Anda di halaman 1dari 6

NO formation in flameless combustion: comparison of different modeling

approaches

A. Parente∗,a , A. Cuocib , C. Gallettia , A. Frassoldatib , T. Faravellib , L. Tognottia


a Dipartimentodi Ingegneria Chimica, Chimica Industriale e Scienza dei Materiali, Università di Pisa, Via Diotisalvi 2,
56123 Pisa, Italy
b CMIC Dipartimento di Chimica, Materiali e Ingegneria Chimica, Politecnico di Milano, Piazza Leonardo da Vinci 32,

20133 Milano, Italy

Abstract
The prediction of NO emissions from industrial burners represents a key goal of Computational Fluid Dy-
namics (CFD) aided design. Simplified NO formation mechanisms are usually desirable, to reduce the
computational effort related to the numerical simulations; however, they must be able to capture the NO
trends with acceptable accuracy. Simplified mechanisms for the thermal and prompt NO formation routes
are generally available within the existing commercial CFD packages and they provide acceptable NO predic-
tions at relatively high temperatures. However, when operating at lower temperatures and with hydrogen,
other mechanisms can be relevant, such as those involving N2 O and NNH intermediates. This can become
particularly relevant in non traditional combustion regimes, such as flameless combustion, characterized by
operating temperature far below the levels observed in traditional burners. The present work shows a nu-
merical and experimental investigation of a flameless combustion burner operating with methane-hydrogen
mixtures with a H2 content up to 50% by wt. The work is focused on the requirements of the CFD model for
the accurate prediction of the NO emissions from the burner. In particular, the influence of the combustion
model and kinetic mechanism on the temperature fields on which the NO prediction is based is thoroughly
discussed, together with the simplified NO formation paths to be included in the model. The approach
based on the direct coupling of simplified NO mechanisms to the CFD calculation is compared to a different
methodology, based on the post-processing of the CFD results with detailed kinetic mechanism for the gas-
phase combustion and pollutants formation. A validation methodology is also implemented to quantitatively
assess the degree of agreement between the numerical results and the experiments and to guide the selection
of the modeling parameters required for predicting NO emissions accurately.

1. Introduction process of determining whether or not the math-


ematical models are implemented into computer
Nowadays, numerical simulations represent a fun- code as the programmer intended) and model val-
damental tool for the design and development of en- idation (the process of determining how well the
gineering systems. The so called Computer Aided computer model matches the physical world). In
Engineering (CAE) for “virtual prototyping” and other words, verification is mainly a mathematical
“virtual testing” is continuously growing, due to the issue, while validation is a physical issue.
improved reliability of computational tools and the
high cost and time associated with laboratory test- The present study shows a validation activity car-
ing of components and complete systems. However, ried out on a semi-industrial burner designed for
users and developers of computational tools face to- heating applications in the steel industry [2]. The
day the critical issue of determining how the confi- peculiarity of the system is that to operate in the
dence in modeling and simulations should be crit- so-called flameless combustion regime, which allows
ically assessed, to effectively help decision making to control NOx emissions, even at the highest air-
in new design and regulation (i.e. safety and envi- preheating [3, 4, 5]. In particular, the objective of
ronmental). Verification and Validation [1] (V&V) the study is to compare different modeling strate-
of computational simulations provides a primary gies for the prediction of NO formation . The burner
method for building and identifying this confidence. under consideration has been extensively studied
The methodology is cyclical and involves the activi- in [2] and [6], to investigate the impact of tur-
ties of model developments, model verification (the bulence/chemistry interactions models and kinetic
mechanisms on the NO predictions. Results from
these studies showed that it is possible to capture
∗ E-mail address: alessandro.parente@ing.unipi.it NO trends from the burner with simple NO forma-
Preprint submitted to Proceedings of the European Combustion Meeting 2009 February 9, 2009
tion models (e.g. thermal and prompt simplified bustion chamber are available.
mechanisms), given an adequate description of the In the next Section, a brief description of the
temperature field in the burner. However, the valid- methodology used to assess the level of agreement
ity of such result largely depends on the relatively between experimental data and the computational
high temperature of the combustion process at the results within the V&V framework is provided, fol-
investigated operating conditions. lowed by the description of the modeling approaches
The present numerical investigation refer to ex- and the discussion of the results.
perimental conditions significantly different from
those characterizing the previous campaigns. In 2. Validation metrics
particular, the flameless combustion of methane-
hydrogen mixtures has been investigated in ultra- The level of agreement between experimental
lean conditions. The addition of hydrogen to lean data and numerical results is assessed by means
hydrocarbon flames is of great interest for combus- of validation metrics [18], based on the definition
tion scientists and engineers, as this could improve of confidence intervals for the experimental data,
the burning stability [7, 8]; however, adverse effects which contains the observed mean value of the mea-
on NOx formation could be observed. Thus, oper- sured population within a defined confidence. For
ating in flameless regime could represent a valid so- an arbitrary number of random experiments, it is
lution for H2 -enriched ultra-lean mixtures [9]. From possible to define a standardized random variable,
−µ
ye √
the modeling perspective, the description of NO for- T , as T = s/ n
, where y e is the estimated sample
Pn
mation at low-temperatures requires the incorpora- mean based on n experiments yei , y e = n1 i=1 yei ,
tion of additional mechanisms, beside the ones typ- µ is the true mean, s is the sample standard de-
ically adopted, i.e. thermal and prompt. In par-
h
1
Pn i
2 i 12
viation, s = n−1 i=1 y e − y e , and n is the
ticular, the N2 O intermediate NO mechanism [10]
and the NNH route have been taken into account. number of experiments. Following [18], it is possible
NNH formation was first proposed by Bozzelli and to define a confidence interval for µ as:
Dean [11], who deduced a large kinetic constant for 
s s

the reaction employing the direct oxidation of NNH µ ∼ x − t α2 ,υ √ , x + t α2 ,υ √ (1)
n n
to NO. Such result was confirmed by Hayhurst and
Hutchinson [12], who performed measurements of where ν = n − 1 represents the degrees of free-
NOx in a laminar, premixed, flat CH4 /H2 flame. dom of the population and t α2 ,υ is the 1− α2 quantile
Such formation route must be included in the over- of t-distribution with υ degrees of freedom. There-
all NO formation scheme when operating at low- fore, it is possible to state that the true mean is
temperatures and with significant amounts of hy- contained in the interval given by Eq. (1) with a
drogen in the fuel, to explain NOx concentration confidence, C, equal to 100 · (1 − α) %. Common
too high to be entirely determined by the Zeldovich values for C are 0.9, 0.95, 0.98 and 0.99 which cor-
mechanism. respond to values of α equal to 0.1, 0.05, 0.02 and
0.01, respectively.
This paper presents a comparison of different
The construction of an error validation metric
modeling strategies for the prediction of NO in
requires the evaluation of two quantities: the es-
ultra-lean conditions and at low temperature. In
timated error and the interval containing the true
particular, the approach based on the direct cou-
error:
pling of simplified NO mechanisms to the CFD cal-
culation is compared to a different methodology, e = ym − y e
E E = ym − µ (2)
based on the post-processing of the CFD results
with detailed kinetic mechanisms for the gas-phase Following Eqs. (1) and (2), the following expres-
combustion and pollutants formation. The latter sion for the confidence interval of the true error is
approach is justified by the fact that NOx chem- obtained:
istry barely influences the overall temperature and  
s e s
flow field. Therefore, the flow field is taken from E ∼ E − t 2 ,υ √ , E + t 2 ,υ √
e α α (3)
the CFD simulation and a detailed kinetic calcula- n n
tion is carried out on a reactor network built from With respect to the metric defined in Eq. (3), it is
the CFD grid, using criteria based on the similari- important to mention that the confidence intervals
ties of local temperature and species concentration are defined solely on the basis of the experimental
[13, 14, 15, 16, 17]. Computational results are com- data, since the true mean is the real unknown of the
pared to experimental data collected on the recu- problem.
perative burner operating in flameless combustion Equations (2)-(3) can be easily extended to the
regime, with hydrogen contents up to 50% (by wt.). case of a system quantity response monitored over a
Data regarding the composition of the flue gases range of the input variable. Moreover, in some ap-
and the temperature on the outer wall of the com- plications, it is desirable to provide global metrics,
2
able to summarize the level of agreement for a large burner under investigation is designed for heating-
number of computational models and experimental up applications, in environment where the oxidizing
data. In particular, an average error metric can be atmosphere must be kept separated from the pro-
defined as: cess atmosphere (e.g. reducing), as it happens in
the steel industry for annealing processes. There-
Pp Pp fore, the code was coupled with a subroutine for
Ee 1 ym (xi )−ȳe (xi ) 1 E(x
e i)
ȳe = p i=1 ȳe (xi ) = p i=1 ȳe (xi ) the evaluation of the radiative heat transfer from
avg
(4) the radiant tube. Details of the subroutine can be
found elsewhere [2].
The confidence interval to be associated with the
average error metric is the normalized average con- The direct coupling of detailed kinetics and fluid
fidence interval: dynamics was compared to a different approach, the
kinetic post-processing (KinPP), based on the gen-

CI t α ,υ Pp

s(xi )
eral concept of “reactor network analysis” [13, 14]
ȳe = p1 √2 n i=1 ȳe (xi ) (5) and already applied to industrial burners [27, 28]
avg
for the prediction of different pollutants (NOx , CO
and unburned hydrocarbons). The first step for the
3. Modeling approaches KinPP application is the CFD calculation of the
flame structure using a simplified kinetic scheme,
Numerical simulations were performed with the based on a small number of species and global re-
commercial code FLUENT 6.3 by Ansys Inc. Favre- actions. The knowledge of the thermo-fluid dy-
averaged Navier-Stokes equations were solved using namic field allows to identify critical and non criti-
the standard k-ε turbulence model. The Eddy Dis- cal zones in the overall reacting system and to lump
sipation Concept (EDC) by Magnussen [19, 20] was (or group) several cells, similar from a kinetic point
taken into account to model turbulence/chemistry of view, into a single lumped reactor, thus reducing
interactions and to include detailed chemistry in the dimensions of the overall system. A fixed av-
the numerical simulation. Two different kinetic erage temperature is assumed in each reactor and
mechanisms were used to describe the oxidation of the rates of all the reactions involved in the ki-
methane/hydrogen mixtures, the KEE58 [21] and netic scheme are evaluated accordingly. However,
the DRM-19 [22]. Thermal NO formation was mod- in turbulent combustion conditions, a proper cor-
eled using a Finite Rate (FR) approach and a sim- rection must be applied to the reaction rates, to
plified one-step kinetic mechanism, available in the take into account the effects of temperature fluctu-
code [23] and obtained from the Zeldovich scheme ations [27]. These corrections are very relevant for
by assuming a steady state for the N radicals and re- thermal NOx formation reactions, characterized by
lating the O radical concentration to that of oxygen high activation energies. Additional details on the
by means of the dissociation reaction [24]. Similarly, Kinetic Post-Processor and its numerical method
prompt NO formation from methane was modeled are reported in [13, 14].
following De Soete [25]. The N2 O intermediate [10] The numerical simulations to be post-processed
and NNH [11] routes were also included in the mod- with the kinetic post-processor were carried out us-
eling approach, to account for low-temperature for- ing a simplified kinetic scheme for the CFD sim-
mation mechanisms and for the presence of H2 frac- ulations. In particular, a 4-steps mechanism for
tions in the fuel. The NNH route is not directly methane-hydrogen mixtures was adopted [29]. The
available in the code; therefore, it was implemented detailed kinetic scheme, used in the post-processing
by means of a user defined subroutine (UDF) follow- step, is able to describe the oxidation of heavy hy-
ing Konnov et al. [26]. The kinetic rates were inte- drocarbon fuels, up to diesel and jet fuels, and its
grated over a probability density function (PDF) for main features were already discussed in the litera-
temperature in order to take into account the effect ture [30]. The chemistry of nitrogen compounds is
of turbulent fluctuations on the mean reaction rates also discussed elsewhere [31, 32]. The thermody-
[23]. Additional detailes regarding the modelin ap- namic properties are taken from CHEMKIN Ther-
proach (i.e. radiative model, spectral properties, ...) modynamic Database [33]. The overall mechanism
can be found in [6]. is freely available at the following web address:
As far as the boundary conditions are concerned, www.chem.polimi.it/CRECKModeling.
two user defined functions were coupled with the A scheme of the conceptual path followed in the
code in order to make the burner model self- present study in shown in Figure.
sufficient. The air inlet temperature was evaluated
as a function of the exhaust gases temperature,
using an empirical correlation available from the 4. Results
burner supplier. Moreover, particular attention was
paid to the boundary condition at the wall delim- The present Section first describes the results of
iting the combustion chamber (radiant tube). The the direct coupling of detailed kinetics and fluid
3
mass flow rate for the combustion of methane fuels.
Figure 3 shows the 98% confidence intervals for
the average temperature at the radiant tube, to-
gether with the comparison between experiments
and model predictions, when increasing the hydro-
gen thermal content from 10% to 70%. The results
provided by the KEE58 and DRM19 models are
very similar, as expected, and in very good agree-
ment with the experimental mean for the 10% and
20% cases, while both the models under predict the
average radiant tube temperature for all the other
cases. The maximum departure is observed for the
70% H2 thermal input and is around 25 K. This
slight under prediction is actually expected, due to
the geometric simplification and the choice of a 2D
model for the burner, instead of a 3D grid, which
determines an over prediction of the recirculation
Figure 1: Conceptual scheme of the numerical simula- degree, thus leading to a reduction of the temper-
tions.
ature levels. Such result can be considered accept-
able, given all the experimental and numerical un-
certainties of the problem.
dynamics within the CFD simulations. Then, the
comparison with the kinetic post-processing is pro-
posed. The contour plots of temperature in the
burner fed with increasing fractions of H2 (from 10
to 70 % of the total thermal input) are shown in
Figure 2.

(a)

Figure 2: Temperature field in the burner fed with


increasing H2 in the fuel. H2 content corresponding to
(a) 10%, (b) 20%, (c) 30%, (d) 50%, (e) 60% and (f)
70% of the total burner thermal input, 10 kW. Kinetic
mechanism: KEE58 [21].

It can be observed how the addition of hydro-


gen results in a combustion region which is progres- (b)
sively shifted toward the burner exit. The maxi-
mum temperature is not observed for the case cor- Figure 3: Experimental and computed average temper-
responding to the highest hydrogen content in the atures at the radiant tube (a) and NO emissions in the
fuel; however, this is due to the fact that the air- flue gases (b), for different values of hydrogen thermal
excess was found to increase almost linearly (from input. Experimental data are presented with associated
120 to 170 %) when increasing the hydrogen ther- 98% confidence intervals.
mal input (from 10 to 70%). This is due to the
setting of the burner, designed to control the air Figure 3 shows the 98% confidence intervals for
4
the average NO emissions from the burner, to- Table 1: Average and maximum global validation met-
gether with the comparison between experiments rics for the average NO emissions from the burner. In
and model predictions. Different variants of the the Table, K and D denote the KEE58 and DRM19
model, including NNH, have been tested, due to an mechanisms.
uncertainty in the activation energy for the NO for-
Ee CI
mation reaction via NNH [12]. It can be observed Model ȳe ȳe
avg avg
that NO formation is very sensible to such parame-
K - Th+Pr 0.992 ±0.013
ter, especially at high H2 concentrations. It is also
K - Th+Pr+N2 O 0.649 ±0.013
noteworthy to mention that the confidence inter-
vals for NO are very narrow, limited to fractions of K - Th+Pr+N2 O+NNH 0.184 ±0.013
ppmv . From Figure 3, it is clear how the role of K - Th+Pr+N2 O+NNH
0.081 ±0.013
the NNH route is critical to explain the NO emis- (Ea +2.5kJ/mol )
sions. Without considering such route the resulting D - Th+Pr 0.991 ±0.013
error would be huge. A considerable improvement D - Th+Pr+N2 O 0.644 ±0.013
in the prediction is also obtained by varying the D - Th+Pr+N2 O+NNH 0.237 ±0.013
activation energy for the NNH route within the un- D - Th+Pr+N2 O+NNH
0.155 ±0.013
certainty range (±5 kJ/mol) indicated by Hayhurst (Ea +2.5kJ/mol )
and Hutchinson [12]. The value of 2.5 kJ/kmol kinPP - No T fluctuations 0.207 ±0.013
seems the one providing, in all cases. better agree- kinPP - T fluctuations 0.077 ±0.013
ment with the experimental data. Moreover, the
KEE58 mechanism appears to provide always bet-
the more accurate is the prediction of the thermal
ter results than the DRM19 mechanism.
field by the CFD calculation, the more reliable the
Such analysis can be quantitatively supported by KinPP predictions are expected. The effects of tem-
the global validation metrics provided in Table 1. perature fluctuations on the formation of NO were
The values confirm that the KEE58 kinetic mech- also investigated. Even if in ultra-lean conditions
anisms is able to better capture the trend of NO the temperature fluctuations are relatively small,
emissions over the range of hydrogen thermal in- their effect on the NO emissions is not negligible.
puts, when NNH is coupled to the NO model. In Figure 3 shows a decrease of 25-30% of the NO
particular the combination of KEE58 and of a NO emissions, when temperature fluctuations are not
model with an activation energy for the NO forma- taken into account. This leads to an increase in the
tion mechanism via NNH increased by 2.5 kJ/mol, average validation metric, as pointed out in Table
results in an average error metric of 8.1%±1.3% 1.
(with 98% confidence). This is a very interest-
ing results, considering the extremely challenging
conditions for the prediction, i.e. ultra-lean Conclusions
and
hydrogen-rich. All the other values for Ee/ȳe avg
A comparison between different modeling ap-
are well above the values indicated for the selected proaches and validation strategies for the prediction
model, which will be referenced in the following sim- of NO emissions from a semi-industrial flameless
ply as complete NO mechanism. burner has been presented. The direct computation
The results of the kinetic post-processing of the of NO in the CFD calculation, by means of detailed
CFD calculation, carried out with a simplified ki- mechanisms for gas-phase oxidation and simplified
netic mechanism (Section 3), are summarized in approaches for NO formation, has been compared to
Figure 3 and Table 1. The comparison with the the kinetic post-processing of CFD simulations, us-
experimental data can be considered satisfactory: ing very detailed schemes for pollutant formation.
the effect of H2 addition is well represented by Results have been compared to the experimental
the model and quantitative agreement is reasonably data using validation metrics for the quantitative
good. In particular, the average validation metrics assessment of the level of agreement. With respect
in Table 1 show a reduction of the average error to the CFD simulations, the analysis showed that
with respect to the results obtained with the direct it is possible to achieve a reasonable agreement be-
coupling of detailed kinetic in the CFD simulation, tween simulations and experiments only by taking
decreasing from the value of 15.5%, given by the into account detailed kinetic schemes for gas-phase
complete NO mechanism, to 7.7%. combustion and including, with simplified one-step
Figure 3 also shows that the predicted NOx emis- mechanisms, all the relevant sources of NO at the
sions are always larger than the experimental values investigated operating conditions. On the other
and this seems to indicate that the CFD simulation hand, the kinetic post-processing allows to adopt
tends to over-predict the temperature. This result is simpler kinetic mechanisms in the CFD simulations,
somehow expected, being the numerical simulation as the pollutant formation is computed using very
carried out with a global kinetic scheme. Of course, detailed schemes.
5
References Reaction in Turbulent Flow, in: 19th AIAA Aerospace
Science Meeting, 1981.
[1] W. L. Oberkampf, T. G. Trucano, Verification and Val- [20] I. R. Gran, B. F. Magnussen, A Numerical Study of
idation in computational fluid dynamics, Progress in a Bluff-Body Stabilized Diffusion Flame. Part 2. Influ-
Aerospace Sciences 38 (3) (2002) 209–272. ence of Combustion Modeling and Finite-Rate Chem-
[2] C. Galletti, A. Parente, L. Tognotti, Numerical and ex- istry, Combustion Science and Technology 119 (1996)
perimental investigation of a mild combustion burner, 191–217.
Combustion and Flame 151 (2007) 649–664. [21] R. Bilger, S. Starner, R. Kee, On Reduced Mechanisms
[3] J. A. Wünning, J. G. Wünning, Flameless oxidation to for Methane-Air Combustion in Nonpremixed Flames,
reduce thermal NO-formation, Progress in Energy and Combustion and Flame 80 (1990) 135–149.
Combustion Science 23 (1997) 81–94. [22] A. Kazakov, M. Frenklach URL http://www.me.
[4] A. Cavaliere, M. de Joannon, Mild Combustion, berkeley.edu/drm/.
Progress in Energy and Combustion Science 30 (2004) [23] A. Inc., Fluent 6.3 User Guide, 2006.
329–366. [24] A. A. Westenberg, Kinetics of NO and CO in lean, pre-
[5] A. K. Gupta, Thermal characteristics of gaseous fuel mixed hydrocarbon-air flames, Combustion Science and
flames using high temperature air, Journal of Engineer- Technology 4 (1971) 59–64.
ing for Gas Turbines and Power 126 (2004) 9–19. [25] G. G. D. Soete, Overall Reaction Rates of NO and N2
[6] A. Parente, C. Galletti, L. Tognotti, Effect of the com- Formation from Fuel Nitrogen, Proceedings of the Com-
bustion model and kinetic mechanism on the MILD bustion Institute 15 (1974) 1093–1102.
combustion in an industrial burner fed with hydrogen [26] A. A. Konnov, G. Colson, J. D. Ruyck, The New Route
enriched fuels, International Journal of Hydrogen En- Forming NO Via NNH, Combustion and Flame 121
ergy 33 (2008) 7553–7564. (2000) 548–550, brief Communication.
[7] H. Guo, G. J. Smallwood, F. Liu, Y. Ju, . L. Gülder, The [27] A. Cuoci, A. Frassoldati, G. Buzzi-Ferraris, T. Faravelli,
effect of hydrogen addition on flammability limit and E. Ranzi, The ignition, combustion and flame structure
NOx emission in ultra-lean counterflow CH4/air pre- of carbon monoxide/hydrogen mixtures. Note 2: Fluid
mixed flames, Proceedings of the Combustion Institute dynamics and kinetic aspects of syngas combustion, In-
30 (2005) 303–311. ternational Journal of Hydrogen Energy 32 (2007) 3486–
[8] G. J. Rørtveit, J. E. Hustad, S. C. Li, F. A. Williams, 3500.
Effects of diluents on NOx formation in hydrogen coun- [28] A. Frassoldati, A. Cuoci, T. Faravelli, E. Ranzi,
terflow flames, Combustion and Flame 130 (2002) 48– S. Colantuoni, P. D. Martino, G. Cinque, Experimental
61. and modeling study of a low NOx combustor for aero-
[9] M. Derudi, A. Villani, R. Rota, Mild combustion of in- engine turbofan, Combustion Science and Technology
dustrial hydrogen-containing byproducts, Industrial & 181 (2009) in press.
Engineering Chemistry Research 46 (2007) 6806–6811. [29] C. Westbrook, F. Dryer, Chemical Kinetic Modelling
[10] P. Malte, D. Pratt, Measurement of atomic oxygen and of Hydrocarbon Combustion, Progress in Energy and
nitrogen oxides in jet-stirred combustion, Proceedings Combustion Science 10 (1984) 1–57.
of the Combustion Institute 15 (1975) 1061–1070. [30] E. Ranzi, M. Dente, A. Goldaniga, G. Bozzano, T. Far-
[11] J. W. Bozzelli, A. M. Dean, O + NNH — A Possible avelli, Lumping procedures in detailed kinetic modeling
New Route for NOx Formation in Flames, International of gasification, pyrolysis, partial oxidation and combus-
Journal of Chemical Kinetics 27 (1995) 1097–1109. tion of hydrocarbon mixtures., Progress in Energy and
[12] A. N. Hayhurst, E. M. Hutchinson, Evidence for a New Combustion Science 27 (2001) 99–139.
Way of Producing NO via NNH in Fuel-Rich Flames [31] T. Faravelli, A. Frassoldati, E. Ranzi, Kinetic Modeling
at Atmospheric Pressure, Combustion and Flame 114 of Mutual Interactions in NO-Hydrocarbon Low Tem-
(1998) 274–279. perature Oxidation, Combustion and Flame 133 (2003)
[13] M. Falcitelli, S. Pasini, L. Tognotti, Modelling practi- 188–207.
cal combustion systems and predicting NOx emissions [32] T. Faravelli, A. Frassoldati, E. Ranzi, Kinetic Modeling
with an integrated CFD based approach, Computers of Mutual Interactions in NO-Hydrocarbon High Tem-
and Chemical Engineering 26 (2002) 1171–1183. perature Oxidation, Combustion and Flame 135 (2003)
[14] M. Falcitelli, S. Pasini, N. Rossi, L. Tognotti, 97–112.
CFD+reactor network analysis: an integrated method- [33] R. J. Kee, F. Rupley, J. A. Miller, The CHEMKIN Ther-
ology for the modelling and optimisation of industrial modynamic Data Base, Report SAND89-8008, Sandia
systems for energy saving and pollution reduction, Ap- National Laboratories, Livermore, CA, 1989.
plied Thermal Engineering 22 (2002) 971–981.
[15] T. Faravelli, L. Bua, A. Fassoldati, A. Antifora, L. Tog-
notti, E. Ranzi, A new procedure for predicting NOx
emissions from furnaces, Computers and Chemical En-
gineering 25 (2001) 613–618.
[16] A. Frassoldati, S. Frigerio, E. Colombo, F. Inzoli,
T. Faravelli, Determination of NOx emissions from
strong swirling confined flames with an integrated
CFD-based procedure, Chemical Engineering Science 60
(2005) 2851–2869.
[17] M. S. Skjøth-Rasmussen, O. Holm-Christensen, M. Øst-
berg, T. S. Christensen, T. Johannessen, A. D. Jensen,
P. Glarborg, H. Livbjerg, Post-processing of detailed
chemical kinetic mechanisms onto CFD simulations,
Computers and Chemical Engineering 28 (2004) 2351–
2361.
[18] W. L. Oberkampf, M. F. Barone, Measures of agree-
ment between computation and experiment: validation
metrics, Journal of Computational Physics 217 (2006)
5–38.
[19] B. F. Magnussen, On the Structure of Turbulence and
a Generalized Eddy Dissipation Concept for Chemical
6

Anda mungkin juga menyukai