Anda di halaman 1dari 6

Computers and Chemical Engineering 33 (2009) 402–407

Contents lists available at ScienceDirect

Computers and Chemical Engineering


journal homepage: www.elsevier.com/locate/compchemeng

Theoretical and numerical investigation on the EDC-model for


turbulence–chemistry interaction at gasification conditions
M. Rehm ∗ , P. Seifert, B. Meyer
Institute of Energy Process Engineering and Chemical Engineering, TU Bergakademie Freiberg, Reiche Zeche/Fuchsmühlenweg 9,
Haus 1, 09596 Freiberg, Germany

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

Article history: The present work describes results of the CFD modelling of a high-pressure gasification process. For the
Received 20 February 2008 CFD modelling of combustion processes in industrial-scale reactors the eddy dissipation concept (EDC)
Received in revised form 7 November 2008 is often used to couple chemical reaction mechanisms to the computed flow field. However, in modelling
Accepted 12 November 2008
gasification processes the EDC-model yields results, which only partly agree with the experimentally
Available online 21 November 2008
gained data. Since the available EDC-model does not accurately represent the important reformation reac-
tions for gasification modelling, a variation of the EDC parameters for the mixing time and the fine scale
Keywords:
fraction of a cell was accomplished. Hence, the experimental values were approached but not reached.
EDC
Eddy dissipation concept
In the present contribution on the basis of an analysis of the EDC-model and the involved time scales
Turbulence–chemistry interaction weaknesses of the EDC-scheme in reforming zones are discussed and a possible solution is demonstrated.
Gasification © 2008 Elsevier Ltd. All rights reserved.
Partial oxidation

1. Introduction and problem formulation Hence, further studies on the accuracy of the POX model were car-
ried out and are the focus of the present paper.
The present work describes results of the CFD modelling of a
high-pressure gasification process in the HP-POX pilot plant (5 MW
2. Model setup
thermal, pressure up to 100 bar) at the Institute of Energy Process
Engineering and Chemical Engineering. High pressure partial oxi-
Since the computational demand of large eddy simulations (LES)
dation (HP-POX) is a multivalent thermal conversion process with
or even direct numerical simulation (DNS) is still very high, in
which, e.g., natural gas, crude oil stranded gas or heavy residues
simulating large scale flow phenomena, the Reynolds-averaged
from the crude oil processing can be converted with oxygen and
Navier Stokes (RNAS) equations in conjunction with turbulence
steam to fuel gases, synthesis gases and others for example for
modelling are often used. In the present work, a steady-state
methanol synthesis. Three different operational modes of the HP-
solver of the commercial CFD code FLUENT (version 6.1.22) was
POX pilot plant are available: non-catalytic partial oxidation of
used. The SIMPLE algorithm for pressure–velocity coupling was
methane (POX), catalytic partial oxidation of methane (ATR) and
selected and the computational domain was discretized by a two-
gasification of liquid fuels (MPG). In the context of the research
dimensional axisymmetric grid of approximately 10,000 cells with
project COORAMENT (Meyer, Seifert, Zeißler, & Walter, 2005) tools
gradual refinement towards the inlets. A scheme of the reactor is
for the detailed modelling of gasification processes are in develop-
shown in Fig. 1. The original reactor stands upright with the burner
ment. First CFD models for POX and ATR were presented by Zeißler
on top. Natural gas, oxygen and steam are fed separately into the
(2006). With the help of these models retention time distribu-
reaction chamber. Subsequently they are heated up and mixed and
tions were computed and verified using radioactive tracers (Zeißler,
ignite under the oxygen-rich conditions close to the burner. Fur-
2006; Zeißler, Meyer, Seifert, Heinzel, & Jentsch, 2006). In the case
ther downstream a recirculation zone is formed. In the recirculation
of non-catalytic partial oxidation the modelled and measured exit
zone and in the tube the endothermic reformation reactions, which
compositions were in reasonable agreement in terms of the main
produce the synthesis gas, dominate the chemical processes. At the
components H2 , CO, CO2 and H2 O. However, significant deviations
outlet of the reactor hot synthesis gas is sampled and the composi-
arose in the computed temperatures and the concentration of CH4 .
tion and temperature is determined.
The coupling of the flow field to the chemical reaction progress
is critical for the simulation of turbulent reactive flows and done
∗ Corresponding author. Tel.: +49 3731 394553; fax: +49 3731 394555. by combustion models. Common approaches for non-premixed
E-mail address: markus.rehm@iec.tu-freiberg.de (M. Rehm). combustion modelling are models based on presumed probabil-

0098-1354/$ – see front matter © 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compchemeng.2008.11.006
M. Rehm et al. / Computers and Chemical Engineering 33 (2009) 402–407 403

Fig. 1. Outline of the reactor geometry (not drawn to scale).

ity density functions (PDFs), on transported PDF and the eddy Table 1
Experimental and modeled synthesis gas compositions at 30 bar from Zeißler (2006)
dissipation concept (EDC). The use of PDF for the calculation of
and the equilibrium point calculated with ATRMech. C2 corresponds to the sum of
turbulence–chemistry interaction goes back to Pope (1985). In the all C2 -species. All numbers for compositions are in [mol/mol].
simplest variant of pre-PDF, equilibrium assumptions are made and
Experiment Model Equilibrium
all mixing and reaction processes can be reduced to the mixture
fraction of fuel and oxidizer and its fluctuation. Transported PDF H2 0.4932 0.4615 0.5269
models use an exact representation for chemistry, but need further CH4 0.0027 0.0320 0.0003
CO 0.2653 0.2546 0.2751
modelling for the mixing. This is usually done by Monte–Carlo sim-
CO2 0.0356 0.0307 0.0285
ulations for a sufficiently large number of particles. Excellent results H2 O 0.2012 0.2091 0.1675
were achieved with this model in the context of turbulent non- N2 0.0020 0.0086 0.0017
premixed flames (Kempf, 2006), though the computational demand C2 0.0000 0.0035 0.0000
T (K) 1663 1783 1615
is very high.
In the simulation discussed here the EDC-model (Magnussen &
Hjertager, 1976) was chosen because equilibrium assumptions of
the pre-PDF model are not fulfilled. Furthermore, simulations with the chemical kinetics package CANTERA (Goodwin, 2002) with the
the transported PDF combustion model are still very expensive but complete GRI 3.0 mechanism and the reduced ATRMech. For the PFR
are considered for future work. In conjunction with ISAT tabula- input a mixture from the CFD simulation at the entry of the tube was
tion (Pope, 1997) detailed chemical reaction mechanisms can be chosen. It was shown that the methane concentration (see Fig. 2)
applied. The EDC is a compromise of accuracy and computational and the temperature fields (not shown here) were nearly identi-
cost and was successfully applied to different regimes of combus- cal in both PFR calculations. The reduction of the mechanism was
tion (Chakraborty, Paul, & Mukunda, 2000; Gran & Magnussen, thus not recognized as the cause of the modelling error. Addition-
1996) and steam cracking (Stefanidis, Merci, Heynderickx, & Marin, ally, with assumption of ideal mixing in the PFR, the reformation
2006) using detailed reaction mechanisms. The EDC-model is more proceeds much faster than in the CFD case with the EDC-model.
thoroughly explained in Section 4. For turbulence modelling the In a further step of preliminary work the effects of the turbu-
two-equation Realizable k–  model (Shih, Liou, & Shabbir, 1995), lence model were investigated for conditions at the reactor outlet. A
which is recommended for axisymmetric jet flows (Fluent, 2005), variation of the parameters of the realizable k–  turbulence model
was used with standard parameters. For radiation the Rosseland was accomplished and in addition comparative calculations with
model was applied together with the weighted-sum-of-gray-gases the more complex Reynolds stress model (Launder, Reece, & Rodi,
model (WSGGM) to compute the absorption and emission coef- 1975) were completed. As a result, changes of the flame length could
ficients of the fluid. A reduced version of the GRI3.0 mechanism be observed, but a considerable influence on the exit condition
(Smith et al., 1999) called ATRMech (28 species, 114 reactions) regarding the critical parameters CH4 concentration and temper-
was used as chemical reaction mechanism. This mechanism was
obtained by analysis of sensitivities, assumption of quasi steady
states and partial equilibria. For more details see the work of Zeißler
(2006). Previously, detailed chemical mechanisms have success-
fully been applied to partial oxidation conditions by Zhu, Zhang,
and King (2001). Konnov, Zhu, Bromly, and Zhang (2003) found that
results for GRI-Mech 3.0 were satisfactory at temperatures above
1000 K.

3. Preliminary work

A summary of modelling results from the previous work of


Zeißler (2006) is shown in Table 1. One can see that the experi-
mental values and the modelling results for the main components
H2 , CO and CO2 are in reasonable agreement but deviations for
temperature and the methane concentration are large. In order to
analyse the cause of the deviations more thoroughly, the validity
of the reduced reaction mechanism, the turbulence model and the
turbulence–chemistry interaction were examined.
In order to analyze the reduced reaction mechanism, the lower
Fig. 2. CH4 conversion for PFR calculations and CFD results in the tube on the sym-
part of the reactor (“tube” in Fig. 1) was represented by an ideal metry axis. The methane concentration is normalized by the initial concentration of
plug flow reactor (PFR). The PFR calculations were performed using CH4 at the inlet.
404 M. Rehm et al. / Computers and Chemical Engineering 33 (2009) 402–407

ature could not be seen. Thus the standard parameters (Shih et al., Here yio is the composition of the mixture entering the reactor and
1995) of the turbulence model were kept for all results presented can be obtained with:
here.
yi =  3 yi ∗ + (1 −  3 )yio , i = 1, . . . , ns (6)
Afterwards, the coupling of chemistry and the turbulence by the
EDC-model was more thoroughly examined and a clear dependence where yi is the cell-averaged composition. Furthermore, ns is the
of the reformation behaviour on the EDC parameters was observed. number of species, ωi ∗ the reaction rate of the ith specie and ∗ the
First results are presented in Section 5. density for a gas with the composition y∗ , which marks the compo-
sition of the mixture at the exit of the ideal reactor of Eq. (5). The
4. Theoretical fundamentals of the EDC-model reaction rate ωi ∗ is calculated with a detailed reaction mechanism.
The mixture exchange of the reactive and the inert portion is
In simulating combustion systems, both physical and chemical modelled by  2 / as proposed by Magnussen (1989), Gran and
processes are of great importance. They can act in vastly different Magnussen (1996). The required expression for the mean reaction
time scales. Thereby the chemical processes cover the entire spec- rate ωi of a specie i over a whole cell is then:
2  ∗ 
trum of time scales of the physical processes (Warnatz, Maas, &
ωi
Dibble, 2001). The formation of turbulent structures is a physical = yi − yio , i = 1, . . . , ns (7)
¯ 
process, for which the concept of the eddy cascade model is the
basis: The largest turbulent structures have a size, which is of the where yi ∗ is the steady-state solution of Eq. (5) and yio is the compo-
magnitude of the system’s dimensions. By interaction among them- sition in the inert part. Using Eq. (6), Eq. (7) can be re-written with
selves, the eddies dissipate to smaller ones. The smallest eddies cell-averaged values:
have an extension:  
ωi 2 yi ∗ − yi
 1/4 = , i = 1, . . . , ns (8)
3 ¯  1 − 3
∼ (1)

This corresponds to the implementation of the EDC-model in the
where  is the Kolmogorov length scale.1 At these scales, dissipation CFD code (Fluent, 2005). It should be noted that the expression ,
of turbulent kinetic energy k takes place with a rate  where  is the which enters in Magnussen’s original formulation of formulas (6)
kinematic viscosity. At a size smaller than  no turbulent structures and (7), is set to unity as proposed by Magnussen (1989) and Gran
exist, because in those regions molecular diffusion is faster than tur- and Magnussen (1996).
bulent transport. From this thought the conception of the EDC was
developed: There is a range in which the reactions can be regarded 5. Variations of EDC-model parameters
as ideally mixed. Thus, chemical reaction kinetics determines the
speed of the process, while outside this range the reactants are not In order to examine the influence of the EDC-model onto the
mixed and do not react. So each computing cell is split into a reac- complete CFD simulation of the gasification process, parameter
tive fraction ( 3 ) and an inert fraction (1 −  3 ). By the previous variations were accomplished in FLUENT and in a post-processing
considerations, the reactive fraction represents the part of the mix- step the CFD data was exported to MATLAB. At first, the effects of the
ture, which is in the smallest turbulence structures. It is also called parameter change are only examined on a single cell in the refor-
fine-scale fraction. The inert fraction is referred to as surrounding mation zone. The regarded cell has an average temperature of 1811
fluid. The model was presented for the first time by Magnussen and K and the composition in mass fractions is: yH2 = 0.064, yH2 O =
Hjertager (1976). However, the basis for the present work was the 0.283, yCH4 = 0.0374, yCO = 0.4916, yCO2 = 0.1, yC2 H2 = 0.0081,
model proposed by Magnussen (1989), where the reactive portion yC2 H4 = 0.0026, yN2 = 0.0141. The influence of the parameters C
 3 is modelled as follows: and C from Eqs. (2) and (4) on the reaction rate Eq. (8) is exem-
  1/4 plarily described for the specie methane in Fig. 3. The star marks
 = C (2) the result calculated by the CFD model with standard parameters
k2 (Eqs. (3) and (4)) for the regarded cell. In the current example the
One recognizes that  is solely obtained from turbulence magni- mean reaction rate is strongly dependant on the parameter C ,
tudes. It is: whereas C has almost no impact. The reason for this is the time
 1/4 scale  which is rather large (≈ 10−2 ) in the observed cell. In the
3CD2
C = 2
= 2.1377, with CD1 = 0.134, CD2 = 0.5 (3)
4CD1

The constants were obtained with the help of the energy cascade
model (Ertesvåg & Magnussen, 2000). The time scale for the mass
transfer from the fine structures to the surrounding fluid is:
  1/2  C 1/2
D2
 = C = C  , with C = = 0.4082 (4)
 3
The chemical reactions are represented by an isobaric, adiabatic,
perfectly stirred reactor. This means, that the radiation from or
to the fine structures is neglected. The mixing rate 1/ charac-
terizes an open reactor vessel with the retention time  (Gran &
Magnussen, 1996). This reactor is described by:
dyi ∗ ω∗ yo − yi ∗
= i∗ + i , with i = 1, . . . , ns (5)
dt  

Fig. 3. Influence of the parameters C and C on the absolute amount of the mean
1 1/2 reaction rate |ωCH4 |.
Another Kolmogorov micro scale is the time scale ( ∼ (/) ).
M. Rehm et al. / Computers and Chemical Engineering 33 (2009) 402–407 405

Fig. 6. Relative error of experimental and calculated values of mole fraction of CH4
and temperature at C = 8 and variation of the parameter C .

Fig. 4. Fine scale fraction on the reactor axis with variation of C . In the background
6. Analysis of the involved time scales
the reactor outline is depicted (not drawn to scale).

In the following section an analysis of the time scales of mixing


and reaction is performed. The time scales of turbulent transport
regions of higher turbulence the time scale may be much smaller
can be obtained with the help of the quantities from the turbulence
and the influence of C will become stronger. So the parameter C
model. The following time scales are used to describe the flow and
is going to be the best candidate to have an impact on the mean
mixing effects:
reaction rate outside the flame zone and is chosen for parameter
variations in the complete CFD model of non-catalytic partial oxi- C k
int = D (9)
dation at 30 bar. For comparison, Table 1 shows the results of the 2/3
work of Zeißler (2006) using standard EDC-parameters. In a first
variation the parameter C was increased to demonstrate the influ-   1/2
ence on the reactive volume fraction per cell. The results on the  = (10)

symmetry axis of the computational domain are shown in Fig. 4.
For the standard value C = 2.1377 a peak for the volume fraction V (1/3)
cell = (11)
directly downstream of the inlet and a second one at the stricture |v|
close to the exit can be observed. Choosing C = 3.2066 doubles
EDC = C  (12)
the reactive fraction. At C = 5.56 it is notable that  3 is limited to
0.75 in the EDC implementation of the CFD code. It should be men- The Kolmogorov time scale  is also called the “turnover time” for
tioned that the parameters of the accomplished variations cannot the smallest eddies and the integral time scale int is also referred to
be deduced from the parameters, which originate from the energy as turbulent time scale. The eddies with the largest kinetic energy
cascade model. Here they only serve to show the influence of the are of integral length scale. Eq. (9) is valid for large turbulent
parameters and to point out variation possibilities and limitations Reynolds numbers with CD = C
0.75 and C = 0.09 (Poruba, 2002).

of the EDC. The effects of variation of C and two different values of Furthermore, cell is a measure for the retention time of the fluid in
C on the results at the outlet are shown in illustrations 5 and 6. The a computational cell of volume V and EDC is the term used for the
observations for the individual cell are confirmed with the help of mixing time scale in Eq. (4).
the complete CFD model: an increase of C leads to lower methane- The chemical time scale is a measure for the duration of the
slip and thus to an approximation towards the experimental results. slowest chemical reactions. Two different values for chemical time
However, the experimentally determined values could not be scale will be introduced. First with the help of a sensitivity analysis
achieved. of the main species: Let ij be according to:

I = {i1 = iH2 , i2 = iH2 O , i3 = iCH4 , i4 = iCO , i5 = iCO2 } (13)

the indices of most abundant components in the process. For the


reaction rate holds:
dy
= f (y) (14)
dt
The expansion of f (y) into a Taylor series and neglecting the terms
of higher order results in:

dy
≈ f (y0 ) + f  (y0 )(y − y0 ) with f  (y0 )
dt
∂fi (y0 )
=J= and i, k = 1, . . . , ns (15)
∂y0k

The components ji,k of J are a measure of the dynamics of the chem-


Fig. 5. Relative error of experimental and calculated values of mole fraction of CH4 ical system at the point y0 in the state space. This is well defined,
and temperature at the reactor exit for C = 0.4082 and variation of C . because pressure and enthalpy are constant in the ideal reactor.
406 M. Rehm et al. / Computers and Chemical Engineering 33 (2009) 402–407

bigger than retention time in the fine scales (EDC ), reactions in the
actually inert surrounding fluid should also be accounted for:
o
ωi =  3 ωk ∗ + (1 −  3 )ωk (19)
o
Here ωk ∗ stands for the reaction rate in the fine scales and ωk
for the reaction rate in the surrounding fluid. So adapting the EDC
in that manner could lead to better results for the representation
of slow reforming reactions in zones where mixing is faster than
chemistry.

7. Discussion and summary

An examination of the EDC-model for the modelling of the


HP-POX reactor showed, that the important endothermic reform-
ing reactions for gasification are not rendered sufficiently enough.
By changing the model parameters the consistency of measured
and modelled results could be improved. However, no agreement
with the experimental values was achieved. A time scale analysis
indicates that a partitioning of the modelling domain in two char-
Fig. 7. Illustration of the time scales obtained from the CFD model plotted on the
symmetry axis of the domain.
acteristic zones can be possible: First, a flame zone characterized
by high temperatures, where mixing of reactants is the retardant
process. And a second part, a reforming zone, where lower tempera-
The reciprocal entries of J correspond to reaction times. The slow-
tures occur and the chemical reactions proceed slower than mixing.
est of the times of a main component is now selected and defines a
A zone-dependent adjustment of the model seems appropriate to
characteristic chemical time scale:
  increase the accuracy of gasification modelling. An implementa-
1 tion of that approach is tested at the moment and will be subject of
chem1 = maxi,k ∈ I (16)
ji,k further proceedings.

With the help of this approach and the indices i and k, it is possible to
tell which species are involved in the slowest processes in chemical Acknowledgements
space.
Another more general approach is to use the real parts of the We thank the German Federal Ministry of Economics and Tech-
eigenvalues i of J to obtain the slowest chemical time scale from: nology (BMWi No. 0327113B), the Saxon Ministry of Science and the
 1
 Fine Arts (SMWK), and the Lurgi GmbH/Air Liquide for supporting
chem2 = maxi (17) our work.
|Re( i )|
In Fig. 7 it can be seen that in the downstream domain of the reactor References
chem1 and chem2 are almost equal. The specie associated to chem1
in that part of the reactor was found to be CH4 . This means that Chakraborty, D., Paul, P. J., & Mukunda, H. S. (2000). Evaluation of combustion models
for high speed H2 /air confined mixing layer using DNS data. Combustion and
decomposition of methane is one of the slowest chemical processes. Flame, 121, 195–209.
All time scales presented here were obtained from the CFD model Ertesvåg, I. S., & Magnussen, B. F. (2000). The eddy dissipation turbulence energy
and can be seen in illustration Fig. 7. cascade model. Combustion Science and Technology, 159, 213–235.
Fluent (2005). The FLUENT 6.2 User’s Guide. Fluent, Inc. http://www.fluent.com.
The relation between the mixing time scale and the chemical
Goodwin, D. (2002). Cantera: Object-oriented software for reacting flows. Tech. Rep.,
time scale is also called the Damköhler number Da: California Institute of Technology. http://blue.caltech.edu/cantera/index.html.
 Gran, I. R., & Magnussen, B. F. (1996). A numerical study of a bluff-body stabilized dif-
Da = mix (18) fusion flame. Part 2. Influence of combustion modeling and finite-rate chemistry.
chem Combustion Science and Technology, 119, 191–217.
Since turbulence dominates the mixing process, a choice for the Kempf, A. M. (Ed.) (2006). The Eighth International Workshop on Mea-
surement and Computation of Turbulent Nonpremixed Flames (TNF8).
mixing time scale would be int Eq. (9) or EDC resulting from the http://www.ca.sandia.gov/TNF/abstract.html.
EDC according to Eq. (4). For the chemical time scale chem2 is chosen Kjäldman, L., Brink, A., & Hupa, M. (2000). Micro mixing time in the eddy dissipation
Eq. (17). concept. Combustion Science and Technology, 154, 207–227.
Konnov, A., Zhu, J., Bromly, J., & Zhang, D. (2003). Non-catalytic partial oxidation of
If Da  1 it can be assumed that the mixing time has no effect methane over a wide temperature range. In Proceedings of the European combus-
on the overall process. The chemical reactions are the limiting pro- tion meeting 2003, Orléans, France
cesses and the system can be regarded as an ideally stirred reactor. Launder, B., Reece, G., & Rodi, W. (1975). Progress in the development of a Reynolds
stress turbulence model. Journal of Fluid Mechanics, 68(3), 537–566.
Likewise, in regions where Da  1, chemical kinetics act much Magnussen, B. (1989). Modeling of NOx and soot formation by the eddy dissipation
faster than mixing processes, thus equilibrium assumptions can be concept. Int. Flame Research Foundation, 1st Topic Oriented Technical Meeting,
made without loss of accuracy. None of the two statements applies 17–19.
Magnussen, B. F., & Hjertager, B. H. (1976). On mathematical models of turbulent
to the whole modelling area which can be seen in Fig.. Both mix-
combustion with special emphasis on soot formation and combustion. In: 16th
ing time scales lead to comparable results: In the flame region a Symp. (Int’l.) on combustion. The Combustion Institute.
large Damköhler number can be observed. Mixing of fuel and oxi- Meyer, B., Seifert, P., Zeißler, R., & Walter, S. (2005). Synthesegaserzeugung durch
Hochdruck-Partialoxidation (HP POX® ). Erdöl, Erdgas, Kohle, 121–125, 190–195.
dizer is the dominating process. In contrast, further downstream in
Pope, S. (1985). PDF methods for turbulent reactive flows. Progress in Energy and
the reformation zone Da is decreasing rapidly. As modelling results Combustion Science, 11(2), 119–192.
for the products of combustion are satisfactory, it is suspected that Pope, S. B. (1997). Computationally efficient implementation of combustion chem-
weaknesses of the EDC stem from the region of low Damköhler istry using in-situ adaptive tabulation. Combustion Theory and Modeling, 1,
41–63.
number. Kjäldman, Brink, and Hupa (2000) mentioned for the EDC- Poruba, C. (2002). Turbulente Flammenausbreitung in Wasserstoff-Luft-Gemischen.
modelling, that in case the chemical time scale (chem ) is much Ph.D. Thesis, Lehrstuhl für Thermodynamik Technische Universität München.
M. Rehm et al. / Computers and Chemical Engineering 33 (2009) 402–407 407

Shih, T.-H., Liou, W. W., & Shabbir, A. (1995). A new k-epsilon eddy-viscosity model Zeißler, R. (2006). Modellierung der Gasphasenreaktion bei der autothermen kat-
for high reynolds number turbulent flows—model development and validation. alytischen Erdgasspaltung unter hohen Drücken. Dissertation, TU Bergakademie
Computers Fluids, 24(3), 227–238. Freiberg.
Smith, G. P., Golden, D. M., Frenklach, M., Moriarty, N. W., Eiteneer, B., Goldenberg, Zeißler, R., Meyer, B., Seifert, P., Heinzel, A., & Jentsch, T. (2006). Untersuchung des
M., et al. (1999). Gri-mech 3.0. http://www.me.berkeley.edu/gri mech. Verweilzeitverhaltens eines autothermen Vergasungsreaktors bei Drücken bis
Stefanidis, G., Merci, B., Heynderickx, G., & Marin, G. (2006). CFD simulationsof 70 bar(Ü). Chemie Ingenieur Technik, 78(Issue 1–2), 72–77.
steam cracking furnaces using detailed combustion mechanisms. Computers and Zhu, J., Zhang, D., & King, K. (2001). Reforming of CH4 by partial oxidation: Thermo-
Chemical Engineering, 30, 635–649. dynamic and kinetic analyses. Fuel, 80(7), 899–905.
Warnatz, J., Maas, U., & Dibble, R. (2001). Combustion: Physical and chemical funda-
mentals, modelling and simulation. Berlin: Springer.

Anda mungkin juga menyukai