Subscriber access provided by CORNELL UNIVERSITY LIBRARY
Article
Artificial Neural Networks (ANN) for Accurate Prediction of Physical Properties of Aqueous Quaternary Systems of Carbon Dioxide (CO2)-Loaded 4-(diethylamino)-2-butanol (DEAB) and Methyldiethanolamine (MDEA) blended with Monoethanolamine (MEA) Fatemeh Pouryousefi, Raphael O. Idem, Teeradet Supap, and Paitoon Tontiwachwuthikul Ind. Eng. Chem. Res., Just Accepted Manuscript • DOI: 10.1021/acs.iecr.6b03018 • Publication Date (Web): 14 Oct 2016 Downloaded from http://pubs.acs.org on October 18, 2016
Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Industrial & Engineering Chemistry Research is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.
Page 1 of 33
6
8 10
MEA-DEAB-CO2-H2O SYSTEM
2
4
BPNN MODEL RBFNN MODEL EMPIRICAL MODEL
0
PREDICTION
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 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Industrial & Engineering Chemistry Research
0
2
4
6
EXPERIMENT
TOC/GRAPHIC
ACS Paragon Plus Environment
8
10
Industrial & Engineering Chemistry Research
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 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Page 2 of 33
1
Artificial Neural Networks (ANN) for Accurate Prediction of Physical
2
Properties of Aqueous Quaternary Systems of Carbon Dioxide (CO2)-
3
Loaded 4-(diethylamino)-2-butanol (DEAB) and Methyldiethanolamine
4
(MDEA) blended with Monoethanolamine (MEA)
5
6
Fatemeh Pouryousefi, Raphael Idem*, Teeradet Supap*, Paitoon Tontiwachwuthikul,
7
Clean Energy Technologies Research Institute (CETRI), University of Regina, Regina, SK
8
S4S 0A2, Canada
9
10 11 12 13 14 15 16 17 18
Corresponding Authors
19
*Phone: +1-(306) 585-4470; Fax: +1-(306) 585-4855; Email:
[email protected]
20
*Phone: +1-(306) 337-2468; Fax: +1-(306) 585-4855; E-mail:
[email protected]
21
1 ACS Paragon Plus Environment
Page 3 of 33
Industrial & Engineering Chemistry Research
22 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 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
23
Abstract
24
Physical and heat transport properties such as density, viscosity, refractive index, heat capacity,
25
thermal conductivity, and thermal diffusivity of aqueous carbon dioxide (CO2)-loaded and unloaded 4-
26
(diethyl amino)-2-buthanol (DEAB) and methyldiethanolamine (MDEA) as single amines and each
27
blended with a primary amine (MEA) were measured at different ranges of temperature (25°C - 60°C),
28
amine concentrations (0.5M-2M for tertiary amine and 5M for primary amine), and CO2 loading (0-0.6
29
mole/mole amine). Results showed an increasing trend of CO2 loading on density, viscosity and
30
refractive index, and a decreasing trend on the heat transport properties. Two Artificial Neural Network
31
techniques, Back Propagation Neural Network (BPNN) and Radial Basis Neural Network (RBFNN) as
32
well as some well-known empirical correlations from literature, were applied to correlate and predict
33
these
34
MEA+MDEA+WATER+CO2. Results from the correlation showed that Artificial Neural Network
35
techniques gave the least deviation for the prediction of all physical properties of both amine systems
36
with less than 1%AAD. The correlation coefficient between the experimental and predicted values in
37
terms of R2 value was in the range of 0.98 - 0.99.
physical
properties
for
two
quaternary
systems;
38 39 40 41 42 43 44 45 46
2 ACS Paragon Plus Environment
MEA+DEAB+WATER+CO2
and
Industrial & Engineering Chemistry Research
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 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Page 4 of 33
47 48
1. Introduction
49 50
Research and development in the area of carbon dioxide (CO2) capture and storage (CCS)
51
technologies for the greenhouse gas mitigation is on the rise because of its potential in allowing
52
continued use of fossil fuels with little or no emissions of CO2 to the atmosphere. Chemical absorption
53
using aqueous amine solution is a promising technology due to its effectiveness in reducing CO2
54
emissions from fossil-fuel power plants, thus helping alleviate the global climate change1. The amine
55
solvent allows the process to reactively capture CO2 from such low pressured flue gases typically found
56
in coal-fired power plants2. Therefore, knowledge of the amine chemical/physical properties including
57
density, viscosity, and thermal property (e.g. heat capacity and thermal conductivity) acquired during
58
the plant run is crucial as they help confirm state of the solvent. Any anomalies can also be early
59
predicted if the amine properties are starting to deviate from their original values. An accurate
60
determination of physical properties of these solvents specifically density, heat capacity, and thermal
61
conductivity must be obtained also for a precise calculation of heat duty during CO2 capture. Normally,
62
density, viscosity, refractive index, and thermal related properties such as heat capacity, thermal
63
conductivity and thermal diffusivity of amines are measured off-line in a laboratory using expensive sets
64
of instrument. Also, prior to measurement, it requires plant personnel to collect the amine samples from
65
various locations throughout the plant. Then, the samples must be properly handled and kept to preserve
66
to as close to their original forms as possible while transporting to a designated laboratory for a detailed
67
analysis. Specifically, sensitive samples such as those containing CO2 withdrawn from lean and rich
68
streams must be given an extra care because CO2, often can slowly evolve from the amine due to
69
temperature and pressure change. Thus density, viscosity, refractive index, and other properties might
70
not even represent true loading from the process.
3 ACS Paragon Plus Environment
Page 5 of 33
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 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Industrial & Engineering Chemistry Research
71
To deal with such a complexity and time consuming process, correlation models for predicting
72
these physical properties can be formulated and used to replace complicated lab measurement.
73
Empirical correlations used extensively are Weiland for density, Nissan Grunburg for viscosity, and
74
Gladston-Dale for refractive index. Redlich-Kister is also used extensively to predict and correlate many
75
physical properties such as heat capacity, refractive index, and thermal conductivity. Literature work
76
correlated density data for MEA+WATER+CO2 at temperatures from 25°C to 80 °C with Weiland
77
correlation.3 The maximum relative deviation was 1.6%. A study on the viscosity of carbonated mixed
78
aqueous MDEA+MEA at 293.15-343.15 K, 0.2-0.5 amine mole fractions, and 0-0.5 mole CO2/ mole
79
amine was also done.4 A modified Grunberg-Nissan equation was also suggested for prediction of the
80
viscosities of carbonated MDEA-MEA aqueous solutions. The correlation of the viscosity data using
81
Weiland and Grunberg-Nissan correlation for un-loaded aqueous amine solutions were in very good
82
agreement between these correlations. For loaded aqueous amine solutions, both Weiland and
83
Grunberg-Nissan correlations showed very good agreement in correlating the viscosity of mixed
84
conventional alkanolamines solutions. From our previous publication, we measured refractive indices of
85
binaries MEA+WATER, MDEA+WATER, and MEA+MDEA and used the Gladston-Dale, Lorentz-
86
Lorenz, Weiner, Heller, and Arago-Biot to correlate the experimental data.5 The results showed that the
87
Gladstone-Dale correlation fitted the data very well (0.03 % to 0.1 % at different temperatures). For
88
ternaries MEA+WATER+CO2 and MDEA+WATER+CO2, also, Gladston-Dale showed the best
89
agreement with the experimental data (0.53 % to 1.6 %). Refractive indices of some binaries
90
Hexadecane and Heptadecane with n-alkanols was measured and correlated with Gladston-Dale, Heller,
91
Lorentz-Lorenz, Arago-Biot, and Weiner equations and they all showed very good agreement with the
92
experimental data for her systems.6 Correlation of excess molar heat capacities of aqueous
93
alkanolamines such as 3-amino-1-propanol, 2-(methylamino)ethanol (MAE), and 1-amino-2-propanol
94
was also done by using the Redlich-Kister equation.7 The maximum average absolute percentage
95
deviation was 0.23 %. Again, correlation of excess molar heat capacities for aqueous mixture of 4
4
ACS Paragon Plus Environment
Industrial & Engineering Chemistry Research
96 1 2 3 97 4 5 98 6 7 99 8 9 10 100 11 12 101 13 14 102 15 16 17 103 18 19 104 20 21 22 105 23 24 106 25 26 107 27 28 29 108 30 31 109 32 33 110 34 35 36 111 37 38 112 39 40 41 113 42 43 114 44 45 115 46 47 48 116 49 50 117 51 52 118 53 54 55 119 56 57 120 58 59 60
Page 6 of 33
MEA+AMP was done by employing the Redlich-Kister equation.8 The average absolute percentage deviations for binaries and ternary system mentioned were 3.6 % and 13.2 % respectively. However, these correlations mentioned earlier have disadvantages especially when dealing with water based liquid mixtures. The non-ideality of these aqueous amines greatly affects the accuracy of these correlations. In comparison to theoretical based models, Artificial Neural Network (ANN) based methods can be more accurate and faster in predicting the physical properties of both polar and nonpolar liquid mixtures due to its ability to learn from data, classification capabilities, generalization and noise tolerance. In details, the primary advantage of ANN over theoretical and empirical models is that as a black box model, it does not pre-require any governing equation that describes engineering phenomena. Instead, ANN can learn the complex transport processes of a system from given observed data.9 Because of these benefits, ANN shows great potential in helping the CO2 capture plant to predict the physical properties of its solvent system. In 1998, density, viscosity, and refractive index of several ternary and quaternary solvent systems containing water, methanol, acetonitrile and tetrahydrofuran were predicted.10 The relative standard error was less than 1 % for density and refractive index, and 15 % for viscosity. ANN was used also to estimate the density of pure oil-based methyl ester biodiesel. The average absolute percent deviation was 0.29 %.11 As proven success of ANN shown earlier is clear, this work was carried out to formulate correlation models based on artificial neural network techniques to predict density, viscosity, refractive index, heat capacity, thermal conductivity, and thermal diffusivity of amine systems used to capture CO2. Back Propagation Neural Network (BPNN) and Radial Basis Neural Network (RBFNN) were used specifically in this work because of their high accuracy in prediction of physical properties for polar liquid mixtures and faster prediction. Blended systems of monoethanolamine (MEA) with methyldiethanolamine (MDEA) and 4-(diethylamino)-2-butanol (DEAB) were used to showcase the ANN technique applicability and prediction accuracy. MDEA was included in this study based on its being a common blend ingredient with MEA used widely. DEAB specifically developed by the Clean 5 ACS Paragon Plus Environment
Page 7 of 33
121 1 2 3 122 4 5 123 6 7 124 8 9 10 125 11 12 126 13 14 15 127 16 17 128 18 19 20 129 21 22 130 23 24 131 25 26 27 132 28 29 133 30 31 134 32 33 34 135 35 36 136 37 38 39 137 40 41 138 42 43 139 44 45 46 47 140 48 49 141 50 51 52 142 53 54 143 55 56 57 58 59 60
Industrial & Engineering Chemistry Research
Energy Technologies Research Institute (CETRI) of the University of Regina was chosen for this work due to its potential in replacing MDEA for a higher CO2 absorption capacity and reduced regeneration energy12. Physical and heat transport property data based on density, viscosity, refractive index, heat capacity, thermal conductivity, and thermal diffusivity of the 2 blends used for correlating the ANN based models were measured under 25⁰C - 60 ⁰C temperature, 0.5M - 2M tertiary amine +5M primary amine concentrations, and 0-0.6 mole CO2/ mole amine. These conditions were typical of the CO2 capture process. High temperature measurement (e.g. regeneration temperature:120°C and higher) of CO2 loaded amine solvent could lead to errors because CO2 could be released from the liquid solution to the gas phase, thus changing the loading of the solution whose property is being measured. To ensure the accuracy of the measurement and prediction of the ANN models, the on-line determination of all properties in this study was done between 25⁰C - 60 ⁰C. Thus, at this range of temperature, CO2 composition was confidently maintained at the same value during the measurement of physical and heat properties. The experimental data were used to develop the prediction models by using nine data inputs with 70% of the data given to training the systems to predict the properties. The predicted values obtained from BPNN and RBFNN correlations were then compared with those of well-known empirical models consisting of Weiland, Nissan-Grunberg, Gladston-Dale, and Redlich-Kister, developed also with the same data set. The comparison was done to confirm the superiority of ANN derived models in terms of higher prediction accuracy for density, viscosity, refractive index, heat capacity, thermal conductivity, and thermal diffusivity of the amine systems. 2. Theoretical Calculations 2.1 Back Propagation Neural Network (BPNN) & Radial Basis Neural Network (RBFNN) Based Correlations
6 ACS Paragon Plus Environment
Industrial & Engineering Chemistry Research
144 1 2 3 145 4 5 146 6 7 147 8 9 10 148 11 12 149 13 14 150 15 16 17 151 18 19 152 20 21 22 153 23 24 154 25 26 155 27 28 29 156 30 31 157 32 33 158 34 35 36 159 37 38 160 39 40 41 161 42 43 162 44 45 163 46 47 48 164 49 50 165 51 52 53 166 54 55 56 167 57 58 59 60
Page 8 of 33
The BPNN is one of the most widely used ANN techniques in chemical applications. Other than input and output layers, The BPNN can have multiple hidden layers as well. The hidden layers relate input and output. First, weights for the Neurons are randomly assigned. When the network is being trained, the given prediction is compared to the actual data. Based on the error computed between the predicted and actual data, the adjustment will be calculated and considered (backpropagated) through the network, which provides the weight adjustment. In an ANN technique, a portion of input/output data selected from the initial data set is used to train the network, and then the network iteratively adjusts its connection weights and bias values according to the computed errors, which are calculated between the network output and actual output. This training: testing ratio depends on the size of data set available. If there is a large data set, it is ok to consider a ratio of 75:25 to train a generalized ANN. If the ANN is not a generalized one, it tends to over-fit the data and it will predict bizarre outputs for some inputs which were not observed with the training set. If the size of data set is small, it will be better to go with a ratio of 90:10 for training: testing. After being well trained, the BPNN can then be used to predict the unseen target parameters13. Figure S1 shows a feedforward ANN with input units (X units), hidden units (Z units) and output units (Y units).13 W0k and V0j represent bias having output 1 on unit Yk and Zj respectively. During feedforward in stage 1, each input unit Xi receives input signal and broadcasts to hidden units Zj. The hidden units compute activation and send signals to output units Yk. The output units compute activation and produce the output signal of the net from the input pattern. In stage 2, calculation and backpropagation of error is performed. Each output unit computes its activation Yk and compares it with the target output tk to determine the associated error. Based on the error, the factor δ k is computed. The factor δ k distributes error to the output layer and the hidden layer. The weights are then updated between the output layer and the hidden layer. The hidden layer then computes error factor δ j , but it is not necessary to distributes error to the input layer. The factor
δ j is only used to update the weights
7 ACS Paragon Plus Environment
Page 9 of 33
168 1 2 3 169 4 5 170 6 7 171 8 9 10 172 11 12 173 13 14 174 15 16 17 175 18 19 176 20 21 22 177 23 24 178 25 26 179 27 28 29 180 30 31 181 32 33 182 34 35 36 183 37 38 184 39 40 41 185 42 43 44 186 45 46 187 47 48 188 49 50 51 189 52 53 190 54 55 191 56 57 58 59 60
Industrial & Engineering Chemistry Research
between the hidden and the input layer. In stage 3, after all the error factors are computed, the weights of all layers are adjusted simultaneously. The adjustment of the weights from the hidden layer to the output layer is based on the error factor and the activation of hidden units. The adjustment of weights from the input layer to the hidden layer is based on the error factor and the activation of the input layer. RBFNN is another type of neural network which has shown better results in compared to BPNN especially in prediction. These networks generally have 3 layers, the input layer, the hidden layer with the RBF non-linearity, and the linear output layer. RBFNN can beat some of the limitation of BPNN because of using rapid training phase, and having a simple architecture. The RBFNN simulates and predicts the aimed phenomena by using Gaussian basis functions in the hidden layer and linear transfer functions in the output layer. The difference between BPNN and RBFNN comes from the hidden layer. In RBFNN, the distance between the input and center is used in the RBFNN learning process. The first layer of the RBFNN collects the input data. Its training process determines the number of hidden neurons, m, which can be larger than that of BPNN to achieve a certain accuracy of prediction.14
For each neuron in the hidden layer, the distance between the input data and the center is activated by a nonlinear radial basis function, as shown in the following equation;13
[
2 Yˆ = exp − ( xi − ci b1)
]
(1)
Where xi is the input vector, and b1 and ci are parameters that represent the bias in the hidden layer and center vector respectively. Each neuron in the hidden layer will produce a value between 0 & 1 according to how close the input is to the location of the center. Therefore, neurons with centers closer to inputs will have more contributions to outputs; however, if neurons have centers far away from inputs, then the outputs would be nullified. Later, the output layer neurons receive the weighted inputs
8 ACS Paragon Plus Environment
Industrial & Engineering Chemistry Research
192 1 2 3 193 4 5 194 6 7 8 195 9 10 11 12 196 13 14 197 15 16 17 198 18 19 199 20 21 200 22 23 24 201 25 26 202 27 28 203 29 30 31 204 32 33 205 34 35 36 206 37 38 207 39 40 208 41 42 43 209 44 45 210 46 47 211 48 49 212 50 51 52 213 53 54 214 55 56 215 57 58 59 60
Page 10 of 33
from the hidden layer and predict the results by using a linear combination, which is of a similar from to that of the BPNN;
m
ˆ = ∑ωiRi + b2 Υ
(2)
i =1
Where Yˆ is the RBFNN simulation results, wi is the optimized connection weight determined through the training process, and b2 is the bias in the output layer. In RBFNN simulations, the proper initial choice of centers and weights should be regarded as key issues. Various methods are proposed to define the center, but in this study random selection was used to define these parameters. Once the centers have been developed, the weights linking the hidden and output layers should be updated during the training procedure. The training procedure in the RBFNN also determines the number of hidden neurons required for the simulation. The training of RBFNN is initiated by first generating a single neuron in the hidden layer, followed by continuously adding neurons to the hidden layer at a time. In RBFNN, it is very important to determine the number of neuron carefully in the hidden layer because it affects the complexity and the prediction ability of the network. Too many neurons in the hidden layer may cause poor generalization or over-fitting, while insufficient neurons cannot learn the data adequately. In addition, the position of the centers in the hidden layer also affects the network performance significantly, while the nonlinear transfer functions used by RBFNN have little influence.
2.2 Semi-Empirical Correlations Different correlations from literatures have been also applied to correlate the physical and thermal properties measured for both blended amine systems. These correlations were short-listed for this work based on their well-known application in the single property prediction for the polar solutions of the same family type which is amine solutions. Weiland and non-additive equations respectively 9 ACS Paragon Plus Environment
Page 11 of 33
216 1 2 3 217 4 5 218 6 7 219 8 9 10 220 11 12 221 13 14 222 15 16 17 223 18 19 224 20 21 225 22 23 226 24 25 26 227 27 28 228 29 30 31 229 32 33 230 34 35 36 231 37 38 232 39 40 233 41 42 43 234 44 45 235 46 47 48 49 50 51 52 238 53 54 55 56 57 58 59 60
Industrial & Engineering Chemistry Research
shown in Equations (3), (4), and (5) were selected for correlation of density data.15 Equations (6) and (7) representing Gladston-Dale16 and Arago-Biot (as used by Sharma et al.17) correlations were chosen for refractive index. Nissan Grunberg18 given in Equation (8) was used to correlate viscosity. Finally, Redlich-Kister19 correlation was applied to model heat capacity, thermal conductivity, and thermal diffusivity which were given in equations 9-15. For density, r
ρ = ∑i =1 xi ρ i 3
∑ ρ
=
( x
i
.M
i
(3)
(4)
)
i = 1
V
Where
V = xAM .vAM + xW .vW + xCO2 .vCO2 + xAM .xW .v* + xAM .xCO2 .v**
(5)
χ i , ρ i , and Μ i are mole fraction, density, and molecular weight of each component respectively in
the amine system. ρ represents the density of the mixture and V is the molar volume of the solution. For an ideal solution, V is the sum of the partial molar volumes of the components multiplied by their respective molar fractions. However, amine solutions loaded with CO2 are not ideal solutions. Thus, extra parameters are needed to account for interactions of amine + H2O and amine + CO2. These parameters are given respectively by ν* and ν** respectively. For refractive index, 236
nm − 1
ρ
r
= ∑i =1
ni − 1
237ωi
(6)
nm = ∑i =1 niφi
(7)
ρi r
10 ACS Paragon Plus Environment
Industrial & Engineering Chemistry Research
Page 12 of 33
240 1 2 3 241 4 5 6 242 7 8 9 243 10 11 12 244 13 14 15 245 16 17 246 18 19 20 247 21 22 23 248 24 25 249 26 27 250 28 29 251 30 31 252 32 33 34 253 35 36 254 37 38 39 255 40 41 256 42 43 44 257 45 46 47 258 48 49 50 259 51 52 260 53 54 55 261 56 57 58 59 60
n
Where; respectively.
φ
m
i
and
φ
i
are volume fraction of pure component i and refractive index of mixture,
is calculated from xi Vi /
∑ xiVi which ni and ωi are refractive index of component i
and weight fraction of component i
For viscosity, ln η mix = χ i ln ηi + χ i χ j d
(8)
Where d is the adjustable parameter independent on operational parameters18, η i and ηmix are the viscosities of component i and the mixture, and X i and
χ j are the mole fractions of component i and
j
For heat capacity, CP
n
E 12
/( J / mol . K ) = χ 1 χ 2 ∑ i =1 Ai ( χ 1 − χ 2 ) i −1
(9)
Where; the temperature dependence of Ai is assumed to follow the temperature relation as follows;
Ai = ai,0 + ai,1(T / K)
(10)
E
It should be noted that the excess experimental molar heat capacities ( CP ) were calculated from the experimental molar heat capacity values by:
11 ACS Paragon Plus Environment
Page 13 of 33
Industrial & Engineering Chemistry Research
1 262 2 3 263 4 5 6 264 7 8 265 9 10 11 266 12 13 267 14 15 16 268 17 18 269 19 20 21 270 22 23 271 24 25 272 26 27 28 29 273 30 274 31 32 275 33 34 35 276 36 37 277 38 39 278 40 41 42 279 43 44 280 45 46 281 47 48 49 282 50 51 52 283 53 54 284 55 56 57 285 58 59 60
C P /( J / mol.K ) = C P − ( χ1C P ,1 + χ 2C P ,2 )
E
n
K E 12 /( KW / m . K ) = χ 1 χ 2 ∑ i =1 Ai ( χ 1 − χ 2 ) i −1
Ai = ai,0 + ai,1(T / K)
n
D E 12 /( m 2 / s ) = χ 1 χ 2 ∑ i =1 Ai ( χ 1 − χ 2 ) i −1
Ai = ai,0 + ai,1(T / K)
(11)
(12)
(13)
(14)
(15)
Where Ai and ai are adjustable and temperature dependent parameters.
3. Experiments 3.1. Chemicals and Equipment DEAB was synthesized according to the procedure described in the literature20. The chemicals used for DEAB synthesis were methyl vinyl ketone (MVK, 95%), sodium borohydride (NaBH4, 98%), methanol (HPLC grade) and diethylamine (99.5 %). All chemicals were obtained from Sigma Aldrich (Sigma- Aldrich Canada Co, Ontario), except MVK from Fisher Scientific (Fisher Scientific Company, Ontario). Structural confirmation of the DEAB were carried out by nuclear magnetic resonance spectrometer (NMR and Varian Mercury plus 300M Spectrometer). The purity of DEAB was also determined by the NMR to be in the range of 98%-99%. The MEA and MDEA were purchased from Sigma Aldrich (99% purity). Desired blend ratios of MEA+MDEA and MEA+DEAB were prepared by mixing predetermined weight of each amine in the blend. The rest of volume was made up using deionized water (DI water). Total concentration of the mixture was confirmed by titration with 1 N Hydrochloric Acid (laboratory 12 ACS Paragon Plus Environment
Industrial & Engineering Chemistry Research
286 1 2 3 287 4 5 288 6 7 289 8 9 10 290 11 12 291 13 14 292 15 16 17 293 18 19 294 20 21 22 295 23 24 296 25 26 297 27 28 29 298 30 31 299 32 33 300 34 35 36 301 37 38 302 39 40 41 303 42 43 304 44 45 305 46 47 48 306 49 50 307 51 52 53 308 54 55 309 56 57 310 58 59 60
Page 14 of 33
grade from Fisher Scientific) to the end point of methyl orange indicator (0, 1 % indicator, from Sigma Aldrich). For CO2 loaded solutions, research grade CO2 (from Praxair Canada Inc.) was passed into the desired amine solution. The exact CO2 loading was measured also by HCl-methyl orange titration and displacement technique described in details in the literature.20,21 Density, viscosity, and refractive index measurement of all amine solutions were straightforward. Density and viscosity were measured simultaneously with a densitomer coupled with a viscometer (from Anton Paar) using approximate 10 mL sample size. Refractive indices at the wavelength of 589 nm were measured on a separate instrument using automatic refractometer purchased from Anton Paar. Thermal conductivity, thermal diffusivity, and heat capacity were determined by liquid thermal conductivity meter (THW-LAMBDA, ThermTest Inc, Ontario) operated by a transient hot wire technique. The instrument used a thin platinum wire immersed in the amine sample which the heat resistance profile of wire is measured with respect to time and a temperature-time profile of amine sample were generated. These plots were then used to determine thermal conductivity, thermal diffusivity, and heat capacity. Thermal conductivity and diffusivity were measured specifically at the beginning when temperature become stable, and then using the correlation of thermal diffusivity and heat capacity (D= K / (ρ Cp)), heat capacity will be computed at different time intervals. It should be noted that measurement of heat capacity required an input of predetermined density values of the sample at 3-4 temperatures for the software to be able to plot the density-temperature profile for later computation of heat capacity at different temperatures. These were provided using density data described earlier. At least 3 repeated measurements of density, viscosity, refractive index, and thermal related properties were taken for each sample which their averaged values were used and reported. Prior to actual test, all physical and heat measurement techniques and instruments were validated at 25⁰C - 60
⁰C range of temperature against literature values using water and pure amines (e.g. MEA, MDEA, and DEAB). The confirmation of validity and accuracy can be viewed in our previous work.20 Detailed information of MDEA measurement (e.g. viscosity, density and refractive index) previously done in the 13 ACS Paragon Plus Environment
Page 15 of 33
311 1 2 3 312 4 5 6 313 7 8 9 314 10 11 12 315 13 14 316 15 16 317 17 18 318 19 20 319 21 22 320 23 24 321 25 26 27 322 28 29 323 30 31 324 32 33 34 325 35 36 326 37 38 327 39 40 41 328 42 43 329 44 45 46 330 47 48 331 49 50 332 51 52 53 333 54 55 334 56 57 335 58 59 60
Industrial & Engineering Chemistry Research
other works5, particularly in the mixture with MEA, CO2 and H2O also confirms the validity of the data presented in this study. Table S1also shows all specifications related to the equipment used to measure these physical properties.
4. Results and Discussions
4.1 Measurement of Density, Viscosity, Refractive Index, Heat Capacity, Thermal Conductivity, and Thermal Diffusivity This section reports density, viscosity, refractive index, heat capacity, thermal conductivity, and thermal diffusivity data of pure amines and water, binary, ternary, and quaternary systems containing DEAB + H2O, DEAB + CO2 + H2O, and DEAB + MEA + CO2 + H2O and MDEA + MEA + CO2 + H2O, respectively. These data were needed later for development of correlation models using ANN techniques and empirical equations described earlier in section 2. For pure amine systems, density, viscosity, and refractive index shown in Figure S2 (a) – (c) were measured at 25 °C from 60 °C, while 25 °C to 50 °C temperature range was used to determine heat capacity, thermal conductivity, and thermal diffusivity, given in Figure S3 (a), (b), and (c), respectively. For binary DEA and water system, density, viscosity, and refractive index in Figure S4 (a) – (c) were measured from 25 °C to 40 °C for aqueous solutions of DEAB prepared at 0.5M, 1M, 1.25M, 1.5M, and 2M corresponding to 0.01 – 0.05 mole fraction range of DEAB in the system. The same range of DEAB mole fraction and temperature were also used to determine all thermal properties of the binary system shown in Figure S5 (a), (b), and (c) respectively for heat capacity, thermal conductivity, and thermal diffusivity. For ternary system containing DEAB, H2O, and CO2, density, viscosity, and refractive index of 1.5M DEAB loaded with CO2 from 0 to 0.70 mole CO2/mole DEAB reported in Figure S6 (a) to (c) were measured at temperatures from 22 °C to 42 °C. Heat transport properties of the same solution but loaded with 0 to 0.50 mole CO2/mole and measured at temperatures ranging from 25 °C to 40 °C were shown in Figure 14 ACS Paragon Plus Environment
Industrial & Engineering Chemistry Research
336 1 2 3 337 4 5 338 6 7 339 8 9 10 340 11 12 341 13 14 342 15 16 17 18 343 19 20 344 21 22 23 345 24 25 346 26 27 28 347 29 30 348 31 32 349 33 34 35 350 36 37 351 38 39 352 40 41 42 353 43 44 354 45 46 47 355 48 49 356 50 51 357 52 53 54 358 55 56 359 57 58 59 60
Page 16 of 33
6S (a), (b), and (c) respectively for heat capacity, thermal conductivity, and thermal diffusivity. Quaternary systems used 5 M aqueous MEA solution mixed with 1.5 M of either DEAB or MDEA as a based solution. Both mixtures were then loaded with various CO2 loading in range of 0 – 0.6 mol CO2/mol amine and measure for all properties between 25 °C to 50 °C. Density, viscosity, and refractive index of quaternary DEAB and MDEA based systems can be found in Figure S7 to S12. Heat capacity, thermal conductivity, and thermal diffusivity data of the same systems can also be viewed in Figure S13 to S18.
4.2 Prediction of Density, Viscosity, Refractive Index, Heat Capacity, Thermal Conductivity, and Thermal Diffusivity ANN based Models and Empirical Correlations Two methods of artificial neural network known as BPNN and RBFNN were used to correlate data obtained from Section 4.1. MEA + DEAB + H2O + CO2 system was used to showcase the applicability and goodness of these ANN based techniques in terms of precision and accuracy in prediction of density, viscosity, refractive index, and thermal transport properties required during the capture process of CO2. Well-known empirical correlations namely, Weiland and non- additive equations for density, Nissan-Grunberg for viscosity, Gladston-Dale and Arago-Biot for refractive index, and Redlich-Kister for heat capacity, thermal conductivity, and thermal diffusivity also used to correlate the experimental data for prediction of the mentioned physical properties were also compared for accuracy to the ANN based methods. DEAB based system never reported was chosen for the discussion in this section due to its high efficiency as a solvent mixture to capture CO2 (e.g. CO2 equilibrium solubility and regeneration heat duty). Also, this solvent has never been reported elsewhere on this aspect, thus introducing new knowledge to the CO2 capture research community. Figure 1 and 2 show architectural diagrams respectively for BPNN and RBFNN, for the quaternary system. Mole fractions and measurement temperature were considered as inputs while physical properties of the system were considered as outputs. Party chart plotted between predicted and 15 ACS Paragon Plus Environment
Page 17 of 33
360 1 2 3 361 4 5 362 6 7 363 8 9 10 364 11 12 365 13 14 366 15 16 17 367 18 19 368 20 21 22 369 23 24 370 25 26 371 27 28 29 372 30 31 373 32 33 374 34 35 36 375 37 38 376 39 40 41 377 42 43 378 44 45 379 46 47 48 380 49 50 381 51 52 382 53 54 55 383 56 57 384 58 59 60
Industrial & Engineering Chemistry Research
experimental density obtained from BPNN and RBFNN prediction is shown in Figure 3. Densities of the same DEAB system predicted using Weiland and non-additive equation based empirical correlations are also plotted in sample Figure for comparison. BPNN and RBFNN could clearly predict density of the amine mixture very accurately, confirmed by very low %AAD averaged to only 0.09 and 0.12%, respectively. The empirical models on the other hand were much inferior to the ANN based models. Specifically, density values predicted by Weiland equation deviated from the actual values by as much as 76.56%. Though, non-additive model gave an acceptable 5.28%AAD. This model was still far less accurate than BPNN and RBFNN that gave very small % error, virtually close to zero. For viscosity prediction shown in Figure 4, BPNN and RBFNN still did superbly well showing %AAD between the predicted and experimental viscosity values of only 3.10 and 2.52%, respectively, compared to 16.27%AAD obtained from commonly used Nissan Grunberg based correlation. Figure 5 is used to compare refractive index predicted also from ANN techniques to those obtained from Gladston-Dale and Arago-Biot correlations. Prediction accuracy indicated by %AAD is in the order of 0.05, 0.04, 0.5, and 4.025%, respectively for BPNN, RBFNN, Gladston-Dale, and Arago-Biot. Very small deviations of refractive index prediction once again, confirm the applicability of the ANN based models generated in this study. Figures 6, 7 and 8 respectively compare heat capacity, thermal conductivity, and thermal diffusivity obtained by BPNN and RBFNN to those of the empirical Redlich-Kister model. It is clear that BPNN and RBFNN gave the most accurate prediction for DEAB based solvent with less than 1% deviation for all thermal properties. Redlich-Kister on the other hand, showed very high inaccuracy of the predicted values clearly indicated by 27.88%, 45.21%, and 21.93% deviations for heat capacity, thermal conductivity, and thermal diffusivity, respectively. ANN based models were also used to predict all properties for MDEA quaternary system (MEA-MDEA-CO2-H2O). Figures 9 and 10 show clearly how accurately ANN based model could predict density and heat capacity of MDEA quaternary based system. Also, viscosity, refractive index, thermal conductivity, and thermal diffusivity were also 16 ACS Paragon Plus Environment
Industrial & Engineering Chemistry Research
385 1 2 3 386 4 5 387 6 7 388 8 9 10 389 11 12 13 390 14 15 391 16 17 18 392 19 20 393 21 22 394 23 24 25 395 26 27 396 28 29 397 30 31 32 398 33 34 399 35 36 400 37 38 39 401 40 41 402 42 43 44 403 45 46 404 47 48 405 49 50 51 406 52 53 407 54 55 408 56 57 58 59 60
Page 18 of 33
predicted accurately as shown in Figures S19, S20, S21, and S22, respectively. For all measurements, the level of accuracy was still exceptional for both BPNN and RBFNN shown by the maximum of 1.6%AAD compared to as high as 65% from the empirical model predictions. The measurement and prediction accuracy of a known amine like MDEA also confirm the validity of DEAB data presented in this study.
5. Conclusions This study presented for the first time experimental data on physical properties of new alcohol amine solvent, 4-(Diethylamino)-2-buthanol (DEAB) in pure and aqueous solution with/without CO2 under CO2 capture range conditions. Results showed that the density and viscosity and thermal properties of this new solvent were lower than those of conventional MEA and MDEA, which could be considered beneficial for the CO2 capture process in terms of easy operation and energy efficient. The Artificial Neural Network (ANN) was used to correlate the experimental data for an accurate prediction of density, viscosity, refractive index, heat capacity, thermal conductivity, and thermal diffusivity. Using DEAB-MEA-CO2-H2O and MDEA-MEA-CO2-H2O systems, the results have confirmed that ANN based BPNN and RBFNN model were very accurate to predict all properties of the quaternary amine systems with an overall deviation of less than 1%. The empirical correlations commonly used, including Weiland and non-additive equation for density, Nissan Grunberg for viscosity, Gladston-Dale, and Arago-Biot for refractive index, and, Redlich-Kister for prediction of heat capacity, thermal conductivity, and thermal diffusivity, were incomparable to the ANN models and far less accurate in predicting these physical properties indicated by as high as 76%AAD value. Thus, ANN based techniques have proven themselves to be a useful tool for an accurate prediction of amine system’s physical properties required during the operation of CO2 capture process, which could not be predicted accurately by the existing empirical correlations commonly presented in the literatures. It must be noted that the effect of impurities such as amine degradation products was also an important factor 17 ACS Paragon Plus Environment
Page 19 of 33
409 1 2 3 410 4 5 411 6 7 412 8 9 10 11 413 12 13 414 14 15 415 16 17 416 18 19 20 417 21 22 418 23 24 25 419 26 27 420 28 421 29 30 422 31 32 33 423 34 35 424 36 37 425 38 39 40 41 426 42 43 427 44 45 428 46 47 429 48 49 430 50 51 431 52 53 432 54 55 433 56 57 434 58 59 60
Industrial & Engineering Chemistry Research
which could alter the predicted physical and heat property values of amine from the actual values. To evaluate this effect thoroughly, experimental plan and schedule have been set up to incorporate various degradation products and flue gas impurities into building a prediction model which will be reported in our future works.
Supporting Information Backpropagation network diagram, data of density, viscosity, refractive index, and thermal properties of pure amines and mixtures of DEAB-H2O, DEAB-CO2-H2O, MEA-DEAB-CO2-H2O and MEA-MDEA-CO2-H2O; parity plot for prediction of viscosity, refractive index, and thermal conductivity and diffusivity in MEA-MDEA-CO2-H2O mixture; equipment for measurement of all physical properties.
Acknowledgements The financial support provided by the Natural Science and Engineering Research Council of Canada (NSERC), Canada Foundation for Innovation (CFI), the Clean Energy Technologies Research Institute (CETRI), and Faculty of Graduate Studies and Research (FGSR), University of Regina is gratefully acknowledged.
18 ACS Paragon Plus Environment
Industrial & Engineering Chemistry Research
435
1 2 436 3 4 5 437 6 7 438 8 9 10 439 11 12 440 13 14 15 441 16 17 18 442 19 20 21 443 22 23 444 24 25 26 27 445 28 29 446 30 31 32 447 33 34 35 448 36 37 449 38 39 40 450 41 42 43 451 44 45 46 452 47 48 453 49 50 51 454 52 53 54 455 55 56 57 58 59 60
Page 20 of 33
References (1) Zhang, Y.; Chen, H.; Chen, C.-C.; Plaza, J. M.; Dugas, R.; Rochelle, G. T. Rate-Based Process Modeling Study of CO2 Capture with Aqueous Monoethanolamine Solution. Ind. Eng. Chem. Res. 2009,
48, 9233. (2) Chakma, A. Separation of CO2 and SO2 from flue gas streams by liquid membranes. Energy Convers. Manage. 1995, 36, 405. (3) Amundsen, T. G.; Eimer, D. A. Density and Viscosity of Monoethanolamine + Water + Carbon Dioxide from (25 to 80) C. J. Chem. Eng. Data. 2009, 54, 3096. (4) Fu, D.; Chen, L.; Qin, L. Experiment and Model for the Viscosity of Carbonated MDEA−MEA
Aqueous Solutions. Fluid Phase Equilib. 2012, 319, 42.
(5) Pouryousefi, F.; Idem, R.O. New Analytical Technique for Carbon Dioxide Absorption Solvents. Ind. Eng. Chem. Res. 2008, 47, 1268. (6) Mehra, R. Application of Refractive Index Mixing Rules in Binary Systems of Hexadecane and Heptadecane with n-Alkanols at Different Temperatures. Proc. Indian Acad. Sci. (Chem. Sci.), 2003,
55(2), 147. (7) Henni, A.; Mundhwa, M. Molar Heat Capacity of Various Alkanolamine Solutions from 303.15 K to 353.15 K. J. Chem. Eng. Data. 2007, 52, 491.
(8) Chen, Y. J.; Shih, T. W.; Li, M. H. Heat Capacity of Aqueous Mixtures of Monoethanolamine with N-Methyldiethanolamine. J. Chem. Eng. Data. 2001, 46, 51.
(9) Hussain, T. Checking the Integrity of Global Positioning Recommended Minimum (GPRMC) Sentences Using Artificial Neural Network (ANN). Master Thesis, University of Gavle, 2009.
19 ACS Paragon Plus Environment
Page 21 of 33
456 1 2 3 457 4 5 6 458 7 8 459 9 10 11 460 12 13 14 461 15 16 462 17 18 19 463 20 21 22 464 23 24 25 465 26 27 466 28 29 30 31 467 32 33 468 34 35 36 469 37 38 39 470 40 41 42 471 43 44 472 45 46 47 473 48 49 50 474 51 52 53 475 54 55 476 56 57 58 59 60
Industrial & Engineering Chemistry Research
(10) Mehlman, S.A.; Wentzell, P.D.; McGuffin, V.L. Prediction of Simple Physical Properties of Mixed Solvent Systems by Artificial Neural Networks. Analytica Chimica-Acta. 1998, 371(2-3), 117.
(11) Baroutian, S.; Aroua; M.K.; Raman, A.A.A.; Sulaiman, N.M.N. Estimation of Vegetable Oil-Based Ethyl Esters Biodiesel Densities Using Artificial Neural Networks. J.Appl.Sci, 2008, 8(17), 3005.
(12) Tontiwachwuthikul, P.; Wee, A. G. H.; Idem, R. O.; Maneeintr, K; Fan, G. J.; Veawab, A.; Aroonwilas, A.; Chakma. A. Method for capturing carbon dioxide from gas streams. U.S. Patent
Application 2008/0050296 AI, 2008. (13) Fausett, L. Fundamentals of Neural Network Architecture, Algorithms, and Applications; PrenticeHall, UK, 1994.
(14) Demuth, H.; Beale, M. Neural Network Toolbox for Use with MATLAB; The Mathworks Inc., MA, 2002.
(15) Weiland, H.; Dingman, J. C.; Benjamin, D. Density and Viscosity of Some Partially Carbonated Aqueous Alkanolamine Solutions and Their Blends. J. Chem. Eng. Data. 1998, 43, 378.
(16) Dale, D.; Gladstone, F. On the Influence of Temperature on the Refraction of Light. Phil- Trans. 1858, 148, 887. (17) Sharma, S., Patel, P. B., Patel, R. S., Vora, J. J.; Density and comparative refractive index study on mixing properties of binary liquid mixtures of eucalyptol with hydrocarbons at 303.15, 308.15 and 313.15K, E-Journal of Chemistry, 2007, 4, 343. (18) Grunberg, L.; Nissan, A.H. Mixture law for viscosity. Nature. 1949, 164, 799.
(19) Redlich, O.; Kister, A. T. Algebric Representation of Thermodynamic Properties and the Classification of Solutions. Ind. Eng. Chem. 1948, 40, 345. 20 ACS Paragon Plus Environment
Industrial & Engineering Chemistry Research
477 1 2 3 478 4 5 479 6 7 8 480 9 10 11 481 12 13 14 482 15 16 17 483 18 19 20 484 21 22 23 485 24 25 26 486 27 28 29 30 487 31 32 33 488 34 35 36 489 37 38 39 490 40 41 42 491 43 44 45 492 46 47 48 493 49 50 51 494 52 53 54 495 55 56 57 58 59 60
Page 22 of 33
(20) Pouryousefi, F. Development of On-Line Analytical Technique for Determination of Composition of CO2-Loaded Formulated Amine Solvents Based on the Liquid Thermo physical Properties for a PostCombustion CO2 Capture Process. PhD Thesis, University of Regina, 2015.
(21) Maneeintr, K.; Idem, R. O.; Tontiwachwuthikul, P.; Wee, A. G. H. Synthesis, solubility, and cyclic capacity of amino alcohols for CO2 capture from flue gas streams. Energy Proc. 2009, 1, 1327.
21 ACS Paragon Plus Environment
Page 23 of 33
496
1 2 3 497 4 5 6 498 7 8 9 499 10 11 500 12 13 14 501 15 16 17 18 502 19 20 21 503 22 23 24 504 25 26 27 505 28 29 30 506 31 32 33 507 34 35 36 508 37 38 39 509 40 41 42 510 43 44 45 511 46 47 48 512 49 50 513 51 52 53 514 54 55 515 56 57 516 58 59 60
Industrial & Engineering Chemistry Research
List of Figures Figure 1 Architecture of BPNN for thermal property prediction of MEA+DEAB (MDEA) +WATER+ CO2 system
Figure 2 Architecture of RBFNN for thermal property prediction of MEA+DEAB (MDEA) +WATER+ CO2 system
Figure 3 Parity chart for density: MEA+DEAB+water+CO2 system Figure 4 Parity chart for viscosity: MEA+DEAB+water+CO2 system Figure 5 Parity chart for refractive index: MEA+DEAB+water+CO2 system Figure 6 Parity chart for heat capacity: MEA+DEAB+water+CO2 system Figure 7 Parity chart for thermal conductivity: MEA+DEAB+water+CO2 system Figure 8 Parity chart for thermal diffusivity: MEA+DEAB+water+CO2 system Figure 9 Parity chart for density of MEA+ MDEA+water+ CO2 system Figure 10 Parity chart for heat capacity of MEA+ MDEA+water+ CO2 system TOC/GRAPHIC
22 ACS Paragon Plus Environment
Industrial & Engineering Chemistry Research
Page 24 of 33
517
1 2 3 518 4 5 519 6 7 520 8 9 521 10 11 522 12 13 14 523 15 16 524 17 18 525 19 20 21 526 22 23 24 527 25 26 27 28 528 29 30 31 529 32 33 34 530 35 36 531 37 38 532 39 40 533 41 42 534 43 44 535 45 46 536 47 48 537 49 50 51 538 52 53 539 54 55 540 56 57 541 58 59 60
Figure 1: Architecture of BPNN technique for thermal property prediction of MEA+DEAB (MDEA) +WATER+CO2 system
23 ACS Paragon Plus Environment
Page 25 of 33
Industrial & Engineering Chemistry Research
542
1 2 3 543 4 5 6 544 7 8 9 545 10 11 12 546 13 14 15 547 16 17 18 19 548 20 21 22 549 23 24 25 550 26 27 28 551 29 30 31 32 552 33 34 35 553 36 37 554 38 39 40 555 41 42 43 556 44 45 557 46 47 558 48 49 50 559 51 52 560 53 54 55 561 56 57 562 58 59 60
Figure 2: Architecture of RBFNN for thermal property prediction of MEA+DEAB (MDEA)+WATER+ CO2 system
24 ACS Paragon Plus Environment
Industrial & Engineering Chemistry Research
563
1 2 3 564 4 5 565 6 7 566 8 9 10 567 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 568 44 45 569 46 47 48 570 49 50 51 571 52 53 572 54 55 56 573 57 58 59 60
Figure 3: Parity chart for density: MEA+DEAB+water+CO2 system
25 ACS Paragon Plus Environment
Page 26 of 33
Page 27 of 33
Industrial & Engineering Chemistry Research
574
1 2 3 575 4 5 576 6 7 577 8 9 10 578 11 12 579 13 14 580 15 16 17 581 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 582 49 50 51 583 52 53 54 584 55 56 585 57 58 59 60
Figure 4: Parity chart for viscosity: MEA+DEAB+water+CO2 system
26 ACS Paragon Plus Environment
Industrial & Engineering Chemistry Research
586
1 2 3 587 4 5 588 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 589 46 47 590 48 49 50 591 51 52 592 53 54 55 593 56 57 594 58 59 60
Figure 5: Parity chart for refractive index: MEA+DEAB+water+CO2 system
27 ACS Paragon Plus Environment
Page 28 of 33
Page 29 of 33
Industrial & Engineering Chemistry Research
595
1 2 3 596 4 5 597 6 7 598 8 9 599 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 600 44 45 46 601 47 48 602 49 50 51 603 52 53 604 54 55 605 56 57 58 59 60
Figure 6: Parity chart for heat capacity: MEA+DEAB+water+CO2 system
28 ACS Paragon Plus Environment
Industrial & Engineering Chemistry Research
606
1 2 607 3 4 5 608 6 7 609 8 9 610 10 11 611 12 13 612 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 613 48 49 614 50 51 52 615 53 54 616 55 56 57 617 58 59 60
Figure 7: Parity chart for thermal conductivity: MEA+DEAB+water+CO2 system
29 ACS Paragon Plus Environment
Page 30 of 33
Page 31 of 33
Industrial & Engineering Chemistry Research
618 1 2 619 3 4 620 5 621 6 622 7 8 9 623 10 11 624 12 13 625 14 15 626 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 627 47 48 49 628 50 51 629 52 53 54 55 630 56 57 631 58 59 60
Figure 8: Parity chart for thermal diffusivity: MEA+DEAB+water+CO2 system
30 ACS Paragon Plus Environment
Industrial & Engineering Chemistry Research
632
1 2 3 633 4 5 634 6 7 8 635 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 636 49 50 51 637 52 53 54 638 55 56 57 639 58 59 60
Figure 9 Parity chart for density of MEA+ MDEA+water+ CO2 system
31 ACS Paragon Plus Environment
Page 32 of 33
Page 33 of 33
Industrial & Engineering Chemistry Research
640
1 2 3 641 4 5 6 642 7 8 643 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 644 47 48 49 645 50 51 52 646 53 54 55 647 56 57 58 59 60
Figure 10 Parity chart for heat capacity of MEA+ MDEA+water+ CO2 system
32 ACS Paragon Plus Environment
