Professional Documents
Culture Documents
5 1
Escuela Superior Politécnica de Chimborazo
6 Abstract: The objective of this research was to design a neural network (ANN) to predict the
7 methanol flux at the outlet of a carbon dioxide dehydrogenation plant. For the development of the
8 ANN, a database was generated, in the open-source simulation software "DWSIM", from the
9 validation of a process described in the literature. The sample consists of 133 data pairs with four
10 inputs: reactor pressure and temperature, mass flow of carbon dioxide and hydrogen, and one
11 output: flow of methanol. The ANN was designed using 12 neurons in the hidden layer and it was
12 trained with the Levenberg-Marquardt algorithm. In the training, validation and testing phase, a
13 global mean square (RMSE) value of 0.0085 and a global regression coefficient R of 0.9442 were
14 obtained. The network was validated through an analysis of variance (ANOVA), where the P-
15 value for all cases was greater than 0.05, which indicates that there are no significant differences
16 between the observations and those predicted by the ANN. Therefore, the designed ANN can be
17 used to predict the methanol flow at the exit of a dehydrogenation plant and later for the
18 optimization of the system.
20 1. Introduccion
25 In order to mitigate CO2 emissions, three methods have been proposed: capturing
26 and storing CO2 underground, biological treatment with algae and chemical [3]
27 conversion. The catalytic hydrogenation of CO 2 is the most promising process of
28 chemical conversion. This process allows to obtain products such as alcohols, formic
29 acid, formaldehyde; it is also possible to obtain light hydrocarbons such as methane
30 depending on the type of the catalyst used[4,5]. Methanol is the chemical that shows
31 higher performance both in the energy field (25.5 Kcal / H 2) and at the hydrogen use
32 level (H2) (67%) [6]. Moreover, methanol can be used both as a fuel for energy
33 obtainment and as an intermediate means for the production of other more valuable
34 chemicals such as dimethyl ether (used in aerosols) or formaldehyde (used in the
35 manufacture of plastics) [7].
2
3 2 of 11
4
46 hydrogenation process of the synthesis gas. Despite presenting favorable results at a lab
47 scale, they have not been industrially tested. Furthermore, only studies related to
48 characterization, process thermodynamics, and kinetic descriptions have been carried
49 out [4,5,8–12].
50
51 1.1. State of art
52
53 This article discusses the use of an ANN tool to predict the catalytic process.
54 Therefore, state of the art includes both concepts. First, the use of catalyst followed by
55 the limited application of ANN in processes in which different catalysts are involved.
56
57 1.1.1. Catalysts and simulation of the hydrogenation process
58
59 Despite the deactivation suffered by copper-based catalysts, they continue to be the
60 object of study to improve their catalytic properties. Fang et al., [1,15] prepared copper-
61 based catalysts with hydrotalcite. The catalysts prepared from Cu / ZnO / ZrO 2 and Cu /
62 ZnO / Al2O3 contained 56.6% and 40% respectively of hydrotalcite in their structure. As a
63 result of the addition of hydrotalcite, the catalysts had a greater dispersion of copper
64 atoms on its surface . After performing several tests at different conditions, comparing
65 the performance of the catalyst with and without hydrotalcite, it was determined that
66 the later shows higher yields and selectivity than the base catalysts (selectivity of 84%
67 for the Cu / ZnO / ZrO2 catalyst and 73.4 % for Cu / ZnO / Al2O3).
68 Furthermore, it was observed that the conditions under which the absorbent
69 enhanced catalyst operates are lower than those under which the base catalyst operates.
70 These observations are due to the fact that hydrotalcite improves the capacity of the
71 catalyst to absorb CO2.
72 Sadeghinia et al., [16] prepared samples of commercial catalysts Cu / ZnO / Al 2O3,
73 to which different proportions of In2O3 were added. Then tests were conducted to
74 measure the catalytic activity using as feed a stream containing H 2, CO, and CO2 and
75 another containing only H2 and CO2. These studies showed that, in the first stream, the
76 use of In2O3 causes a drop in the production of methanol. This is because the improved
77 catalyst has more capacity to absorb CO2, favoring only the hydrogenation of CO2.
78 Due to the high costs associated with the construction of pilot plants, studies were
79 carried out to evaluate the performance of the hydrogenation process at an industrial
80 level; they have been conducted by employing commercial software such as ASPEN
81 PLUS [17].
82 Perez-Fortes et al., [18] ] developed the simulation of a CO 2 capture and use plant.
83 The plant was simulated in the CHEMCAD software [19] in order to obtain the flow
84 value of energy and matter. Based on the results, they developed an economic analysis,
85 obtaining that the plant can produce 440 KTn / year of methanol. This study concluded
86 that the operation of the plant is not economically viable. Kiss et al., [20] proposed the
87 use of a stripping column to process a stream of wet hydrogen. This stream flows
88 countercurrent together with the effluent leaving the reactor. The use of the stripping
89 column allows a more efficient separation of unconverted gases present in the product
90 stream from the methanol-water mixture. This scheme allows the reduction of the
91 amount of water present at the inlet of the reactor, causing a drop in the flow of
92 methanol produced. In turn, it allows the recirculation of a greater quantity of
93 unconverted CO and CO2. As a result, this scheme consumes 550 kW / Tn of methanol
94 produced and 0.48-1.16 Tn of steam for each ton of methanol produced. As the studies
95 [18] and [20] show in the proposed industrial schemes, the energy consumption
96 associated with methanol production is considerably high. To solve this problem, Szima
97 et al., [21] propose the use of a gas turbine and a simple Rankine cycle that operates at
5
6 3 of 11
7
98 low pressure to compensate for the energy requirements of the plant. The plant was
99 simulated in [19]. The supply to this plant was hydrogen that comes from a water
100 electrolysis process, while the CO 2 was captured from a coal plant. The results of the
101 simulation showed that it is not economically profitable for its construction. Through a
102 sensitivity analysis, they determined that the plant would be economically viable if the
103 sales price of methanol was doubled or the costs of electricity decreased by half.
104 Do and Kim [22], to solve the energy problem, conducted a simulation in [17]. In
105 this process, the energy required for hydrogen obtainment from the water
106 decomposition was obtained from solar collectors, while the CO 2 is reduced to carbon
107 monoxide using solar energy. The primary CO 2 reduction yielded methanol using the
108 traditional process. This technology proved to have a high energy efficiency of 15.5%.
109 Moreover, it managed to establish a competitive selling price of methanol.
110 Van Dal and Bouallou [13] carried out a simulation in [17], where the
111 hydrogenation plant is fed with CO 2 captured from a thermoelectric plant. The
112 hydrogen necessary for the production of methanol was obtained from the electrolysis of
113 water. The results determined that to produce one ton of methanol, 1.6 tons of CO 2 can
114 be removed.
115 On the other hand, the studies carried out to analyze the feasibility of
116 implementing an ethanol production process from CO2 and improving the capabilities of
117 the catalyst involved present great limitations. There are models that describe the
118 behavior of a system in a realistic way. However, they lead to the establishment of
119 robust differential equations that involve complex calculations and increase
120 computational time. This procedure is impractical for sensitivity and optimization
121 studies (2). Studies related to kinetic models and characterizations of new catalysts are
122 limited by DTF (density functional theory) calculations that make idealized assumptions
123 (13).
124
125 1.2.2. ANN as a prediction tool in catalysis processes
126
127 Artificial neural networks (ANN) are predictive tools able to learn directly from a
128 process and give short response times; this allows the modeling of systems in a more
129 complex and realistic manner [23,24].
130 ANNs are a computational model that have the ability to handle multiple complex
131 problems of the real world. The versatility of the ANN is due to the ability of
132 information processing, high parallelism, fault tolerance, nonlinearity, noise, and
133 tolerance generalization capabilities [25]. ANNs in their structure are made up of several
134 layers of neurons that connect to each other to share information and are distributed in
135 input, hidden and output layers; each neuron has an activation function that allows
136 information to be reconstructed or predicted [26]. Nowadays, ANNs are ones of the
137 most used tools in different areas of engineering and science to create complex and
138 nonlinear models and to describe the natural behavior of the system. They constitute a
139 good way to find issues of malfunctions inside a system. They have been used for years
140 in different areas of engineering, science, and business to deal with highly complex and
141 nonlinear data sets [27–29].
142 There are several relevant studies in the area of catalysis, and in most cases, the
143 main objectives are: setting the experimental conditions and/or properties of the catalyst
144 system as inputs and the catalytic activities as outputs of the model. In addition,
145 research suggests that automatic learning can be a good option to reduce the
146 computational cost of the catalysis study and promote the discovery of new catalysts.
147 However, ANN in the field of catalysis is not well studied since the acquisition of a
148 database for training models is costly and, at the same time, should take into account too
149 many input variables for creating the model [30].
8
9 4 of 11
10
150 Kito et al., [31] developed an ANN to estimate the acid strength of mixed oxides,
151 catalytic performance, and selectivities to various products in the oxidative
152 dehydrogenation of ethylbenzene in a series of promoted catalysts SnO 2. The results
153 from the study indicated that the ANN has a large capacity of interpolation,
154 extrapolation, prediction of the catalytic performance of multiple components.
155 Liu et al., [32] trained a neural network with the ability to predict the catalytic
156 activity as well as the selectivity of various catalysts used for CO 2 hydrogenation. The
157 objective was for the network to select the appropriate catalyst based on the products to
158 be obtained. The result of the work indicates that ANN can satisfactorily predict the
159 activity and selectivity of each catalyst. However, it is not possible to satisfactorily select
160 the right catalyst for each experiment because of the limited data used for training.
161 Zahedi et al. [2] designed an ANN in conjunction with a mathematical model to
162 study the behavior of an industrial packed bed reactor. The ANN is responsible for the
163 prediction of the reaction kinetics, while the mathematical model is used to determine
164 temperature and pressure data at the outlet of the reactor. Two types of neural networks
165 were trained, a perceptron-type network and a radial-based one. The radial base
166 network obtained a lower mean square error (RMSE) value, indicating it as the most
167 suitable for predicting the data. The results showed that there is a great agreement with
168 the real values of the reactor studied.
169 Sun et al., [33] designed an ANN with the ability to predict the selectivity of a
170 catalyst in a microchannel reactor. The neural network used was of the perceptron type
171 with two hidden layers and was trained with real data taken from a reactor. The ANN
172 determined the optimum operating conditions of the reactor.
188 2. Metodologia
189 The first part of this study deals with simulating the process of Fig.2, taking into
190 account the operating conditions of the process. The next step is to validate the results
191 and proceed with the design of an artificial neural network considering the inputs and
192 outputs determined for the prediction of the process and the restrictions imposed by the
193 simulation. Subsequently, training and validation of ANN using graphical and statistical
194 analysis to evaluate the predictive ability of the neural network are performed. Figure 2
195 illustrates the flowchart of the methodology.
196 This section describes the applied mathematical model, as well as the design model
197 and training ANN.
11
12 Establishment of
Begin DWSIM simulation 5 of 11
13 operating parameters
Validation and
Data training Statistical Analysis End
133 data pairs of RNA
198
200
201 2.1. DWSIM simulation
202 Different thermodynamic models were used, following the recommendations
203 established in [20,35,36] for this type of system. For currents higher pressure 10 bar, the
204 SRK model was used. For the remaining flows, the NRTL model was used. The model
205 was simulated using the flash method Nessted valid Loops for calculating the balance
206 between liquid and vapor phases. Table 4 shows the detailed reactor geometry
207 configuration and the characteristics of the catalyst used for the simulation of the packed
208 bed reactor.
209
14
15 6 of 11
16
221
222
224 The designed ANN is of the simple perceptron type and is trained with the
225 Levenberg Marquardt algorithm. The data distribution was carried out at 70-30. That is,
226 70% of the data were used for network training, and the remaining 30% were used for
227 the network validation and testing process. The training process seeks to adjust the
228 weights of the connections between neurons in such a way that the predictions made by
229 the network are as accurate as possible with respect to those selected as the target.
230 Validation is a process where the error thrown by the network is measured with respect
231 to the objective data and is used as a measure of the performance of the network. ANN
232 testing is equivalent to presenting data to the network that was not used for training;
233 this involves generating new pairs of data, presenting them to the ANN without
234 showing the desired outputs, and comparing how accurate the predictions are.
235 Once the simulation process of the carbon dioxide hydrogenation plant in DWSIM
236 was validated, 133 data pairs (4 inputs and one output) were generated to train the
237 network. Table 5 details the variation range of the input variables selected for the study.
238 To validate the ANN and demonstrate the reliability of the model for the prediction of
239 methanol flow, the general performance indicators were used: mean square error (MSE),
240 regression (R), and additionally a comparative statistical analysis (ANOVA). Similarly,
241 to determine the optimum ANN, a trial and error procedure was carried out until the
242 neural network with the lowest RMSE associated with the network was found and based
243 on the correlation coefficients obtained for the training, validation, and testing process of
244 the network.
17
18 7 of 11
19
251 The analysis and discussion of results include both the description and validation
252 and topology of the ANN and the performance of the model. In this section, both are
253 presented.
255 Before designing the ANN, it is essential to carry out the comparison of the
256 simulation developed in DWSIM with results from the literature. The study developed
257 in [13] was used for validation. The process DWSIM was adapted according to the
258 simulation developed in ASPEN PLUS. Table 6 details the comparison of the results; as
259 it can be seen, the percentage errors do not exceed 5%. The existence of the error
260 between the values obtained and those of the literature is justified by the
261 thermodynamic model used (SRK-MVHS) and by the volume of the reactor that is not
262 defined in [13].
Component Input flow Output flow [6] Output flow Error (%)
Entrada (kg/h) (Kg/h) DWSIM(kg/h)
Methanol 0 59 300 58 297.9952 1.69%
Water 0 33 700 32 559.832 3.38%
Carbon dioxide 88 000 5 820 5 710.5232 1.88%
Carbon
0 510 524.535 2.85%
monoxide
Hydrogen 12 100 870 833.4674 4.19%
264 * Less or greater than the established ranges, the simulation does not run
265
266
268 This section defines how the ANN was structured through the analysis of the
269 correlation coefficient (R) and the mean square error (RMSE).
270
271 3.2.1. Selection of number of neurons in the hidden layer
272 A trial and error test was carried out, varying the number of neurons in the hidden
273 layer, and R and RMSE were evaluated. Figures 4-6 show the evolution of R-values for
274 the training, validation, test, and global phase using 500, 100, and 12 neurons,
275 respectively. The global R-value using 500 neurons is: 0.3189 and the RSME = 1.26 (Table
276 7). These results suggest that the network is not suitable for methanol flux prediction.
277 On the other hand, in 100 neurons, the value of R = 0.49 and the RMSE = 0.18. In this
278 case, the value of R in the training phase is 0.99, indicating an adequate correlation
279 between the observed data and those predicted by the network. However, the value of R
280 in the validation and testing phase drops to 0.25 and 0.076; these values indicate that the
281 ANN is overtrained, losing the ability to generalize its predictions. When 12 neurons
282 were used (Figure 7), it is observed that the value of R both in the training, validation,
283 test, and global phases are: 0.948, 0.940, 0.942, and 0.944, respectively, and according to
284 Table 7, the value of the RSME 0 0.0085; these results indicate that ANN has a good
285 predictive ability.
286
20
21 8 of 11
22
287
288 Figure 4. Regression and correlation coefficient using 500 neurons in the hidden layer for
289 stage a) the training phase, b) validation, c) testing, and d) overall.
290
292 The performance of the ANN designed during training and validation is necessary
293 to evaluate its functionality with another set of data not known by the neural network
294 and to verify its predictive capacity. For this, a set of 10 at random data (P, T, Mass
295 flows) simulated in DWSIM has been generated to collect new observations based on the
296 new operating conditions. To verify the prediction capacity, the ANN designed to
297 predict the methanol flux at the exit of the hydrogenation plant was used. The
298 comparison between the observations and the predictions is shown in Figure 8. The
299 overlap between them allows to deduce that the designed ANN has a good capacity
300 and precision of prediction of the methanol flow at the exit of the carbon dioxide
301 hydrogenation plant. The graphic analysis was complemented with an ANOVA analysis
302 in order to guarantee that the designed ANN is reliable and suitable as a prediction tool.
303
23
24 9 of 11
25
306 In this research, the ANOVA and Kruskal-Wallis tool was used to statistically
307 validate the ANN. The results of the ANOVA are summarized in Table 8. According to
308 the statistical test employed, the P-value of F-ratio is greater than 0.05. ANOVA revealed
309 that the constructed ANN is statistically valid for predicting the methanol flow in the
310 dehydrogenation plant.
Sum of Middle
Source DF F-reason P-value
squares Square
Between
231849 1 231849 0,02 0,8968
groups
Intra groups 2.41261E8 18 1.34034E7
Total (Corr.) 2.41493E8 19
312
313 4. Conclusions
314 In this study, a neural network was designed to predict methanol flux in a carbon
315 dioxide dehydrogenation plant from the simulation process in DWSIM. The designed
316 neural network has a hidden layer with 12 neurons and was trained with a base of 133
317 data pairs with four input variables: pressure (P) and temperature (T) of the reactor,
318 mass flow of CO2 and H2 and is capable of predicting the methanol flux as an output
319 variable.
320 The ANN was trained with the Levenberg-Marquardt algorithm, has an RMSE of
321 0.0085 and a total regression coefficient of 0.9442. The ANN was validated through a
322 comparative statistical analysis (ANOVA) between the observations (DWSIM) and the
323 values predicted by the network. The statistical test indicates that the network
324 adequately predicts the flow of methanol with a significance level of 95%. Based on the
325 results, it is deduced that the ANN designed in this study can be used as a prediction
326 tool to improve the processes for obtaining methane from the dehydrogenation of
327 carbon dioxide. Future research will focus on the development of an ANN coupled with
328 a genetic optimization algorithm to maximize methanol production and plant
329 performance while minimizing economic costs.
330
331 References
332
333 1. Fang, X.; Men, Y.; Wu, F.; Zhao, Q.; Singh, R.; Xiao, P.; Du, T.; Webley, P.A. Moderate-pressure conversion of H2 and CO2
334 to methanol via adsorption enhanced hydrogenation. Int. J. Hydrogen Energy 2019, 44, 21913–21925,
335 doi:10.1016/j.ijhydene.2019.06.176.
336 2. Zahedi, G.; Elkamel, A.; Lohi, A.; Jahanmiri, A.; Rahimpor, M.R. Hybrid artificial neural network - First principle model
337 formulation for the unsteady state simulation and analysis of a packed bed reactor for CO 2 hydrogenation to methanol.
338 Chem. Eng. J. 2005, 115, 113–120, doi:10.1016/j.cej.2005.08.018.
339 3. Azzolina-Jury, F.; Bento, D.; Henriques, C.; Thibault-Starzyk, F. Chemical engineering aspects of plasma-assisted CO2
340 hydrogenation over nickel zeolites under partial vacuum. J. CO2 Util. 2017, 22, 97–109, doi:10.1016/j.jcou.2017.09.017.
341 4. Fornero, E.L.; Chiavassa, D.L.; Bonivardi, A.L.; Baltanás, M.A. CO2 capture via catalytic hydrogenation to methanol:
26
27 10 of 11
28
342 Thermodynamic limit vs. ’kinetic limit. Catal. Today 2011, 172, 158–165, doi:10.1016/j.cattod.2011.02.036.
343 5. Xiong, S.; Lian, Y.; Xie, H.; Liu, B. Hydrogenation of CO2 to methanol over Cu/ZnCr catalyst. Fuel 2019, 256, 115975,
344 doi:10.1016/j.fuel.2019.115975.
345 6. Leonzio, G.; Zondervan, E.; Foscolo, P.U. Methanol production by CO2 hydrogenation: Analysis and simulation of reactor
346 performance. Int. J. Hydrogen Energy 2019, 44, 7915–7933, doi:10.1016/j.ijhydene.2019.02.056.
347 7. Nguyen, T.B.H.; Zondervan, E. Methanol production from captured CO2 using hydrogenation and reforming technologies-
348 environmental and economic evaluation. J. CO2 Util. 2019, 34, 1–11, doi:10.1016/j.jcou.2019.05.033.
349 8. Li, T.; Yu, D.; Liu, J.; Wang, F. Atomic Pt Promoted N-Doped Carbon as Novel Negative Electrode for Li-Ion Batteries. ACS
350 Appl. Mater. Interfaces 2019, 11, 37559–37566, doi:10.1021/acsami.9b10533.
351 9. Jimenez, M.; Herrera, C.; Alemany, L.J. Ni-Ga no soportado como catalizador para la hidrogenación de CO 2 a presión
352 atmosférica. Secat 19 "Catalisis Para el Futur. 2019, 2–3.
353 10. Treviño, J. A., Ruiz, R., Fierro, J. C., Handy, B. E. & Flores, G.A. Hidrogenación de CO2 Utilizando Catalizadores de Rh
354 Soportado en Nanotubos de TiO2. In Proceedings of the Memorias del XXXVIII Encuentro Nacional de la AMIDIQ; 2017.
355 11. Ahmad, K.; Upadhyayula, S. Kinetics of CO2 hydrogenation to methanol over silica supported intermetallic Ga3Ni5
356 catalyst in a continuous differential fixed bed reactor. Int. J. Hydrogen Energy 2020, 45, 1140–1150,
357 doi:10.1016/j.ijhydene.2019.10.156.
358 12. Ye, J.; Liu, C.J.; Mei, D.; Ge, Q. Methanol synthesis from CO2 hydrogenation over a Pd 4/In2O3 model catalyst: A combined
359 DFT and kinetic study. J. Catal. 2014, 317, 44–53, doi:10.1016/j.jcat.2014.06.002.
360 13. Van-Dal, É.S.; Bouallou, C. Design and simulation of a methanol production plant from CO2 hydrogenation. J. Clean. Prod.
361 2013, 57, 38–45, doi:10.1016/j.jclepro.2013.06.008.
362 14. Vanden Bussche, K.M.; Froment, G.F. A steady-state kinetic model for methanol synthesis and the water gas shift reaction
363 on a commercial Cu/ZnO/Al2O3 catalyst. J. Catal. 1996, 161, 1–10, doi:10.1006/jcat.1996.0156.
364 15. Fang, X.; Men, Y.; Wu, F.; Zhao, Q.; Singh, R.; Xiao, P.; Du, T.; Webley, P.A. Promoting CO2 hydrogenation to methanol by
365 incorporating adsorbents into catalysts: Effects of hydrotalcite. Chem. Eng. J. 2019, 378, doi:10.1016/j.cej.2019.122052.
366 16. Sadeghinia, M.; Rezaei, M.; Nemati Kharat, A.; Namayandeh Jorabchi, M.; Nematollahi, B.; Zareiekordshouli, F. Effect of
367 In2O3 on the structural properties and catalytic performance of the CuO/ZnO/Al2O3 catalyst in CO2 and CO
368 hydrogenation to methanol. Mol. Catal. 2020, 484, 110776, doi:10.1016/j.mcat.2020.110776.
369 17. Aspen Technology Inc. Aspen Plus | Leading Process Simulation Software. Aspen Technol. Inc. 2020.
370 18. Pérez-Fortes, M.; Schöneberger, J.C.; Boulamanti, A.; Tzimas, E. Methanol synthesis using captured CO2 as raw material:
371 Techno-economic and environmental assessment. Appl. Energy 2016, 161, 718–732, doi:10.1016/j.apenergy.2015.07.067.
372 19. Chemcad CHEMCAD | Chemical Engineering Simulation Software by Chemstations 2021.
373 20. Kiss, A.A.; Pragt, J.J.; Vos, H.J.; Bargeman, G.; de Groot, M.T. Novel efficient process for methanol synthesis by CO2
374 hydrogenation. Chem. Eng. J. 2016, 284, 260–269, doi:10.1016/j.cej.2015.08.101.
375 21. Szima, S.; Cormos, C.C. Improving methanol synthesis from carbon-free H2 and captured CO2: A techno-economic and
376 environmental evaluation. J. CO2 Util. 2018, 24, 555–563, doi:10.1016/j.jcou.2018.02.007.
377 22. Do, T.N.; Kim, J. Process development and techno-economic evaluation of methanol production by direct CO2
378 hydrogenation using solar-thermal energy. J. CO2 Util. 2019, 33, 461–472, doi:10.1016/j.jcou.2019.07.003.
379 23. Ye, R.P.; Ding, J.; Gong, W.; Argyle, M.D.; Zhong, Q.; Wang, Y.; Russell, C.K.; Xu, Z.; Russell, A.G.; Li, Q.; et al. CO2
380 hydrogenation to high-value products via heterogeneous catalysis. Nat. Commun. 2019, 10.
381 24. Zalazar-oliva, C. Predicción de la resistencia térmica de las incrustaciones en los enfriadores de ácido sulfhídrico on the
382 sulphydric acid coolers. 2018, 345–359.
383 25. Abiodun, O.I.; Jantan, A.; Omolara, A.E.; Dada, K.V.; Mohamed, N.A.E.; Arshad, H. State-of-the-art in artificial neural
384 network applications: A survey. Heliyon 2018, 4, e00938, doi:10.1016/j.heliyon.2018.e00938.
385 26. Santolamazza, A.; Dadi, D.; Introna, V. A Data-Mining Approach for Wind Turbine Fault Detection Based on SCADA Data
29
30 11 of 11
31
386 Analysis Using Artificial Neural Networks. Energies 2021, 14, 1845, doi:10.3390/en14071845.
387 27. Puttige, A.R.; Andersson, S.; Östin, R.; Olofsson, T. Application of Regression and ANN Models for Heat Pumps with Field
388 Measurements. Energies 2021, 14, 1750, doi:10.3390/en14061750.
389 28. Kampelis, N.; Tsekeri, E.; Kolokotsa, D.; Kalaitzakis, K.; Isidori, D.; Cristalli, C. Development of demand response energy
390 management optimization at building and district levels using genetic algorithm and artificial neural network modelling
391 power predictions. Energies 2018, 11, doi:10.3390/en11113012.
392 29. Alade, O.; Shehri, D. Al; Mahmoud, M.; Sasaki, K. Viscosity – Temperature – Pressure Relationship of Extra-Heavy Oil
393 ( Bitumen ): Empirical Modelling. Energies 2019, doi:https://doi.org/10.3390/en12122390.
394 30. Li, H.; Zhang, Z.; Liu, Z. Application of artificial neural networks for catalysis: A review. Catalysts 2017, 7,
395 doi:10.3390/catal7100306.
396 31. Kito, S.; Hattori, T. Neural network as a tool for catalyst development. Catal. Today 1995, 23, 347–355,
397 doi:https://doi.org/10.1016/0920-5861(94)00148-U.
398 32. Liu, Y.; Liu, Y.; Liu, D.; Cao, T.; Han, S.; Xu, G. Design of CO2 hydrogenation catalyst by an artificial neural network.
399 Comput. Chem. Eng. 2001, 25, 1711–1714, doi:10.1016/S0098-1354(01)00714-1.
400 33. Sun, Y.; Yang, G.; Wen, C.; Zhang, L.; Sun, Z. Artificial neural networks with response surface methodology for
401 optimization of selective CO2 hydrogenation using K-promoted iron catalyst in a microchannel reactor. J. CO2 Util. 2018,
402 24, 10–21, doi:10.1016/j.jcou.2017.11.013.
403 34. Borisut, P.; Nuchitprasittichai, A. Process Configuration Studies of Methanol Production via Carbon Dioxide
404 Hydrogenation: Process Simulation-Based Optimization Using Artificial Neural Networks. Energies 2020, 13, 6608,
405 doi:10.3390/en13246608.
406 35. Dimian, A.C.; Bildea, C.S.; Kiss, A.A. Steady-State Flowsheeting. In Integrated Design and Simulation of Chemical Processes;
407 Elsevier, 2014; Vol. 35, pp. 73–125 ISBN 9780444627001.
408 36. Kiss, A.A. Advanced distillation technologies - Design, Control and Applications; Wiley, J., Ed.; First.; Wiley: Noida, India., 2005;
409 ISBN 9781119993612.
410
32