Vol.:(0123456789)
1 3
Topics in Catalysis
https://doi.org/10.1007/s11244-020-01401-0
ORIGINAL PAPER
Radial Basis Function Neural Network Model Prediction ofThermo-
catalytic Carbon Dioxide Oxidative Coupling ofMethane to C
2
-
hydrocarbon
BamideleVictorAyodele
1
· SitiIndatiMustapa
1
· ThongthaiWitoon
2
· RameshKanthasamy
3
·
MohammedZwawi
4
· ChieduN.Owabor
5
Accepted: 17 November 2020
© Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract
Various anthropogenic activities often result in the emission of carbon dioxide (CO
2
), which is one of the principal compo-
nents of greenhouse gases responsible for greenhouse effect. One vital strategy to mitigate the effect of the released CO
2
on
the environment is through sustainable utilization and conversion to value-added chemicals. This study employs the Radial
Basis Function artificial neural network for modeling the prediction of thermo-catalytic CO
2
oxidative coupling of methane
to C
2
-hydrocarbons. The various architecture of the Radial Basis Function ANN was developed, trained, and tested using
the non-linear relationship between the input parameters (reaction temperature, amount of CaO and MnO in the CaO-MnO/
CeO
2
catalysts and the CO
2
/CH
4
ratio) and the output parameters (C
2
hydrocarbon selectivity and yield). The Radial Basis
Function ANN architecture with the topology of 4-20-2, representing the input layer, hidden neurons, and the output layer
offers the best performance with a sum of square error (SSE) of 3.9 × 10
−24
for training and 0.224 for testing. The R
2
of 0.989
and 0.998 obtained for the prediction of the selectivity and the yield of the C
2
hydrocarbon is an indication of the robustness
of the Radial Basis Function ANN model. The sensitivity analysis revealed that the input parameters significantly influence
the model output. However, the reaction temperature has the most significant influence on the model output based on the
level of importance.
Keywords Artificial neural network· Carbon dioxide utilization· Radial basis function· Oxidative coupling of methane·
C
2
hydrocarbon
1 Introduction
Sustainable utilization of carbon dioxide (CO
2
) to mitigate
its effect on the environment has been the focus of research-
ers in the past decades [13]. CO
2
together with methane
are the major constituents of greenhouse gases which are
the main causes of greenhouse effect [4]. The Paris agree-
ment on climate change has given the mandate to reduce
greenhouse gas emissions by at least 40% by 2030 [5]. In
view of this, researchers have investigated various strate-
gies such as the recommendation of appropriate policies to
reduce CO
2
emissions, the use of renewable energy sources,
and the sustainable utilization of greenhouse gases to pro-
duce value-added chemicals [6, 7]. Several technological
processes such as CO
2
reforming of methane, CO
2
hydro-
genation, CO
2
methanation, and CO
2
oxidative coupling
of methane have been investigated [813]. CO
2
reforming
of methane and CO
2
hydrogenation arethermo-catalytic
* Bamidele Victor Ayodele
1
Institute ofEnergy Policy andResearch, Universiti Tenaga
Nasional, Jalan IKRAM-UNITEN, 43000Kajang, Selangor,
Malaysia
2
Center ofExcellence onPetrochemical andMaterials
Technology, Department ofChemical Engineering,
Faculty ofEngineering, Kasetsart University Bangkok,
Bangkok10900, Thailand
3
Chemical andMaterials Engineering Department,
Faculty ofEngineering Rabigh, King Abdulaziz
University, Rabigh Branch, P.O. Box344, Rabigh21911,
KingdomofSaudiArabia
4
Mechanical Engineering Department, Faculty ofEngineering
Rabigh, King Abdulaziz University, Rabigh Branch,
P.O. Box344, Rabigh21911, KingdomofSaudiArabia
5
Department ofChemical Engineering, Federal University
ofPetroleum Resources, Effurun, DeltaState, Nigeria
Topics in Catalysis
1 3
conversion processes that focus on the production of syn-
gas and methanol [7, 14]. Whereas CO
2
oxidative coupling
of methane focuses on the production of C
2
hydrocarbons
which are important chemical intermediates for several pro-
cess industries [15]. As shown in Eqs. (1) and (2), CO
2
is
employed as an oxidant in the methane coupling reaction
to produce C
2
H
6
and C
2
H
4
, respectively. Although the pro-
duction of C
2
hydrocarbons is the main focus of the CO
2
oxidative coupling of methane, side reactions could also lead
to the production of syngas as shown in Eqs. (3) and (4).
Hence, appropriate catalysts are needed to drive the selectiv-
ity of the C
2
hydrocarbon formation.
C
2
hydrocarbons are widely used as chemical build-
ing blocks most especially in the production of rubber
and plastics [16]. Moreover, ethylene which is one of the
C
2
-hydrocarbons can be used as anesthetic, oxy-fuel gas
used for welding and fabrication as well as a refrigerant
[17]. As shown in Fig.1, several catalysts have been inves-
tigated in the CO
2
oxidative coupling of methane. The find-
ings revealed that the catalysts displayed a high selectivity
to C
2
hydrocarbon with a very low yield. To overcome the
challenges of low C
2
hydrocarbon yield in CO
2
coupling
of methane reaction, Istadi, and Amin [18] employed a
response surface technique to optimize the process condi-
tions. At optimum condition, the selectivity of 76.56% and
a yield of 3.74% were obtained from the thermo-catalytic
(1)
2CH
4
+ CO
2
C
2
H
6
+ CO + H
2
O
(2)
2CH
4
+ 2CO
2
C
2
H
4
+ 2CO + H
2
O
(3)
2CH
4
+ 2CO
2
2CO + 2H
2
(4)
CH
4
+ 3CO
2
4CO + 2H
2
O
CO
2
oxidative coupling of methane. Besides using response
surface techniques, a data-driven predictive model can be
employed to investigate how the various input parameters
influence the C
2
selectivity and yield [19]. The parameter
estimates from the model can be used as a basis in deciding
on how to design a CO
2
oxidative coupling reactor that could
maximize the yield.
An artificial neural network (ANN) is a robust technique
that can be employed to develop a data-driven predictive
model [2023]. ANN has been employed in modeling vari-
ous chemical processes resulting in an excellent predic-
tion of the process output [24]. Ehsani etal. [25] employed
Levenberg–Marquardt trained multilayer perceptron neural
network for modeling the prediction of ethylene production
over Mn/Na
2
WO
4
/SiO
2
catalyst based on the non-linear rela-
tionship between the input parameters (reaction temperature,
CH
4
/O
2
ratio, the concentration of N
2
and gas hourly space
velocity) and the output parameters (selectivity and yield of
ethylene). The optimized 4-9-1 and 4-6-1 ANN topology
were robust in predicting the selectivity and yield of the
ethylene produced from the oxidative coupling reaction. In
a similar study, Ehsani etal. [26] employed the use of Lev-
enberg–Marquardt trained multilayer perceptron neural net-
work and genetic algorithm to model the oxidative coupling
of methane for C
2
hydrocarbon production. A minimum
prediction error obtained from the model is an indication
of appropriately to understand the non-linear relationship
between the input parameters and the output parameters.
Also, Huang etal. [27] employed ANN and hybrid genetic
algorithms to model catalyst design for the oxidative cou-
pling of methane. The models efficiently aided the design
of a multi-component catalyst for methane oxidative cou-
pling resulting in a higher yield of C
2
hydrocarbon. Istadi
and Amin [28], employed a hybrid ANN and genetic algo-
rithm for modeling and optimization of non-catalytic plasma
Fig. 1 Selectivity and yield of
C
2
hydrocarbon obtained from
CO
2
coupling of methane over
various catalysts
0
10
20
30
40
50
60
70
80
90
100
C
2
hydrocarbon
Catalysts
selecvity (%) yield (%)
Topics in Catalysis
1 3
reactor used for the production of C
2
hydrocarbon. The find-
ing shows that the hybrid ANN and genetic algorithm was
robust in modeling the effect of CH
4
/CO
2
feed ratio, total
feed flow rate, and discharge voltage on the C
2
hydrocarbon
production. In all these studies, it can be seen that multilayer
perceptron neural network was employed for modeling the
oxidative coupling of methane using oxygen as oxidant. The
use of Radial Basis Function ANN for modeling CO
2
oxi-
dative coupling of methane to C
2
hydrocarbon is lacking in
literature. Hence, the main focus of this study is to employ
Radial Basis Function ANN for modeling C
2
hydrocarbon
production by CO
2
oxidative coupling of methane using the
non-linear relationship between the input parameters (reac-
tion temperature, amount of CaO in the catalyst, amount
of MnO in the catalyst and CO
2
/CH
4
ratio) and the output
parameters (selectivity and yield of the C
2
hydrocarbon).
Compared to the multilayer perceptron neural network, the
radial basis function offers the advantage of better generali-
zation, not a complicated design, strong tolerance to input
noise, and ability to adapt to online learning.
2 Thermo‑catalytic Carbon Dioxide
Oxidative Coupling ofMethane
The Conceptual representation of the Radial Basis Func-
tion ANN modeling of the thermo-catalytic CO
2
oxidative
coupling of methane to C
2
-hydrocarbon is depicted in Fig.2.
The detailed description of the experimental design and
runs have been reported in Istadi and Amin [18]. The CO
2
oxidative coupling of methane was performed over CaO-
MnO/CeO
2
catalyst [18]. The experiment was designed
using four input parameters namely reaction temperature
(700–1000°C), amount of CaO in the catalyst (5–25mol%),
amount of MnO (1–9mol%) in the catalyst, and CO
2
/CH
4
ratio (1–3) and two output parameters namely selectivity and
yield of the C
2
hydrocarbon. The experimental design con-
sists of 26 treatment combinations of the four input param-
eters. For each of the runs, the data of the input parameters
were set to obtain the corresponding output parameters.
These data were subsequently employed for the Radial Basis
Function ANN modeling.
2.1 Radial Basis Function ANN Modeling
andSensitivity Analysis
Radial Basis Function is an ANN technique that works based
on the principle of function approximation [29]. As shown in
Fig.3, the Radial Basis Function ANN consists of three lay-
ers namely the input layer, hidden layer, and the output layer
[30]. In the input layer, each variable is linked with an inde-
pendent neuron. The input layers facilitate the propagation
of the input vector into the hidden layer for detail configu-
ration [22]. Each of the input vector (x) is associated with
corresponding weight (w). At the hidden layer, the intercon-
nections of the Radial Basis Function units as well as biases.
Each of the Radial Basis Function units is associated with a
Gaussian Kernels. The most relevant input neuron to be used
for the hidden layer configuration was determined using the
Softmax activation function which have offers a range value
between 0 and 1. At the hidden layer, a non-linear transfor-
mation mathematically represented in Eq. (5) is perform by
comparing the Euclidean distance of the input vector (x) and
the Gaussian function center (cj).
GC-TFC-FID
CO
2
CH
4
N
2
Catalyst
CO
2
/CH
4
ratio
Temperature
Amount of catalyst
Model Predicted
C
2
hydrocarbon
Fig. 2 Conceptual representation of the neural network model prediction of thermo-catalytic carbon dioxide oxidative coupling of methane to
C
2
-hydrocarbon
Topics in Catalysis
1 3
where h
j
, x, c
j
and represents the j-th neuron in the hid-
den layer, the input vector, Gaussian function center, and
the width.
The output layer consists of the output vector y
k
which
is computed based on the summation of the input vectors
weights (w
kj.
) and the biases (b
k
) as shown in Eq. (6). The
activation function was used to determine the appropriate
hidden neurons that can be used to compute the output
function.
In order to obtain optimum performance of the Radial
Basis Function ANN, the network is often trained to adjust
itself by performing parameter estimates of hidden neu-
rons based on the weights and biases that could result in
the best output. In the present study, to prevent over-fitting,
the 26 data set was partitioned into two in the proportion
of 70% for training and 30% for testing the models. The
performance of the Radial Basis Function ANN model
was measured using the sum of square errors (SSE) which
measures the actual deviation of the predicted values from
observed values as well as the coefficient of determina-
tion (R
2
) which measures the proportion of variance of
the model outputs that can be explained by the input
parameters. The Neural Network toolbox by International
(5)
(6)
y
k
(x)=
nh
j=
1
w
kj.
h
j
(x)+b
k
Business Machine (IBM) corporation was used for config-
uring the Radial Basis Function ANN model.
The influence of the input parameters on the Radial Basis
Function ANN model output in terms of the relative impor-
tance was investigated using Garson Algorithm [31]. The
sensitivity analysis of the model is useful in helping to make
decisions on the most significant input parameters that have
the greatest influence on the output. The Garson Algorithm
employed the absolute value of partitioned input-hidden and
hidden-output weights of the trained Radial Basis Function
ANN model to select the most significant input parameters.
3 Results andDiscussion
3.1 Interaction Eects oftheProcess Parameters
The effect of parameters such as gas hourly space velocity,
CH
4
/O
2
ratio, reaction temperature, N
2
flow rate, and the
amount of catalyst on the selectivity and yield of C
2
hydro-
carbon obtained from the oxidative coupling of methane has
been investigated. Based on the study reported by Istadi and
Amin [18], reaction temperature, CO
2
/CH
4
ratio, amount
of CaO in the catalyst, amount of MnO in the catalyst, and
CO
2
/CH
4
ratio were reported to influence the selectivity and
yields of C
2
hydrocarbon at optimum conditions. The three-
dimensional plots showing the interaction effects of the reac-
tion temperature, CO
2
/CH
4
ratio, amount of CaO in the cata-
lyst, amount of MnO in the catalyst, and CO
2
/CH
4
ratio on
the selectivity and yields of C
2
hydrocarbon are depicted in
Fig. 3 The Radial Basis Func-
tion ANN model architecture
1
2
3
Bias
Topics in Catalysis
1 3
Fig. 4 Interaction effect of (a) temperature and CO
2
/CH
4
ratio (b)
Amount of CaO and CO
2
/CH
4
ratio (c) Amount of CaO and tempera-
ture on C
2
selectivity; Interaction effect of (d) Temperature and CO
2
/
CH
4
ratio (e) CO
2
/CH
4
ratio and Amount of MnO (f) CO
2
/CH
4
ratio
and temperature on C
2
yield [18]
Topics in Catalysis
1 3
Fig.4. It can be seen that there is an existence of a non-linear
relationship between the input parameters and the output
parameters which makes the data most suitable for mod-
eling the process using the Radial Basis Function ANN [32].
As shown in Fig.4a andf, an increase in the reaction tem-
perature of the CO
2
oxidative coupling of methane reaction
resulted in a corresponding increase in the selectivity and
yields of the C
2
hydrocarbon until it peaked at 850°C and
gradually decrease at temperature >850°C. For a thermo-
catalytic reaction, it is expected that the yield and selectivity
of the products increase with an increase in temperature.
However, the deactivation of the catalysts could set in dur-
ing the reaction thereby reducing the catalytic activity [33].
Also, the increase in the amount of catalysts and the CO
2
/
CH
4
ratio depicted in Fig.4b–e were also observed to have
varying effects on the C
2
hydrocarbon selectivity and yields.
The yield and selectivity of the C
2
hydrocarbon increase
with an increase in the CO
2
/CH
4
ratio until 2 and thereaf-
ter decreased. As the concentration of CO
2
increased in the
reactant mixture, a parallel CO
2
disproportion reaction might
set in thereby converting some of the CO
2
to carbon which
could invariably deactivate the catalyst [34]. Increasing the
components of each of the catalysts was also observed to
increase the yield and selectivity of the C
2
hydrocarbon as
shown in Fig.4band c. However, the MnO dopant were
found to have a negative influence at an amount >15%.
3.2 The Radial Basis Function ANN Model
Performance
3.2.1 Determination ofBest Model Architecture
Prior to the modeling of the thermo-catalytic CO
2
oxidative
coupling of methane, various Radial Basis Function ANN
configurations were tested to determine the model with the
best performance. The Radial Basis Function ANN was con-
figured by varying the hidden neurons from 1 to 20. Each of
the configurations was trained and tested to determine their
performance in terms of the SSE and R
2
. The SSE obtained
during the training and testing of each of the Radial Basis
Function ANN topology and the overall R
2
are depicted in
Fig.5. It can be seen that each of the Radial Basis Func-
tion ANN topologies has varying performance during train-
ing and testing. Since the network is adjusted based on the
weights associated with input and the hidden neurons, the
performance of each of the ANN topology will vary based
on the changes in the hidden neuron. The best performance
of the Radial Basis Function ANN models was obtained for
the 4-20-2 topology as indicated by the red arrow in Fig.5.
The 4-20-2 configurations imply the ANN model consists
of 4 input units, 20 hidden neurons, and 2 output units. The
training and testing of the 4-20-2 Radial Basis Function
ANN resulted in SSE of 3.9 × 10
−24
and 0.224, respectively,
and R
2
of 0.990. The use of 4-9-1 and 4-6-1 architecture for
Levenberg–Marquardt trained multilayer perceptron ANN
has been reported by Ehsani etal. [26]. The variation in the
network architectures could be attributed to the peculiarity
and the inbuilt algorithms of the Radial Basis Function ANN
and the Levenberg–Marquardt trained multilayer perceptron
ANN.
Fig. 5 SSE and R
2
values
obtained for training and testing
the various topology of the
Radial Basis Function ANN
Topics in Catalysis
1 3
3.2.2 The Model Performance
The performance of the 4-20-2 configured Radial Basis
Function ANN in terms of predicting the C
2
selectivity and
yields are depicted in Figs.6 and 7, respectively. As shown
in Fig.6a, at every point of the experimental runs, the pre-
dicted C
2
selectivity is in proximity to the observed values
with all the residuals value less than 0.5%. This can further
be substantiated by the linear regression plots depicted in
Fig.6b. With R
2
of 0.989 and residual <0.05%, it is certain
that the Radial Basis Function ANN model is robust in gen-
eralizing 98.9% of the data which is comparable with that
reported in the literature. Huang etal. [27] reported a predic-
tive error of 0.162% for modeling a multi-component catalyst
used in methane oxidative coupling using Levenberg–Mar-
quardt-trained multilayer perceptron ANN. The predicted
C
2
selectivity by the Levenberg–Marquardt-trained multi-
layer perceptron was in close agreement with the observed
values. Also, in a similar study reported by Abdolahi etal.
[35] whereby a feed-forward Levenberg–Marquardt-trained
multilayer perceptron ANN was employed to model the pre-
diction of C
2
hydrocarbon selectivity, C
2
H
4
selectivity, and
CH
4
conversion, the predicted values were consistent with
the observed values with minimal errors. Several authors
have reported the superiority of the Radial Basis Function
ANN for predictive modeling compared to the multilayer
perceptron [3638].
The Radial Basis Function ANN performance in terms
of the prediction of the C
2
hydrocarbon yield is depicted
in Fig.7. The Radial Basis Function ANN architecture
displayed a well-generalized prediction of C
2
hydrocarbon
yield as illustrated in Fig.7a. For all experimental runs,
the predicted C
2
hydrocarbon yields are consistent with the
observed values. This can further be ascertained from the
regression plot in Fig.7b. With an R
2
of 0.998 and residu-
als less than 0.05, 99.8% of the data can be explained and
Fig. 6 (a) Dispersion plot show-
ing the predicted, observed and
the residuals of C
2
hydrocarbon
selectivity (b) Regression plot
showing the predicted, observed
and the residuals of C
2
hydro-
carbon selectivity
-20
0
20
40
60
80
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
C
2
hydrocarbon selectivity (%)
Experimental runs
Observed Predicted Residuals
(a)
(b)
Topics in Catalysis
1 3
generalized by the Radial Basis Function ANN model. Fath
etal. [38] employed both Radial Basis Function ANN and
multilayer perceptron for modeling the prediction of solu-
tion gas-oil ratio of crude oil systems. The findings revealed
that the Radial Basis Function ANN model displayed a
higher accuracy in predicting the solution gas-oil ratio of
the crude oil systems. The robustness of the Radial Basis
Function ANN compared to multilayer perceptron ANN can
be attributed to its unique feature as a universal approxima-
tor whereby it has a high tendency to generalize any non-
linear process. The stability of the network system is often
enhanced due to its strong tolerance to input noise. Moreo-
ver, the simple nature of the three-layer architecture makes
the Radial Basis Function ANN easier to configure and train.
The sensitivity analysis as a function of the normalized
level of importance of the input parameters (temperature,
Fig. 7 (a) Dispersion plot show-
ing the predicted, observed and
the residuals of C
2
hydrocarbon
yield (b) Regression plot show-
ing the predicted, observed and
the residuals of C
2
hydrocarbon
yield
02040608
00
CO2/CH4 ratio
Temperature (oC)
Amount of MnO in catalyst (%)
Amount of CaO in catalyst (%)
Temperature (
o
C)
CO
2
/CH
4
ratio
Fig. 8 Sensitivity analysis showing the normalized importance of the
input parameters on the Radial Basis Function ANN model output
Topics in Catalysis
1 3
amount of CaO in the catalyst, amount of MnO in the cata-
lyst, and the CH
4
/CO
2
ratio) on the Radial Basis Function
ANN model outputs (C
2
hydrocarbon selectivity and yields)
is depicted in Fig.8. It can be seen that the four input param-
eters significantly influenced the predicted C
2
hydrocarbon
selectivity and yields. The normalized level of importance
revealed that the reaction temperature for the oxidative cou-
pling of methane has the most significant influence on the
C
2
hydrocarbon selectivity and yields. Based on the Arrhe-
nius concept, the gas-phase reaction is highly temperature
dependent [39]. The selectivity and yields of products
obtained from a gaseous phase reaction tend to increase with
an increase in temperature. However, increasing the tem-
perature beyond the thermodynamic acceptable limit could
have a negative effect on the catalytic performance. Hence,
a temperature control system could be incorporated into the
reactor design for the CO
2
oxidative coupling of methane. In
addition to the reaction temperature, the reactant (CO
2
/CH
4
)
ratio also has a significant influence on the C
2
hydrocarbon
selectivity and yields. As stated earlier, the C
2
hydrocarbon
selectivity and yields increased with an increase in CO
2
/CH
4
ratio and peaked at a CO
2
/CH
4
ratio of 2. A further increase
in the CO
2
/CH
4
ratio resulted in a corresponding decline in
the C
2
hydrocarbon selectivity and yields. This phenomenon
could be explained in terms of the occurrence of parallel
reactions such as CO
2
disproportion reaction and methane
cracking resulting in the deactivating of the catalysts by car-
bon deposition.
4 Conclusion
The modeling of thermo-catalytic CO
2
oxidative coupling
of methane for prediction of C
2
-hydrocarbon using Radial
Basis Function ANN has been investigated. Using the non-
linear relationship between the input parameters and the
process outputs, the performance of several configurations
of Radial Basis Function ANN models were examined. The
analysis of each of the Radial Basis Function ANN shows
that the model performance in terms of the prediction of the
C
2
hydrocarbon selectivity and yield are strongly dependent
on the number of hidden neurons. The Radial Basis Func-
tion ANN configuration with 20 hidden artificial neurons
displayed the best performance, evidenced from a very low
predictive error and high R
2
values of 0.998. The predicted
C
2
hydrocarbon selectivity and yields by the 4-20-2 Radial
Basis Function ANN architecture was in close agreement
with the observed values. This was further validated by a
strong correlation between the predicted and the observed
values as indicated by the R
2
of 0.989 and 0.998 for the C
2
selectivity and yields, respectively. Based on the sensitiv-
ity analysis, all the input parameters significantly influence
the prediction of C
2
selectivity and yields by the optimized
Radial Basis Function ANN. However, the reaction tem-
perature with the highest level of importance displayed the
most significant influence on the model output. The Radial
Basis Function ANN algorithm could be employed in incor-
porating a control system over the process variables when
considering the design of an optimized process for CO
2
oxi-
dative coupling of methane. Moreover, the outcome of the
sensitivity analysis can be incorporated as basis to designing
a robust catalyst for CO
2
oxidative coupling of methane.
References
1. Bakonyi P, Peter J, Koter S etal (2020) Possibilities for the bio-
logically-assisted utilization of CO
2
-rich gaseous waste streams
generated during membrane technological separation of bio-
hydrogen. J CO2 Util 36:231–243. https ://doi.org/10.1016/j.
jcou.2019.11.008
2. Yaashikaa PR, Senthil Kumar P, Varjani SJ, Saravanan A (2019)
A review on photochemical, biochemical and electrochemical
transformation of CO
2
into value-added products. J CO2 Util
33:131–147. https ://doi.org/10.1016/j.jcou.2019.05.017
3. Parvez AM, Afzal MT, Victor Hebb TG, Schmid M (2020) Uti-
lization of CO
2
in thermochemical conversion of biomass for
enhanced product properties: a review. J CO2 Util 40:101217.
https ://doi.org/10.1016/j.jcou.2020.10121 7
4. Ayodele BV, Khan MR, Cheng CK (2017) Greenhouse gases
mitigation by CO
2
reforming of methane to hydrogen-rich syn-
gas using praseodymium oxide supported cobalt catalyst. Clean
Techn Environ Policy 19:795–807. https ://doi.org/10.1007/s1009
8-016-1267-z
5. Skytt T, Nielsen SN, Jonsson B-G (2020) Global warming poten-
tial and absolute global temperature change potential from carbon
dioxide and methane fluxes as indicators of regional sustainability
– a case study of Jämtland, Sweden. Ecol Indic 110:105831. https
://doi.org/10.1016/j.ecoli nd.2019.10583 1
6. Khan AA, Tahir M (2019) Recent advancements in engineering
approach towards design of photo-reactors for selective photocata-
lytic CO
2
reduction to renewable fuels. J CO2 Util 29:205–239.
https ://doi.org/10.1016/j.jcou.2018.12.008
7. Din IU, Shaharun MS, Alotaibi MA etal (2019) Recent develop-
ments on heterogeneous catalytic CO
2
reduction to methanol. J
CO2 Util 34:20–33. https ://doi.org/10.1016/j.jcou.2019.05.036
8. Zhang Z, Zhang L, Hülsey MJ, Yan N (2019) Zirconia phase effect
in Pd/ZrO
2
catalyzed CO
2
hydrogenation into formate. Mol Catal
475:110461. https ://doi.org/10.1016/j.mcat.2019.11046 1
9. Ye R-P, Li Q, Gong W, Wang T, Razink JJ, Lin L, Qin YY, Zhou
Z, Adidharma H, Tang J, Russell AG (2020) High-performance
of nanostructured Ni/CeO
2
catalyst on CO
2
methanation. Appl
Catal B Environ 268:118474. https ://doi.org/10.1016/j.apcat
b.2019.11847 4
10. Wang Z, Cao XM, Zhu J, Hu P (2014) Activity and coke forma-
tion of nickel and nickel carbide in dry reforming: a deactivation
scheme from density functional theory. J Catal 311:469–480. https
://doi.org/10.1016/j.jcat.2013.12.015
11. Shin SA, Noh YS, Hong GH, Park JI, Song HT, Lee KY, Moon DJ
(2018) Dry reforming of methane over Ni/ZrO
2
-Al
2
O
3
catalysts:
effect of preparation methods. J Taiwan Inst Chem Eng 90:25–32.
https ://doi.org/10.1016/j.jtice .2017.11.032
12. Ayodele BV, Khan MR, Cheng CK (2015) Syngas production
from CO
2
reforming of methane over ceria supported cobalt
catalyst: effects of reactants partial pressure. J Nat Gas Sci Eng
27:1016–1023. https ://doi.org/10.1016/j.jngse .2015.09.049
Topics in Catalysis
1 3
13. Ayodele BV, Khan MR, Lam SS, Cheng CK (2016) Production
of CO-rich hydrogen from methane dry reforming over lanthania-
supported cobalt catalyst: kinetic and mechanistic studies. Int J
Hydrog Energy 41:4603–4615. https ://doi.org/10.1016/j.ijhyd
ene.2016.01.091
14. Mesrar F, Kacimi M, Liotta LF, Puleo F, Ziyad M (2018) Syn-
gas production from dry reforming of methane over ni/per-
lite catalysts: effect of zirconia and ceria impregnation. Int J
Hydrog Energy 43:17142–17155. https ://doi.org/10.1016/j.ijhyd
ene.2018.07.104
15. Li N, Li Y, Jiang R, Zhou J, Liu M (2019) Photocatalytic coupling
of methane and CO
2
into C
2
-hydrocarbons over Zn doped g-C
3
N
4
catalysts. Appl Surf Sci 498:143861. https ://doi.org/10.1016/j.
apsus c.2019.14386 1
16. Ma Y, Harding JD, Tu X (2019) Catalyst-free low tempera-
ture conversion of n-dodecane for co-generation of CO
x
free
hydrogen and C
2
hydrocarbons using a gliding arc plasma. Int J
Hydrog Energy 44:26158–26168. https ://doi.org/10.1016/j.ijhyd
ene.2019.08.067
17. Sundaram KM, Shreehan MM, Olszewski EF (2001) Ethylene.
In: Kirk-Othmer encyclopedia of chemical technology. American
Cancer Society
18. Istadi I, Amin NAS (2006) Optimization of process parameters
and catalyst compositions in carbon dioxide oxidative coupling
of methane over CaO-MnO/CeO
2
catalyst using response sur-
face methodology. Fuel Process Technol 87:449–459. https ://doi.
org/10.1016/j.fupro c.2005.11.004
19. Istadi I, Amin NAS (2006) Hybrid artificial neural network—
genetic algorithm technique for modeling and optimization of
plasma reactor hybrid artificial neural network—genetic algorithm
technique for modeling and optimization of plasma reactor. Ind
Eng Chem Res 45:6655–6664. https ://doi.or g/10.1021/ie060 562c
20. Koschwitz D, Frisch J, van Treeck C (2018) Data-driven heating
and cooling load predictions for non-residential buildings based
on support vector machine regression and NARX recurrent neural
network: a comparative study on district scale. Energy 165:134–
142. https ://doi.org/10.1016/j.energ y.2018.09.068
21. Alsaffar MA, Ayodele BV, Mustapa SI (2019) Scavenging carbon
deposition on alumina supported cobalt catalyst during renewable
hydrogen-rich syngas production by methane dry reforming using
artificial intelligence modeling technique. J Clean Prod 119168.
https ://doi.org/10.1016/j.jclep ro.2019.11916 8
22. Ayodele BV, Alsaffar MA, Mustapa SI, Vo DN (2020) Back-prop-
agation neural networks modeling of photocatalytic degradation
of organic pollutants using TiO
2
-based photocatalysts. J Chem
Technol Biotechnol 95(10):2739–2749. https ://doi.org/10.1002/
jctb.6407
23. Al-Haiqi O, Nour AH, Ayodele BV, Bargaa R (2020) Bayes-
ian regularization-trained multi-layer perceptron neural net-
work predictive modelling of phenol degradation using ZnO/
Fe
2
O
3
photocatalyst. J Phys Conf Ser 1529:052058. https ://doi.
org/10.1088/1742-6596/1529/5/05205 8
24. Li H, Zhang Z, Liu Z (2017) Application of artificial neural
networks for catalysis: a review. Catalysts 7:306. https ://doi.
org/10.3390/catal 71003 06
25. Ehsani MR, Bateni H, Razi Parchikolaei G (2013) Modeling
of oxidative coupling of methane over Mn/Na
2
Wo
4
/SiO
2
cata-
lyst using artificial neural network. Iran J Chem Chem Eng
32:107–114
26. Ehsani MR, Bateni H, Parchikolaei GR (2012) Modeling the oxi-
dative coupling of methane using artificial neural network and
optimizing of its operational conditions using genetic algorithm.
Korean J Chem Eng 29:855–861. https ://doi.org/10.1007/s1181
4-011-0250-y
27. Huang K, Zhan XL, Chen FQ, Lü DW (2003) Catalyst design for
methane oxidative coupling by using artificial neural network and
hybrid genetic algorithm. Chem Eng Sci 58:81–87. https ://doi.
org/10.1016/S0009 -2509(02)00432 -3
28. Istadi I, Amin NAS (2006) Hybrid artificial neural network-
genetic algorithm technique for modeling and optimization of
plasma reactor. Ind Eng Chem Res 45:6655–6664. https ://doi.
org/10.1021/ie060 562c
29. Hossain MA, Ayodele BV, Cheng CK, Khan MR (2016) Artifi-
cial neural network modeling of hydrogen-rich syngas production
from methane dry reforming over novel Ni/CaFe
2
O
4
catalysts. Int
J Hydrog Energy 41:11119–11130. https ://doi.or g/10.1016/j.ijhyd
ene.2016.04.034
30. Hossain MS, Ong ZC, Ismail Z, Khoo SY (2017) A comparative
study of vibrational response based impact force localization and
quantification using radial basis function network and multilayer
perceptron. Expert Syst Appl 85:87–98. https ://doi.org/10.1016/j.
eswa.2017.05.027
31. Garson GD (1991) Comparison of neural network analysis of
social science data. Soc Sci Comput Rev 9:399–434
32. Sunphorka S, Chalermsinsuwan B, Piumsomboon P (2017) Appli-
cation of artificial neural network for kinetic parameters predic-
tion of biomass oxidation from biomass properties. J Energy Inst
90:51–61. https ://doi.org/10.1016/j.joei.2015.10.007
33. Chen D, Lødeng R, Anundskås A etal (2001) Deactivation
during carbon dioxide reforming of methane over Ni catalyst:
microkinetic analysis. Chem Eng Sci 56:1371–1379. https ://doi.
org/10.1016/S0009 -2509(00)00360 -2
34. Ginsburg JM, Piña J, El Solh T, De Lasa HI (2005) Coke forma-
tion over a nickel catalyst under methane dry reforming condi-
tions: thermodynamic and kinetic models. Ind Eng Chem Res
44:4846–4854. https ://doi.org/10.1021/ie049 6333
35. Abdolahi F, Mortazavi Y, Khodadadi A etal (2005) Modeling of
methane oxidative coupling under periodic operation by neural
network. Chem Eng Technol 28:581–586. https ://doi.org/10.1002/
ceat.20040 7041
36. Han H-G, Qiao J-F, Chen Q-L (2012) Model predictive control
of dissolved oxygen concentration based on a self-organizing
RBF neural network. Control Eng Pract 20:465–476. https ://doi.
org/10.1016/j.conen gprac .2012.01.001
37. Alsaffar MA, Ayodele BV, Mustapa SI (2020) Scavenging carbon
deposition on alumina supported cobalt catalyst during renew-
able hydrogen-rich syngas production by methane dry reform-
ing using artificial intelligence modeling technique. J Clean Prod
247:119168. https ://doi.org/10.1016/j.jclep ro.2019.11916 8
38. Hashemi Fath A, Madanifar F, Abbasi M (2020) Implementa-
tion of multilayer perceptron (MLP) and radial basis function
(RBF) neural networks to predict solution gas-oil ratio of crude
oil systems. Petroleum 6:80–91. https ://doi.org/10.1016/j.petlm
.2018.12.002
39. Crowley JN, CampuzanoJost P, Moortgat GK (1996) Temperature
dependent rate constants for the gas-phase reaction between OH
and CH
3
OCl. J Phys Chem 100:3601–3606
Publisher’s Note Springer Nature remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.