Fuel Processing Technology: Wenyuan Guo, Yanzeng Wu, Liang Dong, Caixia Chen, Fuchen Wang

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Fuel Processing Technology 98 (2012) 45–50

Contents lists available at SciVerse ScienceDirect

Fuel Processing Technology


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

Simulation of non-catalytic partial oxidation and scale-up of natural gas reformer


Wenyuan Guo, Yanzeng Wu, Liang Dong, Caixia Chen ⁎, Fuchen Wang
Key Laboratory of Coal Gasification of Ministry of Education, East China University of Science and Technology, Shanghai 200237, China

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

Article history: Computational Fluid Dynamics (CFD) has been applied in the simulation of non-catalytic partial oxidation
Received 13 September 2011 (NC-POX) of methane and scale up of natural gas reformers. An industrial reference reformer and a large
Received in revised form 11 January 2012 scale virtual prototype were considered in this study. Benchmark simulations were performed using the pre-
Accepted 18 January 2012
sumed PDF model and the Eddy-Dissipation-Concept (EDC) model, respectively. The results show that the
Available online 23 February 2012
PDF model is sufficient and more suitable for an engineering level estimation and trend prediction of the
Keywords:
reactions in a NC-POX reformer. Simulations of a large scale prototype reformer were performed using the
Non-catalytic partial oxidation of methane PDF approach. The predicted peak gas temperature and its relation to the burner parameters were examined
Industrial natural gas reformer numerically. The results suggest that decreasing the temperature of input streams does not necessarily delay
Scale-up the ignition of methane/oxygen mixture. It is verified that a relatively small nozzle can effectively shift the
CFD flame downward, and generate a strong recycle flow field in the reformer.
© 2012 Elsevier B.V. All rights reserved.

1. Introduction known technical problems encountered in NC-POX reformers con-


cern the hot-spot temperatures amounting to 1500-2000 K, which
Chemical conversion of natural gas into liquid fuels, known as frequently causes burner damages. Excessive local temperatures
Gas-to-Liquid (GTL) technology, has gained considerable attention also lead to refractory wall failures. Recently, a 2-million ton/year
in recent years due to the surge in need for liquid fuels and the GTL project has been proposed in China. To meet the goal, multiple
increase in crude price. Usually, a GTL process is mainly comprised reformers with single-line capacity of 63,700 Nm 3/h (natural gas)
of three steps. These are the thermochemical reforming of natural has been under development. The reformer is expected to largely
gas into syngas (with or without catalysts), followed by the synthetic exceed the previous scale of industrial references. In order to assess
conversion of syngas into longer chain hydrocarbons on a catalyst and mitigate any risks associated with the scale-up, the new design
surface, and finally the upgrading of chemical products to meet the of the large scale NC-POX reformer need to be evaluated in detail.
market specifications. Among the three steps of conversions, the syn- Modeling and simulation has been used in virtual prototyping of
gas production is the most capital intensive part of a GTL plant. There- the reactor and thus helps in design, optimization and scale up of
fore, there is a strong economic incentive to develop very large scale chemical reactors. 0-dimensional perfectly stirred reactor model [4]
natural gas reformers for the purpose of cost deductions [1]. and 1-dimensional premixed model [5–7] were reported for the sim-
Steam reforming (SR), non-catalytic partial oxidation (NC-POX), ulations of bench scale NC-POX reactors. Similarly, 1-dimensional
and autothermal reforming (ATR) are the three major technologies plug flow model was used in the simulation of catalytic partial oxida-
used in producing syngas from natural gas. Principally, ATR combines tion (CPOX) membrane reactor [8] and ATR process [9]. Recently,
NC-POX and SR in a single process, and provides the most favorable Amirshaghaghi et. al. performed 3-dimensional simulation of meth-
H2/CO ratio for cobalt-based catalyst. Hence, it is ideally proper for ane partial oxidation in the burner and combustion chamber of ATR
GTL application. However, ATR has limited commercial experience process. A simple two-step reaction mechanism was used for the
due to the uncertainty related to the formation of soot particles. gas combustion in conjunction with an Eddy Dissipation model for
These undesired carbon particles formed in the oxidation zone tend the reaction rates [10]. In our previous work [11], detailed kinetic
to deposit on the catalyst beds and result in subsequent coking and simulations were performed for the reactions of natural gas/O2/H2O
deactivation; consequently lead to the damage of equipment and mixture under industrial conditions, and an in-house multi-
heat transfer problems [2,3]. On the other hand, the conventional dimensional model that takes the turbulent mixing into account
NC-POX technique has been successfully commercialized in large was developed for the simulation of NC-POX process [12]. In this in-
scales and utilized in GTL plants worldwide. Similar to ATR, some house model [12], a rather simplified presumed PDF model is used
in the description of interactions between the turbulent mixing and
the multi-step chemical reactions in the reformers. The initial compo-
⁎ Corresponding author. Tel./fax: + 86 21 64251092. sition of the oxides (oxygen and steam) was assumed uniform, and
E-mail address: cxchen@ecust.edu.cn (C. Chen). only a single mixture fraction was used to track the mixing of fuel

0378-3820/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
doi:10.1016/j.fuproc.2012.01.019
46 W. Guo et al. / Fuel Processing Technology 98 (2012) 45–50

and oxides. In practice, however, oxygen and steam are injected Table 1
through different channel and mixed in the furnace. To take this effect Geometry parameter of Reformer A and B.

into account, the much robust commercial CFD software Ansys FLUENT Parameters Reformer A Reformer B
(Ansys Inc., Lebanon, NH) is considered in the present study. FLUENT
Diameter mm 1800.0 3150.0
provides a variety of turbulent reaction models, and offers more choices Height mm 11,580.0 18,000.0
of handling multi-stream mixing and their interactions. In this paper, Burner
we present the benchmark simulation of an industrial reformer using d1 mm (ignition fuel) 25.0 25.0
d2 mm (oxygen) 41.2 63.35
FLUENT 6.3 and compare the results of two different turbulent reaction
d3 mm (natural gas) 58.64 90.98
models. Simulation of a large scale prototype reformer is performed, d4 mm (steam) 65.0 99.48
and the gas velocity and temperature distributions in the prototype Mass flow rates:
large scale reformer are examined. Natural gas Nm3/h 16,708.0 57,920
Oxygen Nm3/h 11,428.0 39,650
Steam Vol. % 9.0 6.3
2. Methods
Angles
Natural gas 12 12
2.1. Reformers and geometry Oxygen 0 0
Steam 15 15.7
Temperature K
Two reformers are investigated in this paper. The first one is an
Natural gas 523 527
industrial NC-POX reformer which has been under operation in Oxygen 503 413
Neimeng province, China, hereinafter referred to as Reformer A. The Steam 653 653
second one is a proposed large scale prototype reformer, and referred Pressure MPa
to as Reformer B. Both reformers are geometrically similar, which Natural gas 6.0 5.5
Oxygen 6.3 5.5
are cylindrical, top-down pressure vessels with spherical/elliptical
Steam 7.0 10.0
domes. Preheated natural gas and the oxidizer (O2 and H2O) are Reactor 6.0 4.5
injected into the vessel through a multi-channel gas burner on the
top of the dome. A small amount of high pressure steam is injected
into the reformer with oxygen. By controlling the O2/CH4 ratio of following physical and chemical processes are taken into account:
feeding gases, the reformer is operated in the temperature range of (1) Turbulent flow of gas and mixing of gaseous reactants; (2) multi-
1473–1773 K to produce the maximum syngas yield with desired step reactions of methane-oxygen-steam mixture; and (3) Convective
compositions (CO, CO2 and H2). The schematic geometry of the burn- and radiation heat transfer. Partial oxidation of methane can be consid-
er is shown in Fig. 1(a). The major geometric parameters and operat- ered as a process of fuel rich methane combustion. From this point of
ing conditions of the two reformers are listed in Table 1. view, the methods of modeling NC-POX reforming are essentially alike
to methane combustion except for its partial oxidation kinetics. In the
2.2. Simulation methods present study, the widely used CFD code FLUENT 6.3 is employed. The
details of the numerical methods and the sub-models can be found in
For comprehensive modeling of the turbulent flow and non- the FLUENT User's Guide [13], and the major methodologies are sum-
catalytic partial oxidation of methane in a top fired reformer, the marized below.

Fig. 1. (a)Schematic diagram of burner; (b)Blocking structure of computational grids.


W. Guo et al. / Fuel Processing Technology 98 (2012) 45–50 47

2.2.1. Generalized governing equations 3. Results and discussion


The gas flow is assumed to be a steady state, reacting, continuum
field that can be described by a general conservation equation as 3.1. Benchmark simulation of Reformer A
follows:
As a benchmark case, reformer A was simulated by using Fluent's
    PDF and EDC models, respectively. The operating conditions listed in
∂ ∂ ∂ ∂ μ ∂Φ ∂ μ ∂Φ Table 1 were used in the computations. In the PDF simulation, two
ðρuΦÞ þ ðρvΦÞ þ ðρwΦÞ ¼ þ
∂x ∂y ∂z ∂x σ Φ ∂x ∂y σ Φ ∂y mixture fractions were used to track the mixing of the primary
  ð1Þ
∂ μ ∂Φ stream of fuel and the secondary stream of steam. A stable flame
þ þ SΦ was obtained in the simulation when the rich flammability limits of
∂z σ Φ ∂z
the primary and the secondary streams were set at 0.42 and 0.2, re-
spectively. 23 equilibrium gas species (CH4, N2, O2, H2O, C2H6, C3H8,
where, Φ represents any quantity of mass, velocity components (u, v H2, CO2, CO, OH, H, O, C2H2, C2H4, HO2, CH3, H2O2, HCO, CHO, C3H6,
and w), turbulent kinetic energy (k) and turbulent kinetic energy O3, HOCO and CH2O) were presented in the PDF table generated
dissipation rate (ε), gas enthalpy (h), mixture fractions (fi) and their with the non-adiabatic assumptions at the pressure 6.0 MPa and min-
variances (gi). SΦ is the source term. The turbulent viscosity μ is imum temperature 503 K. In the EDC simulation, the 16 species (H2O,
modeled with the realizable two-equation k − ε model with the stan- CO2, O2, CH4, CO, H, H2, OH, O, CH3, HCO, HO2, H2O2, CH2O, CH3O and
dard constants of the k − ε model. N2), 43 steps reaction mechanism [14] was used in the benchmark
case.
2.2.2. Turbulent reaction model The axial distributions of gas species, gas velocities and the tem-
FLUENT provides a variety of turbulence-reaction interaction peratures computed by using the PDF and EDC methods are pre-
models. The presumed PDF model and the Eddy-Dissipation-Concept sented in Figs. 2 and 3. 5 major gas species (CH4, H2, CO, CO2, and
(EDC) model are considered in the present work. In the presumed H2O), and 8 intermediate species (OH, H, HO2, H2O2, O, CH3, CH2O,
PDF approach, two mixture fractions are used to track the mixing of and HCO) which have significant mole fractions, are depicted in the
fuel-oxygen, and that of steam-fuel-oxygen, respectively. The variance figures for comparison. As shown in Fig. 2, the axial profiles of the
of the mixture fractions is modeled with a distribution of beta-function. major gas species computed from the two approaches show similar
The EDC model implemented in the FLUENT is an extension of the pattern of evolution. However, the intermediate/minor gas species
eddy-dissipation model to include multiple simultaneous chemical of the two simulations show substantial differences. In either ap-
reactions and detailed Arrhenius chemical kinetics in turbulent proach, the partial oxidation process is characterized by two consec-
flames [14]. The EDC model assumes that reactions occur in small utive steps: a rapid oxidation (Z/D b 0.5) followed by a relatively
eddies or fine scales, with initial conditions taken from the current slow reforming (Z/D > 0.5) processes. The gas composition plots
species concentrations and temperature in the cell. Reactions proceed show initially a rapid reaction rate, which increases until a significant
over the time scale, governed by Arrhenius rates of reaction and are amount of CO2 and H2O are formed. During this period, the concen-
integrated numerically using an In Situ Adaptive Tabulation (ISAT) al- trations of HCO and H2O2 increases steadily and the concentration
gorithm for tabulation [15]. The equations for the size of the fine scale of reactive radicals (O, H, OH, for example) build up. After ignition,
structure and time scale of reaction, along with the methodology of the amount of HCO and H2O2 quickly decrease as a result of reaction
the corresponding implementation of the EDC model can be found toward carbon monoxide and H2O, and the concentrations of reactive
in the reference [13]. radicals decrease accordingly. In the PDF simulation, gas species be-
come steady from the location around Z/D = 1.5. However, this is
2.2.3. Reaction mechanism not the case for the EDC simulation. The major gas component and
FLUENT 6.3 allows up to 50 gas species in a simulation, which the minor intermediate species obtained from EDC simulation change
limits the use of full detailed reaction mechanism in the description along the axial direction throughout the entire reactor height, indicat-
of partial oxidation of methane. A number of reduced mechanisms ing the reaction system did not reach in the equilibrium state. The
for methane combustion have been developed from detailed mecha- PDF approach, on the other hand, is based on equilibrium assumption
nisms and are available in the literature. In the current study, three of the reactions. This is the main reason of different results of the PDF
reduced mechanisms, i.e., the 19-species GRI 1.2 model [16], the 16- and EDC models.
species CH4 skeleton model [17], and the 37-species Leeds 1.5 The axial profiles of gas velocity magnitudes and the temperatures
model [18] are assessed in conjunction with the EDC model described computed from the PDF and EDC methods are shown in Fig. 3. From
above . the figure, the gas velocity magnitudes of the two simulations are
comparable, excepting a small discrepancy in the peak values. The
2.2.4. Grid and numerical procedures gas temperatures, however, show significant difference between the
With the assumption of an axially symmetric flame, 2-dimensional two methods. Because more gas species are included in the equilibri-
axisymmetric grids consisting of 12,000 cells was used for Reformer A um mixture, the PDF model predicted reasonable gas temperature
and 25,000 cells for Reformer B. Fig. 1(b) shows part of the blocking than that in the EDC simulation. The EDC model predicted much
topology of the computational grids. For the simulations presented in higher gas temperature especially for the peak value (≥3500 K). It
this paper, mass flow inlets and pressure outlet were used for the is worth noting that the peak temperatures were reached at the
boundary conditions. The dome wall was set as adiabatic boundary, same location of about Z/D = 0.5 in either simulation.
and other walls were set at constant temperatures of 1073 K to roughly The simulated outlet gas composition and temperature are com-
match the measurements of Reformer A. Converged solutions was de- pared in Table 2. Under the same boundary conditions, the predicted
fined when the global energy balances to within 0.3% of the total com- numbers for the outlet syngas are quite different. The PDF predictions
bustion energy and the normal residual (mean residual/mean inlet on the syngas compositions match the design values very well. The
velocity) for each of the velocity components was less than 0.3%. In EDC model, however, predicted a higher gas temperature and a much
the EDC simulation, an ISAT error tolerance value of 0.1% was used in higher methane concentration in the outlet products. The results sug-
solving the transport equations of gas species. 1-2 CPU hours were re- gest that the reaction mechanism might not work properly in the EDC
quired for a typical case computation of reformer A, and 5-6 CPU simulation. Two other mechanisms [16,18] are also tested, but no im-
hours for reformer B in a 64-bit dual-core Xeon workstation. provement was observed in the outlet methane concentration. Unless
48 W. Guo et al. / Fuel Processing Technology 98 (2012) 45–50

Fig. 2. Comparison of axial distribution of computed gas species.

more suitable reaction mechanism is available, the EDC model of FLU- soot formations. Therefore, it is very important to examine the gas
ENT is not recommended for the simulation of industrial NC-POX temperature and velocity fields of the prototype reformer.
reformers. Based on this finding, the following simulations for Reformer Firstly, the geometric parameters and the operating conditions listed
B were performed using PDF approach only. in Table 1 were used in the simulations of reformer B (referred to as
B-final). Secondly, a series of case studies (B1-B3) were performed. In
these cases studies, the nozzle diameters were determined in such a
3.2. Scale-up of Reformer B way that the outlet areas were 2.6 (=QBPB/QAPA) times of reformer A.
The resulted burner sizes are: d1 = 25.0 mm, d2= 72.15 mm, d3 =
While similar geometric structures to Reformer A are adopted in 98.24 mm and d4 = 106.56 mm. Furthermore, in CaseB, the same inlet
the design of scale up of the prototype reformer, some significant stream conditions of reformer A, were assumed (i.e., a percentage of
changes have been made as listed in Table 1. The ratio of the mass 9.0% (of oxygen volume), and an oxygen temperature of 503 K), In
flow rates of reformer B to reformer A, QB/QA = 3.47. However, the Case B2, the steam was deceased to 6.3%, and in case B3, the inlet
diameter of reformer B is only 1.75 times of the reformer A, which oxygen temperature was set to 413 K in addition to the reduction of
is smaller than the value of the square root of QB/QA. In addition, steam to 6.3%.
the operating pressure was decreased from PA = 6.0 MPa to PB = The 2-dimentional distribution of gas temperature of Reformers
4.5 MPa. Accordingly, the ratio of the volumetric flow rates of the A and the virtual prototypes of reformer B are compared and shown
two reformer is approximated as (PAQB)/(PBQA). As can be found in in Fig. 4. In the figure, the proposed prototype reformer B is referred
Table 1, other major changes include the decrease of injection tem- to as B-final. Comparing to reformer A, the gas temperatures in
perature of oxygen from 503 K to 413 K in order to delay the ignition prototypes of reformer B are more uniform. Although the oxygen
and prevent hot spots within the dome arch, and the reduction of the temperature has been decreased intentionally in Case B3, the gas
percentage of steam from 9% to 6.3% (Vol. of oxygen) for preventing temperature field shows that the flame center does not shift
W. Guo et al. / Fuel Processing Technology 98 (2012) 45–50 49

Fig. 4. (a) Temperature distributions of reformer A; (b) case B1 (oxygen temperature:


503 K,steam: 9.0%VOL.); (c) case B2 (oxygen temperature: 503 K,steam: 6.3%VOL.);
(d) case B3 (oxygen temperature: 413 K,steam: 6.3%VOL.); and (e) case B-final.

Fig. 3. Comparison of axial distribution of computed gas velocity and temperature for and the PDF model was validated to be sufficient for an engineering
Reformer A. level estimation and trend prediction of the reactions. Simulation
of a large scale prototype reformer was performed, and the gas tem-
perature and flow fields of the prototype were compared with the ref-
downward. Compared with B1-B2, significantly higher gas tempera- erence reformer. The results show that decreasing the temperature
ture is predicted in Case B3. It is also observed that the predicted of oxygen input does not necessarily delay the ignition of methane/
gas temperature in the dome arch area increases with a decrease of oxygen mixture. It is verified that a relatively small nozzle can effec-
oxygen temperature from 503 K to 413 K. To identify the cause, the tively shift the flame downward, and generate a strong recycle flow
axial velocity magnitude and the gas temperatures of the four cases field in the upper part of the reformer.
are redrawn and compared with Reformer A in Fig. 5. The axial gas
temperatures and the velocities of case B1 and B2 are almost identi-
cal, indicating the effect of reducing steam input from 9% to 6.3% on
the gas temperature and flow field is ignorable. It is worth noting
that the gas temperatures of Cases B1–B3 all reach their peaks earlier
than that in reformer A, regardless much higher gas velocities are pre-
dicted in cases B1–B3, The simulation results show that only proto-
type B-final shows a delayed temperature peak than reformer A.
Fig. 6 compares the streamlines and the velocity distributions of
reformer A and B (B-final). It is obvious that the recycle flow in re-
former B is stronger than reformer A. In reformer B, the length of re-
cycle zone extends to 2/3 reactor height, which is important in the
promotions of mixing and the reactions for a relatively small reactor.

4. Summary

Benchmark simulations of an industrial reformer have been


performed using the commercial software FLUENT. Two different tur-
bulent reaction models, i.e., PDF and EDC were used in the simulation,

Table 2
Compositions and temperature of syngas.

H2 CO CO2 CH4 N2 Tout


(K)
(vol.%)

Design 62.25 33.52 2.91 0.49 0.83 1551.0


PDF Sim. 61.39 34.04 3.07 0.67 0.83 1559.0
EDC Sim. 60.83 34.13 2.88 2.10 0.05 1639.0
Fig. 5. Comparison of axial distributions of computed gas velocity and temperature.
50 W. Guo et al. / Fuel Processing Technology 98 (2012) 45–50

[2] K. Aasberg-Petersen, T.S. Christensen, C.S. Nielsen, I. Dybkjær, Recent develop-


ments in autothermal reforming and prereforming for synthesis gas production
in GTL applications, Fuel Processing Technology 83 (2003) 253–261.
[3] D.J. Wilhelm, D.R. Simbeck, A.D. Karp, R.L. Dickenson, Syngas production for gas-
to-liquids applications: technologies, issues and outlook, Fuel Processing
Technology 71 (2001) 139–148.
[4] B. Lemke, C. Roodhouse, N. Glumac, H. Krier, Hydrogen synthesis via combustion
of fuel-rich natural gas/air mixtures at elevated pressure, International Journal of
Hydrogen Energy 30 (2004) 893–902.
[5] A.A. Konnov, J. Zhu, J. Bromly, D.K. Zhang, Noncatalytic Partial Oxidation of
Methane into Syngas over a wide Temperature Range, Combustion Science and
Technology 176 (2004) 1093–1116.
[6] C.L. Rasmussen, J.G. Jakobsen, P. Glarborg, Experimental measurements and
kinetic modeling of CH4/O2 and CH4/C2H6/O2 conversion at high pressure, Inter-
national Journal of Chemical Kinetics 40 (2008) 778–807.
[7] J. Zhu, D.K. Zhang, K.D. King, Reforming of CH4 by partial oxidation: thermody-
namic and kinetic analyses, Fuel 80 (2001) 899–905.
[8] Shashi Kumar, Prajapati Surendra, K. Jitendra, Hydrogen production by partial oxida-
tion of methane: Modeling and simulation, International Journal of Hydrogen Energy
34 (2009) 6655–6668.
[9] V.L. Barrio, G. Schaub, M. Rohde, S. Rabe, F. Vogel, J.F. Cambra, P.L. Arias, M.B.
Güemez, Reactor modeling to simulate catalytic partial oxidation and steam
reforming of methane. Comparison of temperature profiles and strategies for
hot spot minimization, International Journal of Hydrogen Energy 32 (2007)
1421–1428.
[10] H. Amirshaghaghi, A. Zamaniyan, H. Ebrahimi, M. Zarkesh, Numerical simulation
of methane partial oxidation in the burner and combustion chamber of autother-
mal reformer, Applied Mathematical Modeling 34 (2010) 2312–2322.
[11] Xinwen Zhou, Caixia Chen, Fuchen Wang, Modeling of Non-catalytic Partial
Oxidation of Natural Gas under Conditions Found in Industrial Reformers, Chemical
Engineering and Processing 49 (2010) 59–64.
[12] Xinwen Zhou, Caixia Chen, Fuchen Wang, Multi-dimensional modeling of non-
catalytic partial oxidation of natural gas in a high pressure reformer, International
Journal of Hydrogen Energy 35 (2010) 1620–1629.
Fig. 6. (a) Streamline and velocity distributions of reformer A; and (b) Reformer B
[13] FLUENT Users Guide, Fluent Inc, September 2006.
(B-final).
[14] I.R. Gran, B.F. Magnussen, A Numerical Study of a Bluff-Body Stabilized Diffusion
Flame. Part 2. Influence of Combustion Modeling and Finite-Rate Chemistry, Com-
bustion Science and Technology 119 (1996) 191–217.
Acknowledgements [15] S.B. Pope, Computationally Efficient Implementation of Combustion Chemistry
using In Situ Adaptive Tabulation, Combustion Theory and Modeling 1 (1997)
41–63.
This work is supported in part by China's National Science Foundation [16] G.P. Smith, D.M. Golden, M. Frenklach, N.W. Moriarty, B. Eiteneer, M. Goldenberg,
(NSFC) under grant 20876049. Caixia Chen gratefully acknowledges et al., http://www.me.berkeley.edu/gri_mech/ (Last accessed December 2010).
[17] Sanjay M. Correa, Turbulence-Chemistry Interactions in the Intermediate Regime
the financial support by the Pujiang Excellent Talents Plan of Shanghai of Premixed Combustion, Combustion and Flame 93 (1993) 41–60.
(Grant Number: 08PJ1403900). [18] The Leeds Methane Oxidation Mechanism version 1.5, http://garfield.chem.elte.
hu/Combustion/methane.htm (Last accessed December 2010).
References

[1] Buping Bao, Mahmoud M. El-Halwagi, Nimir O. Elbashir, Simulation, integration,


and economic analysis of gas-to-liquid processes, Fuel Processing Technology
91 (2010) 703–713.

You might also like