive gas, p is the sum of all model used a rened version of the Jones and Lindstedt 4-step
partial pressures of all absorbing gases, s is the path length. mechanism [15]. The difference between the original JL 4-step
and the rened mechanism is to use the H2 oxidation model
X
I  
proposed by Marinov et al. [55]. It replaces the reversible H2 re-
a;i T 1  eki ps (3)
ji
action in the original JL 4-step mechanism. This rened 4-step
global mechanism was used by Yin et al. [16] and Kim et al. [56].
Two detailed mechanisms, skeletal25 and smooke46 were used
Table 3
for the SFM model. Using the SFM, it is possible to reduce calcu-
Settings for PDF calculation.
lation time from 4 weeks with EDC to 4 days on 8 CPU-cores. The
IFRF/VO2-furnace temperature proles of the IFRF furnace for the different com-
Number of mean mixture fraction points 40 bustion models are plotted in Fig. 7. It can be seen that the shape
Number of mixture fraction variance points 40 of the ame for the EDC and the EDM model are nearly similar. The
Minimum temperature (
C) 20 major difference between these models is that the cold stream of
Number of mean enthalpy points 41
CH4 and O2 extends further into the chamber with the EDC model.
 B. Mayr et al. / Energy 90 (2015) 1478e1489 1483

Fig. 4. Coarse mesh IFRF 0.8 MW furnace.

Fig. 5. Fine mesh IFRF 0.8 MW furnace.

That means that the ignition of the CH4/O2 mixture is delayed. The the EDC. The SFM with the smooke46 mechanism underestimates
SFM shows a totally different temperature prole compared to the the temperature at x 0.22 m in the same range as the SFM with
EDC and the EDM. It can be seen that ignition of the CH4/O2 the skletal25 mechanism. The temperatures calculated at x 0.82
mixture is even more delayed compared to the EDC model. That and 1.42 m are nearly 1000 K less than the measured temperature.
means that the mixing of the CH4 and the O2 stream is not good This indicates that ignition is delayed and that the mechanism is
enough or cannot be calculated by the SFM. The shape and the not suitable for oxy-fuel combustion. It also has to be considered
temperature of the ame calculated with the SFM and the that for the temperature measurements a water-cooled suction
smooke46 mechanism is considerably different to the other pyrometer was used. Due to the cooling of the gas at the tip of the
models and mechanisms. The temperature calculated with the pyrometer the measured gas temperature is a few hundred de-
smooke46 is in a large area that is far too cold compared to the grees lower than the true gas temperature. Therefore calibration
other simulations. The reason for the low temperature is that the of the suction pyrometer was done by coherent anti-Stokes Raman
ignition with the smooke46 mechanism is even more delayed spectroscopy temperature measurements. But the authors gave no
compared to the skeletal25 mechanism and that the smooke46 value for the measurement uncertainty [43]. In the gures the
mechanism is not suitable for oxy-fuel conditions [45]. To corrected temperatures are plotted. In Fig. 9 the species calculated
emphasise this fact, the different combustion models and mech- by the different combustion models and mechanisms are plotted
anism are compared with measurements at 0.22, 0.82 and 1.42 m against measurements. It can be seen that the SFM with the
(see Fig. 1) from the burner in Fig. 8, showing that the EDM smooke46 mechanism clearly fails to predict the CO2 and the CO
strongly overestimates the temperature at all measured lines. This concentration in the IFRF furnace, which indicates once more that
was expected due to the limitation of the EDM to use a 2-step this mechanism is not suitable for oxy-fuel combustions. The EDM
mechanism, where no radical formations and dissociation can be also clearly fails to predict the CO concentration. The SFM with the
accounted for. The EDC model predicts the temperature near the skeletal25 mechanism can predict the trends of the different
burner (x 0.22 m) well but overrates the temperature at x 0.82 species in the same way as the EDC model does. The measured
and 1.42 m. The prediction of the SFM with the skeletal25 values plotted are averaged values over 60 s. The gas analysers had
mechanism underestimates the temperature near the burner and an accuracy of 0.5% for the species CO, CO2, O2 and H2. The sum of
at x 0.82 m due to delayed ignition. The temperature calculated all measured species was between 97 and 108% what is an indi-
by the SFM at x 1.42 m is overestimated but in the same range as cation for the good quality of the measurements [43].
1484 B. Mayr et al. / Energy 90 (2015) 1478e1489

Fig. 6. Species predicted in the near-burner region of the IFRF furnace for different meshes.

The simulation of the VO2 furnace with the SFM and the skel- good agreement between simulation and measurements.
etal25 showed a different ame shape compared to the IFRF Furthermore comparison shows that the computational model and
furnace. The temperature prole of the VO2 furnace calculated by the measurements are in good agreement for all points and shows
the SFM is plotted in Fig. 10. It can be clearly seen that the ignition the same temperature trends. This shows that the SFM with
is not delayed as in the simulation of the IFRF furnace. Further- skeletal25 works well for oxy-fuel combustion simulations. How-
more, the simulation is confronted with measurements as shown ever, the question remains to why the SFM works for the VO2
in Fig. 11. The maximum temperature of 1142
C was measured at furnace and not for the IFRF furnace. The velocity proles for the
T2, which is placed above the ame. Simulation shows only a de- different furnaces are plotted in Fig. 12. It can be clearly seen that
viation of 5 K from the maximum measured temperature. A dif- the velocity difference between the O2 stream and CH4 stream in
ference between T1 and T2 of 43 K was measured. In the the VO2 furnace is far higher than in IFRF furnace. In the IFRF
simulation the temperature difference between T1 and T2 was furnace, the O2 and the CH4 stream have nearly the same velocity.
calculated with 41 K. The error of the thermocouples for this The fraction of the O2 velocity to the CH4 velocity is 0.21 in the VO2
temperature range is 2.8 K. This means that the deviation between furnace, whereas the velocity fraction in the IFRF furnace is 1.01.
simulation and measurements is nearly in the same order of This indicates furthermore that there is nearly no velocity differ-
magnitude as the measurement uncertainty. This can be seen as a ence between the O2 and CH4 stream in the IFRF furnace. Hence,
with regard to the small shear rate between these two streams the
mixing is bad and cannot be calculated by the SFM approach. The
strain rates for the two furnaces are plotted in Fig. 13. The gure
shows that the strain rate for the IFRF furnace is very high near the
burner, but the strain rate then decreases rapidly until there is
hardly any strain rate between the O2 stream and the CH4 stream.
This indicates furthermore that mixing between the two streams is
poor. In the VO2 the strain rate between the O2 stream and CH4
stream is also high at the beginning but it does not decrease as fast
as in the IFRF furnace. Therefore the mixing between the two
streams is better if the O2 inlet in the IFRF furnace is modied by
changing the diameter to get the same velocity fraction of 0.21 as
in the VO2 furnace. The calculation shows a very similar ame
shape to that in the VO2 furnace as depicted in Fig. 14. The SFM can
calculate ames with high shear rates between the O2 and CH4
streams in good accordance with measurements, but if the shear
Fig. 7. Temperature proles of the IFRF furnace for different combustion models; (a) rates are too low, the SFM fails to predict the temperature prole
EDM; (b) EDC; (c) SFM skeletal25; (d) SFM smooke46. and the shape of the ame.
 B. Mayr et al. / Energy 90 (2015) 1478e1489 1485

Fig. 8. Comparison of the temperature predicted by different combustion models in the IFRF furnace.

Fig. 9. Comparison of different species predicted by different combustion models in the IFRF furnace.
1486 B. Mayr et al. / Energy 90 (2015) 1478e1489

Fig. 10. Temperature prole of the VO2 calculated with the SFM and the skeletal25.

Fig. 13. Strain rate at gases inlet for the; (a) IFRF furnace; (b) VO2 furnace.

turbulence models is negligible. The shape of the ame with the s-

k- and RSM is not considerably different from the shape calculated
with the r-k-. In Fig. 16, the temperature plot of the IFRF furnace
calculated with the RSM and s-k- is shown.

4.4. Effect of radiation modelling

The DO (discrete ordination) model was applied to solve the RTE

in all cases. The radiative properties were calculated with the
WSGGM. For the WSGGM, coefcients from three different authors
were used. These coefcients are from Smith et al. [30], Yin et al.
[31] and Kangwanpongpan et al. [36]. The coefcients from Yin
et al. and Smith et al. were derived from the EWB model with the
Fig. 11. Comparison between measurements and simulation of the VO2 furnace. difference that the coefcients from Yin et al. have a broader vali-
dation range. The WSGGM from Smith et al. is valid for a H2O/CO2
ratio of 1 and 2, path lengths up to 10 m and for temperatures
4.3. Effect of turbulence modelling between 600 and 2400 K. In comparison to that, the WSGGM from
Yin et al. is valid for H2O/CO2 ratios between 0.125 and 4 and path
To evaluate the inuence of the turbulence model, further lengths up to 60 m. Kangwanpongpan et al. derived the coefcients
simulations of the IFRF were conducted with three different tur- for the WSGGM from the highly accurate LBL (HITEMP2010) model.
bulence models. These turbulence models are the realizable k- (r- This WSGGM is valid for H2O/CO2 ratios between 0.125 and 4 and a
k-), the standard k- (s-k-) and the RSM (Reynolds stress model). maximum path length of 60 m. Unfortunately Kangwanpongpan
The solution of the simulations with the SFM, the skeletal25 et al. did not derive coefcients for negligible H2O and CO2 con-
mechanism and the different turbulence models are plotted in centrations arising near the burner. Therefore, the coefcients from
Fig. 15. It can be seen that near the burner (x 0.22 m), the tur- Yin et al. were used. The three WSGGMs are implemented in the
bulence models only have a small inuence on the solution and the CFD code via UDF (user dened functions). The calculated tem-
proles are nearly similar. At x 0.82 m the s-k- and the RSM perature with the SFM, the skeletal25 and the different WSGGM are
improves the solution but the temperature is still too low. In plotted in Fig. 17. The gure clearly indicates that the different
x 1.42 m from the burner the difference between the used WSGGM does not make any difference at all. The reason is that the
beam length (about 0.8 m) in the IFRF furnace is too small to make a
big difference. The inuence of the WSGGM is higher in furnaces
with a larger beam length. Yin et al. [16] simulated a 609 MW utility
boiler with a beam length of 11 m. They simulated the boiler with
WSGGM coefcients from Smith et al. and the coefcient from Yin
et al. [31]. The simulation showed that the calculated absorption
coefcient between these two WSGGMs differs by about 120%
resulting in a temperature difference of 200 K.

Fig. 14. Temperature prole of the IFRF furnace for a velocity ratio of 0.21 between the
Fig. 12. Velocity prole calculated with the SFM; (a) VO2 furnace; (b) IFRF furnace. O2 and CH4 stream.
 B. Mayr et al. / Energy 90 (2015) 1478e1489 1487

Fig. 15. Temperature predicted in the IFRF furnace by the SFM with skeletal25 and different turbulence models.

5. Conclusion time from 4 weeks to 4 days on 8 CPU-cores and is therefore very

interesting for simulations of large furnaces. It was shown that the
The aim of this paper was to simulate two furnaces with detailed SFM with the skeletal25 is in good accordance with the measure-
chemical kinetics and with little computational effort under oxy- ments if the shear rate between the O2 stream and CH4 is high.
fuel conditions. Therefore the steady amelet approach with the Deviation between simulation and measurements was at a
skeletal25 and the smooke46 mechanism was used to simulate the maximum of 5 K, which is a very good accordance for these high
furnaces. Furthermore, simulations with the EDC and the EDM temperatures. For a small velocity difference between the O2 and
were conducted. The EDM used a simple 2-step mechanism which the CH4 streams, which results in a small shear rate, the SFM fails
highly overestimated the temperature in the furnace due to neither to predict the ame shape correctly nor to predict the
neglecting dissociation effects and the radical formations, but temperature. Simulations with the SFM and the smooke46 mech-
which calculates the ame shape correctly. In the simulation with anism showed that the temperature in the furnace is highly
the EDC model, a 4-step mechanism was used which showed good
results compared to the measurements. The big drawback of the
EDC is the long calculation time due to the computationally
demanding chemistry integration. Therefore it is difcult to simu-
late big furnaces with the EDC with reasonable accuracy in an
appropriate amount of time. Using the SFM reduced the calculation

Fig. 16. Temperature proles of the IFRF furnace calculated by the SFM with (a) RSM Fig. 17. Temperature prole in the IFRF furnace calculated with the SFM and different
and (b) skeletal. WSGGM.
1488 B. Mayr et al. / Energy 90 (2015) 1478e1489

underestimated and therefore the smooke46 mechanism is not internationl conference on sustainable fossil fuels for further energy; 2009.
Rome.
applicable for oxy-fuel conditions. The inuence of the turbulence
[18] Glarborg P, Bentzen LLB. Chemical effects of a high CO 2 concentration in oxy-
models on the solution with small shear rates and the SFM was fuel combustion of methane. Energy Fuels Jan. 2008;22(1):291e6.
investigated as well. The standard k-, realizable k- and the Rey- [19] Andersen J, Rasmussen CL, Giselsson T, Glarborg P. Global combustion
nolds stress model were used to evaluate the inuence of turbu- mechanisms for use in CFD modeling under oxy-fuel conditions. Energy Fuels
Mar. 2009;23(3):1379e89.
lence. The impact of the different turbulence models on the [20] Modest MF. The treatment of nongray properties in radiative heat transfer:
solution is negligible and the turbulence models cannot improve from past to present. J Heat Transf 2013;135(6):061801.
the solution calculated with the SFM for small shear rates. [21] Stefanidis GD, Merci B, Heynderickx GJ, Marin GB. Gray/nongray gas radiation
modeling in steam cracker CFD calculations. AIChE J Jul. 2007;53(7):1658e69.
Furthermore, the inuence of the radiative properties on the so- [22] Hottel H, Sarom A. Radiative transfer. New York, USA: McGraw-Hill; 1967.
lution was investigated by applying different WSGGMs from [23] Rothman LS, Gordon IE, Barber RJ, Dothe H, Gamache RR, Goldman A, et al.
different authors. The simulations with the different WSGGMs HITEMP, the high-temperature molecular spectroscopic database. J Quant
Spectrosc Radiat Transf Oct. 2010;111(15):2139e50.
showed that they have a negligible impact on the solution for such [24] Rothman L, Wattson R, Gamache R, Schroeder J, McCann A. HITRAN, HAWKS
small furnaces due to the small beam length, and a CO2/H2O frac- and HITEMP: high-temperature molecular database: SPIE. In: Dainty JC, edi-
tion of 2 due to the use of natural gas as fuel. These results show tor. Atmospheric propagation and remote sensing IV, SPIE; 1995.
[25] Rothman LS, Gordon IE, Barbe A, Benner DC, Bernath PF, Birk M, et al. The
that the SFM with the skeletal25 mechanism is applicable for oxy- HITRAN 2008 molecular spectroscopic database. J Quant Spectrosc Radiat
fuel condition with high shear rates and fails to predict the ame Transf 2009;110(9):533e72.
shape for small shear rates correctly; hence this problem should be [26] Rothman LS, Jacquemart D, Barbe A, Chris Benner D, Birk M, Brown LR, et al.
The HITRAN 2004 molecular spectroscopic database. J Quant Spectrosc Radiat
further investigated.
Transf Dec. 2005;96(2):139e204.
[27] Soufani A, Taine J. High temperature gas radiative property parameters of
statistical narrow-band model for H2O, CO2 and CO, and correlated-K model
Acknowledgements for H2O and CO2. Int J Heat Mass Transf 1997;40(4):987e91.
[28] Grosshandler W. RADCAL: a narrow-band model for radiation calculations in a
combustion environment. Gaitherburg: National Institute of Standards and
This work was nancially supported by the Austrian Research Technology; 1993.
Promotion Agency (FFG), Determination of the radiative heat ux [29] Edwards D. Molecular gas band radiation. Adv Heat Transf 1976;12:115e93.
[30] Smith J, Shen Z, Friedman J. Evaluation of coefficients for the weighted sum of
under oxy-fuel and oxygen enrichment conditions for natural gas- gray gases model. J Heat Transf Nov. 1982;104:602e8.
red furnaces (project 842383, eCall 3861952). [31] Yin C, Johansen LCR, Rosendahl LA, Kr SK. New weighted sum of gray gases
model applicable to computational uid dynamics (CFD) modeling of oxy-
fuel combustion: derivation, validation, and implementation. Energy Fuels
Dec. 2010;24(12):6275e82.
References [32] Becher V, Goanta A, Spliethoff H. Validation of spectral gas radiation models
under oxyfuel conditions e part C: validation of simplied models. Int J
[1] Davison J. Performance and costs of power plants with capture and storage of Greenh Gas Control Nov. 2012;11:34e51.
CO2. Energy Jul. 2007;32(7):1163e76. [33] Johansson R, Leckner B, Andersson K, Johnsson F. Account for variations in the
[2] Manickam B, Dinkelacker F, Lobe T, Tertychnyy M. Enriched oxygen com- H2O to CO2 molar ratio when modelling gaseous radiative heat transfer with
bustion simulation for rotary kiln application. In: presented at the 4th Euro- the weighted-sum-of-grey-gases model. Combust Flame May 2011;158(5):
pean combustion meeting; 2009. Vienna. 893e901.
[3] Pak PS, Lee YD, Ahn KY. Characteristics and economic evaluation of a power [34] Bordbar MH, We cel G, Hyppa nen T. A line by line based weighted sum of gray
plant applying oxy-fuel combustion to increase power output and decrease gases model for inhomogeneous CO2eH2O mixture in oxy-red combustion.
CO2 emission. Energy Aug. 2010;35(8):3230e8. Combust Flame Sep. 2014;161(9):2435e45.
[4] Toftegaard MB, Brix J, Jensen PA, Glarborg P, Jensen AD. Oxy-fuel combustion [35] Rehfeldt S, Kuhr C, Ehmann M, Bergins C. Modeling of radiative properties of
of solid fuels. Prog Energy Combust Sci Oct. 2010;36(5):581e625. an oxyfuel atmosphere with a weighted sum of gray gases for variable carbon
[5] Hammond GP, Akwe SSO, Williams S. Techno-economic appraisal of fossil- dioxide and water vapor concentrations. Energy Procedia Jan. 2011;4:980e7.
fuelled power generation systems with carbon dioxide capture and storage. [36] Kangwanpongpan T, Frana FHR, Corre ^a da Silva R, Schneider PS, Krautz HJ.
Energy Feb. 2011;36(2):975e84. New correlations for the weighted-sum-of-gray-gases model in oxy-fuel
[6] Scheffknecht G, Al-Makhadmeh L, Schnell U, Maier J. Oxy-fuel coal conditions based on HITEMP 2010 database. Int J Heat Mass Transf Dec.
combustiondA review of the current state-of-the-art. Int J Greenh Gas Con- 2012;55(25e26):7419e33.
trol Jul. 2011;5:S16e35. [37] Krishnamoorthy G. A new weighted-sum-of-gray-gases model for CO2eH2O
[7] Oliveira FAD, Carvalho JA, Sobrinho PM, de Castro A. Analysis of oxy-fuel gas mixtures. Int Commun Heat Mass Transf Nov. 2010;37(9):1182e6.
combustion as an alternative to combustion with air in metal reheating fur- [38] Galletti C, Coraggio G, Tognotti L. Numerical investigation of oxy-natural-gas
naces. Energy Dec. 2014;78:290e7. combustion in a semi-industrial furnace: validation of CFD sub-models. Fuel
[8] Be
lohradsk P, Skryja P, Huda k I. Experimental study on the inuence of Jul. 2013;109:445e60.
oxygen content in the combustion air on the combustion characteristics. [39] Coppalle A, Vervisch P. The total emissivities of high-temperature ames.
Energy Oct. 2014;75:116e26. Combust Flame Jan. 1983;49(1e3):101e8.
[9] Prieler R, Demuth M, Spoljaric D, Hochenauer C. Numerical investigation of [40] Lallemant N, Breussin F, Weber R, Ekman T, Dugue  J, Samaniego J, et al. Flame
the steady amelet approach under different combustion environments. Fuel structure, heat transfer and pollutant emission characteristics of oxy-natural
Jan. 2015;140:731e43. gas ames in the 0.7-1 MW thermal input range. J Inst Energy
[10] Kutne P, Kapadia BK, Meier W, Aigner M. Experimental analysis of the com- 2000;73(496):169e82.
bustion behaviour of oxyfuel ames in a gas turbine model combustor. Proc [41] Breussin F, Lallemant N, Weber R. Computing of oxy-natural gas ames using
Combust Inst Jan. 2011;33(2):3383e90. both a global combustion scheme and a chemical equilibrium procedure.
[11] Andersson K, Johansson R, Johnsson F, Leckner B. Radiation intensity of Combust Sci Technol 2000;160(1):369e97.
propane-red oxy-fuel ames: implications for soot formation. Energy Fuels [42] Lallemant NJ, Dugue , Weber R. Measurment techniques for studying oxy-
May 2008;22(3):1535e41. natural gas ames. J Inst Energy 2003;76(507):38e53.
[12] Becher V, Clausen S, Fateev A, Spliethoff H. Validation of spectral gas radiation [43] Lallemant N, Dugue  J, Weber R. Analysis of the experimental data collected
models under oxyfuel conditions. Part A: gas cell experiments. Int J Greenh during OXYFLAM-1 and OXYFLAM-2 experiments. Ijmuiden, IFRF Doc No F 85/
Gas Control Jul. 2011;5:S76e99. y/4. 1997.
[13] Johansson R, Andersson K, Leckner B, Thunman H. Models for gaseous radi- [44] Bollettini U, Breussin F, Lallemant N, Weber R. Mathematical modeling of oxy-
ative heat transfer applied to oxy-fuel conditions in boilers. Int J Heat Mass natural gas ames. Ijmuiden, IFRF Doc No F 85/y/6. 1997.
Transf Jan. 2010;53(1e3):220e30. [45] Prieler R, Demuth M, Spoljaric D, Hochenauer C. Evaluation of a steady
[14] Westbrook C, Dryer F. Simplied reaction mechanisms for the oxidation of amelet approach for use in oxy-fuel combustion. Fuel Feb. 2014;118:55e68.
hydrocarbon fuels in ames. Combust Sci Technol Jul. 1981;27(1e2):31e43. [46] Magnussen B, Hjertager B. On mathematical modeling of turbulent combus-
[15] Jones W, Lindstedt R. Global reaction schemes for hydrocarbon combustion. tion with special emphasis on soot formation and combustion. In: 16th
Combust Flame Sep. 1988;73(3):233e49. Symposium (Int) on combustion. The Combust Inst; 1976.
[16] Yin C, Rosendahl LA, Kr SK. Chemistry and radiation in oxy-fuel combustion: [47] Spalding D. Mixing and chemical reaction in steady conned turbulent ames.
a computational uid dynamics modeling study. Fuel Jul. 2011;90(7): In: 13th Symposium (Int) on combustion. The Combust Inst.; 1970.
2519e29. [48] ANSYS uent theory guid 13.0. ANSYS Inc.; 2010.
[17] Frassoldati A, Cuoci A, Faravelli T, Ranzi E, Candusso C, Tolazzi D. Simplied [49] Dryer F, Glassman I. High-temperature oxidation of CO and CH4. In: 14th
kinetic schemes for oxy-fuel combustion. In: Proceeding of S4FE2009-1st Symposium (Int) on combustion. The Combustion Inst.; 1972.
 B. Mayr et al. / Energy 90 (2015) 1478e1489 1489

[50] Magnussen B. On mathematical modeling of turbulent combustion with [54] Porter R, Liu F, Pourkashanian M, Williams A, Smith D. Evaluation of solution
special emphasis on soot formation and combustion. In: 19th AIAA meeting; methods for radiative heat transfer in gaseous oxy-fuel combustion envi-
1981. St. Louis, USA. ronments. J Quant Spectrosc Radiat Transf Sep. 2010;111(14):2084e94.
[51] Peters N. Laminar diffusion amelet models in non-premixed turbulent [55] Marinov N, Westbrook C, Pitz W. Detailed and global chemical kinetics model
combustion. Prog Energy Combust Sci 1984;10(3):319e39. for hydrogen. In: Chan SH, editor. Transport phenomena in combustion; 1996.
[52] Peeters T. Numerical modeling of turbulence natural-gas diffusion ames. Washington, DC.
Delft: Delft Technical University; 1995 [PhD thesis]. [56] Kim J, Schnell U, Scheffknecht G. Comparison of different global reaction
[53] Smooke M, Puri I, Seshadri K. A comparison between numerical calculations mechanisms for mild combustion of natural gas. Combust Sci Technol
and experimental measurements of the structure of a counterow diffusion 2008;180(4):565e92.
ame burning diluted methane in diluted air. In: 21 st Symposium (Int) on
Combustion. The Combust Inst.; 1986. p. 1783e92.

View publication stats