Journal of Cleaner Production 180 (2018) 655e665

Contents lists available at ScienceDirect

Journal of Cleaner Production

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

Multi-objective optimisation of steam methane reforming considering

stoichiometric ratio indicator for methanol production
Hamid Reza Shahhosseini a, Davood Iranshahi b, *, Samrand Saeidi c, Ehsan Pourazadi d,
Ji
rí Jaromír Kleme
s e
a
Department of Chemical and Petroleum Engineering, Sharif University of Technology, Tehran, Iran
b
Department of Chemical Engineering, Amirkabir University of Technology (Tehran Polytechnic), No. 424, Hafez Avenue, Tehran 15914, Iran
c
Technische Thermodynamik, Universita €t Bremen, Badgasteiner Str. 1, 28359 Bremen, Germany
d
Department of Chemical and Process Engineering, University of Canterbury, Christchurch 8140, New Zealand
e
Sustainable Process Integration Laboratory e SPIL, NETME Centre, Faculty of Mechanical Engineering, Brno University of Technology e VUT Brno,
Technicka 2896/2, 616 00 Brno, Czech Republic

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

Article history: This work proposes a novel configuration for steam methane reformers (SMR) in order to improve their
Available online 29 December 2017 syngas stoichiometric ratio (SR). This is a decisive element for methanol producers to increase their
production tonnage. While the optimum theoretical SR value is around 2, many conventional SMRs
Keywords: operate far beyond this value due to thermodynamic equilibrium limitations. In the new SMR design CO2,
Steam reforming which could be an industrial off gas, is injected into the reactor in multiple stages. The corresponding CO2
Methanol synthesis
injection flow rate is determined by a multi-objective optimization method. The optimum flow rate at
Multi-objective optimisation
each stage is chosen to minimise abs (SR-2) while maintaining the CH4 conversion at its highest value
Pareto frontier
Methanol economy
(about 68%). Furthermore, the new design helps to improve the thermodynamic equilibrium conversion
Stoichiometric ratio SR in SMR resulting in 33% more CO production. As well as this, the pressure drop along the new reactor is
proved to be substantially lower than the conventional SMR.
© 2018 Elsevier Ltd. All rights reserved.

1. Introduction 2014). However, a hydrogen-based energy economy requires sub-

stantial technological advancement in storage and transportation,
Abundant amounts of energy have been consumed to continu- as well as addressing socioeconomic and infrastructural barriers.
ously improve humanity's living standards. Natural gas, oil and There are also some safety concerns in this area (Edwards et al.,
coal, still constitute our major energy sources and offer the raw 2008). Meanwhile, methanol (MeOH) is considered to be a more
materials for producing a large variety of derivatives (Cu

cek et al., feasible alternative (compared to H2) since it is a safer liquid and
2015). However, those resources are still not sustainable, even easier to store and distribute (Masih et al., 2010). The role of MeOH
with discoveries of new sources. It has been assessed that mankind in delivering a cleaner energy model has been further explored in
will probably run out of coal, oil and natural gas within the next the ‘Methanol economy’e a novel conceptual framework which has
three centuries (Cu

cek et al., 2012). For this reason, all feasible al- been proposed by Olah et al. (2011). MeOH can be used as fuel by
ternatives have to be investigated to find feasible and long-term blending with gasoline for internal combustion engines (ICE).
solutions. Alongside the problem of power generation (Rozali Problems still are to be addressed with regard to the metal corro-
et al., 2015), an important task has been how to store and effi- sion issue (i.e. especially for aluminium, zinc, and magnesium) and
ciently use energy (Alwi et al., 2013). cold engine start. Mixtures of methanol and gasoline are already
Numerous literature reports have discussed the importance of being marketed in China for internal combustion engines. In Shanxi
H2 and methanol as alternative fuel options to secure the future of province alone, there are more than 1000 petrol stations which
energy market (Van-Dal and Bouallou, 2013; Riaz et al., 2013; Dutta, distribute M15, M85 and M100 (100% methanol) gasoline/meth-
anol blends (Mendes et al., 2014; Siriworarat et al., 2017).
Methanol is also a solvent and building block for producing
* Corresponding author.
intermediates, complex chemical compounds, and synthetic hy-
E-mail address: [email protected] (D. Iranshahi). drocarbons. Primarily it is used for formaldehyde (which consumes

https://doi.org/10.1016/j.jclepro.2017.12.201
0959-6526/© 2018 Elsevier Ltd. All rights reserved.
656 H.R. Shahhosseini et al. / Journal of Cleaner Production 180 (2018) 655e665

about 30% of the total methanol production in the worldwide), two main units (Bozzano and Manenti, 2016):
acetic acid, methyl-tert-butyl ether (MTBE) and dimethyl ether
synthesis (Chuah et al., 2016; Kliopova et al., 2016; Riaz et al., 2013). (i) Production of syngas (i.e. a mixture of H2, CO and CO2) via
It can also be used in on-grid and off-grid power turbines and a steam methane reforming which includes the following re-
hydrogen carrier for fuel cell technology (Nikoo et al., 2015). actions (Bulatov and Kleme
s, 2011; Xu and Froment, 1989b)
Predictions show a steady growth in the methanol market
especially through China as the biggest consumer of methanol for CH4 þ H2 O4CO þ 3H2 (1)
formaldehyde and as an alternative to transportation fuel (Lira-
Barraga n et al., 2015). Using methanol as fuel substantially re- CO þ H2 O4CO2 þ H2 (2)
duces the emissions of greenhouse gases, haze initiators such hy-
drocarbons, NOx, SOx and particulate materials. The forecasted
CH4 þ 2H2 O4CO2 þ 4H2 (3)
global methanol demand as well as its production capacity are
illustrated in Fig. 1 (Alvarado, 2016). A simplified schematic of steam methane reforming (SMR)
process is illustrated in Fig. 2. The refined and desulphurised nat-
1.1. Methanol-economy players and barriers ural gas is sent to the performer. The performer is an adiabatic
reactor which eliminates traces of remaining sulphur and converts
Methanol as a clean and sustainable energy supply can be pro- heavy hydrocarbons to methane (Rajesh et al., 2000). The outlet
duced from variety of feed sources such as natural gas, coal, stream is mixed with steam and fed to reformer where the syn-
biomass and landfill gas (Kravanja et al., 2015). However, natural thesis gas is produced. Afterwards, to reduce the amount of CO, the
gas is considered to be the most viable and primary feedstock for outlet stream is sent to at least two water-gas shift (WGS) adiabatic
syngas production in methanol plants mainly because of its abun- reactors. These two reactors are high temperature (HT) and low
dance, availability of well-established refining and processing temperature (LT) (Saeidi et al., 2014, 2015) WGS reactors which
technologies/infrastructures as well as its environmental merits accommodate an H2-rich gas mixture.
compared to coal and oil (Velasco et al., 2010). Evaluation of around
25 available methanol production technologies by Shell Research (ii) Conversion of the syngas to methanol through two main
and Technology Centre, suggests that although various commer- reactions (Lange, 2001)
cialized reactor designs and catalyst vendor technologies are crucial
factors of the methanol economy and its cost (i.e. mainly due to the CO þ 2H2 4CH3 OH (4)
balance between production and market demand), synthesis gas
manufacturers are playing a more vital role. In conventional CO2 þ H2 4CO þ H2 O (5)
reforming reactors, thermodynamic equilibrium limitations results
in inappropriate syngas composition and stoichiometry ratio (SR) Here, the ideal composition of syngas for methanol synthesis is
which could be a barrier for additional methanol production. usually characterised by the stoichiometric number SR which is the
Improving SR of syngas in traditional reformers can be considered ratio of the difference between moles of H2 and CO2 over the sum of
as the most practical factor to improve the methanol plant pro- CO and CO2 (eq. (6)).
duction capacity while saving the processing and manufacturing
H2  CO2
costs (Lange, 2001). Hence our focus in this work is to optimize SR ¼ (6)
syngas molar ratio in reformers for subsequent downstream
CO þ CO2
methanol plants. Note that the stoichiometric value SR should include the CO2
molar flow rate as well since H2 reacts with CO2 through the reverse
1.2. Methanol production and the importance of syngas wateregas shift reaction (reaction 5). Methanol operators prefer to
stoichiometric ratio (SR) feed their rectors with SR ¼ 2.0e2.1 (Rostrup-Nielsen, 2000) which
results in substantial economic benefits by improving single pass
Major industrial scale methanol production processes include syngas conversion and decreasing purges and unconverted gas

Fig. 1. Global methanol production and market demand forecast by 2020 (Alvarado, 2016).
 H.R. Shahhosseini et al. / Journal of Cleaner Production 180 (2018) 655e665 657

Fig. 2. Diagram for the conventional steam methane reforming (SMR).

recycling (Lange, 2001). Theoretically, SR ¼ 2 is expected to be the

! ! ! !
best ratio for maximum methanol production. Larger values minimise fð x Þ ¼ ½f 1 ð x Þ; f 2 ð x Þ; …; f k ð x ÞT
necessitate additional technologies for hydrogen separation and Subject to :
recycling while SR < 2 yields excessive by-products such as alcohols !!
gðxÞ  0 (7)
with more than two carbons. The synthesis gas which is produced !!
hðxÞ ¼ 0
by one step reforming has a typical SR of 2.8e3 and contains H2 ! !! !! !!
surplus of about 40% (Aasberg-Petersen et al., 2001). This needs x 2Rn ; f ð x Þ2Rk ; g ð x Þ2Rm and h ð x Þ2Rp
expensive external technologies such as membranes or end-of-pipe
where k; n; m and q are the number of objective functions, inde-
treatment processes to separate and recycle excess H2 (Alwi et al.,
pendent variables xi , inequality and equality constraints. In this
2016). !!
!
Another option is to adjust excess H2 ratio by injecting CO2. regard, x 2Rn is a vector of decision variables and f ð x Þ2RK is a
However, based on experimental work conducted by Ryi et al. vector of objective functions. The feasible decision space X and
(2014), the addition of CO2 to the inlet feed stream of SMR feasible criterion space (feasible search region) W are defined as
considerably decreases the methane conversion (specially at follow:
T < 1073 K) due to lower activity of Ni/MgAL2O3 catalyst in the co-

! ! ! 
feeding approach. Hence in this work we will adopt a new strategy X ¼ x gi ð x Þ  0; i ¼ 1; 2; …; m and hj ð x Þ ¼ 0; j ¼ 1; 2; …; p
to inject CO2 and optimize the SR value for methanol units while ! !
W ¼ fFð x Þj x 2Xg
maintaining high degree conversion of methane. Note that the CO2
(8)
with the natural gas (i.e. separated in absorbers in gas refineries) as
well as power plants, cement industries, steel manufacturers, Since a MOO problem requires simultaneous satisfaction of
aluminium smelters and fermenter exhaust streams are feasible different and often conflicting objectives, there is no single global
carbon dioxide sources and may easily be accessible for CO2 in- solution. Therefore, among various objectives, it is necessary to
jection in SMRs. Using purges from these industries will also help to determine a set of points from which the optimal trade-off pa-
alleviate environmental concerns regarding their contribution to rameters are selected. Pareto frontier helps to find such a set of
greenhouse gas emission (Olah et al., 2011). optimal points which result in the maximized conditions. Pareto
frontier is defined as follows:

! !
1.3. Multi-objective optimisation f i ð x Þ  f *i ð x Þ for all i2f1; 2; …; kg
! * ! (9)
f i ð x Þ < f i ð x Þ for alteast once i2f1; 2; …; kg
Chemical process industries are seeking to maximise their
production capacities and products quality. This also applies to Vector f * is a Pareto optimal solution if there is no other f in the
methanol industries and their corresponding SMR which de- feasible search region (Madetoja et al., 2008). Fig. 3 displays Pareto
termines the final methanol throughput (Mohanty, 2006). How- frontier and the feasible search region for four possible combina-
ever, there is usually a trade-off between production capacity and tions of two types of objectives.
quality. A promising decision can be made using multi-objective The solutions in the Pareto set are mathematically equivalent
optimisation (MOO) technique (Kravanja and Cu

cek, 2013; Li and there is no preference among them, but a user should select a
et al., 2009; Tian et al., 2011; Sayyaadi et al., 2011). Using MOO, single final solution. There are two techniques to find the only so-
the problem can be mathematically formulated as follows: lution to MOO problems: (i) techniques with previous preferences
658 H.R. Shahhosseini et al. / Journal of Cleaner Production 180 (2018) 655e665

CH4 conversion using multi-objective optimization (MOO) method.

2 2 To best of our knowledge this is the first time that such strategy has
Feasible Feasible search been employed for SMRs. The basics of multi-objective optimiza-
search region region tion technique and its application in industries including reformers
is explained subsequently.

1 1
2. Results and discussion
Minimise Minimise Minimise Maximise
2.1. Mathematical modelling of conventional SMR and the novel
configuration description
2 2
Feasible Feasible
search region search region Our steady state mathematical modelling of conventional SMR
reactor was validated against Pantoleontos et al. (2012) experi-
mental data. Our major assumptions for this modelling are:

1 1 (i) Ideal gas phase behaviour (The compressibility factor z of the

feed stream is calculated to be 0.990 and 0.991 based on the
Maximise Minimise Maximise Maximise
Peng-Robinson and Soave-Redlich-Kwong equations of state.
Fig. 3. Pareto set for four combinations of two types of objectives. For the product stream the values are 1.002 and 1.001,
respectively).
(ii) Plug flow and negligible axial dispersion of heat and mass.
and (ii) techniques with subsequent preferences (Marler and Arora, (iii) No radial concentration and temperature gradients.
2004). In previous preferences techniques, the relative significance (iv) Uniform particle size and constant bed porosity
of various objective functions can be estimated by employing the (v) Only reactions (1) to (3) from Xu and Froment (1989b) are
higher level information. According to the estimation, the MOO considered to reduce the complexity of modelling and
problem could then be transferred into a single-objective optimi- optimization
zation (SOO) problem to be solved in an evolutionary or a classical
way. Transformation could happen using the Weighted Product The corresponding SMR reaction rates are provided in Table 1.
Method (WPM). In this method, the compound function is intro- Equilibrium constants, Arrhenius kinetic parameters and Van't
duced by multiplication of objectives with a user-defined power as Hoff parameters for species adsorption are presented in Table 2 (Xu
below: and Froment, 1989a).
The consumption or formation rate for species “i”, ri (kmol/
! Y k
! kgcat.h), are evaluated by summing the reaction rates of species i, in
minimise Feq ð x Þ ¼ ½f i ð x Þwi (10)
reaction j, Rj (kmol/kgcat.h), in all three reactions. Effectiveness
i¼1
factors hj are also applied which account for the intra-particle
where w is a vector of powers which is selected by user (Deb, 2005). transport restriction. Regarding the reactions (1) to (3), one can
The attribution of weight factors to various objectives depends on obtain the reaction rate of SMR species as follows:
the decision makers' viewpoint, therefore various solutions may be
rCH4 ¼ h1 R1  h3 R3
attained for identical problems. In the subsequent preferences
rH2 O ¼ h1 R1  h2 R2  2h3 R3
techniques, the decision makers play a critical role and directly
rH2 ¼ 3h1 R1 þ h2 R2 þ 4h3 R3 (15)
perform their priority on Pareto frontier to select one of the rCO ¼ h1 R1  h2 R2
equivalent choices as the only solution point. In this technique, the rCO2 ¼ h2 R2 þ h3 R3
decisive solution exactly refers to decision makers priorities
(Marler and Arora, 2004; Taghdisian et al., 2015). where h1 ¼ 0.07, h2 ¼ 0.7, h3 ¼ 0.06 (De Groote and Froment, 1996).

1.4. Current study objective Table 1

SMR kinetic reaction model.
Although there have been studies and reviews regarding new Reaction rate equation
reactor configurations, gas recovery technologies and emission
policies for methanol industries to enhance production (Saeidi !
k1 p3H2 pCO 1
et al., 2014; Luu et al., 2016; Kleme
s, 2010), there are few model- R1 ¼ pCH4 pH2 O   (11)
p2:5
H2
Ke1 U2
ling studies on SMR reactors for this purpose. The aim of the pre-
sent study is to introduce a new configuration for SMR to achieve  
SR ¼ 2 and eliminate the implementation of downstream H2 sepa- R2 ¼
k2 p p 1
pCO pH2 O  H2 CO2  2 (12)
ration and recycling practises. Here, we propose a multi-step CO2 pH2 Ke2 U
shot along the SMR reactor tubes which is also expected to enhance
!
the reaction rates based on the Le Chatelier's principle. The length k3 p4H2 pCO2 1
R3 ¼ pCH4 p2H2 O   (13)
of the conventional SMR reactor is discretised into several seg- p3:5
H2
Ke3 U2
ments for CO2 injections. Iranshahi et al. (2011) and Arabpour et al.
(2012) integrated this strategy combined with differential evolu- p H2 O
tion (DE) optimization in the naphtha reformer and Fischer-Tropsch U ¼ 1 þ kCO pCO þ kH2 pH2 þ kCH4 pCH4 þ kH2 O (14)
p H2
reactors for conversion enhancement. In this work, the optimal CO2
injection is determined by minimising abs (SR-2), and maximising
 H.R. Shahhosseini et al. / Journal of Cleaner Production 180 (2018) 655e665 659

Table 2
Reaction equilibrium constants, Arrhenius kinetic parameters and Van't Hoff parameters for adsorption of species.

Reaction; j ðsee Table 1Þ Equilibrium constant; kej koj ðkmol=kgcat hÞ Ej ðkJ=kmolÞ

 
1 4:2251015 ðbar0:5 Þ 240100
ðbar2 Þ
3
ke1 ¼ 4:7071012 exp  206:110
8:314T
 
2 1:955106 67130
¼ 1:142102 exp 41:1510
3
ke2 8:314T
 
3 1:0201015 ðbar0:5 Þ 243900
ðbar2 Þ
3
ke3 ¼ 5:3751010 exp  164:910
8:314T
 
E
kj ¼ koj exp  RTj
       
DHCH DH H O DH H
kCH4 ¼ ko;CH4 exp  RT 4 kH2 O ¼ ko;H2 O exp  RT2 kH2 ¼ ko;H2 exp  RT 2 kCO ¼ kCO exp  DRT
HCO

ko;CH4 ¼ 6:65104 ðbar1 Þ ko;H2 O ¼ 1:77105 ko;H2 ¼ 6:12109 ðbar1 Þ ko;CO ¼ 8:23105 ðbar1 Þ
       
DHCH4 ¼ 38:2103 kmol kj
DHH2 O ¼ 88:68103 kj
kmol
DHH2 ¼ 82:9103 kmolkj
DHCO ¼ 70:65103 kmol kj

The developed mathematical model is a one-dimensional het- temperature, reactor dimensions and catalyst characteristics are
erogeneous one and includes mass and energy balance for gas and also shown in Table 4.
solid phases. Conservation balances including the boundary con- Table 5 compares our modelling outcomes with the data from
ditions as well as auxiliary equations are summarised in Table 3. Pantoleontos et al. (2012). As shown, our model predictions are
The pressure drop along the axial direction of the SMR is calculated very consistent with the actual industrial SMR data. Treating the
by the Ergun equation. Here KD and KV are parameters corre- effectiveness factor as a constant value through the entire catalytic
sponding to the viscous and kinetic loss terms (Alpay et al., 1995). bed as well as ignoring axial dispersion of heat and mass could be
The overall wall-bed heat transfer coefficient is selected from Dixon the main sources of relative and absolute errors.
(1996). As discussed before, based on the previous studies, the single
The model consists of a set of ordinary differential mass and shot addition of CO2 to the SMR feed stream may result in
energy balance equations (ODEs), kinetic model and empirical decreasing CH4 conversion. Hence, in this study multiple locations
correlations. The ODEs are solved by finite difference approxima- along the SMR reactor length are considered for CO2 dosing (see
tion. In this method, the length of the reactor is divided into 10,000 Fig. 4). It is obvious that increasing number of CO2 injection points
separate segments and the GaussNewton algorithm is used to will enhance reactor performance (i.e. shifting equilibrium re-
solve the set of nonlinear algebraic equations for each segment actions) by providing an optimal CO2 profile along the reactor
(Iranshahi et al., 2011). The initial feed composition and length (Note: our study illustrates that the effect of injected CO2 is

Table 3
Mass and Energy models with corresponding boundary conditions and empirical correlations.

Mass and energy balances in thegas phase dCi  

us þ kg;i av Ci  Ci;s ¼ 0 i ¼ CH4 ; H2 O; H2 ; CO; CO2 (16)
dz

dT 4U
us rf Cpg ¼ hf av ðTs  TÞ  ðT  Tw Þ (17)
dz dtube;i

 
Mass and energy balances in the solid phase kg;i av Ci  Ci;s ¼ ð1  εb Þrcat ri i ¼ CH4 ; H2 O; H2 ; CO; CO2 (18)

X 
hf av ðTs  TÞ ¼ ð1  εb Þrcat DHrxn;j hj Rj j ¼ 1; 2; 3 (19)

Pressure drop dP
¼ KD us  KV u2s (20)
dz
150mg ð1εb Þ 1:75ð1εb Þrf
KD ¼ 2 KV ¼
dp ε3b ðPa s=m2 Þ dp ε3b ðPa s2 =m2 Þ

 
overall wall  bed heat transfercoefficient 1 1 dtube;o Bi þ 3
¼ þ (21)
U aw 6ler Bi þ 4

1=3
aw dtube;o NRe Npr 0:95ð1  εb Þ
Bi ¼    ler ¼ l0er þ 0:111lg    l0 ¼ εb ðlg þ 0:95aru dp;i Þ þ
2ler 1 þ 46ðdp;i =dtube;o Þ er 2=3ls þ 1=10lg þ ars dp;i
0 !1:5 1
0:8171ðT=100Þ3 A lg N0:59 N1=3
em dtube;i
am ¼     ars ¼ 0:8171 ðT=100Þ3     aw ¼ @1  1:5 pr
1 þ εb =2ð1  εb Þð1  emÞem 2  em dp;i dp;i Re

Boundary conditions At the reformer inlet z ¼ 0:0

Ci ¼ Ci;o i ¼ CH4 ; H2 O; H2 ; CO; CO2 (22)

T ¼ To P ¼ Po
660 H.R. Shahhosseini et al. / Journal of Cleaner Production 180 (2018) 655e665

Table 4
Additional data for the SMR process.

Parameter Value

TðKÞ 700.15
LðmÞ 12
rcat ðkg=m3 Þ 2355.2
dp ðmÞ 0.0084
dtube;i ðmÞ 0.1016
dtube;o ðmÞ 0.1322
FCH4 ;o ðkmol=hÞ 5.17
FH2 O;o ðkmol=hÞ 17.35
FH2 ;o ðkmol=hÞ 0.63
FCO;o ðkmol=hÞ 0.0
FCO2 ;o ðkmol=hÞ 0.29
Po ðbarÞ 25.7
ls ðW=m KÞ 0.3489
em 0.8

gradually decreased along the reactor length). However, using a

large number of injection points is not feasible mainly due to
operation limitations and costs. Considering that the reactor length
is 12 m, here we considered eleven CO2 injection nodes (i.e. with
equal distances from each other) to obtain optimized injection flow
rate at each node. Note: optimization is expected to decrease the
impact of number of injection points. The optimum CO2 dose at
each section is determined by MOO as described below.

2.2. MOO and SOO joined procedure for the new SMR reactor

The joined procedure contains three main steps for creating a

MOO problem: (i) defining a multi-objective model, (ii) producing
optimal Pareto set and (iii) converting the MOO model to a SOO
model for finding the single solution point. Fig. 5 illustrates the
applied joined procedure in this study. Simply, the multi-objective Fig. 4. Proposed SMR catalytic packed bed reactor.
model is formulated by using the SMR mathematical model. Sub-
sequently, the Pareto frontier is produced and the convex nature of
the Pareto set is evaluated. Finally, the WPM is used to transform 2.2.1. Defining objective functions for the new SMR configuration
the problem into SOO. and determining Pareto line
By means of the Prior Articulation of Preferences Approach To obtain the optimum injected CO2 flow rate at each node, the
(PAPA), the authors will select a single solution in MOO. Subse- multi-objective model is formulated as follow:
quently, WPM (i.e. a common technique in PAPA) is applied to
transfer the MOO problem into a single objective optimisation via !
Max f 1 ð x Þ : maximise CH4 conversion 
multiplication of objectives and applying the power factors to ! (23)
Min f 2 ð x Þ : minimise absðSR  2Þ
different objectives. The single final optimal point is determined by
the genetic algorithm (GA) which is a well-placed method in It is clear, these are the same classical objectives in SMR design
computational optimisation. GA does not need gradient informa- which aim to maximise the CH4 conversion (i.e. higher syngas
tion and therefore is an appropriate method regardless of the na- generation) while minimising abs (SR-2) for enhancing methanol
ture of the objective function (Harris et al., 2000). A population size production. For the case of using two objectives, it is possible to
of 80, generations of 100 with two basic operations, i.e., crossover visualise Pareto frontier in a two-dimensional space.
probability of 0.9 and mutation probability of 0.08 are chosen. Visualisation of the Pareto optimal front is forthright for the

Table 5
Modelling results versus industrial data.

Variable Modelling results Experimental data Relative error % Absolute error 

TðKÞ 1; 101:45 1; 106:01 0:41 4:56

PðbarÞ 23:45 23:34 0:47 0:11
FCH4 1:6501 kmol=h 1:6766 kmol=h 1:58 0:0265
F H2 O 12:5965 kmol=h 12:6801 kmol=h 0:66 0:0836
FCO 2:2334 kmol=h 2:2271 kmol=h 0:28 0:0063
F H2 12:3554 kmol=h 12:3237 kmol=h 0:26 0:0317
FCO2 1:5743 kmol=h 1:5659 kmol=h 0:54 0:0084
Ftotal 30:4097 kmol=h 30:4734 kmol=h 0:21 0:0637
FH2 FCO2 2:8314 2:8362 0:17 0:0048
SR ¼ FCO þFCO2
CH4 conversion 68:08% 67:57% 0:75 0:51
 H.R. Shahhosseini et al. / Journal of Cleaner Production 180 (2018) 655e665 661

Optimum Operational
Optimisation by GA
Decision Variables

Transfering MOO
Pareto Single-Objective
SMR Model MOM Problem into SOO
Frontire Model
Problem

Operational decision
Variables

Generating Multi- Producing Single Solution Point

Objective Model Pareto Set (Optimal Decision Variables)

Fig. 5. Illustrative procedure of the joined methodology.

2.8 C
2.6 feasible search
region
2.4
Stoichiometric ratio (S R)

B
2.2 Coordinates relevant
to the conventional SMR
2
performance
1.8

1.6
Pareto
1.4 optimal-front
(without constraint)
1.2
A
1

0.655 0.66 0.665 0.67 0.675 0.68

CH4 conversion

Fig. 6. Feasible search region and Pareto optimal-front of Multi-objective model based on eq. (23).

decision makers to compare different solutions according to their between points A and B. Point C represents the coordinates relevant
location on the Pareto frontier (Madetoja et al., 2008). Fig. 6 in- to the conventional SMR performance (SR ¼ 2.831 and CH4
dicates feasible search region and Pareto curve for the multi- conversion ¼ 68.08%).
objective model based on eq. (23). The feasible search region con- In addition to SR, CO/CO2 ratio of syngas is another important
tains 531,441 data over a wide range of decision variables in which factor for subsequent methanol factories. CO/CO2 ratio of greater
SR versus CH4 conversion is illustrated (Note for a better visual- than one may increase methanol production while jeopardizing the
isation, SR has been shown in this figure instead of abs (SR-2)). catalyst lifetime due to high water formation (Aasberg-Petersen
It can be observed that the Pareto frontier is a convex line et al., 2007). Hence based on the new constraint, a new feasible

2.8 C
2.6 New feasible search
region
2.4
Stoichiometric ratio (S R)

E B
2.2 Coordinates relevant
selected by GA
O to the conventional SMR
2 decision making
performance
method
1.8

1.6
Pareto
1.4 D optimal-front
(without constraint)
1.2 pareto
A
optimal-front
1
(with constraint)

0.655 0.66 0.665 0.67 0.675 0.68

CH4 conversion

Fig. 7. New feasible search region and convex Pareto line of multi-objective model based on eq. (23) and CO=CO2 ratio restriction, GA decision-making method to specify the final
optimal design point.
662 H.R. Shahhosseini et al. / Journal of Cleaner Production 180 (2018) 655e665

Table 6
The optimum CO2 injection values for the new SMR configuration.

CO2 injection  flow rateðkmol=hÞ Coordinate of inlet

nodes along the reactor length ðmÞ

0.4000 0
0.0925 1
0.0810 2
0.1010 3
0.0890 4
0.0680 5
0.0960 6
0.0749 7
0.0890 8
0.0844 9
0.0986 10
0.0787 11

search region, and Pareto curve is determined as shown in Fig. 7.

As demonstrated in Fig. 7, the new Pareto frontier is a convex
line between points D and E with the highest CH4 conversion
(67.13%) at design point E. The SR indicator has its minimum value of
1.45 at point D and increases gradually as the CH4 conversion rises
to around 67%. Further increase in CH4 conversion results in a
drastic rise in the stoichiometric ratio.
In order to determine a single optimal point, the MOO problem
Fig. 9. H2 molar flow rate along the reactor.
is transformed to a SOO problem using WPM (eq. (24)). A single
optimum point is then obtained by using GA as shown below:
2.3. Comparison between performance of conventional and
optimised SMR
! !
minimise  f eq ð x Þ¼absðSR 2Þ=CH4 conversion ¼f ðx1 ;x2 ;…;x12 Þ
s:t: The performance of SMR is evaluated versus new optimized
0xi ðCO2 =H2 O ¼0:078Þ i¼1;2;3;…;12 configuration in Figs. 8e11. Variations in molar flow rates of
X12 reforming species, temperature and pressure profiles along re-
xi ¼ðCO2 =H2 O ¼0:078Þ actors are compared here.
i¼1
(24)
2.3.1. CH4 conversion and H2O molar flow rate
Point O in Fig. 7 represents the coordination relevant to the new CH4 and H2O profiles along conventional and optimized
optimised SMR performance (SR ¼ 2.015 and CH4 conver- configuration are demonstrated in Fig. 8. Injecting a defined
sion ¼ 67.11%) which was selected based on the GA results. The amount of CO2 at various lengths of the reactor shifts the equilib-
corresponding optimum CO2 injection values for eleven nodes are rium of reactions (1) to (3) according to Le ChateliereBrown prin-
therefore listed in Table 6. ciple. Fig. 8 (a) demonstrates the similar performance of SMR and

Fig. 8. The profiles along the reactor (a) CH4 conversion percentage and (b) H2O molar flow rate.
 H.R. Shahhosseini et al. / Journal of Cleaner Production 180 (2018) 655e665 663

Fig. 10. (a) CO and (b) CO2 molar flow rates along the reactor length.

its optimized configuration regarding the methane reforming. reaction (reaction 2).
Although the CH4 conversions in both reactors are approximately
the same, the water formation tends to be a bit more in the case of 2.3.3. Temperature and pressure profiles
new optimized reactor (Fig. 8 (b)). This could be attributed to the Fig. 11 (a) shows a slightly lower temperature profile for the
shift of reverse reforming reactions as a result of the multi-step CO2 second half of the reactor which also improves the pressure drop
injection. along the reactor. About 1% temperature drop in new optimized
SMR, considerably affects the gas properties such as viscosity,
2.3.2. H2, CO and CO2 products density and velocity which subsequently lowers the pressure drop
Fig. 9 represents variation in the H2 molar flow rate in both (Fig. 11 (b)). This will provide further opportunity for process en-
reactors. Hydrogen production rate decreases insignificantly from gineers to use smaller catalysts with higher effectiveness factors
12.35 kmol/h to 12.02 kmol/h in the optimized configuration which improve reaction yields. The lower pressure drop in opti-
mainly through the reverse of reactions 2 and 3. However, the mized SMR is also beneficial due to the enhancement of reaction
proposed configuration can successfully adjust the stoichiometric rates because of higher gas phase concentrations on the catalyst
ratio SR ¼ 2:0  2:1 which is considered to be advantageous for surface.
methanol producers to improve their production rate.
Fig. 10 indicates efficacy of using industrial CO2 purge streams to 3. Conclusions
add additional values to methanol infrastructures. The effect of
booster CO2 on CO flow rate is obvious. CO production rate has The current study investigated the possibility of utilizing CO2-
increased about 0.73 kmol/h following the injection of 1.35 kmol/h rich effluents of industrial sites in SMRs for production augmen-
of CO2 in total. Higher CO output in the optimized configuration tation of methanol units. Here a new reactor configuration was
could mainly be due to the synergistic effect of high temperature proposed, capable of adjusting the stoichiometric ratio SR in SMRs
and additional CO2 concentration which drive the water gas shift while eliminating the expensive and high-energy duty end-of-pipe

Fig. 11. (a) Temperature and (b) Pressure profiles along the reactor length.
664 H.R. Shahhosseini et al. / Journal of Cleaner Production 180 (2018) 655e665

separation-recycling steps. Multiple shots of CO2 were considered KD The parameter corresponding to the viscous loss term
for conventional SMR along its catalytic bed to adjust the SR and ðPa s=m2 Þ
improve the reformer's products. Optimum CO2 injection flow rates L The length of the reactor ðmÞ
were determined by the multi-objective optimization technique Npr Prandtl number
and identifying the Pareto frontier regions. CH4 conversion and abs NRe Reynolds number
(SR-2) were the objectives to be optimized. Consequently, the P Total gas pressure ðbarÞ
stoichiometric ratio of syngas as a critical parameter in methanol pi The partial pressure of gas species i ðbarÞ
production was decreased from SR ¼ 2.831 to SR ¼ 2.015 for the new Po Initial pressure (bar)
configuration which is closer to the optimum theoretical value of ri The rate of consumption or formation of species
SR ¼ 2. This is while the CH4 conversion remained at the same level i ðmol=ðkgcat sÞÞ
as the conventional SMR at about 68%. The new configuration was Rj Rate of reaction j ðkmol=ðkgcat hÞÞ
capable of producing high amounts of H2; matching the 40% in the T Gas-phase temperature ðKÞ
conventional SMR. Based on our modelling outcomes, CO2 injection To Initial temperature ðKÞ
significantly increased the CO generation mainly due to “reverse Ts Solid catalyst temperature ðKÞ
shift reaction”. The above findings may be beneficial to the SMR Tw Wall temperature ðKÞ
operators to redesign and redefine their operations for production
U Overall heat transfer coefficient ðJ m2 s1 K1 Þ
enhancement in methanol industries.
us Superficial gas mixture velocity ðm=sÞ
Z Axial dimension ðmÞ
Acknowledgment !!
gðxÞ The vector of inequality constraints

The authors would like to express their sincere appreciation to

zmany Peter GAGenetic algorithm
Faculty of Information Technology and Bionics, Pa !!
Catholic University in Budapest, Hungary for their support. We hðxÞ The vector of equality constraints
would also like to thank Ms E. Cowey for her kind help in editing F* The vector of Pareto optimal points
the manuscript language. !
f eq ð x Þ Composite function
!
x The vector of decision variables
List of symbols wi The power factor for objective function f i  
(dimensionless)
W Feasible criterion space (dimensionless)
Variables w The vector of powers (dimensionless)
av External catalyst surface area per unit volume of catalyst X Feasible decision space
Q Number of equality constraints
bed ðm2 =m3 Þ
k Number of objectives
aru The parameter used in to calculate static radial thermal !
f ið x Þ ith objective function
conductivity ðkJ m2 h1 K1 Þ !
ars The parameter used to calculate static radial thermal Fð x Þ The vector of objective functions
M Number of inequality constraints
conductivity ðkJ m2 h1 K1 Þ
N Number of independent variables
aw Wall thermal transfer coefficient ðkJ m2 h1 K1 Þ
Bi The Biot number
Greek symbols
Ci;s The concentration of species i in the solid phase
DHi The heat of adsorption of species i ðJ=molÞ
ðmol=m3 Þ DHrxn;j Heat of reaction j ðJ=molÞ
Ci The concentration of species i in the gas phase ðmol=m3 Þ DHo298 The heat of reaction of at STP ðkJ=kmolÞ
Cp The heat capacity ðJ kg1 K1 Þ εb Packing bed porosity (dimensionless)
Cpg The heat capacity of the gas mixture ðJ kg1 K1 Þ hj Effectiveness factor of reaction j
dp Catalyst particle diameter (m) lg Average gas thermal conductivity ðW=m KÞ
dtube;i The inner diameter of the reactor tube ðmÞ ls Solid thermal conductivity ðW=m KÞ
em The emissivity of the solid surface lfz Effective thermal conductivity ðW=m KÞ
Ej The activation energy of reaction j ðJ=molÞ l0er Static radial thermal conductivity ðkJ m2 h1 K1 )
Fi;o The molar flow rate of component i in the feed mg Average gas viscosity ðkg=ðm sÞÞ
ðkmol=hÞ ði ¼ CH4 ; HO2 ; CO; H2 ; CO2 ; N2 Þ rcat The density of the catalyst pellet ðkg=m3 Þ
Ftotal The total flow rate of the mixture in the bed, (kmol/h)
rf The density of the fluid ðkg=m3 Þ
Ftotal ; o The total flow rate of the mixture in the feed, (kmol/h)
U Dominator term in the reaction kinetics
kej Thermodynamic equilibrium constant of the reaction
ki Adsorption constant of species i
Abbreviations
kj Temperature dependent kinetic rate constant of
SR Stoichiometric ratio
reaction j
SMR Steam methane reforming
kg;i The mass transfer coefficient of species i ðm=sÞ
SOO Single objective modelling
koi Reference adsorption constant of species i WPM Weighted product method
koj The reference temperature dependent kinetic rate ODEs Ordinary differential equations
constant of reaction j MOO Multi-objective optimisation
Kv The parameter corresponding to the kinetic loss term MOM Multi-objective modelling
ðPa s2 =m3 Þ GA Genetic algorithm
 H.R. Shahhosseini et al. / Journal of Cleaner Production 180 (2018) 655e665 665

WGS Water gas shift Des. 87, 233e243.

Lira-Barraga n, L.F., Gutie
rrez-Arriaga, C.G., Bamufleh, H.S., Abdelhady, F., Ponce-
APA Prior Articulation of Preferences Approach lez, M., El-Halwagi, M.M., 2015. Reduction of green-
Ortega, J.M., Serna-Gonza
GTL Gas to liquid house gas emissions from steam power plants through optimal integration with
algae and cogeneration systems. Clean Technol. Environ. Policy 17, 2401e2415.
References Luu, M.T., Milani, D., Abbas, A., 2016. Analysis of CO2 utilisation for methanol syn-
thesis integrated with enhanced gas recovery. J. Clean. Prod. 112 (Part 4),
3540e3554.
Aasberg-Petersen, K., Bak Hansen, J.H., Christensen, T.S., Dybkjaer, I., Madetoja, E., Ruotsalainen, H., Monkkonen, V.M., Hamalainen, J., Deb, K., 2008.
Christensen, P.S., Stub Nielsen, C., Winter Madsen, S.E.L., Rostrup-Nielsen, J.R., Visualising multi-dimensional pareto-optimal fronts with a 3D virtual reality
2001. Technologies for large-scale gas conversion. Appl. Catal. A General 221, system. In: International Multiconference on Computer Science and Informa-
379e387. tion Technology, 2008. IMCSIT 2008, pp. 907e913.
Aasberg-Petersen, K., Stub Nielsen, C., Dybkjaer, I., 2007. Very large scale synthesis Marler, R.T., Arora, J.S., 2004. Survey of multi-objective optimisation methods for
gas production and conversion to methanol or multiple products. In: Bellot engineering. Struct. Multidisciplinary Optimisation 26, 369e395.
Noronha, F., M.S., Falabella, S.-A. (Eds.), Studies in Surface Science and Catalysis. Masih, A.M.M., Albinali, K., DeMello, L., 2010. Price dynamics of natural gas and the
Elsevier, pp. 243e248. regional methanol markets. Energy Policy 38 (3), 1372e1378.
Alpay, E., Kershenbaum, L.S., Kirkby, N.F., 1995. Pressure correction in the inter- Mendes, L., de Medeiros, J.L., Alves, R.M., Araújo, O.Q., 2014. Production of methanol
pretation of microreactor data. Chem. Eng. Sci. 50, 1063e1067. and organic carbonates for chemical sequestration of CO2 from an NGCC power
Alvarado, M., 2016. Global Methanol Outlook 2016. Retrieved from: http://www. plant. Clean Technol. Environ. Policy 16, 1095e1105.
methanol.org. Mohanty, S., 2006. Multiobjective optimisation of synthesis gas production using
Alwi, S.R.W., Tin, O.S., Rozali, N.E.M., Manan, Z.A., Kleme
s, J.J., 2013. New graphical non-dominated sorting genetic algorithm. Comput. Chem. Eng. 30, 1019e1025.
tools for process changes via load shifting for hybrid power systems based on Nikoo, M.K., Saeidi, S., Lohi, A., 2015. A comparative thermodynamic analysis and
Power Pinch Analysis. Clean Technol. Environ. Pol. 15, 459e472. experimental studies on hydrogen synthesis by supercritical water gasification
Alwi, S.R.W., Kleme
s, J.J., Varbanov, P.S., 2016. Cleaner energy planning, manage- of glucose. Clean Technol. Environ. Policy 17, 2267e2288.
ment and technologies: perspectives of supply-demand side and end-of-pipe Olah, G.A., Goeppert, A., Prakash, G.S., 2011. Beyond Oil and Gas: the Methanol
management. Clean. Prouction. https://doi.org/10.1016/j.jclepro.2016.07.181. Economy. WILEY-VCH Verlag GbbH & Co KGaA, Weinheim, Germany.
Arabpour, M., Rahimpour, M.R., Iranshahi, D., Raeissi, S., 2012. Evaluation of Pantoleontos, G., Kikkinides, E.S., Georgiadis, M.C., 2012. A heterogeneous dynamic
maximum gasoline production of FischereTropsch synthesis reactions in GTL model for the simulation and optimisation of the steam methane reforming
technology: a discretised approach. J. Nat. Gas Sci. Eng. 9, 209e219. reactor. Int. J. Hydrogen Energy 37, 16346e16358.
Bozzano, G., Manenti, F., 2016. Efficient methanol synthesis: perspectives, tech- Rajesh, J.K., Gupta, S.K., Rangaiah, G.P., Ray, A.K., 2000. Multiobjective optimisation
nologies and optimisation strategies. Prog. Energy Combust. Sci. 56, 71e105. of steam reformer performance using genetic algorithm. Ind. Eng. Chem. Res.
Bulatov, I., Kleme
s, J.J., 2011. Clean fuel technologies and clean and reliable energy: a 39, 706e717.
summary. Clean Technol. Environ. Policy 13, 543e546. Riaz, A., Zahedi, G., Kleme
s, J.J., 2013. A review of cleaner production methods for
Chuah, L.F., Yusup, S., Aziz, A.R.A., Kleme
s, J.J., Bokhari, A., Abdullah, M.Z., 2016. the manufacture of methanol. J. Clean. Prod. 57, 19e37.
Influence of fatty acids content in non-edible oil for biodiesel properties. Clean Rostrup-Nielsen, J.R., 2000. New aspects of syngas production and use. Catal. Today
Technol. Environ. Policy 18, 473e482. 63, 159e164.



Cu cek, L., Kleme
s, J.J., Kravanja, Z., 2012. Carbon and nitrogen trade-offs in biomass Rozali, N.E.M., Alwi, S.R.W., Manan, Z.A., Kleme
s, J.J., Hassan, M.Y., 2015. A process
energy production. Clean Technol. Environ. Policy 14, 389e397. integration approach for design of hybrid power systems with energy storage.



Cu cek, L., Kleme
s, J.J., Varbanov, P.S., Kravanja, Z., 2015. Significance of environ- Clean Technol. Environ. Policy 17, 2055e2072.
mental footprints for evaluating sustainability and security of development. Ryi, S.-K., Lee, S.-W., Park, J.-W., Oh, D.-K., Park, J.-S., Kim, S.S., 2014. Combined
Clean Technol. Environ. Policy 17, 2125e2141. steam and CO2 reforming of methane using catalytic nickel membrane for gas
De Groote, A.M., Froment, G.F., 1996. Simulation of the catalytic partial oxidation of to liquid (GTL) process. Catal. Today 236 (Part A), 49e56.
methane to synthesis gas. Appl. Catal. A General 138, 245e264. Saeidi, S., Amin, N.A.S., Rahimpour, M.R., 2014a. Hydrogenation of CO2 to value-
Deb, K., 2005. Multi-objective optimisation. In: Burke, E.K., Kendall, G. (Eds.), Search added productsdA review and potential future developments. J. CO2 Util. 5,
Methodologies: Introductory Tutorials in Optimisation and Decision Support 66e81.
Techniques. Springer US, Boston, MA, pp. 273e316. Saeidi, S., Talebi Amiri, M., Amin, N.A.S., Rahimpour, M.R., 2014b. Progress in re-
Dixon, A.G., 1996. An improved equation for the overall heat transfer coefficient in actors for high-temperature fischeretropsch process: determination place of
packed beds. Chem. Eng. Process. Process Intensif. 35, 323e331. intensifier reactor perspective. Int. J. Chem. React. Eng. 12, 1e26.
Dutta, S., 2014. A review on production, storage of hydrogen and its utilisation as an Saeidi, S., Nikoo, M.K., Mirvakili, A., Bahrani, S., Amin, N.A.S., Rahimpour, M.R., 2015.
energy resource. J. Ind. Eng. Chem. 20 (4), 1148e1156, 25. Recent advances in reactors for low-temperature Fischer-Tropsch synthesis:
Edwards, P.P., Kuznetsov, V.L., David, W.I.F., et al., 2008. Hydrogen and fuel cells: process intensification perspective. Rev. Chem. Eng. 31, 209e238.
towards a sustainable energy future. Energy Policy 36 (12), 4356e4362. Sayyaadi, H., Babaie, M., Farmani, M.R., 2011. Implementing of the multi-objective
Harris, S.D., Elliott, L., Ingham, D.B., Pourkashanian, M., Wilson, C.W., 2000. The particle swarm optimizer and fuzzy decision-maker in exergetic, exer-
optimisation of reaction rate parameters for chemical kinetic modelling of goeconomic and environmental optimisation of a benchmark cogeneration
combustion using genetic algorithms. Comput. Methods Appl. Mech. Eng. 190, system. Energy 36, 4777e4789.
1065e1090. Siriworarat, K., Deerattrakul, V., Dittanet, P., Kongkachuichay, P., 2017. Production of
Iranshahi, D., Bahmanpour, A., Paymooni, K., Rahimpour, M.R., Shariati, A., 2011. methanol from carbon dioxide using palladium-copper-zinc loaded on MCM-
Simultaneous hydrogen and aromatics enhancement by obtaining optimum 41: comparison of catalysts synthesized from flame spray pyrolysis and sol-
temperature profile and hydrogen removal in naphtha reforming process; a gel method using silica source from rice husk ash. J. Clean. Prod. 142 (Part 3),
novel theoretical study. Int. J. Hydrogen Energy 36, 8316e8326. 1234e1243.
Kleme
s, J.J., 2010. Environmental policy decision-making support tools and pollu- Taghdisian, H., Pishvaie, M.R., Farhadi, F., 2015. Multi-objective optimisation
tion reduction technologies: a summary. Clean Technol. Environ. Policy 12, approach for green design of methanol plant based on CO2-efficeincy indicator.
587e589. J. Clean. Prod. 103, 640e650.
_
Kliopova, I., Baranauskaite-Fedorova, _ M., Stani
skis, J.K., 2016.
I., Malinauskiene, Tian, X., Zhang, X., Zeng, S., Xu, Y., Yao, Y., Chen, Y., Huang, L., Zhao, Y., Zhang, S.,
Possibilities of increasing resource efficiency in nitrogen fertilizer production. 2011. Process analysis and Multi-objective optimisation of ionic liquid-
Clean Technol. Environ. Policy 18, 901e914. containing acetonitrile process to produce 1, 3-Butadiene. Chem. Eng. Tech-



Kravanja, Z., Cu cek, L., 2013. Multi-objective optimisation for generating sustainable nol. 34, 927e936.
solutions considering total effects on the environment. Appl. Energy 101, 
Van-Dal, E.S., Bouallou, C., 2013. Design and simulation of a methanol production
67e80. plant from CO2 hydrogenation. J. Clean. Prod. 57, 38e45.
Kravanja, Z., Varbanov, P.S., Kleme
s, J.J., 2015. Recent advances in green energy and squez, M., Boutonnet, M., Cabrera, S., Ja
Velasco, J.A., Lopez, L., Vela €rås, S., 2010. Gas
product productions, environmentally friendly, healthier and safer technologies to liquids: a technology for natural gas industrialization in Bolivia. J. Nat. Gas.
and processes, CO2 capturing, storage and recycling, and sustainability assess- Sci. Eng. 2 (5), 222e228.
ment in decision-making. Clean Technol. Environ. Policy 17, 1119e1126. Xu, J., Froment, G.F., 1989a. Methane steam reforming, methanation and water-gas
Lange, J.P., 2001. Methanol synthesis: a short review of technology improvements. shift: I. Intrinsic kinetics. AIChE J. 35, 88e96.
Catal. Today 64, 3e8. Xu, J., Froment, G.F., 1989b. Methane steam reforming: II. Diffusional limitations and
Li, C., Zhang, X., Zhang, S., Suzuki, K., 2009. Environmentally conscious design of reactor simulation. AIChE J. 35, 97e103.
chemical processes and products: multi-optimisation method. Chem. Eng. Res.