Professional Documents
Culture Documents
XII edition
XII edition
A Stefano, Cecilia, e.
Acknowledgements
The first person I wish to thank is my wife. Without her sharing this project nothing would
have been possible. I thank her for the patience, for my absence filled with her love, for the
comprehension, and so on. So, I thank her to exist.
My parents. Without them nothing could have been started a long time ago. They could
never imagine where they allowed me to arrive.
Professor Giovanni Di Nicola. Im proud to have had a kind of supervisor. He always
assisted me in difficult moments and trusting me, let allowed me reach some results that I
would have never reached without him. With him I achieved my first publication, with him
I discovered my first equation, with him I can speak every day about our research of our
life, with him At the end of my course, reaping the rewards of this experience, the most
precious of them is without any doubts his friendship. Thanks Giovanni.
Doctor Roman Stryjek. Once told me: Mariano, I dont teach you thermodynamics, but a
method. This is more difficult! If I only think to the possibility Ive to be side by side, I
will can only thank God for this opportunity. Ill bring forever in my mind his hard work,
his scrupulosity, the severity in the research and the fine talent. All this things wont ever
die because Ill try to teach them to my students.
Professor Fabio Polonara. In the few moments I had the possibility to talk with him I
discovered a man in love with research a tireless worker, that teaches the beauty of this
work.
The first time I met him he clearly explained the purpose of any future tasks and he told me
about the effort to spend in any work in order to continue and fulfill my aims. I hope Ive
not disappointed him expectations!
Professor Giovanni Latini. His rigorousness and passion in research astonished me every
time Ive the fortune to see him working.
Roberta Cocci Grifoni. Her passion in research convinced me to throw myself into the
adventure, and I will never stop to thank her.
Mario Vallorani. My insurance for my mathematical problems. Ive no problems until he is
behind me.
Giulio P. Thanks for all the nights passed behind the PC trying to solve problems.
All my new university friends.
First of all Giovanni. Thanks for giving me what I lack. Always, just ask: this is friendship.
Giulio S. I learned from him more than he can ever imagine.
Massimiliano. Whose knowledge overcame me, and I can't see the limit of it.
Luca. Your tenacity, and your suggestions have unlighted me.
Gianluca. I shared with him any problems trying to solve them and this let me grow and
clear my doubts.
Valerio. Every time I spoke with him I discovered some new abilities, all of them excellent.
Working with him in the future is one of my dreams.
At the end all the students I coped with. How much I got involved to give them all I knew,
complementary learning something from them.
Thanks!
ii
Abstract
In the modern multicultural and multidisciplinary society, always adopting more and more
wider prospective than before.
Its well known from the beginning of calculus that Mathematics and physics are sisters.
In fact Leibnitz, and Newton are the fathers of Calculus in the same way.
In this thesis, we try to adopt a multidisciplinary method, which involves Mathematics,
Physics, but also Chemistry, Statistics, and in general the scientific engineering with its
applications in many technical fields.
We have tried to lift our attention, and not keep it fixed only on the calculations, but we try
to find that thin red line which unites the varies involved disciplines.
The aspects explained are thermo physical properties, and Equations of State (EOS) of
gases.
Regarding thermo physical properties have been analysed Surface Tension, Thermal
Conductivity, Viscosity, and the second virial coefficient.
This thesis is based on previous results of studies conducted by Professor Giovanni Di
Nicola.
On this arguments, the work had been subdivided between the gathering of experimental
data, the analysing of data with statistical techniques transforming them to more reliable
data than row.
The second step was to collect the equations of literature starting from the oldest to most
recent in order to have a good panoramic of the state of art.
We then went ahead studying the sensibility of data to find out which physical properties
could have bigger impact to property examined.
The analysis of statistic techniques involved are or graphical analysis or a more advanced
technique such as the factor analysis and Design of Experiment (DOE).
Once the higher influence on experimental data has been examined, we looked for an
equation that could represent experimental data in a better way. To do this has been adopted
classical techniques of regression, such as gradient descent, and more advanced methods
such as Levenberg Marquardt.
Where it was possible we always tried not to forget the chemical and physical aspects of the
investigation of better equation. So we always preferred the scaled equations that respect
chemical and physical aspects, to the empirical ones.
Comparing our results with better equations in literature, our results are always better, in
fact all of the have been published in the best international journals on this subject.
A separate discussion is that of EOS.
As a first step we tried to understand at what point the research had arrived. Every years
many new equations are produced, and new techniques are developed.
The first thing that came to our minds was that to find the best possible equation is
impossible. Or as Martin wrote copying words of the famous fables Snow White: Mirror
mirror on the wall, who is the fairest of them all?.
Thanks to the precious help of doctor Roman Stryjek we tried to find a line of research that
could produce and guarantee good results.
iii
iv
Contents
1.1 Introduction
1.2 Terminology and units
1.3 Methods of measurement
1.4 An overview of existing equations
1.5 Results
1.5.1 A Scaled Equation for the Surface Tension of Pure Fluids
1.5.2 A new scaled equation for the calculation of surface tension of
ketones
1.5.3 A new Scaled Equation and an Artificial Neural Network for
calculation of surface tension of Alcohols
1.5.4 Surface tension prediction for refrigerant binary systems
1.6 Bibliografia
2.1 Introduction
2.2 Terminology and units
2.3 An overview of existing equations
2.3.1 Equation of Weber
2.3.2 First equation of Smith
2.3.3 Equation of Palmer
2.3.4 Equation of Vargaftik
2.3.5 Equation of Robbins and Kingrea
2.3.6 Equation of Bridgmann
2.3.7 Equation of Osida
2.3.8 Equation of Hirschfelder, Curtiss and Bird
2.3.9 Equation of Viswanath
2.3.10 Equation of Kardos
2.3.11 Equation of Sakiadis and Coates
2.3.12 First equation of Badea
2.3.13 Equation of Narasimhan, Swamy and Narayama
2.3.14 Second equation of Smith
2.3.15 Equation of Denbigh
2.3.16 Equation of Sheffy and Johnson
2.3.17 Equation of Missenard
2.3.18 Second equation of Badea
2.3.19 First equation of Pachaiyappan, Ibrahim, and Kuloor
2.3.20 Second equation of Pachaiyappan, Ibrahim, and Kuloor
Introduction
An overview of existing equations
Results
References
4.1
4.2
4.3
Introduction
Terminology and units
Methods of measurement
vi
5.6
Bibliografia
vii
viii
List of Figures
Figure 1.1 Capillary Rise Method...........................................................................................2
Figure 1.2 Box-plot of the collected surface tension data.......................................................8
Figure 1.3 Histogram of the collected surface tension data ....................................................9
Figure 1.4 Scatter plot of the collected surface tension data of organics vs. (1-Tr) ............11
Figure 1.5 Surface tension at the boiling point vs. critical density .....................................31
Figure 1.6 Surface tension at the boiling point vs. radius of gyration .................................32
Figure 1.7 Surface tension at the boiling point vs. critical temperature ...............................32
Figure 1.8 Cross-correlation between radius of gyration and critical density ......................33
Figure 1.9 Cross-correlation between radius of gyration and critical temperature...............33
Figure 1.10 Cross-correlation between critical density and critical temperature..................34
Figure 1.11 Deviations of Equation (1.17)+ (1.18) as function of reduced temperature......35
Figure 1.12. Scatter plot of the collected surface tension data of inorganics vs. (1-Tr). ......36
Figure 1.13 Schematic diagram of the ANN model. Inputs: Tc (K), critical temperature; Gr
(), radius of gyration; T (K), temperature; c (molcm-3), critical density. Output:
(mN/m), surface tension ..............................................................................................40
Figure 1.14. Absolute average percent deviation between data collected and ANN results
vs. number of neurons in the hidden layer ..................................................................41
Figure 1.15. Correlation between the predictions of the validated, trained and tested MLP
and the corresponding experimental surface tension data. ..........................................43
Figure 1.16. Deviations of ANN as function of reduced temperature ..................................43
Figure 1.17: Descriptive statistics of ketones set of experimental data ................................46
Figure 1.18. Scatter Plot of experimental surface tension data versus reduced temperature.
Confidence bands at 1standard deviation, 2standard deviation and 3standard
deviation ......................................................................................................................46
Figure 1.19. Comparison of residuals between different equations......................................49
Figure 1.20. Deviations for the different ketones .................................................................49
Figure 1.21: Data Distribution of surface tension of alcohols ..............................................51
Figure 1.22. Data on the surface tension of alcohols vs. reduced temperature. The
marked/selected symbols are the fluids not considered during the data regression, as
in Table 1.14................................................................................................................62
Figure 1.23. Surface tension at boiling point vs. critical temperature ..................................64
Figure 1.24. Surface tension at boiling point vs. radius of gyration .....................................64
Figure 1.25. Surface tension at boiling point vs. critical density..........................................65
Figure 1.26. Cross-correlation between radius of gyration and critical temperature............66
Figure 1.27. Cross-correlation between critical density and critical temperature.................66
Figure 1.28. Cross-correlation between critical density and radius of gyration ...................67
Figure 1.29. Deviations of equation Eq (1.25) +Eq (1.26) as function of reduced
temperature ..................................................................................................................68
Figure 1.30. Schematic diagram of the ANN model. Inputs: Tc (K), critical temperature; Gr
(), radius of gyration; T (K), temperature; c (molcm-3), critical density. Output:
(mN/m), surface tension ..............................................................................................77
ix
Figure 1.31. Absolute average percent deviation between data collected and ANN results vs
number of neurons in hidden layer ..............................................................................78
Figure 1.32. Correlation between the calculated and the corresponding experimental surface
tension data. .................................................................................................................80
Figure 1.33. Deviations of ANN as function of reduced temperature ..................................80
Figure 1.34. Experimental surface tension of R32+R227ea, R32+125 and R32+R134a
binary systems according to equation (1.27). ..............................................................86
Figure 1.35. Experimental surface tension of R290+R32, R290+152a and R290+R600a
binary systems according to equation (1.27). ..............................................................87
Figure 1.36. Experimental surface tension of R125+R152a, R125+134a and R125+R143a
binary systems according to equation (1.27). ..............................................................88
Figure 1.37. - Experimental surface tension of R290+RE170, R143a+R134a, R134a+R152a
binary and R143a+R227ea systems according to equation (1.27). ............................89
Figure 1.38. Absolute (left) and relative (right) deviation for the total number of
experimental points obtained with equation (1.33) .....................................................91
Figure 2.1 Deviation between experimental and calculated with equation (2.36) thermal
conductivities versus reduced temperature................................................................108
Figure 2.2 Thermal conductivity data vs. reduced temperature..........................................110
Figure 2.3. Thermal conductivity at boiling point vs. acentric factor.................................126
Figure 2.4. Thermal conductivity at boiling temperature versus molecular mass. .............126
Figure 2.5. Thermal conductivity at boiling point vs. enthalpy of fusion...........................127
Figure 2.6. Thermal conductivity at boiling point vs. enthalpy of fusion...........................127
Figure 2.7. Thermal conductivity deviations vs. reduced temperature ...............................130
Figure 2.8. Schematic diagram of the ANN model. Inputs: M (Kg/Kmol) molecular mass,
acentric factor, Tr, reduced temperature. Output: (W m-1 K-1/m), thermal
conductivity ...............................................................................................................131
Figure 2.9. Absolute average percent deviation between data collected and ANN results vs
number of neurons in hidden layer ............................................................................132
Figure 2.10. Correlation between the calculated and the corresponding experimental
thermal conductivity data. .........................................................................................134
Figure 2.11. Deviations of ANN as function of reduced temperature ................................134
Figure 2.12. Histogram (left) and box plot (right) of experimental data on the thermal
conductivity of refrigerants .......................................................................................136
Figure 2.13. Experimental data on the thermal conductivity for refrigerants vs. reduced
temperature ................................................................................................................137
Figure 2.14. Thermal conductivity at the boiling point vs. molecular mass.......................139
Figure 2.15. Thermal conductivity at the boiling point vs. critical temperature.................140
Figure 2.16. Thermal conductivity at the boiling point vs. critical pressure. .....................140
Figure 2.17. Thermal conductivity at the boiling point vs. acentric factor.........................141
Figure 2.18. Thermal conductivity deviations of equation (2.41) vs. reduced temperature.
...................................................................................................................................144
Figure 2.19. Scatter Plot of experimental thermal conductivity data versus reduced
temperature. Confidence bands at , 2 and 3 are also reported..................146
Figure 2.20. Statistical summary for thermal conductivity.................................................146
Figure 2.21. Statistical summary for reduced temperature. ................................................147
xi
xii
List of Tables
Table 1.1. Coefficients for Sastri Rao Equation .....................................................................5
Table 1.2. Coefficients for Xiang Equation ............................................................................6
Table 1.3.A descriptive statistics referring to distribution properties ...................................10
Table 1.4.Summary of the collected data for organic compounds and comparison between
equations......................................................................................................................12
Table 1.5 Phisycal properties for each compounds and coefficients of linear regression ....13
Table 1.6. Summary of the collected data for inorganic compounds....................................35
Table 1.7.Deviations for inorganic compounds and coefficients for Equation (1.17)..........37
Table 1.8.The parameters of the hidden layer and of the output layer of the ANN..............41
Table 1.9.Summary of the deviations of the ANN ...............................................................42
Table 1.10: Studied ketones and their adopted physical properties ......................................45
Table 1.11.Correlation matrix between all the variables. : surface tension, Tc: critical
temperature, Pc: critical pressure, : acentric factor, Tb: temperature at boiling point,
c: critical density, M: molecular weight, Zc: compressibility factor, Ttp: temperature
at triple point, Ptp: pressure at triple point, Gr: radius of gyration .............................46
Table 1.12: Deviations between equations ...........................................................................48
Table 1.13: Deviation for each ketone ..................................................................................48
Table 1.14. Summary of the collected data and calculated surface tension at the boiling
point, @Tb =mT+q.....................................................................................................52
Table 1.15.Summary of deviations for equations .................................................................68
Table 1.16.Summary of deviations for each fluid.................................................................69
Table 1.17.Parameters of hidden and output layers ..............................................................78
Table 1.18.Surface tension experimental data for the refrigerant binary systems available in
the open literature ........................................................................................................81
Table 1.19.Summary of deviations between experimental surface tension and the ones
predicted by eqs. (1.27) and (1.28)..............................................................................83
Table 1.20.Deviations between the experimental surface tension and the ones predicted
with the regressed r value............................................................................................85
Table 1.21 Physical properties for each refrigerant fluid adopted for eqs (1.11) and (1.12)90
Table 2.1.Deviations between experimental and calculated thermal conductivities.
Deviations calculated according to equation (2.38) ..................................................108
Table 2.2.Effect of each term on thermal conductivity calculation ....................................111
Table 2.3. Summary of experimental data and calculated thermal conductivity at the boiling
point...........................................................................................................................112
Table 2.4. Factor analysis results. Effects were calculated according to Equation (2.39) .128
Table 2.5. Deviations between experimental and calculated thermal conductivities.
Deviations were calculated according to Equation (2.38).........................................129
Table 2.6. Parameters of hidden and output layers .............................................................132
Table 2.7. Factor analysis results. Effects were calculated according to Equation (2.40)..136
Table 2.8. Summary of experimental data and calculated thermal conductivity at the boiling
point...........................................................................................................................138
Table 2.9. Summary of deviations for equations. ...............................................................141
Table 2.10. Coefficients for equations (2.40) and (2.41). ...................................................143
xiii
xiv
1 Surface tension
1.1 Introduction
At the molecular level, a boundary between the liquid phase and the gas phase could be
considered as a third phase with distinct properties of a liquid and its vapor.
From a qualitative point of view of the microscopic surface layer, an explanation of this
phenomenon would be that at the interface, there are unequal forces acting upon the
molecules; for example at low densities the surface molecules are attracted sidewise and
toward the bulk liquid, but they are less attracted toward the bulk vapor phase. In this
manner, the surface layer is in tension, and there is a tendency for the surface layer to
minimize its area compatible to the total mass, container restraints, and external forces. The
index of this tension, can be presented in different ways, one of which is surface tension ()
defined as the force exerted in the plane of the surface per unit length.
Surface tension is an important fluid property that correlates and predicts heat transfer
during a phase change; in particular, it plays a fundamental role in the study of phase
transitions and technical processes such as boiling and condensation.
Liquid-vapor interfaces is very important for the behavior of detergent, and for many
chemical engineering fields, such as absorption, and distillation.
Important names in the physics history try to describe the properties of liquid surface.
Names as Gauss, Laplace, Lord Kelvin, Gibbs, Thomas Young, Bohr, Einstein,
Schrodinger, and van der Waals.
This phenomenon provides a simple and accurate technique for the measurement of the
surface tension of a liquid as shown in
This phenomenon is exploited both in differential and singular capillary rise method
(DCRM and SCRM, respectively). The experimental cell generally contains multiple
circular tubes of small section in the case of DCRM and a single pipe in the case of SCRM.
The accuracy of the method is of the order of 0.05-0.1 mN m-1 in most cases.
The well known Young-Laplace equation, states that the capillarity rise h of a liquid in a
capillary tube is given by
(1.1)
where , and are the liquid and vapor density respectively, is the liquid height into the
tube, is the acceleration due the gravity, is the contact angle between the liquid and the
wall of the tube, and is the radius of the capillary tube. The method is only accurate for
= 0. If the radius of tube is appreciable, then r is replaced by b where b refers to the radius
of curvature of the liquid, and a relation between b and r must be found. The tube must be
of uniform bore and the height h must be measured from the flat surface in the reservoir.
A recently adopted method of measurement is the Surface Light Scattering (SLS),
consisting of the Dynamic Light Scattering (DLS) application to fluid surfaces. The DLS
makes it possible to determine, under conditions of balance, various thermodynamic
properties of fluids, including surface tension, without any direct contact. The method is
based on the analysis of light diffusion caused by microscopic fluctuations on the fluid
surface. The accuracy of the surface tension data obtained by the SLS method is of the
order of 0.2 mNm-1.
Other experimental methods adopted for the surface tension measurement are the Vertical
Plate Method (VPM) and the Maximum Bubble Pressure Method (MBPM).
The vertical plate method is based on the natural convection flow of a viscous fluid along a
vertical plate, again giving an accuracy of the order of 0.2 mN m-1.
In the maximum bubble pressure method, a bubble pressure tensiometer produces gas
bubbles (i.e. air) at constant rate and blows them through a capillary of known radius which
is submerged in the sample liquid. The pressure inside of the gas bubble continues to
increase and the maximum value is obtained when the bubble has the completely hemispherical shape whose radius is exactly corresponding to the radius of the capillary. Then,
the maximum pressure of the bubbles is measured. The surface tension can be determined
using the Laplace equation in the reduced form for spherical bubble shape within the liquid.
The measurements obtained by the MBPM presented an accuracy of the order of 1.5e2%.
= K (l v )
(1.3)
The formula suggests the existence of a simple relationship between surface tension and the
difference between the densities of the liquid and of its vapour; K is a temperature
independent constant characteristic of each substance. One year later, Equation (1.3) was
modified by Sudgen [8]:
= [P(l v )]3
(1.4)
Where K=P4, suggesting that the parameter was temperature dependent; he called this
parameter parachor. Parachor values of various compounds were calculated by Quayle [9]
using experimental surface-tension data. Parachor property can be related to the critical
properties of compounds including critical temperature and molar volume as follows [1011]:
(1.5)
where T is the temperature (K), v is the molar volume (m3kmol-1) and the subscript c
denotes the critical value.
Equation (1.4) may be considered as an equation of state [12] of the interface because it has
been shown to be a good predictor of surface tension if equilibrium densities and
experimental data for the parachor are used. As Escobedo and Mansoori suggested [13], its
good performance and extremely easy analytical form have made Equation (1.4) a very
popular method for calculating surface tension, but it has some shortcomings, as the percent
deviation in predicting surface tension increases with increasing complexity of the
molecular structure.
Ferguson and Kennedy [15] and Guggenheim [15] used the corresponding states principle
to correlate the surface tension in the low temperature range; this decision was justified
under the assumption that, since the corresponding states principle was derived from an
equation of state, surface tension is also expected to follow the same principle. Brock and
Bird [16] developed this concept for non-polar liquids and proposed the following formula
applicable to the range of temperature far from the critical point:
2
11
= Pc 3Tc 3 Q (1 Tr ) 9
(1.6)
where Pc (bar) is the critical pressure, Tr is the reduced temperature, T/Tc. Q is the Riedel
parameter developed in terms of Pc and Tbr according to the Miller suggestion,
T ln Pc
br
1.01325
Q = 0.1196 1 +
0.279
1 Tbr
(1.7)
Tbr is the reduced normal boiling temperature, Tb/Tc, and Tb is the normal boiling
temperature (K).
In some cases, the surface tension can be represented in terms of an additional parameter.
One example is the so-called acentric factor, , initially introduced as an empirical
parameter by Pitzer to explain the deviation from the corresponding states principle, as
defined for noble gases, when applied to larger molecules. Pitzers [17] relation in terms of
Tc (K), Pc (bar), and leads to the following corresponding state relationship for :
2
3
= Pc Tc
1
3
1 Tr
0.291 0.08
19.05
( )
11
9
(1.8)
For compounds that exhibit hydrogen bonds, Sastri and Rao [18] proposed a modified
expression for the Brock and Bird correlation:
m
1 T
r
= KP T T
1
T
br
(1.9)
x y z
c b c
where Pc is in bar. Coefficients were regressed for different families separately. The values
for the constant of the equation (1.9) in reported in Table 1.1
Table 1.1. Coefficients for Sastri Rao Equation
Acids
K
0.125
x
0.50
y
-1.5
z
1.85
2.28
0.158
0.25
0.50
0.175
-1.5
0
1.85
0.8
11
9
Alcohols
All
others
11
9
To underline the work done in this research field, other recent equations have been
proposed. Miqueu et al. [19] proposed the following:
2
N
1.26
0.5
= kTc a ( 4.35 + 4.14 ) t (1 + 0.19t 0.487t )
Vc
3
(1.10)
where
, ,
, , , are the Boltzmann constant, Avogadro number, the
critical volume, and the acentric factor, respectively, and it represents the surface tension
for 31 substances up to the critical point with an absolute average deviation of less than
3.5%.
Di Nicola et al. [20-21] proposed two equations specifically oriented to refrigerants:
= 0.658Pc0.618Tc0.340 (1 + )
(1 Tr )
0.770
= 0.813Pc
0.565
Tc
0.364
(1 T )
1.266
(1.11)
1.262
(4.259 0.645 )
*
0.143
(1.12)
Equation (1.11) is extremely easy, and with only five parameters; equation (1.12) has seven
parameters and involves the reduced dipole moment, *, defined as [22]
* =
4300
Tc VC
(1.13)
0.04
1 3
1 1
= [s0 + s1 1 + s2 1] 1
0
Tr
Tr Tr
l,r
v,r
(1.14)
where
. These equations describe the surface tension for non polar, polar
hydrogen-bonding, and associating substances.
The substance-dependent parameters , ,
are:
(1.15)
s00
0.037516
s10
-0.009836
s20
0.0046745
s01
-0.019905
s11
-0.024977
s21
-0.017050
s02
-2.685506
s12
-1.438909
s22
1.597544
1.5 Results
This part contains the study on surface tension. Some new equations for the surface tension
are presented. In particular a new general and scaled equation for pure organic fluids, tested
also on some inorganic fluids, a new scaled equation for the family of ketones, another
equation for the family of alcohols, and at the end a new equation for refrigerant binary
mixtures. This work produced very good results; in fact all the new equations founded
describe the behavior of surface temperature better than the existing equations in literature,
on a wide range of compounds. Besides the new equations are simple, and with a quite
good level of applicability, in fact they contain physical parameter of large use in chemical
engineering.
1.5.1
The first step of the research on surface tension is to discover a new formula to calculate the
surface tension of organic fluids. An analysis of the data available on the surface tension
was made. A total of 2651 data were collected for alkanes, alkenes, cycloalkanes,
cycloalkenes, aromatics, ketones, epoxides, esters, ethers, mercaptans, peroxides and
sulphides. The data were regressed with the most reliable semi-empirical correlation
methods in literature based on the corresponding states theory. The scaled equation
proposed is very simple and gives noticeable improvement with respect to existing
equations for all the studied organic compounds. The equation was also tested for the main
inorganic compounds, giving good results in terms of deviations (AAD%=5.45%). The
same physical parameters considered in the scaled equation were also adopted as input
parameters in a multi-layer perceptron neural network, to predict the surface tension. The
multi-layer perceptron proposed has one hidden layer with 34 neurons, determined
according to the constructive approach. The model developed was trained, validated and
tested for the set of data collected, showing that the accuracy of the neural network model is
very good (AAD%=4.06%).
The raw surface tension data (experimental, smoothed and predicted) accepted by DIPPR
database [2] were analysed. The main advantage of the DIPPR database is that it collects
data from a wide range of sources and evaluates them critically. References, notes, and
quality codes for all data points are given [3]. During the data collection, a fluid by fluid
analysis was performed and data showing deviations higher than three times the standard
deviation were rejected.
Several theories that have been proposed to date can be applied to describe the surface
tension.
The surface tension (mNm-1) of a fluid can be expressed as a function of temperature
using an equation similar to that of van der Waals [4]
(1.16)
where Tc (K) is the critical temperature, 0 (mNm-1) and n are empirical constants
obtained with the least square method in a fit to the experimental data available on surface
tension [5,6].
Because of association, the equations for the surface tension calculation of polar molecules,
and therefore hydrophilic, often pose serious problems in terms of predictability. For this
reason, compounds that exhibit strong hydrogen bonds such as refrigerants, alcohols and
organic acids were separately analysed by us [24-26] and not considered in the present
paper.
As a first step, the distribution of the entire dataset was statistically analyzed. In Figure 1.2
and in Figure 1.3, the box-plot for the outliers identification and the histograms for the
relative frequency of the data collected are reported, respectively. All the data lying outside
the defined bounds can be considered as outliers. From Figure 1.1 a quite significant
number of outliers were detected for peroxydes and mercaptans (outside the higher fence)
and for ketones and aromatics (outside the lower fence). Additionally, a descriptive
statistics referring to distribution properties such as location, dispersion, and shape of the
data is also reported in Table 1.3. However, all data including the outliers were considered
during the regressions.
300
250
Alkanes
Alkenes
Aromatics
Cycloalkanes
Cycloalkenes
Epoxides
Esters
Ethers
200
150
100
Ketones
Mercaptans
Peroxydes
Sulphides
50
10
20
30
40
50
Points Number
Mean
Variance
Standard
Deviation
Skewness
Kurtosis
Median
10 Quantile
90 Quantile
Alkanes
Alkenes
Aromatics
Cycloalkanes
Cycloalkenes
Epoxides
Ethers
Ketones
Mercaptans
Peroxydes
Sulphides
Esters
393
17.82
52.09
90
22.47
34.14
167
24.96
22.20
44
23.10
12.62
38
25.87
12.41
244
23.40
103.50
93
17.71
66.96
154
22.24
17.80
297
21.50
87.92
163
18.91
87.65
799
23.11
119.12
169
22.21
9.11
7.22
-0.74
2.67
19.30
5.59
26.07
5.84
0.70
2.94
21.34
15.43
30.81
4.71
-1.25
6.21
25.13
20.00
30.21
3.55
-0.06
2.04
23.14
18.22
27.84
3.52
0.36
2.54
25.90
21.54
31.61
10.17
-0.18
2.40
24.69
8.00
35.55
8.18
-0.01
2.63
17.62
4.71
27.43
4.22
-2.29
9.77
23.18
18.40
25.70
9.38
0.05
2.95
22.42
7.03
33.90
9.36
0.29
2.29
18.31
6.75
32.12
10.91
-0.02
2.32
24.26
7.12
36.79
3.02
-0.65
2.46
23.22
17.49
25.56
10
Starting from the consideration that experimental data can generally be well described by
adopting a linear dependence on a function of (1-Tr), in Figure 1.4 a scatter plot of surface
tension data of organics divided by family versus (1-Tr) is presented; it shows a rather
positive correlation and a common trend for all fluids, which underlines the importance of
the reduced temperature in calculating the surface tension. This is also confirmed by
NISTs REFPROP program [27], where Equation (1.16) is adopted as a reference [4-5].
Figure 1.4 Scatter plot of the collected surface tension data of organics vs. (1-Tr)
Besides the experimental temperature and the critical temperature, we searched for some
specific additional fluid properties that could be involved in the model to improve the
representation of data, which represent various chemical structures. According to Equation
(1.4) and to the classical parachor approach, the critical density was selected as a parameter.
Although one additional parameter was selected: the radius of gyration.
From the general definition, the radius of gyration can be calculated as the root mean
square distance of the objects' parts from either a given axis or its center of gravity. In
particular, it can be referred to as the radial distance from a given axis at which the mass of
a body could be concentrated without altering the rotational inertia of the body about that
axis. For a planar distribution of mass rotating about the same axis in the plane of the mass,
the radius of gyration can be considered as the equivalent distance of the mass from the axis
of rotation. Radius of gyration plays an important role in chemistry. In a fact, it is usually a
better estimate of the chain dimensions than the root-mean-squared end-to-end distance.
The end-to-end distance is difficult to measure, while the radius of gyration can be
measured by a light scattering technique.
It is worthwhile to investigate this matter, in order to provide a more thorough explanation
as to how surface tension is related to the selected parameters, namely critical temperature,
critical density and radius of gyration. Since the boiling point was generally found to be
within the range of the experimental temperatures, as witnessed in Table 1.4, a linear
regression of surface tension against the reduced temperature of each fluid was made; then
the surface tension at the boiling point was calculated for each fluid; lastly, the surface
tension at the calculated boiling point was plotted for the three parameters considered.
The coefficient used in the linear regression are listed in Table 1.5. For each compound are
listed the properties, the range of measurement and the surface tension calculated at Boiling
temperature.
11
Table 1.4.Summary of the collected data for organic compounds and comparison between equations
Family
N. of
Fluids
N. of
Points
Alkanes
22
393
Alkenes
Aromatic
s
Cycloalk
anes
Cycloalk
enes
90
167
44
38
Epoxides
17
244
Esters
12
169
Ethers
93
Ketones
Mercapta
ns
Peroxyde
s
154
23
297
14
163
58
799
-
Sulphides
Mean
range
(mNm-1)
T range
(K)
0.227.972
12.7338.182
4.568932.88
16.4-29.8
20.0933.55
1.91848.832
14.6827.07
3.2837.22
2.993530.842
2.376348.48
93.15672.43
133.39373.15
273.1785.7
253.15393.15
239.15349.15
130.664.34
273.15392.15
149.95481.15
252.85569.43
133.02651.6
3.277-42.
2.979255.155
213.-788.
156.12813.69
AAD
%
Eq.
(1.9)
AAD
%
Eq.
(1.10)
AAD
%
(1.6)
AAD
%
(1.8)
AAD %
Eq. (1.17)
+
Eq.(1.18)
6.54
2.65
3.15
5.55
2.53
3.63
2.03
1.85
4.81
1.14
4.65
2.92
8.12
8.94
1.80
1.82
2.16
2.29
3.07
3.04
1.68
1.87
3.23
1.58
0.71
6.17
7.87
6.37
7.89
7.31
2.42
5.39
2.78
6.93
9.60
2.47
3.57
1.70
6.64
4.62
2.20
8.97
4.92
9.20
3.60
7.66
8.71
50.38
8.79
5.43
17.70
37.57
43.48 58.16
8.67
6.53
6.29
7.90
8.09
7.63 10.44
12.85 11.27
6.84
5.45
12
Table 1.5 Phisycal properties for each compounds and coefficients of linear regression
name
family
N
Points
range
(mNm-1)
Ethane
Alkanes
27
0.72-27.08
Methane
Alkanes
16
2.7-17.8
N-butane
Alkanes
22
10.-23.31
N-decane
Alkanes
12
14.63-24.75
N-docosane
Alkanes
5.0759-27.877
N-eicosane
Alkanes
20.54-27.21
N-heneicosane
Alkanes
4.9369-27.972
N-heptadecane
Alkanes
11
19.45-27.48
N-heptane
Alkanes
31
1.94-21.12
N-hexadecane
Alkanes
17
18.93-27.8
N-hexane
Alkanes
29
0.58-20.55
T range
(K)
117.8296.1
93.15173.15
203.15313.15
283.15393.15
317.15664.84
313.15393.15
313.35657.22
298.15393.15
283.15503.15
293.15393.15
273.15493.15
Tc(K)
Radius of
ryration (A)
Critical
density
(molcm-3)
305.32
1.83
0.0069
190.56
1.12
0.0101
425.12
2.89
0.0039
617.70
5.15
0.0016
9.65
787.00
8.80
0.0007
5.63
768.00
8.36
0.0007
1.89
778.00
8.54
0.0007
6.62
736.00
7.36
0.0009
4.04
540.20
4.17
0.0023
723.00
7.06
0.0010
507.60
3.77
0.0027
Tb (K)
@T
b
4.45
13
name
family
N
Points
range
(mNm-1)
N-nonadecane
Alkanes
20.21-26.91
N-nonane
Alkanes
12
13.5-23.79
N-octadecane
Alkanes
27
19.87-27.59
N-octane
Alkanes
43
3.8-27.5
N-pentadecane
Alkanes
27
18.5-27.07
N-pentane
Alkanes
11
13.8-20.5
N-tetradecane
Alkanes
13
17.87-27.43
N-tricosane
Alkanes
4.8374-27.809
N-tridecane
Alkanes
18
17.27-27.73
N-undecane
Alkanes
12
15.65-25.56
Propane
Alkanes
20
0.2-20.6
1-decene
Alkenes
11
16.81-25.97
T range
(K)
313.15393.15
283.15393.15
303.15393.15
233.15503.15
293.15393.15
253.15313.15
283.15393.15
320.65672.43
273.15393.15
283.15393.15
193.15366.48
273.15-
Tc(K)
Radius of
ryration (A)
Critical
density
(molcm-3)
758.00
7.98
0.0008
594.60
4.81
0.0018
747.00
7.66
0.0008
568.70
4.55
0.0021
708.00
6.73
0.0010
469.70
3.34
0.0032
693.00
6.43
0.0011
6.27
796.00
9.07
0.0006
5.02
675.00
6.20
0.0012
7.21
639.00
5.50
0.0015
8.80
369.83
2.43
0.0050
616.60
5.23
0.0017
Tb (K)
@T
b
2.64
3.25
5.58
14
name
family
N
Points
range
(mNm-1)
T range
(K)
373.15
154.12353.15
133.39333.15
171.45373.15
283.15373.15
273.15373.15
193.15253.1
Tc(K)
Radius of
ryration (A)
Critical
density
(molcm-3)
537.40
4.08
0.0025
504.00
3.66
0.0029
566.90
4.46
0.0022
708.00
6.64
0.0011
5.77
691.00
6.40
0.0012
6.50
364.85
2.23
0.0054
Tb (K)
@T
b
1-heptene
Alkenes
19
14.35-36.735
1-hexene
Alkenes
12
14.4-38.182
1-octene
Alkenes
14
14.1-36.024
1-pentadecene
Alkenes
11
20.37-28.17
1-tetradecene
Alkenes
11
19.77-28.56
Propene
Alkenes
12
12.73-21.41
Aromatic
s
33
21.3-32.28
273.1373.15
630.30
3.84
0.0027
Aromatic
s
22
19.8-30.88
273.1373.15
617.00
3.94
0.0027
Aromatic
s
19
19.6-28.98
288.15373.15
616.20
3.83
0.0026
Aromatic
11
4.5689-26.792
488.93-
873.00
4.98
0.0018
1,2dimethylbenze
ne
1,3dimethylbenze
ne
1,4dimethylbenze
ne
Antracene
15
name
family
s
Aromatic
Benzene
s
Aromatic
Ethylbenzene
s
Aromatic
Naphthalene
s
Aromatic
Toluene
s
Cycloalk
Cycloepentane
anes
Cycloalk
Cycloheptane
anes
Cycloalk
Cyclohexane
anes
Cycloalk
Cyclohexene
enes
1,2-epoxy-2Epoxides
methylpropane
1,2Epoxides
epoxybutane
1,2-propylene
Epoxides
Oxyde
N
Points
range
(mNm-1)
17
19.8-30.21
22
20.-31.3
21
18.7-32.88
22
19.-30.92
15
18.22-28.4
18
16.4-29.8
11
16.9-26.43
38
20.09-33.55
10
1.918-48.832
11
3.5224-45.363
12
4.5024-45.182
T range
(K)
785.7
283.1363.1
273.1373.15
363.15491.14
273.1373.15
253.15323.15
273.15393.15
283.15363.15
239.15349.15
130.468.9
143.88473.4
161.22434.02
Tc(K)
Radius of
ryration (A)
Critical
density
(molcm-3)
562.05
3.00
0.0039
617.15
3.90
0.0027
748.40
4.05
0.0025
591.75
3.47
0.0032
511.70
2.85
0.0038
604.20
3.56
0.0028
553.80
3.24
0.0032
560.40
3.16
0.0034
521.00
2.96
0.0041
526.00
2.96
0.0039
482.25
2.66
0.0054
Tb (K)
@T
b
16
name
family
N
Points
range
(mNm-1)
T range
(K)
Tc(K)
Radius of
ryration (A)
Critical
density
(molcm-3)
1,3-dioxane
Epoxides
11
3.9716-44.216
228.15531.
590.00
2.97
0.0042
1,3-propylene
Oxyde
Epoxides
11
4.305-27.85
298.-468. 520.00
2.23
0.0053
1,4-dioxane
Epoxides
10
22.32-34.45
587.00
3.02
0.0042
Epoxides
11
4.2252-42.174
524.00
2.60
0.0049
Epoxides
11
4.2779-46.232
542.00
2.58
0.0049
Epoxides
16
4.1114139.773
549.20
3.03
0.0040
Epoxides
11
4.436-35.55
290.-620. 697.00
4.05
0.0025
Epoxides
16
3.8115939.128
713.70
4.49
0.0022
Dibenzofuran
Epoxides
8.0113-33.543
824.00
4.71
0.0020
Ethylene
Oxyde
Epoxides
26
24.3-36.4
469.15
1.94
0.0071
Furan
Epoxides
32
16.-38.
490.15
2.56
0.0046
2,3dihydrofuran
2,5dihydrofuran
2-methyl-1,3dioxolane
2methylbenzofu
ran
3,5dimethylbenzo
furan
288.15373.15
176.471.6
171.487.8
212.1494.28
257.7642.33
355.31664.34
221.15293.15
193.15-
Tb (K)
@T
b
17
name
family
Tetrahydrofura
Epoxides
n
Tetrahydropyr
Epoxides
an
N
Points
range
(mNm-1)
17
20.3-29.7
18
4.2047536.093
Trioxane
Epoxides
12
6.626-35.93
3-methylbutyl
Ethanoate
Esters
18.11-24.96
Butyl Acetate
Esters
14
16.87-26.48
Butyl
Propanoate
Esters
13
17.44-26.38
Ethyl Acetate
Esters
18
14.68-25.13
Esters
25
14.7-26.6
Esters
11
17.38-25.55
Esters
24
14.9-24.7
Esters
10
18.77-26.32
Ethyl Nbutyrate
Ethyl
Propionate
Isobutyl
Ethanoate
Methyl
Propionate
T range
(K)
353.15
273.15348.15
224.514.84
334.65534.65
292.15361.05
283.15373.15
283.15373.15
283.15373.15
273.15392.15
283.15353.15
278.15383.15
283.15343.15
Tc(K)
Radius of
ryration (A)
Critical
density
(molcm-3)
540.15
2.69
0.0045
572.04
3.10
0.0038
604.00
2.82
0.0045
586.10
4.66
0.0023
575.40
4.28
0.0026
594.60
4.74
0.0023
523.30
3.47
0.0035
571.00
4.30
0.0025
546.00
3.93
0.0029
560.80
4.28
0.0026
530.60
3.49
0.0035
Tb (K)
@T
b
18
name
family
N
Points
range
(mNm-1)
N-pentyl
Formate
Esters
13
17.86-27.07
Pentyl Acetate
Esters
11
18.91-26.36
Propyl Acetate
Esters
15
15.4-25.48
Propyl
Formate
Esters
18.94-25.65
Butoxyethene
Ethers
12
3.522-36.09
Ethers
13.6-18.39
Ethers
17
10.56-27.77
Ethers
21.92-27.31
Ethoxyethane
Ethers
30
3.28-34.87
Propoxypropa
ne
Ethers
17
3.84-37.22
2-butanone
Ketones
21.16-23.96
2-hexanone
Ketones
19
18.4-26.54
Di-iso-propyl
Ether
Di-n-butyl
Ether
Di-n-octyl
Ether
T range
(K)
283.15373.15
288.15361.05
283.15373.15
283.15343.15
181.15481.15
287.25333.15
238.03432.65
298.15363.15
156.85421.65
149.95469.95
298.15323.15
288.15-
Tc(K)
Radius of
ryration (A)
Critical
density
(molcm-3)
576.00
4.18
0.0026
599.90
4.77
0.0023
549.73
3.97
0.0029
538.00
3.36
0.0035
536.00
3.81
0.0027
500.05
3.91
0.0026
584.10
4.80
0.0021
707.00
7.36
0.0011
466.70
3.18
0.0036
530.60
3.91
0.0026
535.50
3.14
0.0037
587.61
4.09
0.0026
Tb (K)
@T
b
5.62
19
name
family
N
Points
range
(mNm-1)
2-octanone
Ketones
16
2.9935-30.842
2-pentanone
Ketones
31
17.66-25.3
3-heptanone
Ketones
11
19.55-26.55
3-pentanone
Ketones
33
4.22-26.9
4-heptanone
Ketones
11
21.93-25.47
6-undecanone
Ketones
21.43-27.4
Acetone
Ketones
20
16.2-26.21
1,2ethanedithiol
Benzyl
Mercaptan
Cyclohexyl
Mercaptan
Ethyl
Mercaptan
Mercapta
ns
Mercapta
ns
Mercapta
ns
Mercapta
ns
11
4.857-48.48
14
3.6716-46.325
10
3.5506-44.411
16.32-25.59
T range
(K)
363.15
252.85569.43
288.15363.15
289.25360.35
273.15500.
297.95333.15
292.55360.85
273.15353.15
231.95591.95
243.95646.2
189.64597.6
284.65350.75
Tc(K)
Radius of
ryration (A)
Critical
density
(molcm-3)
632.70
4.89
0.0020
561.08
3.62
0.0033
606.60
4.54
0.0023
560.95
3.58
0.0030
602.00
4.57
0.0023
678.50
5.72
0.0014
508.20
2.75
0.0048
663.00
2.98
0.0048
718.00
4.09
0.0027
664.00
3.83
0.0028
499.15
2.36
0.0048
Tb (K)
@T
b
8.92
20
name
family
Isobutyl
Mercaptan
Isopropyl
Mercaptan
Methyl
Mercaptan
N-butyl
Mercaptan
N-decyl
Mercaptan
N-dodecyl
Mercaptan
N-heptyl
Mercaptan
N-hexyl
Mercaptan
N-nonyl
Mercaptan
N-octyl
Mercaptan
N-pentyl
Mercaptan
N-propyl
Mercapta
ns
Mercapta
ns
Mercapta
ns
Mercapta
ns
Mercapta
ns
Mercapta
ns
Mercapta
ns
Mercapta
ns
Mercapta
ns
Mercapta
ns
Mercapta
ns
Mercapta
N
Points
range
(mNm-1)
14
17.87-24.52
10
18.39-22.5
20.73-26.44
14
18.93-26.36
11
4.6697-33.332
11
3.4938-32.715
16
3.6026-34.483
25
4.2611-38.273
11
4.123-33.08
24
4.0005-35.399
16
19.51-27.26
12
19.75-25.47
T range
(K)
288.15353.15
288.15323.15
282.95316.65
288.15353.15
247.56617.56
265.15651.6
229.92579.92
192.62552.62
253.05603.05
223.95593.95
284.05363.15
288.15-
Tc(K)
Radius of
ryration (A)
Critical
density
(molcm-3)
559.00
3.46
0.0033
517.00
2.99
0.0039
469.95
1.61
0.0069
570.10
3.40
0.0033
696.00
5.71
0.0016
724.00
6.37
0.0014
645.00
4.64
0.0022
623.00
4.25
0.0024
681.00
5.34
0.0018
667.30
5.01
0.0019
598.00
3.80
0.0028
536.60
2.88
0.0039
Tb (K)
@T
b
21
name
family
Mercaptan
Phenyl
Mercaptan
Sec-butyl
Mercaptan
Tert-butyl
Mercaptan
Tert-dodecyl
Mercaptan
Tert-nonyl
Mercaptan
Tert-octyl
Mercaptan
Undecyl
Mercaptan
Allyl-tertbutylperoxide
Benzoyl
Peroxide
Cumene
Hydroperoxyd
e
Cyclohexyl
ns
Mercapta
ns
Mercapta
ns
Mercapta
ns
Mercapta
ns
Mercapta
ns
Mercapta
ns
Mercapta
ns
Peroxyde
s
Peroxyde
s
N
Points
range
(mNm-1)
Tc(K)
Radius of
ryration (A)
Critical
density
(molcm-3)
689.00
3.61
0.0032
554.00
3.36
0.0033
530.00
3.19
0.0033
691.00
6.38
0.0013
659.00
5.49
0.0017
199.-559. 627.00
4.43
0.0019
710.00
6.07
0.0015
563.00
4.35
0.0022
378.-788. 884.00
6.34
0.0014
T range
(K)
333.15
288.15363.15
133.02493.02
274.26474.26
274.26621.9
230.593.1
Tb (K)
@T
b
12
30.59-39.61
12
3.75-47.14
16
2.915-23.
10
2.3763-21.885
11
3.6937-39.828
11
3.335-33.899
11
4.6258-31.583
15
12.247-28.483
11
4.68-35.03
Peroxyde
s
12
12.64-27.87
264.26434.26
605.00
4.53
0.0024
Peroxyde
11
4.441-38.41
253.15-
685.00
3.86
0.0028
270.15630.15
213.27390.
7.66
22
name
family
Hydroperoide
Dicumyl
Peroxide
Di-t-butyl
Peroxide
Ethylbenzene
Hydroperoxide
Mdiisopropylben
zene
Hydroperoxide
Nbutylbenzeneh
ydroperoxyde
Nbutylhydroper
oxide
Npropylbenzene
Hydroperoxyd
e
Pdiisolpropylbe
s
Peroxyde
s
Peroxyde
s
Peroxyde
s
N
Points
range
(mNm-1)
Tc(K)
Radius of
ryration (A)
Critical
density
(molcm-3)
884.00
6.54
0.0011
547.00
4.53
0.0020
737.00
4.49
0.0024
264.-724. 812.00
5.49
0.0015
T range
(K)
613.15
311.15661.15
233.15483.15
290.663.3
Tb (K)
@T
b
11
12.09-38.32
11
3.319-22.45
10
4.3664-37.832
Peroxyde
s
11
3.277-29.99
Peroxyde
s
15
4.39-42.
266.7686.25
762.50
5.53
0.0019
Peroxyde
s
11
3.9114-36.706
213.508.5
565.00
3.52
0.0034
Peroxyde
s
15
3.95-37.1
259.4672.12
746.80
5.09
0.0021
Peroxyde
s
6.3371-31.489
315.67677.33
810.00
5.38
0.0015
8.24
9.79
23
name
nzene
Hydroperoxide
P-menthane
Hydroperoxide
T-butyl
Hydroperoxyd
e
2,3,5trimethylthiop
hene
2,3dimethylbenzo
thiophene
2,5dimethylthiop
hene
2,7dimethylbenzo
thiophene
2butylbenzothio
phene
2-
Radius of
ryration (A)
Critical
density
(molcm-3)
4.60
0.0018
552.00
3.30
0.0034
215.591.12
656.80
3.92
0.0025
4.128-42.12
282.15707.4
786.00
4.49
0.0022
23.45-30.57
293.15358.15
629.90
3.50
0.0030
Sulphide
s
16
4.8308340.463
341.5695.16
772.40
4.52
0.0021
Sulphide
s
15
3.337-21.6
425.65736.83
818.70
5.36
0.0016
15
4.3841-43.01
774.00
4.56
0.0022
family
N
Points
range
(mNm-1)
Peroxyde
s
11
4.87-36.17
Peroxyde
s
11
3.4978-23.336
277.45496.8
Sulphide
s
15
3.7363-39.158
Sulphide
s
15
Sulphide
s
Sulphide
T range
(K)
Tc(K)
250.-630. 711.00
282.35-
Tb (K)
@T
b
24
name
family
ethylbenzothio
phene
2-ethyltetrahydrothio
phene
2ethylthiophene
2-menthyl
Benzothiophen
e
2methylthiacycl
opentane
2-methyl
Thiaindan
2methylthiophe
ne
2-npropylthiophe
ne
2-pentanehiol
N
Points
range
(mNm-1)
T range
(K)
Tc(K)
Radius of
ryration (A)
Critical
density
(molcm-3)
Tb (K)
@T
b
696.6
Sulphide
s
18
3.965-44.556
201.2592.02
657.80
3.74
0.0026
Sulphide
s
23.45-31.1
293.15358.15
626.60
3.59
0.0030
Sulphide
s
16
3.955-36.485
324.65691.83
768.70
4.19
0.0022
Sulphide
s
28.61-30.63
293.15313.15
628.40
3.35
0.0030
Sulphide
s
15
4.0597-42.239
273.3709.74
788.60
4.24
0.0022
Sulphide
s
22.67-31.7
209.79327.07
609.00
3.30
0.0036
Sulphide
s
23.12-29.67
293.15358.15
649.20
4.06
0.0026
Sulphide
s
15
3.68521-43.16
160.8525.9
584.30
3.87
0.0026
25
name
2propylbenzothi
ophene
3,4dimethylthiop
hene
3,5dimethylbenzo
thiophene
3-ethyl-2,5dimethylthiop
hene
3ethylbenzothio
phene
3methylthiophe
ne
4methyldibenzo
thiophene
Benzothiophen
e
family
N
Points
range
(mNm-1)
T range
(K)
Tc(K)
Radius of
ryration (A)
Critical
density
(molcm-3)
Sulphide
s
15
2.9792-21.044
387.2726.75
807.50
4.90
0.0017
Sulphide
s
16
3.68913-39.67
216.578.52
642.80
3.64
0.0029
Sulphide
s
16
4.2001243.415
265.19706.32
784.80
4.66
0.0021
Sulphide
s
15
4.0682-45.198
195.602.1
669.00
4.30
0.0024
Sulphide
s
16
4.9235452.651
252.2698.85
776.50
4.68
0.0021
Sulphide
s
10
24.51-32.62
293.15358.15
613.00
3.33
0.0036
Sulphide
s
15
4.5879-45.194
339.15813.69
904.10
5.06
0.0018
Sulphide
s
16
3.4054247.584
304.5687.6
764.00
3.87
0.0026
Tb (K)
@T
b
26
name
family
Carbon
Disulfide
Dibenzothioph
ene
Dicyclohexyl
Sulfide
Diethyl
Disulfide
Diethyl
Sulfide
Diisopentylsul
fide
Diisopropyl
Sulfide
Dimethyl
Disulfide
Dimethyl
Sulfide
Di-n-butyl
Sulfide
Di-n-octyl
Sulfide
Di-n-propyl
Sulphide
s
Sulphide
s
Sulphide
s
Sulphide
s
Sulphide
s
Sulphide
s
Sulphide
s
Sulphide
s
Sulphide
s
Sulphide
s
Sulphide
s
Sulphide
N
Points
range
(mNm-1)
16
26.45-42.47
11
4.0321-35.664
15
3.1638-36.267
30
4.299-45.39
14
17.3-26.5
17.64-26.6
16
3.09-33.78
16
3.9269-49.19
11
19.87-26.5
12
20.94-27.4
15
7.2851-31.004
13
3.4218-43.3
T range
(K)
230.75333.15
371.82807.3
283.05721.8
171.63571.63
283.15363.15
288.35394.95
195.07526.68
188.44553.5
284.25330.85
291.45360.95
285.604.2
187.66-
Tc(K)
Radius of
ryration (A)
Critical
density
(molcm-3)
552.00
1.57
0.0063
897.00
4.94
0.0020
802.00
5.41
0.0014
642.00
3.62
0.0028
557.15
3.21
0.0031
664.40
5.78
0.0015
585.20
3.98
0.0024
615.00
2.94
0.0040
503.04
2.37
0.0050
652.00
4.89
0.0019
760.00
7.39
0.0011
675.00
4.41
0.0022
Tb (K)
@T
b
9.32
6.41
27
name
Disulfide
Di-n-propyl
Sulfide
Di-tert-butyl
Disulfide
Di-tert-butyl
Sulfide
family
s
Sulphide
s
Sulphide
s
Sulphide
s
Sulphide
Dyallil Sulfide
s
Ethyl Methyl Sulphide
Disulfide
s
Ethyl N-octyl Sulphide
Sulfide
s
Ethyl N-pentyl Sulphide
Sulfide
s
Ethyl Propyl Sulphide
Disulfide
s
Ethyl Propyl Sulphide
Sulfide
s
Ethyl T-butyl Sulphide
Sulfide
s
Ethyl TertSulphide
butyl Disulfide
s
N
Points
range
(mNm-1)
15
6.7657-40.131
21.22-27.33
15
4.34-32.
15
3.6941-41.96
16
16
16
16
16
4.1712146.6678
3.0105733.942
3.517-37.39
4.0250847.492
3.4458541.135
12
16.85-23.54
16
3.72-40.391
Tc(K)
Radius of
ryration (A)
Critical
density
(molcm-3)
608.00
4.09
0.0024
689.70
4.69
0.0019
623.20
4.32
0.0019
653.53
3.94
0.0026
628.90
3.41
0.0032
700.00
5.53
0.0014
633.20
4.41
0.0022
165.-594. 660.00
4.30
0.0024
583.90
3.59
0.0027
588.00
3.92
0.0024
659.90
4.26
0.0021
T range
(K)
607.5
170.44505.34
290.65361.05
264.15560.88
190.15588.18
183.22566.01
241.3630.
210.569.88
156.12525.51
293.15363.15
206.44593.91
Tb (K)
@T
b
28
name
family
N
Points
range
(mNm-1)
T range
(K)
Tc(K)
Radius of
ryration (A)
Critical
density
(molcm-3)
Methyl Ethyl
Sulfides
Methyl
Isopropyl
Sulfide
Methyl Nbutyl Sulfide
Methyl Npropyl Sulfide
Methyl
Penthyl
Sulfide
Methyl Phenyl
Sulfide
Methyl Tbutyl Sulfide
Methyl Tpentyl Sulfide
Tetrahydrothio
phene
Thiacycloprop
ane
Sulphide
s
13
3.7437-42.848
167.23479.7
533.00
2.85
0.0039
Sulphide
s
23.1-24.2
293.15303.15
553.10
3.23
0.0030
18.84-26.4
593.00
3.82
0.0028
13
3.3757-43.383
565.00
3.37
0.0033
17
3.3862939.161
611.90
4.15
0.0023
32.14-40.31
712.80
3.90
0.0026
13
3.1947-35.039
570.00
3.51
0.0028
10
3.1615-36.794
632.00
3.99
0.0023
23
25.-38.4
631.95
2.90
0.0040
11
4.5977-55.155
547.00
2.06
0.0060
16
4.4611-47.778
759.80
3.95
0.0024
Thiaindan
Sulphide
s
Sulphide
s
Sulphide
s
Sulphide
s
Sulphide
s
Sulphide
s
Sulphide
s
Sulphide
s
Sulphide
293.15364.45
160.17508.5
179.15550.71
293.15359.55
190.84513.
195.568.8
273.15373.15
164.15492.3
269.92-
Tb (K)
@T
b
29
name
Thiophene
Trimethylene
Sulfide
family
s
Sulphide
s
Sulphide
s
N
Points
range
(mNm-1)
32
21.-34.
14
5.2691-49.487
T range
(K)
683.82
273.15373.15
199.91541.8
Tc(K)
Radius of
ryration (A)
Critical
density
(molcm-3)
579.35
2.78
0.0046
602.00
2.54
0.0046
Tb (K)
@T
b
30
Figure 1.5-Figure 1.7 show the relationship between the selected physical properties and
the surface tension at boiling point. From Figure 1.6 to Figure 1.7, a general trend is
clearly evident for surface tension at the boiling point when plotted against the selected
parameters. In particular, surface tension at boiling point was found to be decreasing with
radius of gyration with an evident common trend for all the fluids. Surface tension at
boiling point was found to be decreasing with critical temperature and increasing with
critical density, but in these cases points were found to be rather scattered. From the general
analysis, three alkanes, namely n-heneicosane, n-docosane and n-tricosane, were found not
to exactly follow the common trends of other alkanes. In addition, ethyl mercaptan was
found to be always out of common trends: this was probably due to its very high boiling
point temperature, that was found to be out of the range of the experimental temperatures.
The following equation was finally proposed:
Figure 1.5 Surface tension at the boiling point vs. critical density
31
Figure 1.6 Surface tension at the boiling point vs. radius of gyration
Figure 1.7 Surface tension at the boiling point vs. critical temperature
(1.17)
where 0 is a scaled factor = (kTc)/Gr2, Tr is the reduced temperature, k is Boltzmanns
constant, is an adimensional term = NacGr31012, Gr is the radius of gyration (m), Na is
Avogadros number (mol-1), c is the critical density (molm-3) and 1012 is a scaling factor.
In order to deeper analyze the cross-correlations between the selected parameters, in Figure
1.8-Figure 1.10 the critical density versus radius of gyration, the critical temperature versus
radius of gyration and the critical density versus critical temperature are plotted,
respectively.
32
33
Figure 1.8 a noticeable cross-correlation between radius of gyration and critical density for
all the families, suggesting that one of the parameters could be substituted by a correlation,
giving the possibility to remove a physical parameter from Equation (1.17). The following
very simple correlation between critical density and radius of gyration was considered:
c = e 0.4152Gr - 4.2808
(1.18)
The respective AAD% values were reported in Table 1.4. In spite of its simplicity (reduced
temperature and radius of gyration are the only two physical parameters needed), the
proposed equation is generally giving the lowest deviations. To keep the proposed equation
as general and simple as possible, we decided not to regress the coefficients for each family
separately.
The residuals produced by Equation (1.17) + (1.18) are reported in Figure 1.9. The
prediction of surface tension of organics is generally good, clearly increasing with the
reduced temperature. Excluding few fluids that generally show very low experimental
surface tension data (namely 1,2-ethanedithiol, m-diisopropylbenzene hydroperoxide, npropylbenzene hydroperoxide, dibenzothiophene, tert-dodecyl mercaptan, 2-
34
Inorganic Compounds
In order to understand if the equation proposed for organics was extendable for inorganic
compounds, the surface tension data for inorganics were collected, again from the ones
accepted by DIPPR database [2]. A total of 2032 surface tension data were finally analyzed
for the following families named as in the DIPPR database: elements, silanes, halides,
gases, acids, bases and other inorganics.
In Figure 1.12 the surface tension data for inorganics divided by families are shown. Since
the collected data for the family of elements are clearly out of commons order of
magnitudes for surface tension (up to 1450 mNm-1) and reduced temperatures (up to 5764
K), they were not reported in Figure 1.12. The experimental ranges of collected data are
summarized in Table 1.6 From Figure 1.12, it is clear that a common trend for inorganics,
even neglecting the elements, is not easily evident as it was for organics. In particular, there
are many compounds which behavior is out of trend. For this reason, Equation (1.17) was
separately regressed for the different families.
Table 1.6. Summary of the collected data for inorganic compounds.
Family
Acids (inorg.)
Bases
Elements
Gases
N. of
Fluids
7
1
47
8
N. of
Points
79
22
516
93
range
(mNm-1)
1.26-66.816
0.2-43.39
0.-1450
0.06-36.32
T range
(K)
158.97-1276.2
198.15-403.15
1.13-5764
81.15-472.01
35
Family
Halides
Other Inorganics
Silanes
N. of
N. of
range
Fluids Points
(mNm-1)
18
297
1.5-97.01
4
193
0.07-79.77
55
832
1.6253-75.
T range
(K)
121.85-1280.
272.74-652.74
88.5-810.27
Figure 1.12. Scatter plot of the collected surface tension data of inorganics vs. (1-Tr).
A good prediction was obtained for the following families: acids, bases and gases. In
particular for inorganic acids (hydrogen bromide, hydrogen chloride, hydrogen iodide,
sulfaminic acid and sulfuric acid) an AAD% of 10.71% was obtained with the following
coefficients: A = 0.00038, B = 1.34635, C = 0.06488 and D = 23.05590. Poorer results were
obtained only for hydrogen fluoride and (123.37%) nitric acid (52.37%). For inorganic
gases (carbon dioxide, nitric oxide, nitrogen dioxide, nitrogen tetroxide, nitrous oxide,
sulfur dioxide, sulfur trioxide) an AAD% of 6.97% was obtained with the following
coefficients: A = 20.09860, B = 1.20986, C = 0.27195 and D = 9.81645. All gases showed
deviations below 10%. To avoid any division by zero, gases and elements with a null radius
of gyration (namely Helium-3, Helium-4, Argon, Neon, Krypton, Xenon, Sulfur,
phosphorus white, germanium, bismuth, iron, lithium) were not considered during the
calculations. Very good results were obtained for the unique inorganic base available
(ammonia), with an AAD% of 1.90%.
Other inorganics (deuterium oxide, hydrogen peroxide, tin(iv) chloride and water) were
also well predicted (4.64%) with the following coefficients: A = 7.48842, B = 1.08465, C =
0.27053 and D = 11.87120. Deviations for water were found to be 10.23%.
Bad results were obtained, as expected, with elements (17.78%), halides (32.78%) and
silanes (24.78%). However, for these families, the same kind of deviations were obtained
with the equations available from the literature [15-18].
36
As a final step, in order to give reliable coefficients for a selection of relevant compounds,
9 inorganics were selected and regressed separately. Results are given in Table 1.7, where
the coefficients for Equation (1.17), not considering equation (1.18), and the AAD% for
each fluid are also given.
Table 1.7.Deviations for inorganic compounds and coefficients for Equation (1.17)
Name
Sulfuric Acid
Ammonia
Nitrous Oxide
Carbon Dioxide
Sulfur Dioxide
Sulfur Trioxide
Nitric Oxide
Water
Deuterium Oxide
NPoints
5
22
11
25
16
13
5
48
113
AAD%
0.99
0.75
0.66
0.83
0.56
0.29
0.22
0.53
0.63
a
0.35032
1.21393
1.18465
1.28280
1.18333
1.21787
1.55173
1.15617
0.90714
b
-2.16443
-1.09690
-1.30609
-1.33738
-1.91077
-3.04139
-1.24242
-0.90070
-1.05708
c
0.91502
0.89715
0.79176
0.84276
0.91009
0.85797
0.82278
0.91476
0.90652
d
0.06566
1.89821
2.45910
4.11487
0.62518
0.80687
0.06844
7.52403
0.61183
37
summation of the outputs of the neurons in the hidden layer. This summation on the output
is called the output layer.
The required number of training data points and hidden layer neurons are the main goals of
ANN modelling. They can be determined by the constructive approach [29]. It is worth
noting that a network that has only one hidden layer is able to approximate almost any type
of nonlinear mapping [30].
Each neuron performs a weighted summation of the inputs, which then passes through the
nonlinear activation function s, also called the neuron function.
Mathematically, the functionality of a hidden neuron is described by:
j =1
w j x j + b j
(1.20)
where the weights wj, bj are symbolised with the arrows feeding into the neuron.
The number of output neurons equals the number of outputs of the approximation problem.
The output of this network is given by:
nh
n
(1.21)
where n is the number of input signals to the jth neuron, and nh is the number of neurons in
the hidden layer. The variables are the parameters of the network model that are represented
collectively by the parameter vector . In general, the neural network model will be
represented by the compact notation g(,x). The size of the input and output layers are
defined by the number of inputs and outputs of the network and, therefore, only the number
of hidden neurons has to be specified when the network is defined.
In training the network, its parameters are adjusted incrementally until the training data
satisfy the desired mapping; that is, until matches the desired output y as closely as
possible up to a maximum number of iterations.
The activation function can be: the hyperbolic tangent function, the sigmoid function, the
inverse tangent function, the saturated linear function. For this feedforward network, the
sigmoid function was employed as the activation function as follows:
s( x ) =
1
1 + e x
(1.22)
Moreover, keeping the number of hidden layers fixed, it is possible to modulate the
network by changing the number of hidden neurons in each layer. In this way, we tested
each time the network with a different number of neurons trying to limit the error produced
by the model. Too few neurons in the hidden layer prevent the network from better fitting
the experimental values and approximate points maintaining a low deviation. On the other
hand, too many neurons lead to a network that has a very good response when it is tested
38
with the training points, but the structure of this network is too rigid and too adapting to
those values of database provided.
Often it is more convenient to use the RMSE when evaluating the quality of a model during
and after training, because it can be compared with the output signal directly. Different
neural networks were compared adopting the Levenberg-Marquardt algorithm [28] using
their root mean square errors (RMSE), defined as follows:
RMSE =
1
N
(y
i =1
g (, xi ))
(1.23)
39
Hidden Layer
Input Layer
.
.
.
.
Figure 1.13 Schematic diagram of the ANN model. Inputs: Tc (K), critical temperature; Gr (), radius
of gyration; T (K), temperature; c (molcm-3), critical density. Output: (mN/m), surface tension
40
Figure 1.14. Absolute average percent deviation between data collected and ANN results vs. number
of neurons in the hidden layer
Table 1.8.The parameters of the hidden layer and of the output layer of the ANN
hidden layer
Neurons
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
bias
-13.0
3.7
-4.7
22.7
-1.4
30.1
-3.7
-3.9
-1.6
14.1
-5.7
-28.9
-22.7
19.0
8.8
-15.1
2.5
24.6
-10.9
-4.6
-13.8
6.2
Tc
6.3
10.4
7.6
-23.2
-15.7
-9.4
12.1
5.2
21.3
26.4
0.3
33.4
24.7
-11.3
-2.6
7.5
4.7
-25.6
18.1
-14.1
1.7
-9.6
Pc
11.4
-18.7
6.5
3.5
20.1
-12.7
-14.1
7.8
-23.4
-74.3
2.2
12.9
9.9
-16.7
-22.5
-0.3
-20.0
5.1
-16.5
21.5
6.6
6.0
Gr
-1.2
0.1
-0.6
-0.5
0.0
-33.8
1.8
-0.6
-0.2
1.4
-0.4
-2.9
-4.7
2.0
3.3
0.3
27.9
-0.6
0.0
-0.1
9.9
-0.3
35.0
-17.0
-10.7
-25.2
-4.3
-9.4
-17.3
-8.7
14.1
-24.7
7.6
1.1
16.2
-43.8
12.2
58.4
-13.4
-27.7
10.8
6.1
8.6
-4.1
output layer
output neurons
bias -19.7141
5.2
3.7
7.3
-16.1
22.4
0.0
2.7
-12.6
6.5
0.4
24.4
7.8
-0.5
2.3
-0.3
-0.3
0.0
15.0
4.7
-15.8
0.5
10.6
41
23
24
25
26
27
28
29
30
31
32
33
34
-17.1
25.4
-30.8
-24.4
48.9
-26.8
25.7
26.8
13.6
-10.0
6.0
-9.5
3.7
-43.7
34.7
35.4
-55.2
23.6
1.0
-43.9
-2.3
6.8
-1.6
4.2
7.5
13.6
14.5
5.8
-7.8
1.7
-6.3
13.0
-23.5
-8.2
-17.3
7.3
11.4
0.5
-3.0
-1.2
-0.8
-1.4
-4.3
0.3
-14.2
-0.5
1.0
0.2
6.9
28.5
3.4
-41.3
0.2
5.1
-26.1
22.5
-11.6
14.1
0.6
-21.1
-0.3
-9.9
-6.8
-0.5
-0.3
7.5
3.8
10.3
0.1
-12.0
-5.6
-1.6
AAD%
Entire dataset
Training Set
Validation Set
Test Set
4.06
3.43
5.11
6.39
RMSE
(mNm-1)
0.0093
0.0092
0.0092
0.0106
Sum of Residuals
(mNm-1)
0.000012
-0.000043
-0.000149
0.000713
Because of the discussed cross-correlation between reduced density and radius of gyration,
a configuration with a one-hidden-layer feedforward network without critical density as an
input parameter was also tried, but results in terms of AAD% for the entire dataset were a
little poorer (AAD%=4.6%) and this configuration was finally neglected.
Figure 1.15 shows the correlation between the predictions of the validated, trained and
tested MLP and the corresponding experimental surface tension data. From the figure a
linear fit is evident, witnessing a good correlation between predictions and data.
42
Figure 1.15. Correlation between the predictions of the validated, trained and tested MLP and the
corresponding experimental surface tension data.
Figure 1.16 illustrates the effect of the reduced temperature on the network accuracy.
According to the MLP configuration selected, with the exception of few points, residuals
are well within 0.05 mNm-1.
To test the goodness of Equation (1.17) along with equations in the literature, deviations
between the surface tension data collected and calculated are reported in Table 1.4 for each
family.
43
As summarised in Table 1.4, the equation proposed gave the best results in terms of
deviations when compared to the equations existing in literature.
Different ANN approaches were adopted for surface tension prediction in literature [31-35].
The use of artificial neural networks was coupled with the group contribution method to
determine surface tension of pure compounds [31]. The compounds investigated belong to
78 chemical families containing 151 functional groups. A good compromise was reached
by considering one hidden layer and 10 neurons. The method gave noticeable results, with
an overall AAD%=1.7%. In other papers, the parachor was determined [32-33]. A larger set
of data for surface tension was considered adopting a quantitative structureproperty
relationship strategy [34]. In this case, the required parameters of the model are temperature
and the numbers of 19 molecular descriptors in each molecule investigated. A three layer
feedforward artificial neural network was finally optimised. In another paper [35], a wide
variety of materials such as alkanes, alkenes, aromatics, and sulfur, chlorine, fluorine, and
nitrogen containing compounds were used in an ANN with one hidden layer of 20 neurons.
The AAD% obtained for 1048 data points related to 82 compounds is 1.57%. It is worth
noting that here [35], the ANN adopted about half of our data and an additional input
parameter (a total of five): critical pressure, acentric factor, reduced temperature, reduced
normal boiling temperature and specific gravity at normal boiling point. In addition, about
70% of data was employed as training data, while the remaining 30% was applied for
network verification.
At the end of this part, a new scaled equation and a multi-layer perceptron neural network
are proposed to predict surface tension of organics and 9 inorganics. The equation is very
simple, containing only reduced temperature and radius of gyration as parameters. The
multilayer perceptron proposed adopts as input parameters the same parameters proposed
for the scaled equation and has one hidden layer with 34 neurons, determined according to
the constructive approach. The model was validated, trained and tested for a wide set of
data.
1.5.2
In this part is presented a new formula for the surface tension calculation of ketones.
As a first step, an analysis of the available data of the experimental surface tension data for
ketones was performed. The experimental data were collected for the following pure fluids:
acetone, 2-butanone, 2-pentanone, 3-pentanone, 2-hexanone, 3-heptanone, 4-heptanone, 2octanone, 6-undecanone. Then, the experimental data were regressed with the most reliable
semi-empirical correlating methods based on the corresponding states theory existing in the
literature.
The finally proposed equation is very simple and gives a noticeable improvement with
respect to the existing equations
Ketones are organic compounds with a carbonyl group (C=O) bonded to two other carbon
atoms. Many ketones are known and they are of great importance in industry (acetone) and
in biology.
44
To provide a reliable database of surface tensions experimental data for ketones, as a first, a
literature review was performed. The list of analyzed ketones is summarized in Table 1.10,
together with the physical properties adopted for the calculation.
From Table 1.10, it is possible to observe that the measurements were performed over a
fairly large temperature (from 252.85 to 499.99 K) and surface tension (from 2.99 to 30.84
mN/m) ranges. After the data selection, a total of 154 reliable experimental points were
considered in the regressions.
The main goal of this study was to find a new equation that minimizes the deviations
between the predicted data and the experimental data for ketones. For this purpose, we
based the new equation on the well known corresponding states principle, according to the
Van der Waals theory.
Table 1.10: Studied ketones and their adopted physical properties
Compound
Acetone
2-butanone
2-pentanone
3-pentanone
2-hexanone
3-heptanone
4-heptanone
2-octanone
6-Undecanone
T Range
[K]
273.15353.15
298.14323.15
288.15363.15
273.15499.99
288.14363.15
289.25360.35
297.95333.15
252.85569.43
292.55360.85
Zc
Radius of
Gyration [m]
Critical
Density
[mol/cm
3]
508.2
0.233
2.7510-10
0.00478
535.5
0.249
3.1410-10
0.00374
17.66-25.3
31
561.08
0.238
3.6210-10
0.00332
4.22-26.9
33
560.95
0.269
3.5810-10
0.00297
18.4-26.54
19
587.61
0.254
4.0910-10
0.00264
19.55-26.55
11
606.6
0.251
4.5410-10
0.0023
21.93-25.47
11
602
0.253
4.5710-10
0.0023
2.99-30.84
16
632.7
0.249
4.8910-10
0.00201
21.43-27.4
678.5
0.252
5.7210-10
0.00144
Range
[mN/m]
N. of
data
Tc [K]
16.2-26.21
20
21.16-23.96
45
family of ketones. Then, to analyze the linear correlation between all the variables and the
surface tension, a correlation matrix was created, as reported Table 1.11. From the table,
the strong inverse correlation between the reduced temperature and the surface tension was
confirmed. All others variables showed a rather low linear correlation.
20
10
0
10
15
20
25
30
25
20
15
10
5
0
30
40
50
30
25
20
15
10
5
0
30
r 2 = 0.925351
2- butanone
20
2- hexanone
2- octanone
2- pentanone
25
3- heptanone
15
3- pentanone
10
4- heptanone
6- undecanone
acetone
0.5
0.6
0.7
0.8
0.9
Figure 1.18. Scatter Plot of experimental surface tension data versus reduced temperature. Confidence
bands at 1standard deviation, 2standard deviation and 3standard deviation
Table 1.11.Correlation matrix between all the variables. : surface tension, Tc: critical temperature,
Pc: critical pressure, : acentric factor, Tb: temperature at boiling point, c: critical density, M:
molecular weight, Zc: compressibility factor, Ttp: temperature at triple point, Ptp: pressure at triple
point, Gr: radius of gyration
Tc
Pc
Tr
Tb
Zc
Ttp
Ptp
Gr
1.00
0.02
-0.03
0.04
-0.96
0.04
-0.03
0.04
0.03
0.01
-0.04
0.03
Tc
0.02
1.00
-0.99
0.95
-0.27
0.99
-0.95
0.99
0.28
0.88
-0.16
0.99
Pc
-0.03
-0.99
1.00
-0.91
0.29
-0.97
0.97
-0.97
-0.30
-0.86
0.15
-0.99
0.04
0.95
-0.91
1.00
-0.28
0.98
-0.84
0.97
0.17
0.845
-0.22
0.95
46
Tr
-0.96
-0.27
0.29
-0.28
1.000
-0.29
0.27
-0.29
-0.07
-0.21
0.10
-0.29
Tb
0.043
0.992
0.982
-0.29
1.000
-0.91
0.999
0.233
0.87
-0.20
0.99
-0.03
-0.95
0.975
0.973
-0.84
0.279
-0.91
1.000
-0.50
-0.90
-0.03
-0.94
0.045
0.992
-0.97
0.979
-0.29
0.999
-0.92
0.927
1.000
0.25
0.88
-0.17
0.99
Zc
0.030
0.287
-0.30
0.172
-0.07
0.233
-0.50
0.254
1.00
0.63
0.82
0.25
Ttp
0.010
0.880
-0.86
0.845
-0.21
0.872
-0.90
0.881
0.63
1.00
0.29
0.87
Ptp
-0.04
-0.16
0.158
-0.22
0.108
-0.20
-0.03
0.82
0.29
1.00
-0.19
Gr
0.039
0.996
-0.99
0.952
-0.29
0.992
-0.94
0.175
0.992
0.25
0.87
-0.19
1.00
47
Equations
i
i
exp
calc
i =1
Eq. 7
Eq. 4
Eq. 6
Eq. 8
Eq. 15
-23.55
-162.20
-308.47
-300.87
0.15
i
i
exp
|
| calc
i
i
1 n calc exp
i 100
n i =1
exp
66.37
162.20
308.47
300.98
48.62
-0.75
-4.63
-8.37
-8.01
-0.05
i =1
1
n
i
i
| calc
exp
|
i =1
i
exp
100
2.14
4.63
8.37
8.03
1.57
Compound
i =1
Acetone
2-butanone
2pentanone
3pentanone
2hexanone
3heptanone
4heptanone
2-octanone
6Undecano
ne
i
calc
i
exp
|
i =1
i
calc
i
|
exp
i
i
i
i
1 n calc exp
1 n | calc exp |
100
100
i
i
exp
n i =1
n i =1
exp
-1.25
0.98
1.25
8.76
0.98
12.14
0.00
0.03
0.02
0.09
0.03
0.08
-2.71
6.83
0.03
0.07
3.80
3.80
0.03
0.03
1.35
1.35
0.02
0.02
-1.28
1.28
-0.02
0.02
-7.45
5.46
8.02
5.46
-0.52
0.12
0.53
0.12
48
49
1.5.3
This part presents a new formula to calculate the surface tension of alcohols.
As a first step, an analysis of the data available on the surface tension of alcohols was
made. A total of 1643 data were collected for n-alcohols, aromatic alcohols, cycloaliphatic
alcohols, 2-alkanols and methyl alkanols. The data were then regressed with the most
reliable semi-empirical correlation methods in the literature based on the corresponding
states theory. The scaled equation proposed is very simple and gives noticeable
improvement with respect to existing equations.
The same physical parameters considered in the scaled equation were also adopted as input
parameters in a multi-layer perceptron neural network, to predict the surface tension of
alcohols. The multilayer perceptron proposed has one hidden layer with 29 neurons,
determined according to the constructive approach. The model developed was trained,
validated and tested for the set of data collected, showing that the accuracy of the neural
network model is higher than that of the methods proposed in the literature.
An alcohol is an organic compound in which the hydroxyl functional group (-OH) is bound
to a carbon atom. Due to the presence of the hydroxyl functional group, alcohols can
hydrogen bond. In addition, because of the difference in electronegativity between oxygen
and carbon, they are polar, and therefore hydrophilic, showing the ability of dissolving both
polar and non-polar substances. For this reason, between the several uses of alcohols, of
particular interest are the fields of solvents and reagents. Other important uses are in fuels,
antifreezes, preservatives and beverages. In organic synthesis, alcohols serve as versatile
intermediates.
Surface tension is an important fluid property for the study of industrial applications. In
particular, it plays a fundamental role in the study of phase transitions and technical
processes such as boiling and condensation, absorption and distillation, microscale channel
flow processes and detergents. Liquidvapour interfaces are also important for particular
naturally occurring phenomena, such as the performance of biological membranes.
The raw surface tension data (experimental, smoothed and predicted) available from DIPPR
database [2] were analysed for 94 alcohols. During the data collection, a fluid by fluid
analysis was performed and data showing deviations higher than three times the standard
deviation were rejected. All the fluids containing the alcohol functional group were
considered and divided into the following subfamilies: n-alcohols, aromatic alcohols
(phenols), cycloaliphatic alcohols and other aliphatic alcohols (2-alkanols and methyl
alkanols). Although phenols, whose hydroxyl group is attached to an unsaturated aromatic
ring system, are more reactive than alcohols and act more like acids, they were still
considered in the data treatment.
Despite there being no surface tension calculation/estimation methods specifically oriented
to alcohols, several theories that have been proposed to date can be applied to describe the
surface tension of alcohols. It is necessary to emphasise that, because of association, the
equations for the surface tension calculation of polar molecules, such as alcohols or acids,
often pose serious problems in terms of predictability.
50
In Figure 1.21 the box-plot for the outliers identification and the histograms for the relative
frequency of the data collected are reported. All the data lying outside the defined bounds
can be considered as outliers. From Figure 1.21, it is evident that 47 outliers for n-alcohols
(almost all outside the lower fence), 32 for other aliphatic (both outside the higher and
lower fence), 2 for cycloaliphatic (outside the higher fence) and 11 for aromatic alcohols
(outside the higher fence) were detected. A descriptive statistics referring to distribution
properties such as location, dispersion, and shape of the data is also reported in Table 1.14
Figure 1.21: Data Distribution of surface tension of alcohols
700
600
500
Aromatic
Cycloaliphatic
n- Alcohols
OtherAliphatic
400
300
200
100
0
20
40
60
80
51
Table 1.14. Summary of the collected data and calculated surface tension at the boiling point, @Tb =mT+q.
Alcohol
1-methyl-3-hydroxy-5isopropyl Benzene
1-methyl-3-hydroxy-6isopropyl Benzene
Family
Aromatic
Aromatic
N
Points
15
15
Aromatic
1-phenyl-1-propanol
18
Aromatic
1-phenyl-2-propanol
11
Aromatic
2,3-xylenol
11
Aromatic
2,4-xylenol
13
Aromatic
2,5-xylenol
7
Aromatic
2,6-di-tert-butyl-p-cresol
11
Aromatic
2,6-xylenol
11
Aromatic
2-phenyl-1-propanol
11
Aromatic
2-phenyl-2-propanol*
11
Aromatic
2-phenylethanol
24
range
T range
(mNm-1) (K)
3.6189725.183
3.562419.85
26.6234.93
4.599236.769
4.47928.41
17.231.23
19.7229.92
5.33528.152
4.00528.59
4.432642.363
9.83461.97
6.470646.94
321.65654.12
384.15650.25
283.15373.15
285.610.2
345.71645.71
313.15473.15
353.15473.15
344.-644.
318.76628.76
257.616.5
309.15589.15
246.15615.6
@Tb
Tb
AAD
%
Tc
Gr
(K)
()
(molcm-3)
(mNK- (mNm1m-1)
1)
(mNm
(K)
-1)
514.15
-0.065
45.485
12.18
2.43
511.15
-0.061
43.132
11.79
1.56
677
4.434 0.00227
492.15
-0.092
61.050
15.58
0.13
678
4.457 0.00227
493.15
-0.098
63.743
15.27
3.69
490.07
-0.079
55.597
16.67
0.84
484.13
-0.087
58.495
16.18
0.19
484.33
-0.085
59.938
18.77
0.00
724
5.682 0.00132
541.15
-0.076
53.980
13.08
0.35
474.22
-0.079
53.397
15.92
1.46
685
4.483 0.00227
497.65
-0.105
67.726
15.62
5.39
660
4.29
0.00227
475.15
-0.185
118.28
1.48
684
4.279 0.00258
492.05
-0.106
71.869
30.32
19.65
6
1.35
52
Alcohol
Family
N
Points
Aromatic
3,4-xylenol
16
Aromatic
3,5-xylenol
16
Aromatic
3-phenyl-1-propanol
12
Aromatic
4-hydroxystyrene
Alpha-methylbenzyl
Alcohol
11
Aromatic
24
Aromatic
Benzyl Alcohol
50
Aromatic
Dinonylphenol*
10
Aromatic
M-cresol
25
Aromatic
M-ethylphenol
11
Aromatic
M-tolualcohol
11
Aromatic
Nonylphenol
11
range
O-cresol
25
Aromatic
Gr
(mNm-1) (K)
(K)
()
(molcm-3)
(mNK- (mNm1m-1)
1)
(mNm
(K)
-1)
17.5529.02
17.9628.42
5.064247.852
3.488425.644
4.6141.61
4.6143.33
15.66387.566
21.3738.01
4.17440.499
3.500728.454
5.228835.489
348.15473.15
347.15473.15
255.15624.6
346.65662.4
293.15629.1
257.85648.14
729.9
5
4.156 0.00286
715.6
5
4.318 0.00208
500.15
-0.091
60.663
15.08
0.06
494.89
-0.082
56.553
16.09
0.27
350.-746.
288.15453.15
269.15644.8
252.612.9
279.15679.15
313.15533.15
269.84632.7
O-ethylphenol
11
13.-36.7
3.963842.256
@Tb
Tb
AAD
%
Tc
Aromatic
T range
694
4.737 0.00227
508.15
-0.115
75.634
17.05
5.11
736
4.103 0.00267
502.00
-0.070
49.479
14.35
2.01
699
4.079 0.00256
720.1
3.8
0.00262
5
477.15
-0.110
72.605
20.24
3.39
478.60
-0.100
20.59
2.78
902
11.82 0.00081
705.8
5
3.87 0.00321
716.4
5
4.278 0.00258
735.00
-0.181
68.246
146.87
6
14.04
5.82
475.43
-0.096
65.229
19.49
1.85
491.57
-0.096
65.083
17.80
4.86
681
0.00258
490.15
-0.068
45.055
11.49
2.64
770
6.929 0.00132
697.5
5
3.787 0.00355
590.76
-0.075
55.484
11.14
3.17
464.15
-0.102
67.777
20.24
1.33
703
477.67
-0.105
68.648
18.55
7.60
4.23
4.043 0.00285
53
Alcohol
Family
N
Points
Aromatic
O-tolualcohol
11
Aromatic
P-cresol
53
Aromatic
P-cumylphenol
11
Aromatic
P-ethylphenol
11
Aromatic
Phenol
16
Aromatic
P-isopropenyl Phenol*
15
Aromatic
P-tert-amylphenol*
11
Aromatic
P-tert-butylphenol
11
Aromatic
P-tert-octylphenol
11
Aromatic
P-tolualcohol
11
Aromatic
Thymol
1,4cyclohexanedimethanol
1-methylcyclohexanol
23
Cycloaliph
atic
Cycloaliph
atic
11
11
range
T range
(mNm-1) (K)
4.986440.067
309.15621.9
303.1514.-37.44 533.15
4.4478346.340.961
750.6
4.827318.2334.74
638.23
323.1523.6-38.2 453.15
10.17356.6580.25
668.43
10.782366.-666.
66.361
371.564.6-27.41 651.56
5.12358.5534.12
678.55
26.32343.1534.09
433.15
273.1517.9-34.2 484.15
5.0512320.6535.795
651.6
3.7215299.1532.893
617.4
Gr
(K)
()
(molcm-3)
(mNK- (mNm1m-1)
1)
(mNm
(K)
-1)
691
4.107 0.00258
704.6
5
3.762 0.00361
497.15
-0.112
73.290
17.83
4.31
475.13
-0.094
64.637
19.74
0.71
834
5.739 0.00142
716.4
4.13 0.00267
5
694.2
3.415 0.00437
5
608.15
-0.090
70.702
16.08
4.90
491.14
-0.093
63.656
17.94
2.41
454.99
-0.108
23.66
0.77
513.70
-0.225
42.27
4.57
752
4.913 0.00183
535.65
-0.185
72.845
157.63
1
132.60
7
33.69
2.35
734
4.614 0.00203
512.88
-0.081
57.174
15.45
2.00
765
5.703 0.00142
563.60
-0.090
65.554
14.69
3.37
681
4.148 0.00258
698.2
5
4.679 0.00229
490.15
-0.088
64.345
21.12
0.36
505.65
-0.078
55.147
15.71
1.50
724
4.639 0.00225
556.15
-0.092
64.296
13.07
3.48
686
3.67
441.15
-0.092
59.609
19.10
3.84
0.00267
@Tb
Tb
AAD
%
Tc
54
Alcohol
Beta-cholesterol*
Cis-2methylcyclohezanol
Cis-3methylcyclohexanol
Cis-4methylcyclohexanol
Cyclohexanol
L-menthol
Sitosterol*
Stigmasterol*
Trans-2methylcyclohexanol
Trans-3methylcyclohexanol
Trans-4methylcyclohexanol
1-butanol
1-decanol
Family
Cycloaliph
atic
Cycloaliph
atic
Cycloaliph
atic
Cycloaliph
atic
Cycloaliph
atic
Cycloaliph
atic
Cycloaliph
atic
Cycloaliph
atic
Cycloaliph
atic
Cycloaliph
atic
Cycloaliph
atic
n-Alcohol
n-Alcohol
N
Points
11
14
12
16
12
11
11
15
14
12
16
38
21
range
T range
(mNm-1) (K)
5.20347.189
5.080233.835
5.955534.26
3.993431.625
25.6734.23
4.75424.66
6.135958.764
8.458554.122
5.333334.637
5.047931.718
4.435731.773
14.226.28
22.429.61
421.65863.1
280.15552.6
267.65547.65
263.95559.8
289.35373.15
315.65585.65
413.15857.7
443.15830.57
269.15555.3
272.65564.3
265.559.8
273.15413.15
283.15373.15
Gr
(K)
()
(molcm-3)
(mNK- (mNm1m-1)
1)
(mNm
(K)
-1)
959
3.852 0.00074
771.00
-0.095
84.817
11.91
7.47
614
3.821 0.00267
438.15
-0.104
62.403
16.80
1.45
625
3.931 0.00267
446.15
-0.100
60.218
15.76
1.57
622
3.869 0.00267
444.15
-0.094
55.869
14.25
3.28
434.00
-0.104
64.243
19.22
0.80
658
4.754 0.00188
489.55
-0.073
11.80
0.33
953
8.288 0.00068
778.00
-0.118
12.73
8.67
0.00069
778.00
-0.117
47.639
104.28
3
103.66
6
12.46
5.28
617
3.824 0.00267
440.15
-0.101
61.370
16.99
0.81
627
3.937 0.00267
447.15
-0.091
56.290
15.74
0.42
622
3.854 0.00267
444.15
-0.092
55.488
14.82
1.74
391.90
-0.085
49.568
16.37
1.21
688
503.00
-0.076
50.910
12.90
0.76
953.6 8.23
5.499 0.00155
@Tb
Tb
AAD
%
Tc
55
Alcohol
Family
N
Points
n-Alcohol
1-dodecanol
18
n-Alcohol
1-eicosanol
1-heptadecanol
8
n-Alcohol
n-Alcohol
1-heptanol
10
17
n-Alcohol
1-hexadecanol
10
n-Alcohol
1-hexanol
25
1-nonadecanol
n-Alcohol
10
1-nonanol
n-Alcohol
n-Alcohol
22
1-octadecanol
11
n-Alcohol
1-octanol
39
n-Alcohol
1-pentadecanol
1-pentanol
1-propanol
10
n-Alcohol
n-Alcohol
12
17
range
T range
(mNm-1) (K)
23.829.75
6.927128.767
2.876129.588
293.15373.15
338.55641.53
327.05702.
273.1520.2-28.6 373.15
3.168322.3529.331
693.
18.23273.1528.08
388.15
2.8791334.8529.071
719.1
273.1521.7-29.8 373.15
5.0965331.0529.3
668.78
273.1521.1-29.1 373.15
3.256317.0530.033
683.1
18.8283.1526.67
373.15
18.27283.1524.48
353.15
@Tb
Tb
AAD
%
Tc
Gr
(K)
()
(molcm-3)
(mNK- (mNm1m-1)
1)
(mNm
(K)
-1)
537.10
-0.076
52.166
11.19
0.21
645.50
-0.072
51.982
5.681
4.34
779.2 7.59
0.00086
611.30
-0.071
50.984
7.818
10.13
632.3 4.38
0.00225
448.60
-0.087
52.648
13.63
0.70
768.6 7.31
0.00093
598.00
-0.070
50.563
8.668
7.47
429.90
-0.083
50.798
14.90
0.39
635.10
-0.068
50.018
7.116
10.04
485.20
-0.078
51.218
13.13
0.60
789.3 7.93
0.00081
623.60
-0.072
52.220
7.045
6.20
467.10
-0.080
50.956
13.58
0.22
583.90
-0.073
51.723
9.309
7.22
410.90
-0.086
51.067
15.55
0.21
370.35
-0.083
48.191
17.38
0.70
56
Alcohol
Family
N
Points
n-Alcohol
1-tetradecanol
8
n-Alcohol
1-tridecanol
1-undecanol
11
n-Alcohol
n-Alcohol
2-butanol
2-heptanol
11
8
n-Alcohol
n-Alcohol
2-hexanol
12
27
n-Alcohol
2-methyl-2-butanol
7
n-Alcohol
2-pentanol
2-propanol
5
n-Alcohol
n-Alcohol
3-butanol
12
6
n-Alcohol
3-heptanol
20
n-Alcohol
Ethanol
Methanol
48
n-Alcohol
44
range
T range
(mNm-1) (K)
7.891129.94
3.462930.779
310.65591.93
293.15660.6
293.1522.9-29.2 373.15
22.14288.1523.89
303.15
273.1519.-27.2 373.15
14.15273.1526.5
413.15
18.51288.1523.22
338.15
22.96288.1524.42
303.15
20.46288.1521.79
303.15
16.36298.1519.96
338.15
3.68291- 203.1537.4885 544.86
273.0971.1-24.4 503.052
273.10.34-24.5 508.15
Gr
(K)
()
(molcm-3)
(mNK- (mNm1m-1)
1)
(mNm
(K)
-1)
0.00107
569.00
-0.078
53.366
8.955
2.98
553.40
-0.076
52.473
10.62
5.27
520.30
-0.079
52.265
11.28
0.10
372.90
-0.088
49.151
16.22
1.38
608.3 4.51
0.00224
432.90
-0.082
49.642
14.12
0.15
585.3 4.09
0.00260
412.40
-0.087
50.357
14.45
0.79
543.7 3.42
0.00309
375.20
-0.095
50.570
15.00
0.14
3.655 0.00307
392.20
-0.083
48.477
15.77
0.81
0.00450
355.30
-0.085
46.378
16.09
0.87
355.57
-0.086
45.363
14.89
0.56
429.15
-0.098
55.999
13.83
5.00
351.39
-0.100
51.933
16.96
4.56
337.63
-0.100
52.208
18.42
12.83
745.3 6.73
561
508.3 2.76
@Tb
Tb
AAD
%
Tc
57
Alcohol
2,2-dimethyl-1-propanol
2,6-dimethyl-4-heptanol
2-butanol
2-buthyl-1-nonanol
2-buthyl-1-octanol
2-butyl-1-decanol
2-heptanol
2-hethyl-1-butanol
2-hethyl-1-hexanol
2-hexanol
2-methyl-1-butanol
2-methyl-1-dodecanol
2-methyl-1-hexanol
Family
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
N
Points
15
11
8
15
16
15
12
15
10
27
17
15
11
range
T range
(mNm-1) (K)
2.88514.94
4.18635.3
22.1423.89
4.14733.476
3.823338.935
4.019533.062
19.-27.2
15.2925.9
9.25336.01
14.1526.5
4.751535.41
4.080734.43
4.434335.623
328.15497.43
208.-538.
288.15303.15
271.2613.35
193.15600.57
277.7636.93
273.15373.15
278.15418.15
203.15499.15
273.15413.15
195.517.86
271.614.25
220.535.5
Gr
(K)
()
(molcm-3)
(mNK- (mNm1m-1)
1)
(mNm
(K)
-1)
386.25
-0.071
38.298
10.75
0.32
603
4.915 0.00186
451.00
-0.093
53.615
11.46
4.22
372.90
-0.088
49.151
16.22
1.38
538.00
-0.085
55.278
9.461
4.91
520.00
-0.085
53.488
9.109
6.67
563.00
-0.080
54.031
8.818
5.21
608.3 4.51
0.00224
432.90
-0.082
49.642
14.12
0.15
420.15
-0.073
46.007
15.49
0.75
457.75
-0.090
53.339
12.28
2.17
585.3 4.09
0.00260
412.40
-0.087
50.357
14.45
0.79
401.85
-0.094
53.314
15.52
0.73
538.80
-0.088
56.874
9.656
4.74
595
437.15
-0.098
56.403
13.52
2.56
4.796 0.00231
@Tb
Tb
AAD
%
Tc
58
Alcohol
2-methyl-1-pentanol
2-methyl-1-propanol
2-methyl-1-tridecanol
2-methyl-1-undecanol
2-methyl-2-butanol
2-methyl-2-propanol
2-nonanol
2-octanol
2-pentanol
3-heptanol
3-hethyl-1-heptanol
3-hexanol
3-methyl-1-butanol
Family
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
N
Points
40
15
30
16
7
6
24
29
5
20
14
15
53
range
T range
(mNm-1) (K)
273.15408.15
283.15373.15
277.7636.93
264.3599.31
288.15338.15
298.15338.15
273.1520.5-28.6 373.15
19.8273.1527.14
373.15
22.96288.1524.42
303.15
3.68291- 203.1537.4885 544.86
1.97007- 240.136.4491 608.37
14.14278.1525.74
409.15
273.1515.1-25.3 403.15
15.79-27.
16.5823.73
3.866533.062
4.017332.115
18.5123.22
16.3619.96
@Tb
Tb
AAD
%
Tc
Gr
(K)
()
(molcm-3)
(mNK- (mNm1m-1)
1)
(mNm
(K)
-1)
421.15
-0.082
49.325
14.86
0.15
380.81
-0.080
46.433
15.94
0.11
707.7 6.96
0.00117
563.10
-0.081
54.126
8.764
5.04
521.00
-0.084
53.433
9.521
4.91
543.7 3.42
0.00309
375.20
-0.095
50.570
15.00
0.14
355.57
-0.086
45.363
14.89
0.56
471.70
-0.081
50.830
12.49
0.08
452.90
-0.079
49.180
13.62
0.47
561
3.655 0.00307
392.20
-0.083
48.477
15.77
0.81
429.15
-0.098
55.999
13.83
5.00
636.7 5.28
0.00186
480.15
-0.093
57.255
12.44
9.74
406.15
-0.088
50.164
14.54
0.43
404.15
-0.076
46.106
15.24
1.56
59
Alcohol
3-methyl-1-pentanol
3-methyl-2-butanol
3-methyl-3-pentanol
3-pentanol
4-methyl-1-octanol
4-methyl-2-pentanol
4-methyl-cyclohexanemethanol
5-methyl-1-hexanol
6-methyl-1-octanol
Allyl Alcohol
Alpha-terpineol
Beta-terpineol
Isopropanol
Family
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
Other
Aliphatic
N
Points
52
24
60
17
40
71
10
22
40
16
16
16
43
range
T range
(mNm-1) (K)
15.0526.28
0.430835.254
14.8725.4
7.880536.706
4.057634.763
13.6524.67
2.6285428.89
3.10926.12
4.121336.382
278.15423.15
188.546.1
273.15388.15
204.15447.68
240.1568.53
273.15405.15
303.15582.84
293.15543.15
240.1570.06
273.1519.2-27.6 368.15
2.8892293.1520.975
607.5
3.0126291.1521.18
598.68
16.59273.1522.9
353.15
Gr
(K)
()
(molcm-3)
(mNK- (mNm1m-1)
1)
(mNm
(K)
-1)
588
4.18
0.00263
425.55
-0.077
47.725
15.06
0.52
383.88
-0.097
52.358
15.03
24.80
394.06
-0.089
49.748
14.73
0.46
388.45
-0.117
59.272
13.92
2.32
631.7 5.31
0.00172
477.90
-0.093
55.764
11.49
4.08
404.85
-0.082
47.185
13.83
0.13
463.09
-0.096
57.777
13.18
13.16
605
445.15
-0.092
51.876
11.13
6.02
0.00186
479.15
-0.097
58.270
11.84
4.48
370.23
-0.089
51.972
18.85
0.27
675
4.653 0.00197
492.95
-0.057
37.254
8.93
3.26
483.15
-0.059
37.898
9.38
2.74
508.3 2.76
355.30
-0.079
44.479
16.45
0.43
4.563 0.00231
633.4 5.28
0.00450
@Tb
Tb
AAD
%
Tc
60
Alcohol
Propargyl Alcohol
Family
Other
Aliphatic
N
Points
5
range
T range
Tc
Gr
(mNm-1) (K)
(K)
()
(molcm-3)
(mNK- (mNm1m-1)
1)
(mNm
(K)
-1)
30.9736.05
386.65
74.404
293.15333.15
-0.130
@Tb
Tb
24.00
AAD
%
0.27
61
Starting from the consideration that experimental data can generally be well described by
adopting a linear dependence on a function of (1-Tr), in Figure 1.22 a scatter plot of
surface tension data of alcohols divided by family versus (1-Tr) is presented; it shows a
rather positive correlation for all fluids, which underlines the importance of the reduced
temperature in calculating the surface tension, as confirmed by Equazione (1.16)
Figure 1.22. Data on the surface tension of alcohols vs. reduced temperature. The marked/selected
symbols are the fluids not considered during the data regression, as in Table 1.14
From Figure 1.22, it is also evident that three cycloaliphatics (namely sitosterol,
stigmasterol and beta-cholesterol) and four aromatics (namely 2-phenyl-2-propanol,
dinonylphenol, P-isopropenylphenol and P-tert-amylphenol) are not following the common
trend. In particular the three cycloaliphatics were found to have a much higher critical
temperature value than the other fluids. The same thing was noticed for dinonylphenol,
while the other aromatic alcohols (2-phenyl-2-propanol, P-isopropenylphenol and P-tertamylphenol) were found to have higher surface tension values than the other fluids, as it
can be clearly noticed also from Table 1.14.
Besides the experimental temperature and the critical temperature, we searched for some
specific additional fluid properties that could be involved in the model to improve the
representation of data for all the alcohols, which represent various chemical structures.
According to equation (1.5), and to the classical parachor approach, the critical density was
selected as a parameter. Furthermore, according to equation (1.24) the radius of gyration
was also considered.
62
From the general definition, the radius of gyration can be calculated as the root mean
square distance of the objects' parts from either a given axis or its center of gravity. In
particular, it can be referred to as the radial distance from a given axis at which the mass of
a body could be concentrated without altering the rotational inertia of the body about that
axis. For a planar distribution of mass rotating about some axis in the plane of the mass, the
radius of gyration can be considered as the equivalent distance of the mass from the axis of
rotation. Radius of gyration plays an important role in polymers chemistry: it is well known
that for a polymer in a poor solvent, the radius of gyration is smaller than in a good solvent
such as alcohols are. In fact, the radius of gyration is usually a better estimate of the chain
dimensions than the root-mean-squared end-to-end distance, as it accounts for the
hydrodynamic interactions between the polymer chains and the solvent. The end-to-end
distance is difficult to measure, while the radius of gyration can be measured by a light
scattering technique.
It is worthwhile to investigate this matter in order to provide a more thorough explanation
as to how surface tension for alcohols is related to the selected parameters, namely critical
temperature, critical density and radius of gyration. Since the boiling point was generally
found to be within the range of the experimental temperatures, as witnessed in Table 1.14,
a linear regression of surface tension against the reduced temperature of each fluid was
made; then the surface tension at the boiling point was calculated for each fluid; lastly, the
surface tension at the calculated boiling point was plotted for the three parameters
considered. The regression coefficients m and q are reported in Table 1.14 along with the
range of data collected for surface tension and temperatures, the number of points, and the
calculated surface tension at the boiling point for each alcohol.
From Figure 1.23 to Figure 1.25 show the relationship between the selected physical
properties and the surface tension at the boiling point. From Figure 1.23 to Figure 1.25, a
general trend is clearly evident for surface tension at the boiling point when plotted against
the selected parameters. In particular, surface tension at the boiling point was found to
decrease with critical temperature and radius of gyration and increase with critical density:
a common trend was evident for all the alcohols, excluding the already pointed out seven
fluids: sitosterol, stigmasterol, beta-cholesterol, 2-phenyl-2-propanol, dinonylphenol, Pisopropenylphenol and P-tert-amylphenol. Such fluids were marked in Figures 8-10. Since
these seven fluids presented a very limited number of data (a total of 84 points), we decided
not to consider these data in the regression. Aromatic alcohols show slightly different
behaviour when surface tension at boiling point is plotted versus critical temperature, radius
of gyration and critical density (From Figure 1.23 to Figure 1.25)
63
64
Starting from equation (1.24) and not considering the term containing the critical
compressibility factor, the following equation for alcohols was finally proposed:
(1.25)
where 0 is a scaled factor = (kTc)/Gr2, Tr is the reduced temperature, k is Boltzmanns
constant, is an nondimensional term = NacGr31012, Gr is the radius of gyration (m), Na
is Avogadros number (mol-1), c is the critical density (molm-3) and 1012 is a scaling
factor. To optimize the coefficients, the Levenberg-Marquardt curve-tting method was
adopted [28]. This method is actually a combination of two minimization methods: the
gradient descent method and the Gauss-Newton method. The following coefficients were
found after regression: A=0.13823, B=1.16570, C=0. 144755, D=16.5905.
65
66
In order to further analyze the cross-correlations between the selected parameters, from
Figure 1.26 to Figure 1.28 the radius of gyration versus critical temperature, the critical
density versus critical temperature and the critical density versus radius of gyration are
plotted, respectively. Again, aromatic alcohols exhibit slightly different behaviour in
Figure 1.26 and Figure 1.27.
Figure 1.28 shows a noticeable cross-correlation between radius of gyration and critical
density for all the families including aromatics, suggesting that one of the parameters could
be substituted by a correlation, giving the possibility to remove a physical parameter from
equation (1.25). The following correlation between critical density and radius of gyration
was considered:
(1.26)
Substituting equation (1.26) into equation (1.25), again adopting the Levenberg-Marquardt
curve-tting method [28], the following new values for the coefficients were found:
A=0.151271, B=1.17126, C=0.148516, D= 16.5579.
In order to compare the equations, deviations were calculated as follows:
The respective AAD values are reported in Table 1.15, while deviations for each fluid are
reported in Table 1.16. Deviations are generally rather similar for n-alcohols,
cycloaliphatic and other aliphatics, while aromatic alcohol prediction clearly improves with
the correlation found in equation (1.26). Comparing equation (1.25) and equation (1.25)
with equation (1.26) it has to be considered that the equation obtained has one physical
67
parameter less, but four additional coefficients. In both cases, to keep the proposed
equations as general and simple as possible, we decided not to regress the coefficients for
each sub-family separately.
Table 1.15.Summary of deviations for equations
AAD%
AAD% AAD% AAD% AAD% AAD%
Eq
Eq
Eq
Eq
Eq
Eq.
N. of
(1.25)+Eq AAD%
Family
(1.6)
(1.8)
(1.10) (1.25)
Points (1.9)
(1.26)
ANN
Cycloaliphatic 129
17.2
26.1
35.1
19.9
9.4
8.0
3.5
Other
487
11.4
27.4
35.1
29.2
5.2
5.8
1.5
Aliphatic
n-Alcohol
528
20.3
26.9
32.6
23.2
15.0
9.7
2.6
Aromatic
893
17.4
28.8
38.3
24.9
7.2
8.4
2.1
Mean
16.7
27.8
35.9
25.1
8.9
8.1
2.1
The residuals produced by equation (1.25) are reported in Figure 1.29. The prediction of
surface tension of alcohols is good, especially at high reduced temperatures. The behaviour
of n-alcohols, cycloaliphatic and other aliphatic is generally well described by the equation
(within 3 mNm-1), while aromatic alcohols are, as expected, poorly predicted at low
reduced temperatures (up to 10 mNm-1). However, prediction of aromatic alcohols
drastically improves with the increase of the reduced temperature.
Figure 1.29. Deviations of equation Eq (1.25) +Eq (1.26) as function of reduced temperature
68
Alcohol
1-methyl-3-hydroxy5-isopropyl Benzene
1-methyl-3-hydroxy6-isopropyl Benzene
1-phenyl-1-propanol
1-phenyl-2-propanol
2,3-xylenol
2,4-xylenol
2,5-xylenol
2,6-di-tert-butyl-pcresol
2,6-xylenol
2-phenyl-1-propanol
2-phenylethanol
3,4-xylenol
3,5-xylenol
3-phenyl-1-propanol
4-hydroxystyrene
AAD
AAD
AAD%
AAD%
%
%
Eq
Eq.
Eq
Eq
(1.8)
(1.25)
(1.6)
(1.10)
AAD%
Eq
(1.25)+Eq
(1.26)
AAD
%
ANN
Deviations
exp- calc
(mNm1)
Eq
(1.25)+Eq
(1.26)
Deviatio
ns
expcalc
(mNm1)
ANN
Family
N
Point
s
AAD
%
Eq
(1.9)
Aromatic alcohols
15
38.78
50.61
58.93
40.91
28.68
25.40
4.29
-3.68
-0.08
Aromatic alcohols
15
47.83
52.47
60.65
41.82
25.73
29.78
4.63
-3.56
0.03
Aromatic alcohols
Aromatic alcohols
Aromatic alcohols
Aromatic alcohols
Aromatic alcohols
18
11
11
13
7
11.52
18.14
17.94
7.02
10.26
31.53
25.17
48.24
40.24
39.40
39.99
33.24
54.85
46.10
45.74
22.05
15.63
34.70
28.41
26.93
7.65
8.99
19.11
9.12
8.58
6.71
9.11
12.12
8.55
2.84
0.37
3.92
3.27
3.83
0.35
2.07
1.82
-2.06
-2.18
0.60
0.02
0.00
0.49
0.18
-0.06
Aromatic alcohols
11
12.49
8.31
6.61
11.85
22.78
9.19
4.50
1.18
0.20
Aromatic alcohols
Aromatic alcohols
Aromatic alcohols
Aromatic alcohols
Aromatic alcohols
Aromatic alcohols
Aromatic alcohols
11
11
24
16
16
12
11
19.70
20.49
21.78
4.69
7.60
20.62
33.64
35.83
23.08
15.36
63.94
20.06
12.09
62.88
41.53
33.23
22.28
71.93
25.14
21.70
69.54
26.16
15.95
10.76
50.44
11.01
7.07
53.65
12.78
9.09
16.49
28.03
6.54
13.10
36.05
15.14
10.04
21.72
13.55
7.40
21.39
31.36
8.81
4.53
3.04
1.41
0.87
1.51
3.24
-2.45
2.51
6.47
-3.28
-1.82
5.86
-4.39
-0.05
0.70
0.07
-0.03
0.04
0.05
-0.21
69
Alcohol
Alpha-methylbenzyl
Alcohol
Benzyl Alcohol
M-cresol
M-ethylphenol
M-tolualcohol
Nonylphenol
O-cresol
O-ethylphenol
O-tolualcohol
P-cresol
P-cumylphenol
P-ethylphenol
Phenol
P-tert-butylphenol
P-tert-octylphenol
P-tolualcohol
Thymol
1,4cyclohexanedimethano
AAD
AAD
AAD%
AAD%
%
%
Eq
Eq.
Eq
Eq
(1.8)
(1.25)
(1.6)
(1.10)
AAD%
Eq
(1.25)+Eq
(1.26)
AAD
%
ANN
Deviations
exp- calc
(mNm1)
Eq
(1.25)+Eq
(1.26)
Deviatio
ns
expcalc
(mNm1)
ANN
Family
N
Point
s
AAD
%
Eq
(1.9)
Aromatic alcohols
24
25.27
3.25
1.80
2.38
12.76
10.96
4.43
3.30
0.34
Aromatic alcohols
Aromatic alcohols
Aromatic alcohols
Aromatic alcohols
Aromatic alcohols
Aromatic alcohols
Aromatic alcohols
Aromatic alcohols
Aromatic alcohols
Aromatic alcohols
Aromatic alcohols
Aromatic alcohols
Aromatic alcohols
Aromatic alcohols
Aromatic alcohols
Aromatic alcohols
Cycloaliphatic
alcohols
50
25
11
11
11
25
11
11
53
11
11
16
11
11
11
23
26.87
20.64
20.74
45.52
21.60
17.99
23.66
18.57
14.85
20.75
17.28
28.06
18.33
17.36
21.27
4.26
9.03
21.16
26.91
85.01
11.55
22.20
21.27
19.82
33.49
5.07
24.25
28.41
15.17
11.29
16.58
32.14
11.48
25.20
32.43
98.62
23.89
26.63
26.66
27.93
39.05
12.75
29.70
33.24
20.63
5.58
24.78
41.36
3.69
25.88
19.04
74.52
12.24
27.37
20.95
13.52
38.04
4.81
19.48
33.93
11.12
11.34
10.33
29.53
10.29
12.10
7.49
35.32
10.05
16.92
10.04
13.65
20.53
13.72
5.67
16.94
1.28
17.55
18.07
6.81
2.24
1.96
2.52
29.01
4.93
3.69
5.76
13.82
2.22
7.75
1.22
8.64
7.11
10.62
19.29
3.03
2.25
1.66
3.77
8.16
2.38
2.29
4.11
1.32
0.85
1.44
2.90
0.69
2.44
1.08
1.43
3.18
0.16
0.23
0.23
-4.73
0.79
0.97
0.46
3.16
0.29
2.03
-0.07
2.89
-1.02
2.18
5.86
0.58
0.08
-0.13
0.00
0.15
0.02
-0.18
1.05
0.01
0.11
-0.16
-0.30
0.03
-0.16
0.02
-0.25
-0.02
11
23.40
59.48
85.33
46.57
3.47
8.62
3.81
1.54
0.50
70
Alcohol
Family
AAD
AAD
AAD%
AAD%
%
%
Eq
Eq.
Eq
Eq
(1.8)
(1.25)
(1.6)
(1.10)
AAD%
Eq
(1.25)+Eq
(1.26)
AAD
%
ANN
Deviations
exp- calc
(mNm1)
Eq
(1.25)+Eq
(1.26)
Deviatio
ns
expcalc
(mNm1)
ANN
N
Point
s
AAD
%
Eq
(1.9)
11
21.00
1.30
4.66
2.79
3.92
13.98
2.77
-2.02
0.03
14
13.77
17.17
24.42
9.93
20.20
14.01
4.93
2.70
-0.05
12
13.18
25.67
34.75
18.63
11.17
8.28
2.75
1.53
-0.01
16
16.77
39.88
49.03
30.95
3.40
3.02
5.16
-0.41
-0.51
12
23.34
9.65
11.81
10.59
4.52
2.98
1.82
0.90
-0.08
11
28.61
27.08
36.71
20.39
6.72
5.58
1.47
-0.67
-0.07
14
12.72
16.59
23.69
11.16
20.40
14.31
1.93
2.69
0.13
12
11.01
25.33
33.42
18.61
10.57
7.88
3.16
1.01
0.20
16
12.77
35.12
44.65
26.95
7.12
3.11
5.86
-0.01
0.24
38
21
18
2.27
3.36
2.49
42.61
6.42
1.09
47.86
14.25
8.88
39.22
6.10
0.85
3.79
4.27
7.71
3.99
1.41
0.87
1.03
0.89
1.52
-0.21
-0.24
-0.23
0.03
-0.05
-0.13
l
1-methylcyclohexanol
Cis-2methylcyclohezanol
Cis-3methylcyclohexanol
Cis-4methylcyclohexanol
Cyclohexanol
L-menthol
Trans-2methylcyclohexanol
Trans-3methylcyclohexanol
Trans-4methylcyclohexanol
1-butanol
1-decanol
1-dodecanol
Cycloaliphatic
alcohols
Cycloaliphatic
alcohols
Cycloaliphatic
alcohols
Cycloaliphatic
alcohols
Cycloaliphatic
alcohols
Cycloaliphatic
alcohols
Cycloaliphatic
alcohols
Cycloaliphatic
alcohols
Cycloaliphatic
alcohols
n-Alcohols
n-Alcohols
n-Alcohols
71
Alcohol
Family
N
Point
s
AAD
%
Eq
(1.9)
1-eicosanol
1-heptadecanol
1-heptanol
1-hexadecanol
1-hexanol
1-nonadecanol
1-nonanol
1-octadecanol
1-octanol
1-pentadecanol
1-pentanol
1-propanol
1-tetradecanol
1-tridecanol
1-undecanol
2-butanol
2-heptanol
2-hexanol
2-methyl-2-butanol
n-Alcohols
n-Alcohols
n-Alcohols
n-Alcohols
n-Alcohols
n-Alcohols
n-Alcohols
n-Alcohols
n-Alcohols
n-Alcohols
n-Alcohols
n-Alcohols
n-Alcohols
n-Alcohols
n-Alcohols
n-Alcohols
n-Alcohols
n-Alcohols
n-Alcohols
8
10
17
10
25
10
22
11
39
10
12
17
8
11
11
8
12
27
7
43.02
54.70
3.50
45.05
2.79
59.53
3.58
37.22
3.61
39.91
2.51
9.79
24.06
28.75
2.33
3.18
2.40
3.34
3.56
AAD
AAD
AAD%
AAD%
%
%
Eq
Eq.
Eq
Eq
(1.8)
(1.25)
(1.6)
(1.10)
6.56
4.78
20.26
2.21
25.30
4.47
10.38
4.63
15.89
2.80
36.34
28.44
1.76
2.44
4.15
36.89
17.86
20.09
22.78
7.68
11.24
27.21
7.50
31.85
9.39
17.85
7.04
23.12
6.39
42.38
63.31
8.17
5.87
11.51
43.52
24.89
26.53
27.38
4.38
4.74
19.34
2.23
24.38
4.08
9.82
3.85
15.01
2.75
31.92
54.29
1.84
2.81
3.55
36.70
15.73
18.62
19.95
6.10
7.05
1.19
4.96
3.58
6.86
3.39
6.46
1.03
5.44
2.22
2.85
5.70
7.56
5.37
2.99
1.59
2.58
8.41
AAD%
Eq
(1.25)+Eq
(1.26)
AAD
%
ANN
27.03
18.90
3.18
12.13
1.93
25.50
2.24
16.93
3.02
8.06
2.08
3.02
6.57
3.44
1.57
1.61
1.71
2.56
1.06
2.99
2.25
0.99
5.66
1.40
0.54
0.77
2.06
0.39
7.11
1.49
1.12
0.71
1.57
1.24
1.38
0.73
1.23
0.74
Deviations
exp- calc
(mNm1)
Eq
(1.25)+Eq
(1.26)
-3.93
-1.96
-0.80
-1.40
-0.24
-2.87
-0.58
-2.42
-0.80
-0.83
-0.34
-0.13
-0.88
-0.24
-0.42
0.35
0.26
0.36
-0.17
Deviatio
ns
expcalc
(mNm1)
ANN
0.26
0.17
-0.07
-0.36
-0.04
0.00
0.04
0.00
0.04
-0.09
0.18
0.06
0.07
-0.13
-0.07
0.08
0.09
-0.02
-0.16
72
Alcohol
Family
2-pentanol
2-propanol
3-butanol
3-heptanol
Ethanol
Methanol
2,2-dimethyl-1propanol
2,6-dimethyl-4heptanol
n-Alcohols
n-Alcohols
n-Alcohols
n-Alcohols
n-Alcohols
n-Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
2-butanol
2-buthyl-1-nonanol
2-buthyl-1-octanol
2-butyl-1-decanol
2-heptanol
2-hethyl-1-butanol
0.90
0.90
1.33
3.81
1.62
1.04
Deviations
exp- calc
(mNm1)
Eq
(1.25)+Eq
(1.26)
0.06
0.17
0.56
0.70
0.40
-0.46
Deviatio
ns
expcalc
(mNm1)
ANN
-0.02
0.04
-0.06
0.06
0.02
0.00
33.32
9.89
-2.98
0.08
11.85
7.68
0.89
1.45
0.00
36.70
2.99
1.61
1.38
0.35
0.08
30.36
10.11
3.85
13.32
1.59
2.36
0.05
18.78
36.07
16.70
2.45
7.22
0.97
1.60
0.00
37.46
13.08
33.97
12.20
2.23
8.17
0.51
1.45
-0.05
12
2.40
17.86
24.89
15.73
1.59
1.71
0.73
0.26
0.09
15
2.14
29.37
37.53
26.34
8.94
5.51
3.26
-0.51
0.14
N
Point
s
AAD
%
Eq
(1.9)
AAD
AAD
AAD%
AAD%
%
%
Eq
Eq.
Eq
Eq
(1.8)
(1.25)
(1.6)
(1.10)
AAD%
Eq
(1.25)+Eq
(1.26)
AAD
%
ANN
5
12
6
20
48
44
3.63
1.30
5.43
21.14
8.09
11.42
29.41
49.63
34.37
14.13
58.58
73.86
35.74
64.65
42.03
20.21
66.53
81.03
28.28
55.54
32.35
11.11
62.12
88.29
0.86
1.86
10.72
3.83
8.70
12.29
0.75
1.20
3.21
2.76
8.91
11.38
15
69.48
78.82
88.11
61.84
19.04
11
24.24
20.29
31.43
12.25
3.18
36.89
43.52
15
32.04
11.30
16
31.38
15
73
Alcohol
2-hethyl-1-hexanol
2-hexanol
2-methyl-1-butanol
2-methyl-1-dodecanol
2-methyl-1-hexanol
2-methyl-1-pentanol
2-methyl-1-propanol
2-methyl-1-tridecanol
2-methyl-1-undecanol
2-methyl-2-butanol
2-methyl-2-propanol
2-nonanol
Family
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
AAD
%
ANN
Deviations
exp- calc
(mNm1)
Eq
(1.25)+Eq
(1.26)
Deviatio
ns
expcalc
(mNm1)
ANN
3.06
1.92
-0.59
-0.22
2.58
2.56
1.23
0.36
-0.02
22.12
11.50
5.07
2.13
-0.12
-0.22
27.21
7.65
4.45
15.24
1.46
2.82
0.00
19.27
29.89
14.39
3.66
16.26
1.42
2.93
0.09
2.66
23.11
29.01
22.31
2.05
4.32
1.11
-1.02
-0.03
15
1.60
42.55
49.43
40.48
4.80
3.37
3.93
-0.14
-0.03
30
38.29
13.58
34.66
12.47
5.15
7.30
1.25
1.35
0.01
16
32.61
17.49
36.10
15.12
4.04
9.95
1.10
1.78
0.03
3.56
22.78
27.38
19.95
8.41
1.06
0.74
-0.17
-0.16
5.43
34.37
42.03
32.35
10.72
3.21
1.33
0.56
-0.06
24
2.43
13.46
21.20
12.07
2.08
1.35
0.59
0.31
0.04
N
Point
s
AAD
%
Eq
(1.9)
AAD
AAD
AAD%
AAD%
%
%
Eq
Eq.
Eq
Eq
(1.8)
(1.25)
(1.6)
(1.10)
AAD%
Eq
(1.25)+Eq
(1.26)
10
12.40
16.35
23.02
16.93
5.49
27
3.34
20.09
26.53
18.62
17
11.08
26.93
34.55
15
29.93
8.96
11
17.03
40
74
Alcohol
2-octanol
2-pentanol
3-heptanol
3-hethyl-1-heptanol
3-hexanol
3-methyl-1-butanol
3-methyl-1-pentanol
3-methyl-2-butanol
3-methyl-3-pentanol
3-pentanol
4-methyl-1-octanol
Family
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
AAD
AAD
AAD%
AAD%
%
%
Eq
Eq.
Eq
Eq
(1.8)
(1.25)
(1.6)
(1.10)
AAD%
Eq
(1.25)+Eq
(1.26)
AAD
%
ANN
Deviations
exp- calc
(mNm1)
Eq
(1.25)+Eq
(1.26)
Deviatio
ns
expcalc
(mNm1)
ANN
N
Point
s
AAD
%
Eq
(1.9)
29
1.92
15.84
23.29
12.67
1.49
1.53
1.81
0.22
-0.02
3.63
29.41
35.74
28.28
0.86
0.75
0.90
0.06
-0.02
20
21.14
14.13
20.21
11.11
3.83
2.76
3.81
0.70
0.06
14
27.93
15.77
26.53
13.24
8.64
15.26
4.08
2.50
-0.13
15
3.81
17.78
22.85
15.34
4.05
1.89
0.39
-0.07
0.00
53
3.23
43.81
51.32
38.12
4.51
6.45
2.77
-1.44
-0.05
52
3.11
31.55
41.26
26.79
4.53
7.16
0.51
1.36
0.01
24
40.94
27.73
32.68
20.10
7.35
4.51
14.00
-0.48
0.28
60
3.68
17.30
22.42
12.53
8.60
4.92
0.65
-1.03
-0.06
17
13.55
23.88
29.02
22.42
5.63
3.65
3.78
0.43
0.33
40
23.88
19.56
31.83
10.24
11.58
11.42
1.98
2.10
0.00
75
Alcohol
4-methyl-2-pentanol
4-methylcyclohexane-methanol
5-methyl-1-hexanol
6-methyl-1-octanol
Allyl Alcohol
Alpha-terpineol
Beta-terpineol
Isopropanol
Propargyl Alcohol
Mean
Family
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
Other Aliphatic
Alcohols
-
AAD
AAD
AAD%
AAD%
%
%
Eq
Eq.
Eq
Eq
(1.8)
(1.25)
(1.6)
(1.10)
AAD%
Eq
(1.25)+Eq
(1.26)
AAD
%
ANN
Deviations
exp- calc
(mNm1)
Eq
(1.25)+Eq
(1.26)
Deviatio
ns
expcalc
(mNm1)
ANN
N
Point
s
AAD
%
Eq
(1.9)
71
6.77
31.41
38.37
24.75
4.05
4.32
1.21
-0.93
0.04
10
91.71
57.48
67.53
55.62
33.73
35.09
12.82
-1.58
-0.09
22
46.86
46.61
59.57
41.60
13.87
8.25
3.32
-0.48
0.14
40
22.25
15.84
28.26
13.13
7.23
13.72
0.98
2.69
0.00
16
8.02
49.96
56.05
47.37
5.23
3.44
0.54
0.51
0.06
16
75.24
80.59
92.07
71.29
37.95
48.52
1.25
-5.66
-0.04
16
68.00
74.02
86.31
63.61
31.85
42.68
2.13
-5.20
0.00
43
2.78
51.25
60.05
51.15
4.40
3.88
0.48
0.58
0.00
31.05
20.98
25.70
27.16
11.96
14.85
1.24
5.07
0.00
16.73
27.82
35.84
25.13
8.88
8.10
2.19
0.16
0.02
76
.
.
.
.
Figure 1.30. Schematic diagram of the ANN model. Inputs: Tc (K), critical temperature; Gr (), radius
of gyration; T (K), temperature; c (molcm-3), critical density. Output: (mN/m), surface tension
77
Figure 1.31. Absolute average percent deviation between data collected and ANN results vs number
of neurons in hidden layer
During the modeling, the same set of data used for the scaled equation was considered,
again excluding the 7 fluids not following the common trend. From the figure it is evident
that, as expected, AAD decreases with the increase of the number of neurons and that a
good configuration is reached when the number of neurons in the hidden layer is 29. This
configuration is selected as the best network architecture. The AADs for the proposed ANN
are reported in Table 1.15, and Table 1.16. The parameters of the hidden layer and of the
output layer are reported in Table 1.17.
Table 1.17.Parameters of hidden and output layers
hidden
layer
Neuron
number
1
2
3
4
5
6
7
8
9
10
11
12
13
output
layer
bias
Tc
Pc
Gr
8.58
24.21
-1.45
13.70
12.93
36.32
-48.15
0.07
-10.41
7.02
48.23
-15.35
41.35
15.79
8.78
-14.68
-0.78
2.97
-50.87
30.18
-34.06
-6.79
18.77
-26.10
1.44
-60.19
-47.87
15.42
20.05
-38.69
21.57
27.23
84.02
0.25
22.17
22.54
-95.53
20.78
29.52
-0.08
0.15
-0.10
0.43
1.05
0.22
0.81
-1.23
0.02
-41.05
-0.94
-0.03
0.20
9.76
-49.98
19.04
1.76
-45.14
-33.59
-50.51
40.99
29.26
-41.66
56.62
29.64
-31.75
Neuron
number
1
2
3
4
5
6
7
8
9
10
11
12
13
bias
-2.35
30.43
58.44
-5.64
-70.53
85.04
22.33
40.87
90.36
0.08
52.61
-43.44
-42.33
output
bias
-13.55
78
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
102.73
-7.42
4.51
-18.78
-38.03
-68.60
31.94
102.32
-15.07
12.90
13.35
100.63
123.00
-42.58
71.06
222.68
-49.79 24.47
109.20 105.55
-1.83
-1.31
151.71
131.01
-15.33 46.45
20.97 -6.85
176.86
143.35
2.24
-8.85
34.26
-0.74
371.16
14
15.35
-12.32
-22.58
-26.68
-39.03
-156.93
26.06
-1.12
0.05
-0.16
6.51
-0.56
0.24
42.11
-21.95
33.41
-69.15
107.05
-36.63
15
16
17
18
19
20
-108.26
106.59
-67.01
0.31
-0.79
-42.23
133.95
-0.42
36.49
21
1.20
105.03
1.05
-61.69
22
30.68
41.44
-0.67
382.92
23
-14.98
3.03
1.85
0.32
24
-4.18
76.98
-0.85
-85.78
25
-1.17
-31.76
23.60
4.02
0.16
-23.80
-39.12
26
27
-0.85
-94.51
117.46
-3.44
-147.67
28
0.61
10.89
-0.56
0.079
29
-8.60
Because of the discussed cross-correlation between reduced density and radius of gyration,
a configuration with a one-hidden-layer feedforward network without critical density as an
input parameter was also tried, but results in terms of AAD were much poorer
(AAD=4.3%) and this configuration was finally neglected.
Figure 1.32 shows the correlation between the predictions of the validated, trained and
tested MLP and the corresponding experimental surface tension data. From the figure a
linear fit is evident, witnessing a good correlation between predictions and data. It is worth
noticing that the training dataset, even if linear, shows a slightly different trend compared to
the dataset tested and validated.
Figure 1.33 illustrates the effect of the reduced temperature on the network accuracy.
According to the MLP configuration selected, with the exception of few points, deviations
are well within 0.05 mNm-1.
79
Figure 1.32. Correlation between the calculated and the corresponding experimental surface tension
data.
80
1.5.4
This part of research presents a literature survey of the available data of the experimental
surface tension data for refrigerant binary systems. The experimental data were collected
for the following binary systems: R32+R227ea, R32-R125, R32-R134a, R290-R32, R125R152a, R290-R152a, R290-R600a, RE170-R290, R143a-R134a, R125-R134a, R125R143a, R134a-R152a and R143a+R227ea. Experimental data were analyzed in terms of
mole and mass fraction and discussed.
A new equation for the prediction of the refrigerant binary systems surface tension based on
the corresponding states theory was proposed.
The recent environmental issues lead to consider refrigerant blends as attractive option for
the refrigeration industry. However, a limited number of experimental works covering wide
ranges of temperature were published for the refrigerant mixtures and available in the open
literature.
After a literature survey, a total of 1844 experimental data on 13 different refrigerant binary
systems were found. Surface tension data were measured especially for environmental
friendly alternative refrigerants, with particular attention to mixtures of partially
halogenated hydrocarbons, propane, isobutane and dimethyl ether.
The surface tension data of refrigerant binary systems available were measured by four
different labs. They all adopted the Differential Capillary Rise Method (DCRM).
In Table 1.18, the surface tension experimental data for the refrigerant binary systems
available in the open literature are summarised. The table gives the references to the
original work and to the year, the number of experimental points per data set, the claimed
accuracy, the temperature and surface tension ranges. In addition, the mole fraction ranges
were reported.
From Table 1.18, it is possible to notice that the measurements were performed over wide
temperature (from 223.15 to 333.53 K) and surface tension (from 0.78 to 19.70 mNm-1)
ranges.
Table 1.18.Surface tension experimental data for the refrigerant binary systems available in the open
literature
Binary
System
Ref.
Number
of
points
R32-R227ea
[36]
334
[37]
236
[38]
21
[39]
18
[40]
294
R32-R125
R32-R134a
Accuracy
0.15
mNm-1
0.15
mNm-1
1%
0.2 mNm1
0.15
mNm-1
T range,
K
range,
252.66333.55
252.93333.41
223.15333.15
273.33313.31
254.87333.53
12.771.63
12.770.78
15.071.26
mNm-1
Mole
fraction
range
0.54150.8975
0.41660.8222
0.413-0.863
8.73-3.08
0.647-0.694
13.973.07
0.34860.8598
81
[38]
21
[41]
97
[42]
54
[38]
21
R290-R152a
[42]
51
R290-R600a
[43]
43
R290-R32
1%
0.2 mNm1
0.2 mNm1
R125-R152a
RE170-R290
1%
0.2 mNm1
0.2 mNm1
0.2
mNm-
[44]
114
[45]
210
0.15
mNm-1
[38]
21
1%
[38]
21
1%
[38]
18
1%
[39]
R134a-R152a
[38]
21
1%
R143a+R227ea
[46]
242
0.15
mNm-1
R143a-R134a
R125-R134a
R125-R143a
0.2 mNm1
223.15333.15
280.00300.00
243.08328.16
223.15333.15
248.15328.15
278.28320.00
243.40333.14
256.91329.44
223.15333.15
223.15333.15
223.15323.15
273.32303.09
223.15333.15
252.56333.42
19.492.94
0.363-0.835
9.70-4.00
0.004-0.999
17.0-4.3
0.056-0.191
15.284.17
0.162-0.694
15.3-3.8
0.27340.5921
12.4-4.6
0.00-0.95
17.3-3.7
12.601.80
15.192.77
14.812.69
14.021.70
7.29-3.74
19.703.98
11.591.27
0.29160.6856
0.28680.7856
0.265-0.755
0.213-0.715
0.216-0.719
0.419-0.419
0.211-0.677
0.390-0.848
The analysis of data was carried out both considering the mass and the mole fraction,
according to the following equations:
mr =(1x1)r+(2x2)
(1.27)
mr =(1y1)r+(2y2)r
(1.28)
and
where r is a parameter and xi and yi are the mole and the mass fraction, respectively. The
pure fluid surface tensions, 1 and 2, were calculated according to equation:
=0 (1-T/Tc)n
(1.29)
82
and the coefficients 0 and n were adopted as reported from original literature sources.
As a first step, r was considered to be equal 1, as suggested by Hadden for most
hydrocarbon mixtures [47], which would predict linear behaviour in surface tension versus
composition.
The deviations for both equations (1.27) and (1.28) are reported in Table 1.2. For the
deviations, the following equations were adopted:
(1.30)
(1.31)
(1.32)
From Table 1.19 it is possible to notice that both equations (1.27) and (1.28) show very
similar deviations in almost all cases, witnessing that both approaches for the data analysis
are possible. In this paper, for the further analysis, the mole fraction approach was
followed. In addition, deviations were found to be very high for all systems containing R32,
excluding R32+R134a. This could be probably attributed to the strong polarity of the R32
molecule. In particular, R290-R32 system showed deviations up to almost 30%. The same
trend is confirmed also from Figure 1.37 to Figure 1.34, were the experimental data
coming from different sources were regressed together and showed at a fixed temperature,
T = 303.15 K, for each binary system versus mole fraction.
Table 1.19.Summary of deviations between experimental surface tension and the ones predicted by
eqs. (1.27) and (1.28)
Dev.
d
d%
|d%|
d
d%
|d%|
d
d%
|d%|
d
d%
Binary systems
R32-R227ea
R32-R125
R32-R134a
R290-R32
Eq. (1.27)
-0.83
-12.12
12.12
-0.86
-16.94
16.94
-0.23
-1.97
2.55
-1.71
-28.51
Eq (1.28)
-0.76
-13.75
13.75
-0.22
-4.34
4.34
-0.36
-4.72
4.72
-1.71
-28.54
83
|d%|
d
d%
|d%|
d
d%
|d%|
d
d%
|d%|
d
d%
|d%|
d
d%
|d%|
d
d%
|d%|
d
d%
|d%|
d
d%
|d%|
d
d%
|d%|
d
d%
|d%|
R125-R152a
R290-R152a
R290-R600a
RE170-R290
R143a-R134a
R125-R134a
R125-R143a
R134a-R152a
R143a+R227ea
Avg.
28.51
-0.45
-5.83
5.83
-0.81
-10.21
10.21
-0.07
-5.90
1.56
-0.49
-5.45
5.45
-0.11
-2.20
3.51
-0.47
-8.90
8.90
-0.03
-0.34
0.85
-0.34
-3.18
3.18
-0.01
0.01
0.73
-0.50
-7.85
8.23
28.54
0.08
1.06
1.06
-1.05
-13.10
13.10
-0.18
-2.47
2.68
-0.53
-5.90
5.90
-0.23
-4.29
5.64
-0.32
-5.91
5.91
0.02
0.83
1.14
-0.20
-1.55
1.75
-0.36
-7.64
7.64
-0.47
-7.80
8.11
In order to find the optimal r values for equation (1.27), all the experimental data were also
regressed keeping each binary system as a separate set of data. The r values together with
deviations are reported in Table 1.20.
From Figure 1.37 to Figure 1.34, and from the r value reported in Table 1.20 it is
possible to notice that the behaviour of the surface tension binary systems shows a
generally very linear trend in terms of mole fraction. This is also confirmed by the
regressed r values, always close to unity for the following binary systems: R32+R134a,
R290+R600a, RE170+R290, R143a+R134a, R125+R134a, R125+R143a, R134a+R152a
and R143a+R227ea. Slightly deviating from linearity were found to be R125+R152a and
R290+R152a. Higher deviations were found for R32+R227ea and R32+R125 (r=1.27 and
84
r=1.29, respectively). Strong deviation from linearity was found for R32+R290 binary
system (r=3.36), falling below the linear relation.
Table 1.20.Deviations between the experimental surface tension and the ones predicted with the
regressed r value
Binary
System
R32-R227ea
R32-R125
R32-R134a
R290-R32
R125-R152a
R290-R152a
R290-R600a
RE170-R290
R143a-R134a
R125-R134a
R125-R143a
R134a-R152a
R143a+R227ea
Tot. number
of points
334
275
315
97
75
51
43
114
231
21
25
21
242
d %
|d%|
1.27
1.29
1.06
3.36
1.13
1.14
1.01
1.08
1.03
1.09
1.01
1.05
1.00
-0.03
-0.13
0.04
-0.32
-0.13
-0.07
-0.02
-0.03
0.00
-0.10
0.01
0.01
0.01
-0.17
-5.15
1.35
-5.20
-2.75
-2.03
-0.40
-0.86
-0.08
-4.58
0.31
-0.00
0.12
1.99
6.47
2.24
6.12
3.29
3.34
1.45
1.57
1.06
5.22
0.75
0.98
0.79
85
8
R32+R227ea
R32+R125
R32+R134a
/ (mN.m-1)
3
0.0
0.2
0.4
0.6
0.8
1.0
x1
Figure 1.34. Experimental surface tension of R32+R227ea, R32+125 and R32+R134a binary systems
according to equation (1.27).
86
10
R290+R32
R290+R152a
R290+R600a
/ (mN.m-1)
8
7
6
5
4
3
2
0.0
0.2
0.4
0.6
0.8
1.0
x1
Figure 1.35. Experimental surface tension of R290+R32, R290+152a and R290+R600a binary
systems according to equation (1.27).
87
10
R125+R152a
R125+R134a
R125+R143a
/ (mN.m-1)
8
7
6
5
4
3
2
0.0
0.2
0.4
0.6
0.8
1.0
x1
Figure 1.36. Experimental surface tension of R125+R152a, R125+134a and R125+R143a binary
systems according to equation (1.27).
88
11
R290+RE170
R143a+R134a
R134a+R152a
R143a+R227ea
10
/ (mN.m-1)
9
8
7
6
5
4
3
0.0
0.2
0.4
0.6
0.8
1.0
x1
Figure 1.37. - Experimental surface tension of R290+RE170, R143a+R134a, R134a+R152a binary
and R143a+R227ea systems according to equation (1.27).
Searching for an equation for the binary systems surface tension prediction, as a first step,
in order to test the validity of the recently proposed equations for the pure fluid refrigerants,
equation (1.27) was used adopting equations (1.11) and (1.12) for the pure fluids surface
tension prediction in place of equation (1.29). Equation (1.11) gave the following average
deviations d = -0.02 mNm-1, d% = -1.45 % and |d%| = 7.20 %. Equation (1.11) gave
the following average deviations: d= -0.47 mNm-1, d% = 7.50 % and |d%| = 9.19 %.
The obtained deviation were of the same order of magnitude of the ones presented in Table
1.19, where the binary system surface tensions were calculated adopting equation (1.29),
directly regressed from the pure fluid surface tension experimental data, clearly
demonstrating the validity of the two predictive equations based on the corresponding states
principle.
In addition, it is worth noting that slightly better results were obtained with equation (1.12).
This was probably due to fact that equation (1.12) operates with five different parameters
(Tc, Pc, Vc, and ), while equation (1.11) operates only with three different independent
89
parameters (Tc, Pc, and ). The adopted parameters for each pure fluid are reported in Table
Table 1.21
Table 1.21 Physical properties for each refrigerant fluid adopted for eqs (1.11) and (1.12)
Refrigerant
R32
R125
R134a
R143a
R152a
R227ea
R290
R600a
RE170
Tc [K]
351.26
339.17
374.21
345.86
386.41
374.9
369.89
425.13
401.1
Pc [MPa]
5.782
3.6177
4.0593
3.761
4.5168
2.925
4.2512
3.796
4.872
Vc [m3kg-1]
0.002358
0.001743
0.001953
0.002320
0.002717
0.001682
0.004536
0.004386
0.004839
0.2769
0.3052
0.3268
0.2615
0.2752
0.357
0.1521
0.201
0.2002
[Debye]
1.978
1.563
2.058
2.34
2.262
1.456
0.084
0.05
0.0
According to this consideration and the similar results in terms of deviations, the simplest
way was preferred and equation 1 as a starting point for data fitting was considered. For the
mixtures critical properties definition, since they are often not a linear mole fraction
average of the pure component critical properties, two different ways were followed.
As a first, the most commonly used empirical literature methods were followed: Li method
for Tcm, [48], Kreglewski and Kay method for Pcm and m [49] and Chueh and Prausnitz
method for Vcm [50]. As a second, Tcm, Pcm, and m were simply approximated by mole
fraction averages of the pure component properties Tci, Pci, and i.
Since the two different approaches gave the same results in terms of deviation, again the
simplest way was chosen and the mole fraction averages of the pure component properties
Tc, Pc, and for the prediction of the Tcm, Pcm, and m parameters were adopted.
Finally, the proposed equation is the following:
=6.09810-8m0.203(3.285106+Pcm3.449)Tcm (1 Tr)1.258
(1.33)
Equation (1.33) is rather similar to equation (1.11) operating also only with three different
independent parameters (Tcm, Pcm, and m). The coefficients were found defining the
objective function according to the chi-square minimization.
Equation (1.33) gave the following average deviations d = -0.07 mNm-1, d% = -0.96 %
and |d%| = 5.51 %. Improvements in terms of deviations for the proposed equation are
evident. For this equation, deviations are showed in Figure 1.5
90
-
(mN.m-1)
calc
exp
-2
-4
-6
0.5
R143a+R227ea
R32+R227ea
R290+R32
R290+R600a
R125+R134a
R32+R125
R125+R152a
RE170+R290
R125+R143a
R32+R134a
R290+R152a
R143a+R134a
R134a+R152a
0.7
0.8
0.6
0.9
1.0
Tr
Figure 1.38. Absolute (left) and relative (right) deviation for the total number of experimental points
obtained with equation (1.33)
Excluding R290+R32 and R290+R600a binary systems that showed higher deviations,
deviations were found to be generally within 1 mNm-1. In addition, deviations were found
to be very low near the critical point, slightly increasing with the lowering of reduced
temperature. In general, it can be stated that equation (1.33) is well predicting the surface
tension of refrigerant binary systems.
In conclusion this work is important because none of the literature methods were
specifically oriented to refrigerant binary systems and their surface tension prediction, in
this part of research an equation explicitly oriented to refrigerant binary systems was
presented.
The presented generalized equation is a very simple equation, containing only three
physical parameters, and well predicting the surface tension of refrigerant binary systems
91
1.6 Bibliografia
[1.] Everett, D.H., IUPAC Manual of symbols and terminology for physiochemical
quantities and units Appendix II: Definitons, Terminology and Symbols in Coloid
and Surface Chemistry, Part I, London, 1972.
[2.] DIPPR-801, Project 801, Evaluated Process Design Data, Public Release
Documentation, Design institute for Physical Properties (DIPPR), American
Institute of Chemical Engineers, AIChE, 2013.
[3.] A. Mulero, M.I. Parra, I. Cachadia, The Somayajulu correlation for the surface
tension revisited, Fluid Phase Equilib. 339 (2013) 81 88.
[4.] J.D. van der Waals, The thermodynamic theory of capillarity under the hypothesis of
a continuous variation of density, J. Stat. Phys. 20 (1879) 200-244.
[5.] A. Mulero, I. Cachadia, M.I. Parra, Recommended Correlations for the Surface
Tension of Common Fluids, J. Phys. Chem. Ref. Data 41 (2012) 043105 1:13
[6.] A. Mulero, I. Cachadia, Recommended Correlations for the Surface Tension of
Several Fluids Included in the REFPROP Program, J. Phys. Chem. Ref. Data (2014)
http://dx.doi.org/10.1063/1.4878755.
[7.] D.B. Macleod, On a relation between surface tension and density, Trans. Farad. Soc.
19 (1923) 38-41.
[8.] S. Sugden, The Variation of Surface Tension with Temperature and some Related
Functions, J. Chem. Soc. 125 (1924) 32.
[9.] O.R. Quayle, The Parachors of Organic Compounds. An Interpretation and
Catalogue, Chem. Rev. 53 (1953) 439-589.
[10.] J.R. Fanchi, Calculation of parachors for compositional simulation, J. Pet.
Technol. 85, (1985) 20492050.
[11.] J.R. Fanchi, Calculation of parachors for compositional simulation, an update. SPE
Res. Eng. 5 (1990) 433436.
[12.] M.E. Boudh-Hir, G.A. Mansoori, Statistical mechanics basis of Macleod's
formula, J. Phys. Chem. 94 (1990) 8362-8364.
[13.] J. Escobedo, G.A. Mansoori, Surface tension prediction for pure fluids, AIChE J.
42 (1996) 1425-1433.
[14.] A. Ferguson, S.J. Kennedy, Free and total surface energies and related quantities,
Trans. Farad. Soc. 32 (1936) 1474-1481
[15.] E.A. Guggenheim, The Principle of Corresponding States, J. Chem. Phys. 13
(1945) 253-261.
[16.] J.R. Brock, R.B. Bird, Surface tension and the principle of corresponding states,
AIChE J. 1 (1955) 174-177.
[17.] K.S. Pitzer, Thermodynamics, McGraw Hill, New York, 1995.
[18.] S.R.S Sastri, K.K. Rao, A simple method to predict surface tension of organic
liquids, Chem. Eng. J. Biochem. Eng. J. 59 (1995) 181-186.
[19.] C. Miqueu, D. Broseta, J. Satherley, B. Mendiboure, J. Lachaise, A. Graciaa, An
extended scaled equation for the temperature dependence of the surface tension of
pure compounds inferred from an analysis of experimental data, Fluid Phase
Equilib. 172 (2000) 169-182.
[20.] G. Di Nicola, C. Di Nicola, M. Moglie, A New Surface Tension Equation for
92
93
[39.] Okada, M., Shibata, T., Sato, Y., Higashi, Y., 1999. Surface Tension of HFC
Refrigerant Mixtures. International Journal of Thermophysisics 20 119127.
[40.] Lin, H., Duan, Y.Y., Wang, Z.W., 2003c. Surface Tension Measurements of
Difluoromethane (R-32) and the Binary Mixture Difluoromethane (R-32) + 1,1,1,2Tetrafluoroethane (R-134a) from (253 to 333) K. Journal of Chemical and
Engineering Data 48, 1068-1072..
[41.] Tanaka, K., Higashi, Y., 2009. Measurements of the Surface Tension for the R290
+ R32 Mixture. Journal of Chemical and Engineering Data 54, 16561659.
[42.] Bi, S., Zhao, G., Wu, J., 2009. Surface tension of pentafluoroethane + 1,1difluoroethane from (243 to 328) K. Fluid Phase Equilibria 287, 2325.
[43.] Tanaka, K., Higashi, Y., 2007. Measurements of the surface tension for R290,
R600a and 290/R600a mixture. International Journal of Refrigeration 30. 13681373.
[44.] Bi, S., Li X., Zhao, G., Wu, J., 2010. Surface tension of dimethyl ether + propane
from 243 to 333K. Fluid Phase Equilibria 298, 150153.
[45.] Lin, H., Duan, Y.Y., 2004. Surface Tension for the 1,1,1-Trifluoroethane (R-143a)
+ 1,1,1,2-Tetrafluoroethane (R-134a) System. Journal of Chemical and Engineering
Data 49, 372-375.
[46.] Lin, H., Duan, Y.Y., 2003b. Surface Tension of 1, 1, 1-Trifluoroethane (HFC143a), 1, 1, 1, 2, 3, 3, 3-Heptafluoropropane (HFC-227ea), and Their Binary
Mixture HFC-143a/227ea. International Journal of Thermophysics 24, 1495-1508.
[47.] Hadden, S. T., 1966. Surface tension of hydrocarbons. Hydrocarbon Processing
45, 161-164.
[48.] Li, C.C., 1971. Critical Temperature Estimation for Simple Mixtures. The
Canadian Journal of Chemical Engineering 19, 709-711
[49.] Kreglewski A., Kay W.B., 1969. The critical constants of conformal mixtures. The
Journal of Physical Chemistry 73, 3359-3371.
[50.] Chueh, P.L., Prausnitz, J.M, 1967. Vapour-Liquid Equilibria at High Pressures.
Calculation of partial Molar Volumes in Nonpolar Liquid Mixtures. AIChE Journal
13, 1099-1107.
94
2 Thermal conductivity
2.1 Introduction
The thermal conductivity as defined by Fouriers equation (2.1), is a materials ability to
transmit heat by means of conduction. For this reason, thermal conductivity is essential for
heat transfer knowledge and its estimation has been the subject of many studies.
(2.1)
where
and T are the local heat flux (W m2) and temperature gradient (K m1),
respectively.
A number of theories have appeared in the literature for the estimation of liquid thermal
conductivity. There are equations based upon a materials specific property such as density
and/or heat capacity [1,2]. These are all dependent upon temperature thus making them
impossible to use for some fluids whereby not enough data are available; there are
theoretical calculations which consider the intermolecular distances [3] or the degree of
association of liquids [4]; there are a lot of equations that try to describe the thermal
conductivity behavior. In this work some of them are reported, and analyzed.
The two variables that mostly influence the value of the thermal conductivity of liquids are
temperature and pressure. Thermal conductivity of liquids usually decreases with the
increase of temperature, even if water and some aqueous solutions differ from this trend
showing a different behavior. The variation of thermal conductivity is usually linear with
the variation of temperature under the normal boiling point or near it, and in this range
can be represented as [5]
= 0 BT
(2.2)
95
12 6
(r ) = 4
r
r
(2.3)
where and are the characteristic Lennard- Jones energy and length, and r is the
intermolecular distance.
Other than Lennard-Jones, different models can be used, simpler ones, such as the square
well potential, or very complicated and also angle-dependent potentials. The potential
function is always one of the input data for the formulation of a liquid state model. The
existing theories, based on gas like or solid like models, or on statistical mechanics,
usually not allow to easily find a value for thermal conductivity and in any case empirical
data are necessary to link the expression derived from the theory with the characteristics of
the investigated liquid. Due to these difficulties these models are rarely employed for
engineering purposes.
Empirical or semiempirical estimation methods are often used to evaluate the thermal
conductivity of a liquid at particular conditions of temperature and pressure.
While temperature always appears in the empirical formulas, pressure is only considered
when its value exceeds 5-6 MPa, according to what was previously stated. There is a wide
variety of very different formulas, although it is possible to divide them into groups based
on the physical approximations [5].
96
2.3.1
Equation of Weber
Weber [9] carried out a simplification of the general expression of Debye [10] valid for
dielectric solids. Weber replaced the specific heat at constant pressure Cp with the specific
heat at constant volume Cv, and supposed that the product between the wave velocity w
and the mean free path l is proportional to the inverse of the mean molecular distance
(/M)1/3.
1/ 3
= 3.5808 10
Cp
M
(2.4)
where is the density and M the molecular weight. The Eq. (2.4) can be used at
atmospheric pressure and at temperature of about 20 C.
2.3.2
Smith [11] improved the Webers Eq. (2.4), splitting up the factors and proposed:
8
11+ 8.7885 10 C p 1884.06
4
= 4.1868 10
1 / 3
1 / 9
+125 + 464.16
M
)
3
(2.5)
The equation (2.5) does not present remarkable improvements, if compared to Eq. (2.4),
and requires the knowledge of the dynamic viscosity .
2.3.3
Equation of Palmer
Palmer [12], after some simplifications, introduced in Eq. (2.4) a corrective factor and
proposed:
1/3
= 3.9642 10 C p
M
3
Tb
M H vb
(2.6)
97
where Hvb is the vaporization latent heat at the normal boiling point and Tb the normal
boiling point temperature.
2.3.4
Equation of Vargaftik
87922.8 Tb
= 4.2713 10 C p
M H vb
M
8
(2.7)
Equation (2.7) does not significantly improves the Eq. (2.6) of Palmer.
2.3.5
= 4.1868 10
88.0 4.94 H
H
vb + R ln 273
M
T
Tb
b
(2.8)
0.55 1/3
Cp
Tr M
where R is the gas constant, H depends on the molecular structure, and N on the density at
20C. Equation (2.8) yields satisfactory results (deviations usually less than 8%) for
temperatures around 20C.
2.3.6
Equation of Bridgmann
Bridgmann [15-17] assumed for the liquid state a simple cubic structure with a distance l
between the centers of two adjacent molecules, and obtained:
= 3 k
w
l
(2.9)
where k is the Boltzmann constant, w the sound velocity. The application of Eq. (2.9) is
quite difficult, and its use is possible only for a few groups of liquids.
98
2.3.7
Equation of Osida
= 1.2979 10
2/3
Vm
Tm
(2.10)
where Tm is the melting point and Vm the molar volume. The results of Eq. (2.10) are
similar to those obtained through Eq. (2.9) of Bridgmann.
2.3.8
Hirschfelder, Curtiss, and Bird [19] propose the following equation based on more
sophisticated considerations, which take into account the speed of sound in the liquids wliq:
5
= 2.3875 10 w
liq M
2/3
(2.11)
The Eq. (2.11) yields better results compared to those obtained by Eq. (2.9) of Bridgmann,
but its use is not possible for several families of compounds.
2.3.9
Equation of Viswanath
Viswanath [20], assuming Eq. (2.11) as starting point, and using some physical
assumptions and empirical simplifications, proposed:
= 3.6 10
0.38
1 T
r
2/3
1/2
C Vm T
1 Tbr
M
1/2
H vb
(2.12)
where Hvb is the latent heat of vaporization at the normal boiling point, Tr the reduced
temperature T/Tc, Vm the molar volume, Trb the reduced normal boiling point, and C a
particular packing factor depending on the molecular structure. C is commonly considered
constant for a given compound, but in this case the results are unsatisfactory.
99
= L w Cp
(2.13)
w Cp
= 1
f Tc 2 c Tf
(2.14)
where is the distance between the centers of two adjacent molecules (an additive quantity
difficult to be calculate by group contributions methods), Tf the freezing point, and Tc the
critical temperature.
= 1.0798 10
Cp
2/3
1/2
1/3
(2.15)
1/2
(M C p 8314.32)
100
= 2.8379 10
Cp
0.83
293
0.38
(2.16)
Equation (2.16) is simple, but the temperature dependence does not appear correct. Cp and
are estimated at the temperature of 20C.
= 1.283 10
13
2.15
1.55
Cp
0.192
(2.17)
0.12
Cp
=
Log
2.4055 105
(H v / T )
1.8
(2.18)
where Hv is the latent heat of vaporization. Equation (2.18) is generally useless for
organic compounds.
= 1.951
1 0.00126 T Tm
0.216
Tm
(2.19)
0.300
101
Eq. (2.19) is certainly simple and it provides correct values for the liquid thermal
conductivity when the reduced temperature is less than 0.7.
= 2.846 10
Tb
1/2
Cp
N
(2.20)
1/4
where N is the number of atoms composing the molecules. Usually the mean deviations
between estimated and experimental thermal conductivity data are less than 20%.
= 3.3062 10
10
1/3
2/3
C p H
1/2
vb
1 T
r
1
T
rb
(2.21)
= 8.3455 10
11
C p H v M
1.30
(2.22)
2/3
Vm
The authors assume the values of Hv and Vm at the normal boiling point and propose their
correlation at the temperature of 20C.
102
=C
(2.23)
Vm
where C and n vary with homologous groups. Equation (2.23) is very simple, but the
authors do not present the values of C and n for all the organic families. Furthermore Vm
estimated at the normal boiling point.
= 3.0090 10
10
1/2
1/3
C p H vb Vb
(2.24)
where Vb is the molar volume at the normal boiling point. No particular improvements are
obtained with respect to Eq (2.23).
T
10
1/3
= 2.27956 10 C p H vb c
T
(2.25)
where Cp and are taken at the temperature of 20C. The temperature dependence does not
appear completely correct.
= 0.884 10
Cp
(2.26)
The absence of a term containing explicitly the temperature leads to large deviations
between estimated and experimental thermal conductivity data.
103
= 7.5 10
T
Cp b
T
(2.27)
where Cp and are at the temperature of 20C. The calculations are very simple , but the
correct dependence on the temperature appears to be poor.
(2.28)
where = 0.6 + A, 0.6 is the thermal conductivity at Tr=0.6 and A is an additive constant
depending on functional groups. The method appears to be rather complicated for
engineering purposes.
( )
( )
1.1053 3 + 20 1 Tr
=
1/2
M
3 + 20 1 Tbr
2/3
(2.29)
2/3
Equation (2.29) appears simple and useful for engineering purposes in reduced temperature
ranges around 0.6.
104
= MI
3 + 20 1 Tr
2/3
293
3 + 20 1
(2.30)
2/3
where MI is the value of the thermal conductivity calculated according to Eq. (2.20) of
Missenard. Equation (2.29) is reliable for reduced temperature ranges around 0.7.
1 T
m
r
L = b a , where m = 1
1 Trb
(2.31)
In equation (2.31) b is the thermal conductivity at the normal boiling point and is
determined by a group contribution method. Sastri [5] reports an average deviation of 8%
for 186 points that were tested. For alcohols and phenols, a = 0.856 and n = 1.23. For other
compounds, a= 0.16 and n = 0.2. The subscript b indicates the normal boiling point.
105
1.009 B 2
10
+
2P
2T
+
4
+1.908(T
+
)
c
b
2
4
M
= 110
3.9287 M 4 A
+
+
B4
B8
(2.32)
where
(2.33)
(2.34)
and
T
0.1
= A1.951 r
1/6
1.1 Tr
(
(
)
)
(2.35)
where the A values were studied and tabulated for each family in several past papers [42
47 Atene]. The Latini equation is generally recommended [5] and it was proposed for
several chemical families of compounds including olefins, satu- rated hydrocarbons,
cycloparaffins, aromatics, alcohols, ketones, carboxylic acids, refrigerants and esters and its
main drawback is that it may be inaccurate for reduced temperatures above 0.65.
2.4 Results
106
In this part there are the results of our study of thermal conductivity. There are 4 new
equations for describing the thermal conductivity. Two of them are general and suitable for
most compounds. Another one is dedicated for refrigerants including recent refrigerants
used in the modern applications. The last one is a modification of one famous equation of
Latini, dedicated for the family of aromatic.
2.4.1
The main goal of the present work was to find a new physically grounded fitting equation
that minimizes the deviation between the predictions and the experimental data. For the
coefficients optimization, the Levenberg-Marquardt curve-tting method was adopted [48].
The method is actually a combination of two minimization methods: the gradient descent
method and the Gauss-Newton method. The Levenberg-Marquardt method acts more like a
gradient-descent method when the parameters are far from their optimal value, and acts
more like the Gauss-Newton method when the parameters are close to their optimal value
and the solution typically converges rapidly to the local minimum. In this way, it was
possible to identify the parameters of the proposed equation, that guarantee the lowest
deviation of the predicted thermal conductivity.
Following the same approach of surface tension, we proposed a very simple equations that
gave a noticeable improvement with respect to the existing. These equations were based on
the corresponding states principle, containing a limited number of constants.
In this work, having selected a set of reliable experimental data for each compound, has
been presented two version of an equation.
One version was obtained using four terms with
Reduced temperature
Acentric factor
Molecular mass
Enthalpy of fusion
The second version is obtained adding 438 data points, and was cutted of the term with
Enthalpy of fusion. In this way the second formula is more general, and more simply than
the first one.
First equation
The first attempt to describe the behavior of thermal conductivity is reported below in
equation (2.36)
f
e
= a + b Tr + c h fus + d +
0
M
(2.36)
107
were reported as a function of reduced temperatures for all the studied compounds.
Figure 2.1 Deviation between experimental and calculated with equation (2.36) thermal
conductivities versus reduced temperature
To test the goodness of the existing equations, deviations between the experimental thermal
conductivities and the ones predicted by equations (2.29) (2.19), (2.35), and (2.32) were
calculated for and reported as averages in Table 2. Deviations were calculated according to
equation (2.38).
(2.38)
108
Alcohols
Alkanes
Alkenes
Aromatic
Carboxylic
acids
Cycloalkanes
Cycloalkenes
Esters
Ethers
Ketones
Refrigerants
Average
Eq.
Eq.
Eq.
Eq.
Eq.
(2.29)
(2.19)
(2.35)
(2.32)
(2.36)
13.3
13.9
7.4
6.0
10.9
21.6
13.3
8.9
8.3
5.9
7.4
6.4
13.0
8.1
6.5
4.8
7.7
5.4
3.0
3.4
11.3
9.9
6.1
9.5
6.6
3.1
13.2
11.6
11.4
17.6
9.9
5.0
12.4
3.7
57.9
26.5
9.6
3.2
11.7
8.0
4.3
6.9
9.7
7.9
10.7
5.9
3.2
10.7
7.1
5.5
32.9
16.0
9.8
3.4
3.3
8.1
5.4
6.1
9.2
7.1
Total
number
of
points
775
1025
135
414
318
Number
of
rejected
points
27
56
-
Total
number
family
35
10
236
111
185
1340
-
7
7
18
-
3
1
12
5
7
37
-
of
20
20
7
8
11
As a general comment, it could be stated that the present equation provides the lowest
deviations for almost all the selected families. When compared with other equations,
particularly low deviations were found for alkanes, cycloalkanes, alkenes, cycloalkenes,
and aromatics. Higher deviations were found for refrigerants, alcohols, carboxylic acids,
following the trends of the other models. In particular, high negative deviations (up to 86%) were observed for methane (here considered as refrigerant, R50) and ethane (up to 57%), while high positive deviations were found for R114 (up to 66%), methanoic acid (up
to 46%) and 2-methyl-2-propanol (up to 36%). As is evident from Table 2.1, greater
deviations for acids and alcohols, are probably due to the associating phenomena, and there
is poorer representation for methane and ethane. For this reasons, and to made equation
(2.36) more easy, and usable, has been decided to modify this equation, and to build a new
equation.
Second equation
In order to provide a reliable database of thermal conductivity for pure compounds, the first
step was to perform a careful literature survey. For many compounds, thermal conductivity
data can be found in reviews [49]. A complete and updated source of data can be found in
the DIPPR 801 database [50], where experimental, predicted and smoothed data are
reported. Furthermore, the DIPPR database collects data from a wide range of sources and
evaluates them critically, giving quality codes for all data points. Both sources were used to
elaborate the data.
109
In addition, a critical analysis and a selection of the published experimental data were
performed. All of the data collected are reported in Figure (2.1) as a function of the reduced
temperature.
Comparing with the previous data collection, in this step 438 data points were added to the
databank for families with a limited amount of experimental data. In particular, new data
were added for condensed rings, HC chains, monocyclic compounds, cycloalkanes,
cycloalkenes, esters, and ketones. Considering that a tenth of the compound data comes
from one source only and that when different sources were found, strong differences among
the data were sometimes observed, a fluid-by-fluid analysis was performed and data
deviating more than three sigma from the mean were rejected.
To build the new equation, as a first step, we start from the consideration that experimental
data can generally be well described by adopting a linear dependence with reduced
temperature, as shown by the solid line in Figure 2.2 where a common trend with reduced
temperature is also evident for all compounds. Following this observation, we searched for
some specific fluid property that could be involved in the model to improve the
experimental data representation for all species studied, representing various chemical
structures. Then, considering the corresponding states principle parameters typically
adopted, statistical analysis suggests that, besides the critical temperature and experimental
temperature, an important parameter with good statistical effect could be the molecular
mass. This is as expected, since it is well related to the movement of molecules, being
based on molecular kinetic energy. In addition, a good statistical effect was also found for
the acentric factor, an important parameter characterizing the molecular shape. This
analysis was also confirmed by the fact that the same parameters have been adopted by
other authors in the literature.
In this work, the contribution of the four terms in equation (2.36) was analyzed in three
ways: first by quantifying the contribution of each of them to the experimental thermal
conductivity for three different compounds chosen as samples: toluene, propane, and nhexane; second by plotting the thermal conductivity at the boiling point as a function of the
physical parameters adopted (critical temperature, enthalpy of fusion, acentric factor, and
110
molecular mass); third performing the factor analysis of the physical parameters adopted
for the new set of data.
As can be seen in Table 2.2 the contribution of the enthalpy of fusion term in the
calculation of thermal conductivity is numerically marginal. On the other hand, the reduced
temperature term, the acentric factor term, and the molecular mass term were found to have
far more numerical significance in the calculation.
Table 2.2.Effect of each term on thermal conductivity calculation
kexp
0.1350
0.1320
Fluid
Toluene
A
-0.5694
Propane -0.5694
0.1134 n-hexane -0.5694
bTr
chfus
(e/M)f
kcalc
AAD%
-0.0687
3.64310-3
0.0134
0.7548
0.1337
0.96
-0.0889
1.93410-3
0.0077
0.7902
0.1414
7.15
-0.0985
7.18010-3
0.0153
0.7579
0.1125
0.79
Figure 2.3-Figure 2.6 provide a more thorough explanation as to how thermal conductivity
is related to each parameter. To create the following graphs, a linear regression of data
against the experimental temperature of each fluid was performed (since the data showed a
linear trend, as shown in Figure 2.2; then, the thermal conductivity at the boiling point was
calculated for each fluid; lastly, the calculated thermal conductivity was plotted for each
family for the parameters considered. From Figure 2.3-Figure 2.6, a general trend is
evident for the molecular mass for all families.
The coefficients od the linear regression are listed in Table 2.3. In addition in Table 2.3 are
listed the physical properties used in the new equation.
111
Table 2.3. Summary of experimental data and calculated thermal conductivity at the boiling point.
family
Brderivatives
Brderivatives
Brderivatives
Carboxylic
acids
Carboxylic
acids
Carboxylic
acids
Carboxylic
acids
Carboxylic
acids
Carboxylic
acids
Carboxylic
acids
Carboxylic
acids
N
Points
range
(mNm-1)
T range
(K)
Tc (K)
Radius of
ryration
(A)
Critical
density
(molcm-3)
Tb (K)
sigma@Tb
17
0.08388-0.103
280.373.15
650.15
2.833
0.00382263
404.51
-0.0002
0.1497
0.0826
25
0.0935-0.112
200.-327.
503.8
1.976
0.004901961 311.49
-0.0001
0.1372
0.0995
25
0.0926-0.112
535.5
2.518
0.003802281 344.12
-0.0002
0.1542
0.0915
43
0.1244-0.1547
615.7
3.61
0.00341
436.42
-0.0001
0.1869
0.1261
19
0.11132-0.15
722.1
5.932
0.00156
543.15
-0.0002
0.2007
0.1075
57
0.1417-0.2001
591.95
2.61
0.00564
391.05
-0.0002
0.2345
0.1431
13
0.122-0.146
677.3
4.842
0.00214
496.15
-0.0002
0.1932
0.1072
27
0.113780.14969
660.2
4.404
0.00245
478.85
-0.0001
0.1863
0.1152
40
0.223-0.298
588
1.847
0.008
373.71
-0.0001
0.3049
0.2596
16
0.11811-0.15
710.7
5.732
0.00171
528.75
-0.0002
0.1901
0.1104
41
0.11827-0.153
639.16
3.965
0.00285
458.95
-0.0001
0.1830
0.1187
230.333.15
259.29438.08
310.514.84
274.09424.65
280.423.15
261.12497.33
285.15430.
259.444.28
248.15468.68
112
family
Carboxylic
acids
Carboxylic
acids
Carboxylic
acids
Clderivatives
Clderivatives
Clderivatives
Clderivatives
Clderivatives
Clderivatives
Condensed
Rings
Condensed
Rings
Cycloalkan
N
Points
range
(mNm-1)
40
0.1246-0.1633
17
0.12585-0.162
0.149-0.155
89
0.0651-0.114
35
0.0821-0.1159
31
0.0753-0.14528
31
0.100620.13716
11
0.1128-0.1224
24
0.12995-0.1848
11
0.149-0.157
0.12608-0.132
0.10588-
Tc (K)
Radius of
ryration
(A)
Critical
density
(molcm-3)
Tb (K)
sigma@Tb
600.81
3.107
0.00425
414.32
-0.0001
0.1872
0.1309
763
7.265
0.00112
599.35
-0.0002
0.2241
0.1020
310.-350.
732
6.351
0.00141
557.15
-0.0002
0.2048
0.1157
250.-470.
556.35
3.482
0.003623188 349.79
-0.0002
0.1665
0.0913
645
4.044
0.003076923 418.25
-0.0002
0.1601
0.0918
561.6
2.827
0.004545455
356.59
4
-0.0004
0.2312
0.0895
536.4
3.249
0.0041841
334.33
-0.0002
0.1850
0.1085
503.15
2.771
0.004115226 319.67
-0.0002
0.1822
0.1115
510
2.322
0.005405405
312.9
-0.0003
0.2424
0.1345
873
4.9801
0.0018
615.18
-0.0001
0.1878
0.1375
748.4
4.045
0.00245
491.14
3
-0.0001
0.1531
0.1227
727
4.917
0.001672241 512.19
0.0000
0.1106
0.1042
T range
(K)
259.56438.57
320.483.15
280.473.15
223.15363.15
213.15348.15
273.15313.15
178.15323.15
380.-480.
353.63410.
335.93-
113
family
es
Cycloalkan
es
Cycloalkan
es
Cycloalkan
es
Cycloalkan
es
Cycloalkan
es
Cycloalkan
es
Cycloalken
es
Cycloalken
es
N
Points
range
(mNm-1)
T range
(K)
0.10639
377.04
293.15430.93
223.15320.
273.15370.
293.15353.15
298.15398.15
293.15430.93
223.15347.15
223.15298.15
270.363.15
270.373.15
277.75680.
0.10276-0.1151
16
0.12-0.1581
10
0.108-0.1262
0.103-0.1463
0.1069-0.1203
0.10276-0.1151
10
0.1158-0.1471
0.1435-0.1666
Esters
23
0.1141-0.1383
Esters
33
0.1053-0.143
Esters
21
0.0661-0.1397
Tc (K)
Radius of
ryration
(A)
Critical
density
(molcm-3)
Tb (K)
sigma@Tb
703.6
4.206
0.002083333
468.96
5
-0.0001
0.1364
0.0999
511.7
2.85
0.00382
322.4
-0.0004
0.2398
0.1196
604.2
3.555
0.002832
391.94
-0.0002
0.1792
0.1041
553.8
3.242
0.0032438
353.89
-0.0004
0.2439
0.1039
647.2
4.125
0.002439024 424.29
-0.0001
0.1606
0.1034
687
4.314
0.002083333 460.46
-0.0001
0.1364
0.1006
560.4
3.157
356.12
-0.0003
0.2093
0.1163
507
2.728
0.004081633 317.38
-0.0003
0.2358
0.1370
586.1
4.656
0.00226
415.15
-0.0002
0.2010
0.1022
575.4
4.28
0.00257
399.26
-0.0002
0.2062
0.1078
594.6000
1
4.738
0.00226
418.26
-0.0002
0.1921
0.1120
0.00343
114
family
N
Points
Esters
41
Esters
14
Esters
17
Esters
Esters
8
16
Esters
Esters
Esters
Esters
Esters
Esters
28
18
9
Ethers
37
F+Cl+Brderivatives
F+Cl+Brderivatives
range
(mNm-1)
T range
(K)
270.504.15
0.14107277.150.16889
354.15
277.450.1187-0.1412
350.
0.12-0.1477
293.-350.
0.097-0.134
260.-373.
273.150.1845-0.20083
313.15
280.0.1326-0.1469
341.15
278.250.12-0.1421
383.15
0.0854-0.14447 273.-398.
0.1211-0.14549 280.-370.
0.129-0.1615 270.-350.
183.0.0745-0.1721
433.15
0.099-0.1598
Tc (K)
Radius of
ryration
(A)
Critical
density
(molcm-3)
Tb (K)
sigma@Tb
523.3
3.468
0.00349
350.21
-0.0003
0.2190
0.1305
508.4
2.963
0.004366812 327.46
-0.0003
0.2571
0.1508
571
4.298
0.00248
394.65
-0.0003
0.2121
0.1116
546
560.8
3.934
4.276
0.00289
0.00257
372.25
389.8
-0.0005
-0.0003
0.2917
0.2183
0.1088
0.0899
487.2
2.387
0.005813953
304.9
-0.0003
0.2706
0.1869
530.6
3.49
0.00354
352.6
-0.0002
0.2041
0.1320
576
4.176
0.00257
405.45
-0.0002
0.1961
0.1145
599.9
549.73
538
4.772
3.969
3.364
0.00226
0.00289
0.0035
421.15
374.65
353.97
-0.0003
-0.0002
-0.0003
0.2114
0.2078
0.2433
0.0917
0.1219
0.1287
466.7
3.177
0.00357
307.58
-0.0004
0.2344
0.1239
77
0.0489-0.12
173.-438.
471.11
3.289
0.004033
296.86
-0.0003
0.1680
0.0882
85
0.0412-0.09
193.-453.
487.25
3.791
0.0029887
320.74
-0.0002
0.1345
0.0658
115
family
F+Cl+Brderivatives
F+Cl+Brderivatives
F+Cl+Brderivatives
F+Cl+Brderivatives
F+Cl+Brderivatives
F+Cl+Brderivatives
F+Cl+Brderivatives
F+Cl+Brderivatives
F+Cl+Brderivatives
F+Cl+Brderivatives
F+Cl+Brderivatives
F+Cl+Br-
Tc (K)
Radius of
ryration
(A)
Critical
density
(molcm-3)
Tb (K)
sigma@Tb
418.83
3.602
0.0033932
276.74
-0.0002
0.1246
0.0709
353.1
3.587
0.00398
233.9
-0.0003
0.1476
0.0717
384.95
3.052
0.0046728
243.36
-0.0003
0.1733
0.0885
456.83
3.748
0.0035964
300.97
-0.0003
0.1601
0.0765
461.6
3.613
0.003597122 301.15
-0.0002
0.1389
0.0752
395.43
3.523
261.19
-0.0003
0.1613
0.0808
203.-373.
426.15
3.105
0.004065041 269.14
-0.0002
0.1403
0.0775
0.048-0.134
100.293.15
302
2.894
191.67
-0.0005
0.1790
0.0921
0.084-0.135
170.-310.
430
3.2513
0.004201681 279.25
-0.0004
0.1963
0.0954
22
0.0605-0.1119
240.-392.
478.85
3.214
0.003921
305.2
-0.0003
0.1960
0.0899
33
0.0488-0.1124
410.26
2.9
0.004438
264.03
-0.0003
0.1848
0.0937
41
0.0573-0.15942
451.48
3.015
0.0051108
282.01
-0.0004
0.2053
0.1066
N
Points
range
(mNm-1)
67
0.0412-0.09449
33
0.0366-0.0927
84
0.0418-0.13
64
0.0625-0.1154
0.0666-0.0734
22
0.0493-0.0901
17
0.0545-0.0945
28
T range
(K)
178.15453.
170.343.65
118.-366.4
159.34354.15
308.05341.55
234.2366.2
223.15403.
148.15-
0.0041033
0.00558
116
family
derivatives
F+Cl+Brderivatives
Fderivatives
Fderivatives
Fderivatives
Fderivatives
Fderivatives
Fderivatives
Fderivatives
Fderivatives
Fderivatives
Fderivatives
N
Points
range
(mNm-1)
T range
(K)
Tc (K)
Radius of
ryration
(A)
Critical
density
(molcm-3)
427
2.081
293.03
3.423
339.17
3.282
0.004779
374.21
2.944
227.51
Tb (K)
sigma@Tb
0.006493506 264.06
-0.0006
0.3103
0.1472
0.004464286 194.95
-0.0004
0.1596
0.0784
225.06
-0.0004
0.1818
0.0921
0.0050171
247.08
-0.0005
0.2183
0.1061
2.582
0.0071094
145.1
-0.0008
0.2077
0.0933
345.86
2.767
0.0051285
225.91
-0.0005
0.2174
0.1029
386.41
2.5143
0.0055715
249.13
-0.0005
0.2484
0.1276
433.
9
0.116-0.203
170.-290.
13
0.043-0.08523
43
0.0471-0.1116
133
0.0501-0.143
14
0.0873-0.1361
0.0663-0.0994
37
0.0753-0.17117
22
0.0373-0.1
130.-340.
345.02
3.999
0.00334
236.36
-0.0003
0.1395
0.0670
28
0.072-0.18
123.-280.
299.29
2.454
0.00752
191.13
-0.0007
0.2689
0.1341
13
0.0621-0.1071
398.07
3.847
0.003759
272.45
-0.0003
0.1786
0.0944
33
0.0962-0.24765
351.26
1.95
0.0081501
221.5
-0.0008
0.3585
0.1892
178.15280.
172.74338.15
169.87367.95
91.35152.75
233.65298.85
163.15363.15
237.14377.25
148.15344.35
117
family
N
Points
Fderivatives
HC chains
47
HC chains
58
HC chains
13
HC chains
58
HC chains
38
HC chains
67
HC chains
29
HC chains
44
HC chains
HC chains
31
HC chains
18
range
(mNm-1)
T range
(K)
263.15333.15
198.150.121-0.172
480.
0.097-0.1658 290.-580.
220.0.0976-0.145
433.15
218.150.0935-0.16736
573.15
163.150.1037-0.161
363.15
203.550.09-0.16318
574.65
143.150.1076-0.166
333.15
273.150.0965-0.1687
578.65
307.750.0932-0.1174
407.55
273.150.096-0.1681
570.15
200.0.0995-0.151
393.15
0.0702-0.1047
Tc (K)
Radius of
ryration
(A)
Critical
density
(molcm-3)
Tb (K)
sigma@Tb
460
3.783
0.0032
313.3
-0.0005
0.2356
0.0816
563.1
3.225
0.003663004
391.9
-0.0002
0.2081
0.1352
688
5.499
0.001550388
503
-0.0002
0.2296
0.1170
616.6
5.227
0.001712329 443.75
-0.0002
0.1963
0.0928
632.3
4.38
0.002252252
448.6
-0.0002
0.2112
0.1226
537.4
4.083
0.002487562 366.79
-0.0003
0.2119
0.1041
611.3
4.144
0.002617801
429.9
-0.0002
0.2113
0.1248
504
3.66
0.002873563 336.63
-0.0003
0.2186
0.1083
670.9
5.108
0.001736111
485.2
-0.0002
0.2309
0.1184
593.1
4.873
0.001908397
420.01
8
-0.0002
0.1906
0.0898
652.3
4.787
0.001964637
467.1
-0.0002
0.2197
0.1188
566.9
4.457
0.002155172 394.41
-0.0002
0.2007
0.1035
118
family
N
Points
range
(mNm-1)
T range
(K)
Tc (K)
Radius of
ryration
(A)
HC chains
12
0.0945-0.135
300.-520.
708.0000
1
6.637
HC chains
47
588.1
HC chains
HC chains
HC chains
50
11
11
HC chains
13
HC chains
HC chains
HC chains
HC chains
21
10
HC chains
HC chains
51
HC chains
12
HC chains
29
HC chains
27
198.15500.
0.124-0.182
160.-460.
0.0977-0.134 310.-500.
0.0922-0.13
300.-500.
307.750.093-0.1227
448.65
293.150.1286-0.136
338.15
273.150.1397-0.1397
273.15
0.119-0.148
230.-430.
0.125-0.152
200.-370.
293.150.11504-0.1239
328.15
173.150.112-0.157
470.
0.1073-0.118 280.-360.
0.086206293.150.12131
480.
0.097-0.157
160.-520.
0.115-0.171
Critical
density
(molcm-3)
Tb (K)
sigma@Tb
0.001068376 541.61
-0.0002
0.1898
0.0895
3.679
0.003067485
410.9
-0.0002
0.2075
0.1309
536.8
691
674.8
2.7359
6.398
6.136
0.00456621 370.35
0.001156069 524.25
0.001257862 505.99
-0.0002
-0.0002
-0.0002
0.2090
0.1931
0.1870
0.1413
0.0928
0.0901
637.8
5.485
0.00152207
465.82
-0.0002
0.1878
0.0888
535.9
3.183
0.003703704
372.9
-0.0002
0.1804
0.1227
535.9
3.183
0.003703704
372.9
608.3
585.3
4.51
4.09
0.002237136
0.002597403
432.9
412.4
509.8
3.616
0.002898551 341.02
-0.0003
0.1991
0.1133
547.8
3.332
0.003703704 380.81
-0.0001
0.1793
0.1229
543.7
3.42
0.00308642
375.2
-0.0001
0.1598
0.1060
506.2
3.067
0.003636364 355.57
-0.0001
0.1427
0.1084
561
3.655
0.003067485
-0.0002
0.1813
0.1186
392.2
{0.1397}
0.1397
[[2]]
-0.0001 0.1828
-0.0002 0.1865
0.1397
0.1188
0.1187
119
family
N
Points
range
(mNm-1)
HC chains
HC chains
HC chains
62
4
4
0.1098-0.161
0.13-0.139
0.114-0.126
HC chains
24
HC chains
23
HC chains
HC chains
43
87
HC chains
HC chains
18
HC chains
HC chains
59
38
HC chains
19
HC chains
71
HC chains
29
HC chains
34
T range
(K)
190.-470.
310.-340.
310.-340.
235.460.0931-0.1259
335.94
174.10.1054-0.137
392.65
0.067643-0.24 100.-304.
0.14-0.207
160.-420.
0.073316192.20.16895
395.2
133.150.1176-0.1758
303.15
0.0709-0.21551 95.-190.
0.162-0.243
180.-480.
140.0.116-0.178
273.15
240.0.095842-0.146
433.15
330.0.1177-0.1521
473.15
310.150.0992-0.1522
613.15
508.3
605.4
536
Radius of
ryration
(A)
2.76
4.381
3.806
500.05
3.906
0.00259
341.45
-0.0003
0.1977
0.0939
584.1
4.797
0.00205
414.15
-0.0001
0.1657
0.1074
305.32
513.9
1.826
2.259
0.0068569
0.005991
184.57
351.39
-0.0008
-0.0003
0.3220
0.2478
0.1661
0.1531
417.9
2.875
0.0041841
266.25
-0.0004
0.2421
0.1227
484
3.213
0.003623188
-0.0004
0.2299
0.1179
190.564
512.5
1.118
1.552
-0.0015
-0.0003
0.3500
0.2826
0.1880
0.1921
425.12
2.886
0.0039228
272.66
-0.0005
0.2479
0.1164
617.7
5.148
0.00164
447.27
-0.0002
0.2015
0.0962
787
8.799
0.000675
641.75
-0.0002
0.2164
0.0886
768
8.364
0.000746
616.93
-0.0002
0.2004
0.0961
Tc (K)
Critical
Tb (K)
density
(molcm-3)
0.004504505 355.3
0.002304147 429.15
0.00274
366.97
307.20
5
0.010141988 111.66
0.0086
337.63
sigma@Tb
-0.0002
-0.0003
-0.0004
0.1928
0.2320
0.2500
0.1275
0.1033
0.1032
120
family
N
Points
range
(mNm-1)
HC chains
17
0.115-0.149
HC chains
65
0.0989-0.22
HC chains
90
0.0905-0.1562
HC chains
50
0.1008-0.21
HC chains
57
0.0964820.1597
HC chains
67
0.0998-0.23
HC chains
66
0.096-0.1494
HC chains
71
0.1017-0.18
HC chains
62
0.105-0.149
HC chains
47
0.098-0.1447
HC chains
56
0.0586-0.1695
HC chains
55
0.0979-0.1432
T range
(K)
330.473.15
295.13573.15
183.15453.15
291.31533.15
183.15366.48
305.04593.15
223.15420.
301.31573.15
220.-390.
283.15520.
146.27454.15
284.25509.8
Tc (K)
Radius of
ryration
(A)
Critical
density
(molcm-3)
Tb (K)
sigma@Tb
778
8.535
0.000746
629.65
-0.0002
0.2195
0.0820
736
7.355
0.0009
575.3
-0.0002
0.2104
0.0943
540.2
4.173
0.0023153
371.53
-0.0003
0.2038
0.1045
723
7.063
0.000961
560.01
4
-0.0002
0.2119
0.0872
507.6
3.769
0.0027059
341.86
-0.0003
0.2230
0.1033
758
7.983
0.000793
603.05
-0.0002
0.2110
0.0965
594.6
4.814
0.00181
423.91
-0.0003
0.2102
0.0949
747
7.655
0.00084
589.86
-0.0002
0.2025
0.0975
568.7
4.546
0.0020564
-0.0003
0.2088
0.0990
708
6.729
0.00103
398.77
543.83
5
-0.0002
0.2022
0.0899
469.7000
1
3.337
0.0032156
309.21
-0.0004
0.2215
0.1105
693
6.427
0.00111
526.72
7
-0.0002
0.1982
0.0908
121
family
N
Points
range
(mNm-1)
HC chains
15
0.1195-0.148
HC chains
31
0.089-0.142
HC chains
24
0.0945-0.142
HC chains
29
0.0696670.20535
HC chains
14
0.0818-0.219
HC chains
23
0.08989-0.1669
Ketones
34
0.125-0.1736
Ketones
10
0.11002-0.1432
Ketones
15
0.0948-0.153
Ketones
25
0.1226-0.154
Ketones
12
0.1096-0.1426
Ketones
27
0.123-0.1736
T range
(K)
350.473.15
280.501.45
293.15460.
93.15357.3
100.340.49
159.18419.18
184.45374.56
273.15413.15
273.15440.
248.15373.15
273.15413.15
187.86-
Tc (K)
Radius of
ryration
(A)
Critical
density
(molcm-3)
Tb (K)
sigma@Tb
796
9.065
0.000641
653.35
-0.0002
0.2184
0.0842
675
6.198
0.00121
-0.0002
0.1999
0.0914
639
5.496
0.00145
-0.0002
0.2062
0.0976
369.83
2.431
0.005
231.11
-0.0006
0.2652
0.1334
364.85
2.2311
0.005405405 225.45
-0.0006
0.2872
0.1481
530.6000
001
3.914
0.00261
363.23
-0.0003
0.2024
0.1086
535.5
3.135
0.00374
352.79
-0.0002
0.2199
0.1325
611.4
4.498
0.002304147 424.18
-0.0002
0.2108
0.1064
587.61
4.093
0.00264
400.7
-0.0003
0.2400
0.1083
561.08
3.618
0.00332
375.46
-0.0003
0.2221
0.1208
606.6
4.539
0.0023
420.55
-0.0002
0.2106
0.1072
560.95
3.583
0.00297
375.14
-0.0003
0.2231
0.1230
508.61
6
469.07
8
122
family
N
Points
range
(mNm-1)
Ketones
13
0.116-0.1505
Ketones
52
0.08-0.2034
Ketones
17
0.14351-0.292
Ketones
0.1528-0.1528
Ketones
0.12542-0.142
Ketones
0.1375-0.1527
Monocyclic
20
0.1003-0.1399
Monocyclic
40
0.0809-0.1264
Monocyclic
20
0.099-0.1439
Monocyclic
20
0.1042-0.1505
Monocyclic
20
0.099-0.1553
T range
(K)
373.15
243.15400.
183.15510.
279.4363.77
303.15303.15
273.15398.15
273.15373.15
247.79449.27
352.38613.15
229.33442.53
228.42437.89
265.75511.15
Tc (K)
Radius of
ryration
(A)
Critical
density
(molcm-3)
Tb (K)
sigma@Tb
602
4.567
0.0023
417.15
-0.0002
0.2057
0.1108
508.2
2.746
0.00478
329.44
-0.0004
0.2692
0.1484
709.6
4.09
0.002590674 475.26
0.0009
0.0883
0.3245
646.65
3.722
0.003095975 410.15
{0.1528}
0.1528
[[2]]
0.1528
653
3.511
0.003215434 428.58
-0.0001
0.1830
0.1193
624.5
3.165
0.003875969
403.8
-0.0002
0.1938
0.1324
664.5
4.127
0.002415459 449.27
-0.0002
0.1886
0.1003
676
4.502
0.002074689 469.99
-0.0002
0.1878
0.1059
649.1
4.199
0.002325581 442.53
-0.0002
0.1922
0.0990
637.3
4.375
0.002309469 437.89
-0.0002
0.2009
0.1042
696.6
5.964
0.00135318
0.0002
0.0379
0.1553
511.15
123
family
N
Points
range
(mNm-1)
T range
(K)
Monocyclic
79
0.099-0.154
562.05
Monocyclic
20
0.0992-0.1485
631.1
4.322
Monocyclic
20
0.1015-0.1574
617.15
3.897
0.00267
Monocyclic
46
617.15
3.897
Monocyclic
20
758
Monocyclic
41
0.0755-0.153
0.1061740.107448
0.0762-0.1474
Monocyclic
20
0.0923-0.1499
Monocyclic
Monocyclic
9
9
0.113-0.14
0.104-0.134
Monocyclic
20
0.0713-0.1526
273.1-450.
177.14413.15
178.2413.1
180.-580.
438.65536.6
225.3-580.
185.3473.15
280.-440.
280.-430.
173.55583.15
Radius of
ryration
(A)
3.004
Monocyclic
0.097-0.132
280.-440.
Monocyclic
43
0.0721-0.1429
Monocyclic
20
0.1056-0.1238
Monocyclic
Monocyclic
11
41
0.126-0.129
0.076-0.1325
247.98600.
327.45504.55
380.-480.
286.41-
Tc (K)
Critical
density
(molcm-3)
0.003902
Tb (K)
sigma@Tb
353.22
-0.0003
0.2359
0.1272
0.002304147 425.56
-0.0002
0.1855
0.0966
409.35
-0.0002
0.1999
0.1024
0.00267
409.35
-0.0002
0.1920
0.1046
4.894
0.001686341
536.6
0.0000
0.1132
0.1062
617
3.937
0.00266
-0.0002
0.1934
0.1051
660.5
4.849
0.002012072
-0.0002
0.1871
0.0956
698
679.9
5.63
5.189
0.001686341
0.001818182
-0.0002
-0.0002
0.1866
0.1900
0.1030
0.0943
638.35
4.344
0.002272727
-0.0002
0.1871
0.1013
668
4.398
0.001992032
412.27
456.45
5
499.26
478.61
432.39
1
456.60
8
-0.0002
0.1953
0.0929
630.3
3.836
0.0027
417.58
-0.0002
0.1943
0.1047
719.15
4.702
0.001858736 504.55
-0.0001
0.1574
0.1056
869
616.2
4.961
3.831
0.001805054 610.03
0.00264
411.51
0.0000
-0.0002
0.1393
0.1901
0.1221
0.1044
124
family
N
Points
range
(mNm-1)
Monocyclic
0.099-0.128
Monocyclic
20
0.0957-0.1303
Monocyclic
244
0.076-0.1614
T range
(K)
580.
280.-420.
215.27442.3
178.18550.
Tc (K)
Radius of
ryration
(A)
664.54
4.564
660
591.75
Critical
density
(molcm-3)
Tb (K)
sigma@Tb
0.002012072 446.48
-0.0002
0.1868
0.0933
4.318
0.00203252
442.3
-0.0002
0.1630
0.0957
3.472
0.003164
383.78
-0.0002
0.2025
0.1107
125
For the critical temperature, the acentric factor, and the enthalpy of fusion, a slight trend
was evident for HC chains, monocyclic compounds, and carboxylic acids. F+Cl+Brderivatives and F-derivatives generally showed random behavior.
126
A factor analysis was performed for the data selected in order to understand the structure of
an instrument and the relationships among variables [51].
Factor analysis is used for understanding the underlying conceptual structure of an
instrument and relationships among variables, as well as for refining instruments. It
involves analyzing relationships among items of an instrument or large numbers of
variables in order to obtain a smaller number of variables or factors. In this way, having a
smaller number of variables facilitates theory development and testing. Thus, factor
analysis offers not only the possibility of gaining a clear view of the data, but also the
possibility of using the output in subsequent analyses.
Let
and
x x
D +j = x D | x j >
2
lb
j
be the
127
And
lb
x ub
j xj
D = x D | xj <
is:
Eff (x j )=
xD +j
y (x )
( )
card D
+
j
xD j
y (x )
(2.39)
( )
card D
where
denotes the cardinality of the set .
For all the experimental data, the factor analysis was performed, so as to identify which of
the fluids parameters were more likely to influence the thermal conductivity
In this part, the factor analysis was performed for the following physical parameters:
reduced temperature, acentric factor, molecular mass, enthalpy of fusion at the melting
point. All physical parameters were collected from the DIIPR database [50]. For our
calculations, only values accepted by the database were considered. Results of the factor
analysis are reported in Table 2.4.
Table 2.4. Factor analysis results. Effects were calculated according to Equation (2.39) .
Physical property
Reduced temperature
Acentric factor
Molecular mass
Enthalpy of fusion at melting point
Eff(xj)
-0.045
0.022
-0.012
0.005
As a first comment, even if the databank was enriched with 438 data points, one can see
that the factor analysis confirmed, for the selected properties, the same trend of the results
obtained for the first equation (2.36). In particular, a very low statistical effect was obtained
for the enthalpy of fusion at melting point; Eff(xj) = 0.005.
As a result of the previous observations, and considering the factor analysis results in which
the enthalpy of fusion was found to have a lower statistical effect, it was decided to remove
this property when formulating the new equation. Moreover, the enthalpy of fusion is
generally significant for transitions from the crystal to fluid phases and is not so relevant
for the liquid state.
The resulting equation is:
f
e
= a + b Tr + d +
0
M
(2.40)
128
1].
67
AAD%
Refs. [4247]
8.32
AAD%
Ref.
[37]
13.26
AAD%
Ref.[ 2829]
10.92
318
7.12
15.74
221
8.38
20
AAD%
Eq.
AAD
% Eq.
13.05
7.86
9.83
AAD
%
ANN
3.25
28.37
9.53
6.63
4.67
2.78
7.76
13.71
6.39
4.00
4.12
2.70
6.68
6.12
7.44
5.22
3.98
4.34
2.77
56
13
245
37
9.64
3.99
11.73
3.23
11.27
8.86
6.08
0.85
11.38
14.15
9.86
4.03
10.74
8.69
3.23
10.16
10.16
4.90
2.94
10.52
13.34
5.21
3.31
11.96
2.43
4.82
5.53
7.28
594
5.61
8.39
5.25
11.33
8.51
8.87
5.63
348
2075
216
-
4.28
8.50
8.61
7.70
6.56
4.55
11.31
10.96
12.38
5.15
54.46
24.39
7.09
6.21
33.49
15.06
5.47
6.91
8.49
7.01
5.70
7.63
9.56
7.48
4.30
3.82
4.62
3.52
Family
NPoints
Br-derivatives
Carboxylic
Acids
Cl-derivatives
Condensed
Rings
Cycloalkanes
Cycloalkenes
Esters
Ethers
F+cl+brderivatives
F-derivatives
Hc chains
Ketones
Mean
AAD%
Ref.[40]
(2.36) (2.40)
The Latini equation is probably the most commonly known and adopted equation for
organic liquids [42-47] and is generally recommended [5]. It was proposed for several
chemical families of compounds including olefins, saturated hydrocarbons, cycloparaffins,
aromatics, alcohols, ketones, carboxylic acids, refrigerants and esters, but it has a
considerable drawback in that it is inaccurate for reduced temperatures above 0.65.
129
A recent equation that overcomes prediction limitations near the critical point was proposed
by Gharagheizi et al. [40], who adopted a GEP [41] mathematical strategy. This equation
also considered amines, silanes/siloxanes, inorganic compounds, sulfides/thiophenes,
mercaptanes, epoxides, peroxides, nitriles, elements, and aldehydes. On the other hand, the
main drawback of this work is that it was based on an empirical approach, with a huge
number of coefficients being correlated.
To test the goodness of the literature equations and Equation (2.22), deviations between the
experimental thermal conductivities were calculated and are reported as averages in Table
2.3 and Figure 2.7.
As a general comment, Equations (2.36) and (2.40) give the best results in terms of average
deviations. Equation (2.40) is a bit less precise than Equation (2.36), especially for few
families such as carboxylic acids and condensed rings. The main advantage of Equation
(2.36) lies in its simplicity, with only three physical parameters, namely molecular mass,
critical temperature, and acentric factor. However, the deviations produced by Equations
(2.36) and (2.40) are rather similar for all families. It is worth pointing out that for HC
chains, which form the most numerous set of data, deviations produced by Equation (2.40)
are even lower than those produced by Equation (2.36) .When compared with other
equations, the same order of deviations of Equations (2.36) (2.40) was obtained with the
Latini Equation [42-47], in which the coefficients were fixed for each family.
An Artificial Neural Network
In the same way described for the surface tension in par 1.5.1, For the modelling of thermal
conductivity, the Wolfram Mathematica artificial neural network toolbox is employed.
Figure 2.8 illustrates a diagram of a one-hidden-layer feedforward network adopting the
following inputs: molecular mass, M, acentric factor, , and reduced temperature, Tr, The
output is thermal conductivity, . Each arrow in the figure symbolises a parameter in the
network.
130
Hidden Layer
Input Layer
.
.
.
Figure 2.8. Schematic diagram of the ANN model. Inputs: M (Kg/Kmol) molecular mass, acentric
factor, Tr, reduced temperature. Output: (W m-1 K-1/m), thermal conductivity
Figure 2.9 illustrates the AAD for the training, validation and test, and the overall set of
data obtained for various neural network configurations with one hidden layer, achieved
after 1000 iterations during the modelling of the thermal conductivity. From the figure it is
evident that, as expected, AAD decreases with the increase of the number of neurons and
that a good configuration is reached when the number of neurons in the hidden layer is 39.
This configuration is selected as the best network architecture.
131
Figure 2.9. Absolute average percent deviation between data collected and ANN results vs number of
neurons in hidden layer
The AADs for the proposed ANN are reported in Table 2.5. The parameters of the hidden
layer and of the output layer are reported in Table 2.6.
Table 2.6. Parameters of hidden and output layers
hidden layer
neuron
number
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
bias
6.29
19.21
-3.18
-14.57
5.63
15.77
4.51
-20.37
36.68
-4.49
5.76
-10.63
-8.10
-3.98
-0.73
-1.34
20.38
-0.95
output layer
neuron
number
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
bias
-39.69
0.12
0.81
-0.61
29.58
2.17
0.23
-14.61
0.06
-21.99
32.91
-3.76
-15.27
-42.92
-30.08
-22.57
-15.13
-23.82
output bias
-15.00
132
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
-24.90
-2.93
-1.47
46.82
-5.57
-38.93
-6.59
-1.23
5.12
-23.18
0.86
-13.01
-9.49
-17.99
19.11
-5.91
-6.01
29.85
-53.57
-3.57
18.84
25.82
-8.65
151.74
-53.11
-0.15
68.42
6.16
120.26
-1.15
29.68
12.40
-4.84
-1.63
25.19
33.49
1.18
8.13
-50.06
-0.93
-9.31
-22.33
14.70
1.28
-0.27
0.64
5.73
1.04
-0.38
0.12
-8.08
7.43
-0.19
3.31
0.67
-0.56
-0.34
15.56
0.04
9.11
0.37
0.22
0.54
-15.91
4.18
-143.83
-28.16
2.05
20.47
31.50
-114.97
-2.41
-8.22
-22.88
17.50
17.97
7.39
-159.91
8.71
2.40
-1.64
81.81
26.56
-10.87
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
0.34
33.05
12.77
-2.21
-1.44
0.15
24.28
9.93
1.68
-0.60
7.04
2.34
8.98
5.07
-0.21
-1.21
42.67
0.21
1.01
6.71
12.27
Figure 2.9 shows the correlation between the predictions of the validated, trained and
tested MLP and the corresponding experimental thermal conductivity data. From the figure
a linear fit is evident, witnessing a good correlation between predictions and data. It is
worth noticing that the training dataset, even if linear, shows a slightly different trend
compared to the dataset tested and validated.
133
Figure 2.10. Correlation between the calculated and the corresponding experimental thermal
conductivity data.
Figure 2.10 illustrates the effect of the reduced temperature on the network accuracy.
According to the MLP configuration selected, with the exception of few points, residuals
are well within 0.02 Wm-1K-1
In concluision, thermal conductivity data were collected for 164 pure compounds. All of the
data obtained were analyzed so as to identify which of the fluid parameters were most
likely to influence the thermal conductivity. After the sensitivity analysis, starting with a
134
recently proposed equation, a new equation related to the most significant parameters was
proposed. The new correlation is very simple and gives a total average deviation of AAD =
7.48%. Experimental data were also compared with equations from the literature.
The multilayer perceptron proposed adopts as input parameters the same parameters
proposed for the scaled equation and has one hidden layer with 39 neurons, determined
according to the constructive approach. The model was validated, trained and tested for a
wide set of data, showing that the accuracy of the neural network model is very high.
2.4.2
135
quite different from non-polar compounds. This feature is also confirmed in the general
equation proposed in equation (2.36), which gives the highest deviations in terms of
thermal conductivity for refrigerants (AAD = 8.9%). This has suggested a possible
improvement of the general equation, specifically orienting the study to refrigerants.
All the refrigerants found were considered and ASHRAE notation was maintained. In the
end, a total of 41 fluids and 1372 points were selected. Statistical information on the
thermal conductivity of refrigerants is summarized in Figure 2.12, where the distribution of
experimental data is shown together with information regarding the distribution of outliers.
The mean value of the thermal conductivity experimental data was 0.0906 W m-1 K-1, with
a variance of 0.0008 and a standard deviation of 0.0285.
Br derivatives
Cl derivatives
F Cl Br derivatives
F derivatives
Relative Frequency
200
150
100
50
0.05
0.10
0.15
exp
0.20
0.25
W
m K
Figure 2.12. Histogram (left) and box plot (right) of experimental data on the thermal conductivity of
refrigerants
As a second step, we analyzed the statistical effects for the refrigerants, considering them to
be an entire set of data. The analysis showed the following statistical results: the largest
effects were obtained for the critical density (Eff = 0.034), critical pressure (Eff = 0.032),
reduced temperature (Eff = -0.031), and molecular mass (Eff = -0.027). A nearly negligible
effect was found for the enthalpy of fusion (Eff = 0.0033). Complete results of the factor
analysis for the full set of refrigerants are reported in Table 2.7. By analyzing the statistical
effects, it is expected that thermal conductivity should increase with critical density and
critical pressure, while it should decrease with reduced temperature and molecular mass.
Table 2.7. Factor analysis results. Effects were calculated according to Equation (2.40)
Physical property
Critical density
Critical pressure
Reduced temperature
Molecular mass
Triple point pressure
Reduced boiling temperature
Radius of gyration
Effects
0.0345
0.0319
-0.0316
-0.0269
-0.0268
-0.0263
-0.0238
136
-0.0126
0.0123
0.0119
0.0086
0.0086
-0.0053
-0.0053
0.0033
0.0011
Figure 2.13. Experimental data on the thermal conductivity for refrigerants vs. reduced temperature
The regression coefficients a and b are reported in Table 2, along with the range of
experimental data for thermal conductivities and temperatures, the number of points, and
the calculated thermal conductivity at the boiling point for each refrigerant. From Figures
3-6, a general trend for all refrigerants is clearly evident for thermal conductivities at the
137
boiling point when plotted against molecular mass and critical pressure, while the points
seem to be rather scattered when plotted versus critical temperature and acentric factor.
Table 2.8. Summary of experimental data and calculated thermal conductivity at the boiling point.
Refrigera
nt
R10
R11
R12
R12B1
R13
R13B1
R14
R20
R21
R22
R23
R30
R31
R32
R112
R113
R114
R114B2
R115
R116
R123
R123a
R124
R125
R130
R133a
R134a
R141b
R142b
R143a
R150
R150B2
R152a
R160B1
N. of
Point
s
89
77
84
17
28
20
14
31
41
43
28
24
9
54
3
85
63
7
33
13
64
8
22
43
35
8
133
22
33
4
31
17
37
25
T min
(K)
T max
(K)
min
(Wm-1K-1)
max
(Wm-1K-1)
250
173
118
203
100
113.15
91.35
213.15
148.15
148.15
123
178.15
170
148.15
303.15
193
178.15
168.15
170
178.15
159.34
308.05
234.2
172.74
280
170
169.87
240
223.15
233.65
223.15
280
163.15
200
470
438
366.4
373
293.15
303.15
152.75
348.15
433
350
280
323.15
290
344.35
373.15
453
413
323.15
343.65
280
354.15
341.55
366.2
338.15
473.15
310
367.95
392
403
298.85
363.15
373.15
363.15
327
0.0651
0.0489
0.0418
0.0545
0.048
0.053
0.0873
0.10062
0.0573
0.064
0.072
0.12995
0.116
0.0931
0.0695
0.0412
0.0414
0.0588
0.0366
0.043
0.0625
0.0666
0.0493
0.0471
0.0821
0.084
0.0501
0.0605
0.0488
0.0663
0.0753
0.08388
0.0753
0.0935
0.114
0.12
0.13
0.0945
0.134
0.117
0.1361
0.13716
0.15942
0.162
0.18
0.1848
0.203
0.24765
0.08105
0.09
0.09449
0.08329
0.0927
0.08523
0.1154
0.0734
0.0901
0.1116
0.1159
0.135
0.143
0.1119
0.1124
0.0994
0.14528
0.103
0.17117
0.112
@Tb
(Wm-1K1)
0.091
0.088
0.088
0.077
0.092
0.080
0.093
0.109
0.107
0.116
0.134
0.135
0.147
0.189
0.071
0.066
0.071
0.060
0.072
0.078
0.077
0.075
0.081
0.092
0.092
0.095
0.106
0.090
0.094
0.103
0.090
0.083
0.128
0.099
R2
-0.0002
-0.0003
-0.0003
-0.0002
-0.0005
-0.0003
-0.0008
-0.0002
-0.0004
-0.0005
-0.0007
-0.0003
-0.0006
-0.0008
-0.0002
-0.0002
-0.0002
-0.0002
-0.0003
-0.0004
-0.0003
-0.0002
-0.0003
-0.0004
-0.0002
-0.0004
-0.0005
-0.0003
-0.0003
-0.0005
-0.0004
-0.0002
-0.0005
-0.0001
0.1665
0.1680
0.1733
0.1403
0.1790
0.1533
0.2077
0.1850
0.2053
0.2213
0.2689
0.2424
0.3103
0.3616
0.1311
0.1345
0.1292
0.1094
0.1476
0.1596
0.1601
0.1389
0.1613
0.1818
0.1601
0.1963
0.2183
0.1960
0.1848
0.2174
0.2312
0.1497
0.2484
0.1372
0.9
0.9
0.9
0.9
0.9
1.0
1.0
0.9
0.9
0.9
1.0
0.9
0.9
1.0
1.0
0.7
0.8
1.0
1.0
1.0
1.0
1.0
0.9
0.9
0.9
1.0
0.9
1.0
0.9
1.0
0.4
0.8
0.9
0.8
138
R218
R236fa
R280
R280B1fb
R365mfc
R1234yf
R1234ze(
E)
22
13
11
25
8
24
130
237.14
273.15
230
263.15
283.15
340
377.25
313.15
333.15
333.15
353.15
0.0373
0.0621
0.1128
0.0926
0.0702
0.04941
0.1
0.1071
0.1224
0.112
0.1047
0.06824
0.067
0.094
0.111
0.092
0.082
0.078
-0.0003
-0.0003
-0.0002
-0.0002
-0.0005
-0.0003
0.1395
0.1786
0.1822
0.1542
0.2356
0.1438
1.0
1.0
0.9
0.9
0.9
0.9
24
283.15
353.15
0.05656
0.07919
0.089
-0.0003
0.1714
1.0
Considering both the trends in Figures Figure 2.14 Figure 2.17 and the statistical effects
reported in Table 2.7, Equation (2.36) was consequently modified expressly for the
refrigerants. It should be noted that the enthalpy of fusion was removed when formulating
the new equation. In fact, this quantity: a) is generally significant only for transitions from
the crystal to fluid phase and is not so relevant for the liquid state; b) was found to have a
lower statistical effect in the factor analysis; c) is not always available for fluids, as
confirmed by the recently synthesized propene derivatives (i.e., R1234yf and R1234ze(E)).
Figure 2.14. Thermal conductivity at the boiling point vs. molecular mass.
139
Figure 2.15. Thermal conductivity at the boiling point vs. critical temperature.
Figure 2.16. Thermal conductivity at the boiling point vs. critical pressure.
140
Figure 2.17. Thermal conductivity at the boiling point vs. acentric factor.
In spite its high statistical effect, reduced density was found to be not as relevant in the
regression procedures, so it has not been considered as a parameter. On the other hand, the
critical pressure term effectively improved the calculation of thermal conductivity.
Following these considerations, the general equation (2.36) was finally modified, removing
the enthalpy of fusion term and adding a term containing the critical pressure as follows:
where is the dipole moment reported in Table 2.9. The respective coefficients were then
regressed again with the Levenberg-Marquardt curve-tting method and are reported in
Table 2.10. Equation produces deviations of AAD = 6.1% when the coefficients are kept
constant. Deviations are reported in Table 2.9.
Table 2.9. Summary of deviations for equations.
AAD
Eq
*AAD
Eq.
Series
(D)
Numb
er of
Points
R10
Cl-
0.000
89
5.5
3.4
R11
F+Cl+Br-
0.446
77
2.0
1.9
R12
F+Cl+Br-
0.506
84
3.0
R12B1
F+Cl+Br-
0.839
17
5.1
Fluid
*AAD
Eq.
AAD
[40]
AAD
[37]
1.8
17.0
3.2
6.0
2.2
26.1
6.3
7.6
92.8
3.4
3.9
41.3
11.4
7.2
117.1
5.2
3.4
14.1
9.2
5.3
77.7
(2.40 (2.41)
(2.41)v
)
f
AAD
[52]
AAD
[28-29]
29.5
141
R13
F+Cl+Br-
0.506
28
5.0
5.3
5.7
26.9
15.8
4.8
79.6
R13B1
F+Cl+Br-
0.646
20
3.5
4.0
3.5
37.4
10.9
13.9
44.5
R14
F-
0.000
14
3.7
4.0
2.0
20.5
19.2
11.9
19.4
R20
Cl-
1.003
31
2.3
2.1
2.9
38.8
5.7
4.1
9.9
R21
F+Cl+Br-
1.280
41
3.6
3.3
4.7
24.4
3.4
5.5
37.5
R22
F+Cl+Br-
1.447
43
4.7
4.5
6.2
35.6
3.8
2.8
48.4
R23
F-
1.637
28
8.3
8.2
5.9
20.6
3.0
3.4
23.5
R30
Cl-
1.589
24
10.3
10.9
3.8
49.2
11.2
2.8
17.1
R31
F+Cl+Br-
1.807
15.7
15.8
19.1
48.8
12.4
5.9
20.8
22.2
R32
F-
1.964
54
11.1
10.2
8.3
40.5
25.0
2.9
R112
F+Cl+Br-
0.735
2.7
2.5
2.0
59.0
10.4
8.8
92.3
R113
F+Cl+Br-
0.797
85
10.6
10.7
9.8
40.9
22.9
10.3
124.1
R114
F+Cl+Br-
0.664
63
9.4
9.5
7.9
36.8
18.8
6.9
105.6
R114B2
F+Cl+Br-
0.878
9.3
8.5
7.3
76.0
16.0
5.0
112.7
R115
F+Cl+Br-
0.516
33
3.9
3.5
2.1
37.0
20.1
3.5
128.3
R116
F-
0.000
13
5.7
2.1
6.6
25.5
15.0
13.7
81.0
R123
F+Cl+Br-
1.346
64
5.7
4.8
2.4
24.5
15.5
4.4
51.5
R123a
F+Cl+Br-
1.293
1.8
1.4
1.6
39.3
17.5
1.1
75.7
R124
F+Cl+Br-
1.458
22
4.7
3.6
2.6
38.0
13.8
2.3
79.8
R125
F-
1.530
43
3.1
3.0
4.5
24.1
9.4
12.2
63.8
R130
Cl-
1.280
35
9.3
11.2
2.7
26.8
5.7
4.7
24.3
R133a
F+Cl+Br-
1.637
6.5
8.4
11.7
29.4
4.3
9.1
30.5
R134a
F-
2.043
133
4.6
4.4
3.2
23.0
4.6
8.4
47.7
R141b
F+Cl+Br-
1.999
22
4.5
3.0
5.0
16.0
12.4
6.4
40.8
R142b
F+Cl+Br-
2.125
33
7.8
2.9
2.4
63.2
16.2
3.9
96.3
R143a
F-
2.303
7.4
4.6
8.4
20.5
18.4
2.2
48.5
R150
Cl-
1.429
31
20.3
21.3
12.0
28.9
18.8
16.6
41.1
R150B2
Br-
1.003
17
9.1
7.0
1.5
6.8
2.8
20.5
18.5
R152a
F-
2.246
37
4.8
4.6
8.7
26.1
5.8
6.4
23.6
R160B1
Br-
2.015
25
14.1
10.0
1.7
2.2
11.5
15.9
31.3
R218
F-
0.000
22
7.8
8.7
16.7
46.1
17.6
33.5
114.6
R236fa
F-
1.968
13
17.6
18.1
14.3
25.1
13.3
5.7
59.9
R280
Cl-
2.134
11
0.7
1.8
1.6
25.1
14.0
20.3
10.9
R280B1fb
Br-
2.134
25
10.3
7.3
1.1
4.4
11.7
19.0
27.2
R365mfc
F-
3.780
4.9
10.4
11.9
9.8
7.9
21.8
57.0
R1234ze(E)
F-
1.270
24
4.5
7.3
3.0
13.4
13.0
102.0
R1234yf
F-
2.480
24
5.4
4.3
5.4
69.7
18.4
178.7
Mean
6.5
6.1
4.9
11.1
11.5
8.0
56.1
*f is fixed coefficients
v is variable coefficients.
142
(2.40)
Eq.
(2.41)f
Eq.
(2.41)v
Number
of
Fluids
41
0
(Wm1K-1)
b
(bar-1)
e
(kg kmol1)
0.5147
-0.2537
0.0017
0.1501
-0.299
0.6542
0.1108
0.0025
0.0775
0.5094
-0.2034
-1.1193
-63.582
-1.1594
-0.2377
0.0013
0.0128
1.6465
-0.117
0.0017
0.1714
-2.5399
-39.291
-26.692
0.1877
0.3539
-0.071
-0.721
-0.459
0.3149
1
1
1
1
1
41
ClFBrF+Cl+Br-
6
13
3
19
f
(D)
0.0070
0.1607
1.4770
2.6639
-0.008
In addition, in the literature on thermal conductivity, the refrigerant family has been the
object of particular attention in recent papers [52-55]. In some studies, the three main
groups of hydrocarbon derivatives, i.e., the methane, ethane, and propane series, were
considered separately [52, 55] and the coefficients for the equations proposed were
regressed expressly for each separate group [52]. In this paper, a different approach has
been followed and experimental data for the thermal conductivity of the refrigerants have
been divided into four subgroups according to the different chemical halogens forming the
compounds: namely Br- derivatives, Cl- derivatives, F- derivatives, and F+Cl+Brderivatives, when more than one halogen is present in the refrigerant molecule. For
Equation (2.41), the coefficients were again regressed using the Levenberg-Marquardt
curve-tting method, but the regression was performed by grouping the refrigerants
according to the different chemical halogen constituent, as explained above. Deviations
produced by applying this approach to Equation (2.41) are reported in Table 2.9, while the
respective coefficients are reported in Table 2.10.
As expected, by analyzing the deviations, it is evident that the new approach improves the
calculation of thermal conductivity, producing an AAD = 4.9% when coefficients are
regressed for each subgroup separately. It should be pointed out that an AAD% = 5.4% can
be obtained with Equation (2.41) when the refrigerants are regressed divided into four
subgroups according to the classical division of groups of hydrocarbons, namely the
methane, ethane, propane, and propene series, as found in the literature [52, 55]. The
propane and propene series were regressed, even though they present a very limited number
of fluids for which experimental data have been collected; note also that R365mfc, a butane
derivative, was not considered in this regression.
Equations (2.40) and (2.41) were also compared to the most widely applied, simple,
general, semi-empirical correlations for the calculation of liquid thermal conductivities:
Reid-Prausnitz-Poling [37], Sheffy and Johnson [28-29] Latini et al. [42-47], and
Gharagheizi et al. [40]. The Reid-Prausnitz-Poling [37], is a rather simple equation that
requires knowledge of experimental temperature, boiling point, molecular mass, and critical
temperature. The Sheffy and Johnson equation [28-29] is also a very simple equation,
containing only experimental temperature, the melting point, and molecular mass as
physical parameters. The Latini equation is probably the most commonly known; it has
been adopted for organic liquids [42-47] and is generally recommended [5]. It has been
proposed for several families of chemical compounds including olefins, saturated
143
Figure 2.18. Thermal conductivity deviations of equation (2.41) vs. reduced temperature.
As a general comment, inserting the dipole moment into Equation (2.41) slightly improves
the calculation of thermal conductivity, especially for fluids where the dipole moment is
higher than 2 Debye (i.e., R142b, R143a, R160B1, R280B1fb). In spite of its high dipole
moment, surprisingly poor results were obtained for R365mfc. The best results were
obtained from Equation (2.41)when the coefficients are regressed for each subgroup.
However, in spite of its simplicity, Equation (2.40)also gives satisfactory results in terms of
144
deviations. For this reason, we recommend using Equation (2.41) adapted for each
subgroup whenever possible, but Equation (2.40) represents a valid choice for all other
cases.
Statistical overview
In Figure 2.19, all the collected data are reported as a function of the reduced temperature.
From this figure, excluding few points for alcohols and carboxylic acids, a general common
trend for all compounds as function of reduced temperature is evident. The whole database
has been studied according to the an explanatory data analysis [57] with a statistical chart;
this approach allows to suggest reliable models fitting the data, taking into account the
structure and the outliers.
145
0.30
Alcohols
0.25
r 2 = 0.392987
0.20
0.15
0.10
Ethers
Alkanes
Ketones
Alkenes
Refrigerants
Aromatics
Carboxylic acids
Cycloalkanes
Cycloalkenes
0.05
Esters
0.4
0.6
0.8
1.0
Figure 2.19. Scatter Plot of experimental thermal conductivity data versus reduced temperature.
Confidence bands at , 2 and 3 are also reported.
In Figure 2.20 and Figure 2.21, the statistical summary for thermal conductivity and
reduced temperature for the entire dataset are reported.
On the left, the histogram that underlines the smooth distribution of the values of thermal
conductivity and reduced temperature is reported; it allows to see where the majority of
values falls in a measurement scale, and how much variation present is. The middle chart is
the box whisker chart, that helps in understanding if the data points are clustered around
some central value, where are the lower and upper quartile ad the median; this chart
contains also the representation of some outliers signed out of the whiskers. The right chart
shows the distribution chart that draws a representation of the distribution of values in each
point. Then, the summary of the number of points, the mean, the variance and the standard
deviation are also reported.
1200
1000
800
Exp
600
400
0.30
0.30
0.25
0.25
0.20
0.20
0.15
Exp
Relative Frequency
1400
0.15
200
0.10
0.10
0.05
0.05
0.05
0.10
0.15
Exp
0.20
0.25
0.30
146
Relative Frequency
800
1.0
400
0.8
0.8
200
0
0.4
0.6
0.8
1.0
Tr
1.2
1.0
Tr
1.2
600
0.6
0.6
0.4
0.4
0.2
0.2
Tr
After the general analysis, a more detailed family by family analysis was performed, as
shown in Table 2.11
Alcohols
50
Alcohols
0.25
0.25
0.20
0.20
0.15
0.15
100
0.10
0.10
0.15
0.20
0.10
0.25
Alkanes
Alkanes
400
200
100
0.10
0.15
0.20
0.25
0.20
0.15
0.10
0.05
Alkanes
300
0.25
0.20
0.15
0.10
0.05
0.25
Alkenes
40
Alkenes
Alkenes
0.20
20
0.15
0.15
0.10
10
0
0.20
30
0.10
147
Aromatics
120
Aromatics
Aromatics
80
60
40
20
0.08
0.10
0.12
0.14
0.16
0.14
0.12
0.10
0.08
100
0.16
0.14
0.12
0.10
0.08
0.06
0.16
Carboxylic acids
Carboxylic acids
0.15
0.20
0.25
0.30
0.25
0.20
0.15
0.10
Carboxylic acids
120
100
80
60
40
20
0.30
0.25
0.20
0.15
0.10
0.30
Cycloalkanes
12
10
8
6
0.150
0.145
0.140
0.135
0.130
0.125
0.120
0.115
4
2
Cycloalkanes
0.16
0.14
0.12
0.10
Cycloalkanes
0.16
0.15
0.14
0.13
0.12
0.11
0.10
0.16
0.15
0.14
0.13
0.12
0.11
0.10
Cycloalkenes
6
Cycloalkenes
Cycloalkenes
5
4
3
2
1
0.11 0.12 0.13 0.14 0.15 0.16
Esters
0.08
0.10
0.12
0.14
0.16
0.14
0.12
0.10
0.08
0.06
Esters
Esters
70
60
50
40
30
20
10
0
0.16
0.14
0.12
0.10
0.08
0.06
0.16
148
Ethers
25
5
0.08
0.10
0.12 0.14
0.16
10
0.18
0.16
0.14
0.12
0.10
0.08
0.18
0.16
0.14
0.12
0.10
0.08
0.06
15
Ethers
Ethers
20
0.20
0.18
0.16
0.14
0.12
0.10
0.08
0.18
Ketones
40
Ketones
30
20
10
0
0.20
0.18
0.16
0.14
0.12
0.10
0.08
Ketones
Refrigerants
200
100
50
0.05
0.10
0.15
0.20
0.25
0.20
0.15
0.10
0.05
Refrigerants
Refrigerants
150
0.25
0.20
0.15
0.10
0.05
0.25
From Table 2.11, it is possible to point out that the main part of the experimental data is
distributed from 0.03 to 0.18 W/mK and that the maximum number of experimental data
was obtained at around = 0.12-0.13 W/mK, with the exception of refrigerants, ketones,
aromatics, alcohols and carboxylic acids.. It is also evident that the families that show the
higher number of experimental data are refrigerants, alkanes, alcohols and aromatics.
Furthermore, all the distributions are quite regular excluding points at high value of thermal
conductivity for carboxylic acids.
The new equation for aromatic
After the general statistical analysis, the research was specifically oriented to aromatic
family containing a benzene ring of six carbon atoms, exception doing for naphtalene and
phenanthrene that contain a double ring.
To find potential outliers, the scatter plot for each compound was reported in Figure 2.22.
The experimental data of 1,3,5 Tris (1-Methylethyl) clearly showed anomalous behaviour
if compared with the rest of the experimental data.
149
0.10
0.15
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1,2,4,5- Tetramethylbenzene
Pentamethylbenzene
1,2,4- Trimethylbenzene
Naphtalene
Phenanthrene
1,3,5- Trimethylbenzene
n- Butylbenzene
n- Hexylbenzene
sec- Butylbenzene
Benzene
n- Pentylbenzene
tert- Butylbenzene
Cumene
n- Propylbenzene
Ethylbenzene
o- Diethylbenzene
0.05
Hexamethylbenzene
1,2,3- Trimethylbenzene
Toluene
Tr
Figure 2.22. Scatter Plot of experimental thermal conductivities as a function of reduced temperature.
Name
1-2-3Trimethylbenzene
1-2-4-5Tetramethylbenzene
1-2-4Trimethylbenzene
1-3-5Trimethylbenzene
1-3-5-Tris(1Methylethyl)benzene
Benzene
N
points
20
40
20
20
20
37
T range
247.79449.27
352.38613.15
229.33442.53
228.42437.89
265.75511.15
278.68- 450
177.14413.15
Cumene
20
Ethylbenzene
41
178.2- 580
Hexamethylbenzene
20
438.65- 536.6
35
225.3- 580
370- 410
n-Butylbenzene
20
185.3- 473.15
n-Hexylbenzene
n-Pentylbenzene
9
9
n-Propylbenzene
20
o-Diethylbenzene
o-Xylene (1-2
9
36
280- 440
280- 430
173.55583.15
280- 440
247.98- 600
m-Xylene (1-3
dimethyl benzene)
Naphtalene
range
[W/mK]
0.10030.1399
0.08090.1264
AAD%
0.5
0.107
0.58
0.4
0.114
0.85
0.099- 0.1439
0.5
0.108
0.42
0.10420.1505
0.5
0.113
0.56
0.099- 0.1553
-0.5
0.135
1.25
0.099- 0.1492
0.09920.1485
0.07550.1574
0.1061740.107448
0.07620.1474
0.128- 0.132
0.09230.1499
0.113- 0.14
0.104- 0.134
0.07130.1526
0.097- 0.132
0.0721-
0.5
0.129
0.86
0.5
0.104
0.43
0.5
0.110
1.23
0.1
0.107
0.63
0.5
0.110
1.34
0.5
0.116
0.93
0.5
0.105
0.45
0.4
0.5
0.115
0.106
0.34
0.13
0.5
0.108
1.11
0.4
0.5
0.106
0.111
2.68
1.33
150
dimethyl benzene)
0.1429
0.10560.1238
0.126- 0.129
0.3
0.111
0.56
0.1
0.124
0.36
Pentamethylbenzene
20
Phenanthrene
p-Xylene (1-4
dimethyl benzene)
sec-Butylbenzene
11
327.45504.55
380- 480
35
286.41- 580
0.076- 0.1325
0.5
0.111
1.28
280- 420
0.5
0.102
0.59
tert-Butylbenzene
20
215.27- 442.3
0.4
0.101
0.19
Toluene
95
178.18- 550
0.099- 0.128
0.09570.1303
0.076- 0.1614
0.5
0.115
1.29
Tr )
(
= A 5
+ Tr )
(
where the coefficients A and a are tuned for each compound. The obtained coefficients are
listed in the Table 2.12, where the AAD % according to equation (2.38) are also reported
Since the two parameters a and A are tuned for each compound, Equation (2.42) shows very
low deviations. The aim of this research was to find a version of Eq (2.42) without losing in
terms of prediction capability.
At the beginning, a non linear regression of the coefficient a, varying A, was performed, as
reported in Table 2.12,. In this way, having just one varying parameter, slightly higher
deviations were observed. The coefficients are reported in Table 2.13, attempt 1.
Table 2.13. Summary of the coefficients.
Attempt 1
Attempt 2
Attempt 3
Attempt 4
a
0.434
0.410
0.469
0.434
A
as reported in Table 1
0.156
as reported in Table1
0.158
As second step, both a and A were regressed for the complete set of data. The exponents a
and the factors A are reported in Table 2.13, attempt 2.
Deviations for the equation (2.42) attempts 1 and 2 are summarized in Table 2.14.
Table 2.14. AAD% comparison between Equation(2.42), Eq. (2.42) Attempt 1 and Eq. (2.42)
Attempt 2
151
Fluid
1,2,3-Trimethylbenzene
1,2,4,5-Tetramethylbenzene
1,2,4-Trimethylbenzene
1,3,5-Trimethylbenzene
1,3,5-Tris(1Methylethyl)benzene
Benzene
Cumene
Ethylbenzene
Hexamethylbenzene
m-Xylene (1,3 dimethyl
benzene)
Naphtalene
n-Butylbenzene
n-Hexylbenzene
n-Pentylbenzene
n-Propylbenzene
o-Diethylbenzene
o-Xylene (1-2 dimethyl
benzene)
Pentamethylbenzene
Phenanthrene
p-Xylene (1,4 dimethyl
benzene)
sec-Butylbenzene
tert-Butylbenzene
Toluene
Mean
0.58
0.85
0.42
0.56
AAD% Eq
(2.42)
Attempt 1
1.71
0.86
1.54
1.91
1.25
26.46
20.35
0.86
0.43
1.23
0.63
1.99
1.96
2.51
3.76
11.80
5.39
3.13
3.82
1.34
2.32
2.81
0.93
0.45
0.34
0.13
1.11
2.68
1.43
1.98
1.17
1.51
1.83
2.10
4.86
4.94
1.53
4.17
4.36
6.56
1.33
2.65
3.60
0.56
0.36
2.63
10.55
2.68
2.34
1.28
1.92
2.82
0.59
0.19
1.29
0.95
1.62
1.19
1.97
3.02
8.66
12.00
3.00
4.74
AAD% Eq.
(2.42)
From Table 2.14, it is evident, as expected, that the higher value of AAD% is for 1,3,5
Tris (1-Methylethyl). Then, regressions were repeated after removing this set of data.
Attempts 1 and 2 became respectively attempts 3 and 4. The respective coefficients are also
reported in Table 2.13.
To test the validity of the new equation, for comparison three different literature equations
were considered: The Reid-Prausnitz-Poling equation (2.29), the Sheffy and Johnson
equation (2.19), and the Gharagheizi et al. equation (2.32)To test the goodness of the
existing equations, deviations between the experimental thermal conductivities and the
predicted ones were calculated and reported in Table 2.15.
152
N.
Poin
ts
AAD
%
Eq.(2.
42)
AAD%
Eq. (2.42)
Attempt 3
AAD%
Eq.
(2.42)
Attemp
t4
AAD
%
Eq.
(2.29
)
AAD
%
Eq.(2
.19)
AAD
%
Eq.
(2.32
)
20
0.58
0.95
2.88
2.11
2.59
6.66
40
0.85
1.82
2.61
13.16
2.12
3.71
20
0.42
0.62
2.56
2.69
2.45
5.41
20
0.56
1.09
2.73
1.85
2.28
0.95
37
20
41
0.86
0.43
1.23
1.37
0.74
1.64
12.37
5.66
2.33
2.16
7.56
5.36
12.54
4.41
4.64
3.24
8.10
3.93
20
0.63
4.08
3.74
10.24
3.15
4.66
35
1.34
1.71
2.12
5.72
7.63
4.76
5
20
9
9
20
9
0.93
0.45
0.34
0.13
1.11
2.68
0.92
0.83
2.02
0.76
0.75
1.74
4.85
5.03
1.50
4.13
3.85
6.45
10.69
0.86
9.93
2.46
4.56
1.29
6.02
1.71
14.26
6.24
1.15
2.62
1.03
4.51
3.33
2.71
3.60
5.16
36
1.33
1.77
2.83
5.60
9.23
6.56
20
0.56
3.23
2.87
6.03
3.69
4.53
11
0.36
11.68
2.50
14.42
16.37
7.32
35
1.28
1.41
2.15
6.15
13.82
9.40
8
20
95
0.59
0.19
1.29
1.00
2.27
1.31
8.62
12.13
2.80
Total
550
18.15
43.71
96.72
Mean
0.95
1.75
4.06
1.60
4.61
5.58
124.6
3
5.69
2.08
6.57
8.48
134.0
3
6.59
5.01
7.83
4.80
107.2
1
5.03
Fluid Name
1,2,3Trimethylbenzene
1,2,4,5Tetramethylbenzen
e
1,2,4Trimethylbenzene
1,3,5Trimethylbenzene
Benzene
Cumene
Ethylbenzene
Hexamethylbenzen
e
m-Xylene
(1,3
dimethyl benzene)
Naphtalene
n-Butylbenzene
n-Hexylbenzene
n-Pentylbenzene
n-Propylbenzene
o-Diethylbenzene
o-Xylene
(1-2
dimethyl benzene)
Pentamethylbenzen
e
Phenanthrene
p-Xylene
(1,4
dimethyl benzene)
sec-Butylbenzene
tert-Butylbenzene
Toluene
153
From Table 2.15 it is evident that the simplification of Eq. (2.42)by Eq (2.42)attempt 3 and
4 was successfully obtained with a minimal lost in terms of deviations. Other equations
existing in the literature generally showed higher deviations for aromatics. However, in as
to be considered that literature equations were not specifically oriented to aromatics, but to
a generic organic compound. Furthermore, the Eq. (2.42) attempt 3 and 4 are very simply,
containing only reduced temperature as physical property.
In conclusion, after a careful literature survey, the available data of the experimental
thermal conductivity data were collected and statistically analyzed. Particular attention was
then focused on aromatic compounds. Experimental data for aromatics were regressed with
the most reliable semi-empirical correlating methods based on the corresponding states
theory existing in the literature. A new and simple equation was developed for the thermal
conductivity calculation of aromatics.
154
2.6 Bibliografia
[1.] Weber The thermal Conductivity of Liquids (in German). Report Phys.
1886;22:116-22.
[2.] Dick MF, Cready DW. The thermal conductivities of some organic liquids, Trans.
ASME. 1954;76:831.
[3.] Bridgman PW. Thermal Conductivity of Liquids under Pressure. Proc Amer Acad
Arts Sci. 1923;59:141-69.
[4.] Vargaftik NB. Izvestiya VTI. 1949;8:6.
[5.] B. E. Poling, J. M. Prausnitz.and J. P. O'Connel, in The Properties of Gases and
Liquids, 5th edn. (McGraw-Hill, New York, 2001), pp.10.1-10.70.
[6.] E. McLaughlin, Chemical Review, 64:389 (1964).
[7.] IUPAC, , in Reccommended Reference Materials for the Realization of
Physicochemical
Properties,
K.N
Marsh.ed.
(Blackwell
Scientific
Publications,London 1987), p.322.
[8.] J. P. Lennard-Jones, Physica, 4:941 (1937).
[9.] H. F. Weber, Rept. Phys., 22:116 (1886).
[10.] P. Debye, in Kinetische Theorie der Materie, (Teubner, Leipzig, 1914).
[11.] J. F. D Smith, Trans. ASME, 58:719 (1936).
[12.] G. Palmer, Ind. Eng. Chem., 40:89 (1948).
[13.] N. B. Vargaftik, Proc. Joint Conf. on Thermodynamic Transport Prop. of Fluids,
publ. by Inst. Mech. Engrs., London, p.142, (1958).
[14.] L. A. Robbins and C. L. Kingrea, Hydrocarbons Processing and Petroleum
Refiner, 41(5):133 (1962).
[15.] P. W. Bridgmann, Proc. Amer. Acad.,57:75, (1922).
[16.] P. W. Bridgmann, Proc. Amer. Acad.,59:141 (1923).
[17.] P. W. Bridgmann., Amer. J. Sci.,7:81 (1924).
[18.] I. Osida, Proc. Phys. -Math. Soc. Japan, 21:353 (1939).
[19.] J. O. Hirschfelder, C. F. Curtiss and R. B. Bird, in Molecular Theory of Gases and
Liquids, 2nd printing (John Wiley & Sons, New York, 1964) section 9.1.
[20.] D. S. Viswanath, Am. Inst. Chem. Eng. J., 13(5):850 (1967).
[21.] A. Kardos, Forsch. Gebiete Ingenieurw, 5B:14 (1934).
[22.] B. C. Sakiadis and J. Coates, A.I.Ch.E. J., 1(3):275 (1955).
[23.] B. C. Sakiadis and J. Coates, A.I.Ch.E. J., 3(1):121 (1957).
[24.] L. Badea, Revista de Chimie, 26(3):209 (1975).
[25.] K. S. Narasimhan., K. M. Swamy and K. L. Narayama, Chem. Eng., 82(8):83
(1975).
[26.] J. F. D. Smith, Ind. Eng. Chem., 22:1246 (1930).
[27.] K.G., J. Denbigh, Soc. Chem. Ind., 65:61 (1946).
[28.] W. J. Scheffy, Tech.Rept. PR-85-R, Princeton Univ., Princeton, N.Y., (1958).
[29.] W. J. Scheffy and E. F. Johnson, J. Chem. Eng. Data, 6(2):245 (1961).
[30.] F. A Missenard, Rev. Gn. Therm., Fr. 141:751 (1973).
[31.] V. Pachaiyappan, S. H. Ibrahim and N. R. Kuloor, J. Chem. Eng. Data, 11(1):73
(1966).
[32.] V. Pachaiyappan, S. H. Ibrahim and N. R. Kuloor, Chem. Eng., 74:140 (1967).
155
156
157
Awhere B is the second virial coefficient, C is the third virial coefficient and D is the fourth
virial coefficient. Even when truncated at the second coefficient, the virial equation of state
gives a remarkable estimate of the PVT relationship of real gases at low and moderate
pressures (generally less than 1 MPa). Because of this, estimation of second virial
coefficients is an important task in thermodynamics. In addition, virial coefficients are an
important factor because they form a link between the microscopic and macroscopic points
of view, experimental results, and the knowledge of molecular interactions. In fact, the
second virial coefficient represents the deviation from perfection due to interactions
between pairs of molecules, the third virial coefficient reflects the effects of interactions of
molecular triplets, and so on.
158
(3.3)
He also introduced a corrective function to estimate the second virial coefficients of polar
gases:
(3.4)
where Tr is the reduced temperature, and aT and bT are tabulated parameters related to the
reduced dipole moment, r , of the compound studied. The final form of the Tsonopoulos
equation is:
(3.5)
where Pc is the critical pressure (Pa) and is the acentric factor.
Vetere [3] proposed a new version of the Pitzer-Curls correlation [2]. To this equation he
added a new factor, V, and an additional term,
, where:
(3.6)
(3.7)
Tb and M in equation (3.6) are, respectively, the normal boiling point temperature and
molar mass. He obtained a correlation that describes the temperature dependence of B for
polar and non-polar gases:
(3.8)
OConnell and Prausnitz [4] adopted both the first and the second terms,
and
, of the Pitzer-Curls correlation for non-polar gases. For polar gases, they
introduced two additional functions,
and
, based on the extended theory
of corresponding states in order to consider the polarity of compounds and their selfassociation trend. Their equation is shown below:
(3.9)
159
(3.10)
The OConnell-Prausnitzs complete equation reads as follows:
(3.11)
where H is the acentric factor of the polar components homomorph and is a constant of
association for the compound.
Black [5] suggested a Van der Waals-type equation of state
(3.12)
where expresses the effects of temperature and pressure on the molar cohesive energy.
Starting from this equation of state, Black proposed a correlation for B:
(3.13)
Where
(3.14)
(3.15)
as follows:
(3.16)
[7].
function:
(3.17)
160
In light of new data appearing in the open literature, especially for polar fluids such as
haloalkanes, detailed analysis on second virial coefficients predictions by the Tsonopoulos
and Weber equations were performed [8-9]. Important modifications to the polar term were
recently proposed [10], with a particular attention to associated fluids, such as alcohols,
amines and water, and quantum fluids [11]. A different approach to finding a correlation
capable of estimating the second virial coefficients is based on the group contribution
method. The basic principle of this method is that every compound can be divided into
simpler subgroups and that these groups contribute to the property examined. McCann and
Danner [12] proposed an equation capable of predicting the second virial coefficients based
on group additivity. Group contributions were derived by analyzing available second virial
coefficient data, and are represented by the equation:
(3.18)
For most groups, only the first four terms were required. Second virial coefficients for any
organic compound can be calculated from these group contributions and critical
temperatures, for reduced temperatures from 0.5 to 5, by summing up the products of the
group contribution (primary and secondary) and their respective number of occurrences, n:
(3.19)
The secondary group was defined only for the following groups: C-C2H2, C-C2F2, Cb-F.
In any case, the Bi term, (cm3/mol), was calculated with equation (3.18) both for the
primary and the secondary group.
161
3.3 Results
Techniques most commonly used for the estimation of the second virial coefficients are
based on the corresponding states principle. They are generally semi-empirical correlating
methods and their validity is usually limited to non-polar gases or to small polar molecules.
In this work, an estimation technique for the second virial coefficients based on a modified
group contribution method is presented. This method can be applied to a wide range of
organic compounds. In fact, the method depends on the knowledge of reduced temperature
and acentric factor, and is based only on summing the products of the group contributions
and their respective number of occurrences. Each group contribution is derived by analysis
of the second virial coefficient data available in the literature.
All estimation techniques generally need a large set of experimental data. In fact, through
experimental data analysis, it is possible to find a function that describes the correlation
between the second virial coefficients and the physical/molecular properties. Therefore, a
data bank was created by collecting experimental data on the second virial coefficients
available in the literature [13-14]. It includes 3441 experimental data of second virial
coefficients and their respective experimental temperatures. Data were collected for 157
compounds belonging to 19 families. In Figure 3.1, a statistical survey of the data
distribution of second virial coefficients is reported. From Figure 3.1, information on the
dataset distribution and on the presence of the outliers inside the data set are reported. Each
datum lying outside the defined bounds can be considered as outlier.
400
Relative Frequency
300
200
100
- 1000
- 3000
- 3000 - 2500 - 2000 - 1500 - 1000
- 500
- 2000
Figure 3.1. Histogram, data distribution, and statistical summary of the second virial coefficient
experimental data.
In our case, it is evident that the considered data are only out of the lower fence. In
particular, 55 outliers were detected. All the experimental data by chemical family are
shown in Figure 3.2, where the scatter plot of the second virial coefficients versus the
reduced temperature is reported. In Figure 3.2, a common behaviour of the second virial
coefficients for all compounds present in the data bank is clearly evident. In particular, at
low reduced temperatures, all values of B are widely negative, growing rapidly for reduced
temperatures close to unity.
162
Figure 3.2. Scatter plot of the complete data set of second virial coefficients vs. reduced temperature.
Then, the second virial coefficients values slowly increase up to Boyles temperature.
Information on all data in the database is presented in Table 3.1, grouped by families. For
each group, the number of experimental data, the range of second virial coefficients, and
the respective reduced temperature range are summarized.
Table 3.1. Summary of experimental data for each family.
Name
N. of
Points
N. of
Compounds
Maximum
Molecular
Mass
Range
B Range
(cm3/mol)
Acetates
66
102.13
0.61 0.85
-2200 -500
Alcohols
179
10
102.17
0.55 1.22
-3500 -755
Aldehydes
25
72.10
0.56 1.02
-2210 -260
Alkanes
1121
16
114.22
0.55 2.04
-2710 -17.9
Alkenes
415
13
112.21
0.48 1.67
-2179 -42.9
Alkynes
62
54.090
0.59 1.17
-1100 -133
Amines
100
10
101.19
0.55 1.28
-2000 -118
Aromatics
500
13
186.055
0.48 1.21
-2760 -247
Cycloalkanes
100
200.03
0.56 1.60
-2050 -82
Cycloalkenes
68.12
0.66 0.74
-820 -630
Dialkenes
40
68.12
0.56 0.90
-1513 -260.9
163
Epoxides
25
88.10
0.52 0.78
-1430 -375
Ethers
79
130.23
0.58 0.93
-2650 -306
Formates
56
102.13
0.60 0.81
-1738 -440
Haloalkanes
442
24
338.04
0.54 1.31
-2320 -106
Haloalkenes
22
64.034
0.89 1.40
-299 -85.8
Ketones
118
100.16
0.58 0.93
-2860 -400
Mercaptans
28
116.22
0.56 0.67
-1960 -839
Sulfides
58
12
112.19
0.53 0.67
-2410 -660
This work presents an estimation technique based on a modified group contribution method
for the second virial coefficients developed by McCann and Danner,12 following the
Benson group method [15]. In fact, the proposed method starts from the Benson group
contribution method based on summing the products of the groups and their respective
number of occurrences as in equation (3.19), replacing equation (3.18) with a new equation
containing the reduced temperature and the acentric factor, , as follows:
(3.20)
According to equation (3.20), the coefficients ai, bi, ci, di and ei were calculated for each
group by a non-linear regression fitting procedure. Since the families of alkanes and
aromatics were found to be the most numerous in terms of data, their groups were regressed
separately. The obtained coefficients for alkanes and aromatics were then kept as fixed
values during the regression of the complete set of data. Comparing to the McCann and
Danner approach, the present method, besides the new reduced temperature dependence
and the acentric factor introduction as a parameter, considered larger number of fluids. In
addition, four new groups were added, namely cyclopentene ring correction, C-(CO)(C)3,
Cb-(NI)(Cb)(H) and Cb-(NI)(Cb)(C). To optimize the coefficients, the Levenberg-Marquardt
curve-tting method was adopted [16]. This method is a combination of two minimization
methods: the gradient descent method and the Gauss-Newton method. The LevenbergMarquardt method acts more like a gradient-descent method when the parameters are far
from their optimal value, and more like the Gauss-Newton method when the parameters are
close to their optimal value and the solution typically converges rapidly to the local
minimum. This way, it was possible to identify the parameters of the proposed equation
which guarantee the lowest deviation of the predicted virial coefficients. The results of the
calculations are reported in Table 3.2.
164
Table 3.2. Coefficients ai, bi, ci, di, and ei for each group.
Group
N. of
occurrences
C-(C)(H)3
3549
C-(C)2(H)2 Primary
C-(C)3(H)
2896
198
C-(C)4
56
Cd-(H)2
675
Cd-(C)(H)
Cd-(C)2
243
24
Cd-(Cd)(H)
102
C-(Cd)(H)3
215
C-(Cd)(C)(H)2
96
Ca
12
Ct-(H)
81
Ct-(C)
43
C-Ct(H)3
36
C-Ct(C)(H)2
Cb-(H)
Cb-(C)
C-(Cb)(H)3
2385
209
219
C-(Cb)(C)(H)2
CIS correction
16
19
ai
(cm3/mol
)
bi
(cm3/mol)
1052.5
1
1251.82
-98.93
-830.61
1454.8
9
2076.0
2
868.42
-887.81
2233.9
2
371.61
1282.04
915.80
1968.8
0
737.32
744.51
2234.6
2
-856.57
65260.
20
290.34
278.52
193.97
4607.5
4
2149.2
1
559.23
2435.02
1231.99
-265.20
512.06
1083.83
-642.97
518.51
2996.13
-689.11
7293.91
6657.09
142735.
41
-289.49
-175.02
100.85
10549.6
7
5140.50
-67.91
ci
(cm3/mol)
di
(cm3/mol)
202.33
-217.55
-351.73
-87.42
-83.74
-89.22
-1.92
1.35
2.67
-605.08
-104.88
4.71
166.57
-27.47
-181.07
-165.02
-164.98
-47.25
-0.86
-1.82
-0.94
412.18
74.01
-427.88
-156.35
-6.35
-0.83
-133.82
121.00
1.48
929.91
-80.02
76.77
-65.19
-3.29
1.23
4054.82
109.93
29.53
374.94
-25.24
3564.01
43659.3
0
-7.99
-0.99
-149.76
16117.70
ei
(cm3/mol)
-52.77
-92.61
-120.37
315.00
-0.14
-1.41
0.34
3391.19
1398.25
-30.78
1870.13
-981.22
-368.83
71.49
14.59
15.77
165
Group
N. of
occurrences
22
14
48
Cyclopentene ring
cor
CO-(C)2
118
CO-(C)(H)
25
O-(C)2
111
O-(C)(H)
179
O-(CO)(C) + CO(O)(C)
66
O-(CO)(C) + CO(O)(H)
56
C-(CO)(C)2(H)
C-(CO)(C)(H)2
72
C-(CO)(H)3
243
C-(CO)(C)3
C-(O)(C)3
C-(O)(C)2(H)
29
C-(O)(C)(H)2
347
C-(O)(H)3
116
C-(N)(H)3
76
ai
(cm3/mol
)
4003.7
5
6302.9
4
1192.1
7
5836.9
1
-59.93
7690.0
0
179.05
4241.6
6
5718.6
8
3109.7
0
1185.1
1
bi
(cm3/mol)
1236.70
6353.29
2920.10
14373.6
0
1187.49
1823.27
ci
(cm3/mol)
di
(cm3/mol)
ei
(cm3/mol)
395.64
-1424.35
-27.19
1687.11
-174.33
-9.15
801.89
322.73
-9.14
5130.11
1144.68
-41.93
1349.75
-994.99
-32.34
861.31
-2523.06
-17.44
2086.34
-600.86
-1103.02
7.92
5241.98
1704.81
7.48
13.33
8389.04
2486.16
-1300.81
25.83
7595.32
2649.52
-1308.30
21.93
-481.96
-363.34
-303.57
-534.71
380.24
571.12
11.49
11.12
-31.47
-745.78
-181.78
11.09
-381.08
-273.61
18520.9
7
421.76
10.71
-6437.45
796.36
1032.83
314.78
-34.86
868.94
333.44
-9.43
1466.35
2018.10
677.11
-20.75
-1555.91
19.62
616.26
2440.1
5
1288.5
7
10463.
94
1431.0
6
2111.4
4
2972.2
9
37816.6
7
2466.50
3057.22
5209.52
818.47
5077.12
166
Group
N. of
occurrences
C-(N)(C)(H)2
110
N-(C)(H)2
40
N-(C)2(H)
51
N-(C)3
26
NI-(Cb) 2
75
Cb (NI)(Cb)(H)
131
Cb (NI)(Cb)(C)
19
C-(F)3(C)
228
C-(F)2(C)(H)
70
C-(F)(C)(H)2
65
C-(F)2(C)2 primary
192
Cd-(F)(H)
Cd-(F)2
17
C-(Cl)(C)(H)2
107
C-(Cl)(C)2(H)
C-(Cl)(C)3
C-(Cl)2(C)(H)
19
ai
(cm3/mol
)
4195.6
4
7861.1
9
6633.3
8
7605.9
3
5404.3
8
1256.0
9
16233.
96
8205.3
9
18886.
55
2821.5
3
5390.0
4
-110.25
-611.89
1113.4
7
3829.0
9
22194.
86
4973.0
1
bi
(cm3/mol)
8582.00
6186.13
13078.9
4
23496.2
4
2350.46
ci
(cm3/mol)
di
(cm3/mol)
ei
(cm3/mol)
3210.28
-867.04
32.42
2300.56
-838.19
-33.19
4872.40
1032.95
-55.55
8784.95
3529.01
-88.61
600.37
-1113.07
-0.70
918.41
15441.1
9
-253.34
-564.66
1.50
4627.48
-775.33
-43.81
408.01
-495.81
1973.27
6.88
610.40
-465.42
4983.63
0.99
1247.30
386.65
1671.48
-8.97
277.66
3168.46
-15.75
3.59
7.77
-83.36
6.73
314.72
-167.43
-6.95
4115.13
760.28
-32.95
24436.0
8
4249.18
-201.05
65.76
-1130.97
-4.29
1148.50
-712.53
1251.47
1308.01
11268.0
6
62939.7
5
1018.20
-473.61
1279.36
167
Group
N. of
occurrences
ai
(cm3/mol
)
C-(Cl)3(C)
11744.
64
C-(Cl)2(F)(C)
24
C-(Cl)(F)2(C)
104
Cb-(F) primary
C-(S)(H)3
172
30
C-(S)(C)(H)2
49
C-(S)(C)2(H)
13
C-(S)(C)3
Cd-(S)(H)
26
Cd-(S)(C)
20
S-(C)(H)
28
S-(C)2
32
S-(S)(C)
S-(Cd)2 (in thiophene
ring)
C-(C)2(H)2 secondary
2891
C-(F)2(C)2 secondary
192
Cb-(F) secondary
Legend:
Ca: Allenic Carbon
=C=
Cd: Double bounded
carbon >C=
Ct: Triple bounded
carbon -C
Cb: Aromatic carbon
CIS: CIS
172
23
-56.35
2071.6
5
-28.81
967.31
4623.6
4
2803.5
4
9127.9
3
2336.7
5
2867.3
7
-716.00
3378.8
5
237.02
435.55
9.25
3054.2
8
15.85
bi
(cm3/mol)
ci
(cm3/mol)
di
(cm3/mol)
ei
(cm3/mol)
28871.9
7
2482.41
8769.30
2563.29
-52.98
432.68
329.66
-3.25
-968.64
-60.68
-838.50
60.93
-129.13
-113.05
-479.89
-32.16
72.30
-1.53
1.99
0.55
1222.56
-558.53
4.19
400.94
-300.29
0.28
1427.49
-565.33
-2.44
-695.13
35.46
4.83
85.07
275.22
-172.02
64.78
78.94
2.38
-0.86
5230.15
-398.44
1337.16
-173.43
-725.05
48.81
5.30
3.18
-642.22
-46.84
-181.94
20.12
72.98
1.15
5.41
-0.08
464.84
-26.58
-142.70
18.47
-1311.17
6.27
-0.57
-0.61
5191.69
1659.33
5749.50
2049.09
2436.51
168
Group
N. of
occurrences
ai
(cm3/mol
)
bi
(cm3/mol)
ci
(cm3/mol)
di
(cm3/mol)
ei
(cm3/mol)
configuration
NI: Pyridine group
The final equation proposed, even if very simple, produced only minor deviations, as
presented in Figures 3 and 5, where absolute and relative deviations are reported as a
function of reduced temperature, respectively, for all the families excluding alkanes and
aromatics.
Figure 3.3. Scatter plot of absolute deviations vs. reduced temperature for all the families excluding
alkanes and aromatics.
169
Figure 3.4. Scatter plot of absolute deviations vs. reduced temperature for alkanes and aromatics.
Figure 3.5. Scatter plot of relative deviations vs. reduced temperature for all the families excluding
alkanes and aromatics.
170
Figure 3.6. Scatter plot of relative deviations vs. reduced temperature for alkanes and aromatics.
Deviations for alkanes and aromatics are presented in Figure 3.4 and Figure 3.6. From
Figure 3.3 and Figure 3.4, it is evident that deviations decrease with increasing
temperature, while from Figure 3.5 and Figure 3.6 it is clear that most deviations are
within 5%. In order to compare the results obtained with the equations discussed above, the
second virial coefficients were recalculated with equations, (3.5) (3.8), (3.11), (3.16), (3.18)
and (3.20). Because of the four groups added, AAD% was not calculated for eight
fluids for equation (3.18). The results are reported in Table 3.3 in terms of AAD%. In
Table 3.3, improvements of the proposed equation on the calculation of the second virial
coefficients are clearly evident, with a lower AAD% for almost all of the chemical families
studied.
Table 3.3. Summary of deviations for Equations (3.5) (3.8), (3.11), (3.16), (3.18) and (3.20).
Family
Eq. (3.5)
AAD%
Eq. (3.8)
AAD%
(3.11)
AAD%
Acetates
Alcohols
Aldehydes
Alkanes
Alkenes
Alkynes
Ammines
Aromatics
14.05
12.85
10.95
3.14
4.37
370.60
10.92
12.34
6.18
92.61
68.12
3.48
5.97
27.40
37.55
83.71
51.90
49.02
19.73
3.30
13.38
54.50
46.56
11.57
Eq.
(3.16)
AAD%
18.16
14.94
15.02
4.06
4.70
38.36
10.67
13.67
Eq. (3.18
AAD%
Eq.(3.20)
AAD%
10.80
15.03
18.97
5.99
4.77
13.85
14.40
16.47
2.92
6.61
10.14
2.93
1.79
1.47
2.32
3.87
171
Family
Eq. (3.5)
AAD%
Eq. (3.8)
AAD%
(3.11)
AAD%
Eq. (3.18
AAD%
Eq.(3.20)
AAD%
8.39
5.28
21.21
32.59
49.97
43.20
30.84
46.68
21.35
54.35
42.63
Eq.
(3.16)
AAD%
3.63
7.48
9.22
6.64
12.81
12.27
4.34
7.13
15.60
9.64
16.29
Cycloalkanes
Cycloalkenes
Dialkenes
Epoxides
Ethers
Formates
Haloalkanes
Haloalkenes
Ketones
Mercaptans
Sulfides
5.55
3.95
8.23
5.33
12.50
10.33
4.57
6.87
15.88
3.72
12.07
20.97
29.68
21.31
129.41
9.08
12.79
6.72
6.69
38.63
40.48
83.44
22.64
30.65
9.22
39.98
16.70
19.85
11.79
8.57
6.89
8.38
22.62
1.34
0.18
2.12
8.85
8.42
2.87
2.83
3.16
3.71
2.07
5.02
Avg.
13.51
26.38
18.67
8.26
10.71
3.28
172
3.4 References
[1.] Tsonopoulos, C. An empirical correlation of second virial coefficients. A.I.Ch.E. J.
1974, 20, 263.
[2.] Pitzer, K.S.; Curl, R.F. The volumetric and thermodynamic properties of fluids. III
Empirical equation for the second virial coefficient. J. Am. Chem. Soc. 1957, 79,
2369.
[3.] Vetere A. An improved method to predict the second virial coefficients of pure
compounds. Fluid Phase Equilib. 1999, 164, 49.
[4.] OConell, J.P.; Prausnitz, J.M. Empirical correlation of second virial coefficients for
vapour-liquid equilibrium calculations. I.&E.C. Proc. Des. Dev. 1967, 6, 245.
[5.] Black, C. Vapor phase imperfections in vapor-liquid equilibria. Ind. Eng. Chem.
1958, 50, 391.
[6.] Weber, L.A. Estimating the Virial Coefficients Of Small Polar Molecules. Int. J.
Thermophys. 1994, 15, 461.
[7.] Tsonopoulos, C. Second Virial Coefficients of Polar Haloalkanes. A.I.Ch.E. J. 1975,
21, 827.
[8.] Dymond, J.H. Second virial coefficients and liquid transport properties at saturated
vapour pressure of haloalkanes. Fluid Phase Equilib. 2000, 174, 13.
[9.] Tsonopoulos, C. Second Virial Coefficients of Polar Haloalkanes-2002. Fluid Phase
Equilib. 2003, 211, 35.
[10.] Meng, L.; Duan, Y.Y.; Li, L. Correlations for second and third virial coefficients
of pure fluids. Fluid Phase Equilib. 2004, 226, 109.
[11.] Meng, L.; Duan, Y.Y. An extended correlation for second virial coefficients of
associated and quantum fluids. Fluid Phase Equilib. 2007, 258, 29.
[12.] McCann, D.W.; Danner R.P. Prediction of Second Virial Coefficients of Organic
Compounds by a Group Contribution Method. I.&E.C. Proc. Des. Dev. 1984, 23,
529.
[13.] Dymond, J.H.; Marsh, K.N.; Wilhoit, R.C.; Wong, K.C. The Virial Coefficients of
Pures Gases and Mixtures. Landolt-Brnstein, Springer: New York, 2002.
[14.] DIPPR-801, Project 801, Evaluated Process Design Data, Public Release
Documentation, Design institute for Physical Properties (DIPPR), American
Institute of Chemical Engineers, A.I.Ch.E., 2006.
[15.] Benson, S.W.; Buss, J.H. Additivity Rules for the Estimation of Molecular
Properties. Thermodynamic Properties. J. Chem. Phys. 1958, 29, 546.
[16.] Marquardt, D.W. An algorithm for least-squares estimation of nonlinear
parameters. J. Soc. Ind. Appl. Math. 1963, 11, 431.
173
174
4 Viscosity
4.1 Introduction
Viscosity is an important characteristic and transport property of liquids that is related to
the strength of the forces acting between its molecular units. Viscosity is required for many
engineering applications, for example, viscosity is investigated because of its importance in
relation to the diffusion analysis. Furthermore, viscosity has a direct effect on heat transfer
coefficients, which are especially important. Additionally, it plays a very significant role for
some refrigerants used in refrigeration systems, heat pumps, foam bowling and other
applications; therefore, an adequate knowledge of viscosity is of importance.
Many of the experimental viscosities for liquids, that are available in literature, are rather
sparse along the saturation line, especially near the melting and above the normal boiling
points and they are limited to mostly simple and more common components so that a
reliable method of estimating viscosity is required. Many prediction models have been
developed over the years with many different approaches to estimate liquid viscosities.
Also, numerous correlations of the viscosity exist which employ the interrelationships
between viscosity and various other thermodynamic and transport properties.
so
(4.1)
175
The classic instruments to measure viscosity of liquids is the viscometer. There are many
kinds of it, but the most important are: capillary viscometers, orifice viscometers, rotational
viscometers, and vibrational viscometers. [2,3]
4.3.1
Capillary viscometers
This model is the most used for measuring viscosity of Newtonians liquids [2], and it is
developed by George Gabriel Stokes.
It consist of a fine capillary tube in which a liquid is placed and measurements made of the
time for a fixed volume of liquid to flow through a tube. The liquid may flow through the
capillary tube either under the influence of gravity or external force. This method is used
for the measurement of the water and for the Newtonian fluids. It is not applicable for fluids
with high viscosity or with solid particles. The calculation odf viscosity for the data
measured using glass capillary viscometer is based on the Hagen-Poiseuille equation
(4.2)
Where
V is is the volume of the liquid transferred as a function of time
v is mean fluid velocity along the length of the tube
x is distance in direction of flow
R is the internal radius of the tube
is the pressure difference between the two ends
is the dynamic fluid viscosity (pascal-second (Pas))
L is the length of the tube.
General the measured viscosity is compared with the viscosity of a reference fluid, so it is
used the related viscosity given by
(4.3)
With
4.3.2
, and
Orifice viscometers
They are very simply to use, and they are also known as efflux viscometer. Orifice
viscometers consist in a reservoir, an orifice and a receiver. The orifice length does not
exceed 10 times the diameter of the orifice. The friction loss at the orifice entrance was
found to be a function of cross section area ratio of cup to orifice, velocity of fluid, and
176
shape of orifice entrance. Because of these reasons the efflux time no longer remains
proportional to viscosity. [2].
4.3.3
Rotational viscometer
Rotational viscometers use the idea that the torque required to turn an object in a fluid is a
function of the viscosity of that fluid. They measure the torque required to rotate a disk or
bob in a fluid at a known speed. [3]. Although rotational viscometer are more elaborate that
capillary viscometer, and often less accurate for the measurement of viscosity of
Newtonians liquids, but it can be used to measure the viscosity of non Newtonians
materials. [2].
4.3.4
Vibrational viscometers
177
T =
R P
a
2
b + (b + c ) bc
(4.4)
T = T Td ,
Td = Tc
(4.5)
In Equation (4.5), indicates the constant coefficient and subscript c is the critical state.
By applying the similarities between the PvT and TP diagrams, the parameters of PT
viscosity equation of state are determined by following equations:
a = ac
(rc Pc )2
Tc
(4.6)
178
In Eq. (5),
b = b
rc Pc
Tc
(4.7)
c = c
rc Pc
Tc
(4.8)
c = 1 3
(4.9)
is determined as follows:
(4.10)
3b + (2 3 ) b2 + 3 2 b 3 = 0
(4.11)
ac = 3 2 + 3(1 2 ) b + b2 + (1 3 )
(4.12)
Based on the similarities between the PvT and TP diagrams and relations, the critical
compressibility factor for viscosity equation of state can also be expressed in terms of
critical properties:
c Tc
(4.13)
rc Pc
rc =
c Tc
Pc
(4.14)
179
Similar to previous works [38-42], the critical viscosity, , can be determined by applying
empirical correlation. In this work, this correlation is done for critical viscosities of 11
refrigerants containing R23, R32, R123, R124, R125, R134a, R142b, R143a, R152a,
R236fa and R245ca. The critical viscosities are given from (4.14).The following expression
in terms of critical temperature (Tc), critical pressure (Pc) and molecular mass (M) is
determined for the critical viscosity.
1
6
1
2
c
c = 8.043Tc P M
2
3
(4.15)
R = (P )rc
(4.16)
(P ) = e0 1
In Equation (4.15),
1
Pr
e1
1
1
+
0.8 +
Pr Pr + 0.25
Pr
(4.17)
The present model contains three fitting parameters: , e0, e1. These parameters are fitted
based on the minimization of the average absolute deviation (AAD) of viscosity data. This
average absolute deviation is expressed as follows:
AAD =
1
N
i ,exp i ,calc
i ,exp
(4.18)
100
The constant coefficients of the present model are listed in Table 4.1.
Table 4.1. Summary of collected data, constant coefficients of the model, and AAD% deviations, and
, e0, e1 coefficients
Name
R10
R11
NPoints
53
49
T Range (K)
263.15-363.15
209.07-353.15
P Range
(bar)
1.-1.492
1.-5.233
0.4011
0.3709
e0
-0.0002601
-0.02211
e1
0.165
0.1279
AAD %
1.36
2.35
180
Name
R112
R113
R114
R115
R12
R123
R123a
R125
R13
R130
R133a
R134a
R13b1
R141b
R142b
R143a
R150
R150b2
R152a
R160b1
R170
R20
R21
R22
R23
R236fa
NPoints
3
84
22
1
53
48
30
102
27
15
9
332
41
54
49
16
13
34
157
43
87
19
40
19
8
44
R245ca
R245fa
R280
R280b1fb
R290
R30
R31
R32
R365mfc
R50
84
65
19
48
56
22
10
133
15
22
T Range (K)
298.15-323.15
241.-388.71
273.15-353.15
298.15-298.15
201.68-343.15
233.-418.41
273.15-353.15
176.-336.15
100.-300.
273.15-353.15
223.93-342.45
175.-374.
180.-340.15
175.-353.15
239.35-353.15
255.62-323.15
288.15-313.15
273.15-404.51
133.15-383.15
173.15-393.15
95-300
210.08-353.2
207.78-346.71
273.15-333.15
189.83-256.93
233.15-373.15
248.544333.232
250.09-315.05
273.15-423.15
273.15-393.15
88.15-363.15
208.38-373.93
192.2-314.55
200.28-347.15
253.15-373.15
93.15-190.
P Range
(bar)
1.-1.
0.402
1.-6.201
0.4299
1.-9.313
0.3491
9.147-9.147 0.6855
1.-18.978
0.3126
1.-19.41
0.4132
1.-4.674
0.4295
0.769-33.934 0.4351
1.-29786.
0.3018
1.-1.
0.3664
1.-6.906
0.3989
1.-40.425
0.4286
1.-39.75
0.4169
1.-4.215
0.3453
1.-13.895
0.4323
3.448-23.285 0.4095
1.-1.
0.3991
1.-1.037
0.3651
1.-42.595
0.3481
1.-9.007
0.3022
1-43.787
0.2601
1.-1.791
0.3328
1.-7.324
0.3466
4.966-24.392 0.03218
1.-15.763
0.05063
1.-19.58
0.4622
1.-2.774
1.-2.617
1.-12.86
1.-3.785
1.-37.835
1.-5.848
1.-5.501
1.-53.266
1.-5.844
1.-45.24
0.4771
0.4794
0.3722
0.3548
0.2513
0.3344
0.3527
0.4697
0.4549
0.3095
e0
e1
AAD %
0.05736
0.4252
2.08
-0.0001564
0.2522
2.58
-0.000254
0.3194
2.31
-0.0001815
0.265
0.00
-0.0003431
0.1727
6.98
-0.0002413
0.1862
3.95
0.00006653 0.2299
4.11
-0.251
0.09847
8.25
0.0002128
0.1607
5.79
-0.01777
0.2981
2.11
-0.02407
0.2047
2.73
-0.1841
0.1057
2.76
-0.3153
-0.04594
5.05
0.567
0.4055
7.21
-0.2817
-0.001558
3.79
-0.163
0.06886
2.83
-0.1454
0.0966
0.95
-0.1508
0.09897
2.84
-0.0705
0.1469
9.60
-0.1007
0.07287
2.50
-0.05346
0.06007
2.66
-0.01501
0.1154
2.38
-0.4251
-0.04595
7.82
-0.0003269
0.214
1.53
0.0001198
0.2848
2.98
-0.2206
0.1069
2.64
-0.03668
-0.05829
-0.06779
-0.1097
0.06421
-0.0007145
0.001854
-0.1251
-0.04649
-0.0008829
0.2251
0.1583
0.07135
0.05869
0.2235
0.1061
0.1162
-0.03114
0.1801
0.01392
1.67
0.72
0.59
1.56
9.01
1.70
1.89
2.88
5.12
5.15
181
One of the advantages of this model in comparison with the applied model in [38-43] is that
this model does not need the viscosity correction so that the number of parameters of the
present model is reduced.
In this study, the modeling investigation has been conducted for 38 pure liquid refrigerants,
namely R10, R11, R112, R113, R114, R115, R12, R123, R123a, R125, R13, R130, R133a,
R134a, R13b1, R141b, R142b, R143a, R150, R150b2, R152a, R160b1, R170, R20, R21,
R22, R23, R236fa, R245ca, R245fa, R280, R280b1fb, R290, R30, R31, R32, R365mfc,
R50. A total of 3064 viscosity data points were collected with temperatures ranged (88.15580.0) K and viscosities ranged (0.0131-7.4173) cP, and pressure ranged (0.24602-500.)
The viscosity data were taken from data accepted (experimental, smoothed and predicted)
by DIPPR database [45].
As a preliminary step, a statistical overview for the experimental viscosities of refrigerants
has been performed. In order to understand if some data had to be rejected, the distribution
of the entire dataset has been statistically analyzed. In Figure 4.1, the descriptive statistics
of data has been reported. In particular, the histograms for the relative frequency of the
experimental data (left), the box-plot for the outliers identification (middle) and a
descriptive statistics referring to distribution properties such as location, dispersion, and
shape of the data (right) are reported. From the box plot, 222 outliers were identified.
However, all the available data were kept during the modeling.
2000
Points Number
3064.00
1500
Mean
0.40
1000
Variance
0.20
Standard Deviation
0.44
Skewness
5.56
Kurtosis
62.52
Median
0.28
10 Quantile
0.10
90 Quantile
0.83
500
Furthermore, in Figure 4.2, the scatter plot of the viscosity data versus reduced temperature
for each liquid refrigerant has been reported, showing a rather negative correlation of all
fluids.
182
Subsequently, the Patel-Teja viscosity equation of state has been applied to reproduce the
viscosities of pure liquid refrigerants. For saturate states, at prevailing temperature, the
vapor pressure of the refrigerant has been computed by applying Patel-Teja equation of
state (PT EOS). The three fitting parameters of the proposed model (, e0, e1) for each pure
liquid refrigerant have been determined through the minimization of Equation (2.38) For
each pure liquid refrigerant (at the pressures that are equal to or greater than the saturated
vapor pressure at given temperature), the maximum real root corresponds to viscosity of
pure liquid refrigerant. The required physical properties for the present model are presented
in Figure 4.2. The results of the proposed model and obtained parameters of the model are
listed in Table 4.1. The data and modeling results of all refrigerants have been compared
and reported, Figure 4.3. One can see the modeling results agree well with DIPPR data,
being badly predicted at low reduced temperatures.
183
184
Figure 4.3. The experimental and calculated viscosity versus reduced temperature
Residuals produced by Equation (2.38) are also reported in Figure 4.4. Figure 4.5 shows
the comparison between AAD% of the single compounds of liquid refrigerants obtained
when viscosity is calculated with Equation (4.4). This comparison proves that the
reproduction of viscosity for liquid refrigerants are good, especially at high reduced
temperatures.
185
186
10
R10
R11
R112
R113
R114
R115
R12
R123
R123a
R125
R13
R130
R133a
R134a
R13b1
R141b
R142b
R143a
R150
R150b2
R152a
R160b1
R170
R20
R21
R22
R23
R236fa
R245ca
R245fa
R280
R280b1fb
R290
R30
R31
R32
R365mfc
R50
In conclusion, experimental viscosity data for pure liquid refrigerants have been collected.
The refrigerants family has been selected due to the importance of viscosity for applications
with this chemical family. Based on similarities of the PvT and T(viscosity)P models and
Patel-Teja EOS, an equation of state has been proposed for modeling the viscosities of the
pure liquid refrigerants. The proposed model is also needless of the viscosity correction,
and the number of parameters in the proposed model have been reduced. According to
results of the present model, this proposed model correctly represents the data available in
the literature and deviations have been found to be low for all the analyzed compounds.
187
4.6 References
[1.]
[2.]
[3.]
[4.]
[5.]
[6.]
[7.]
[8.]
[9.]
[10.]
[11.]
[12.]
[13.]
[14.]
[15.]
[16.]
[17.]
[18.]
[19.]
[20.]
188
189
[38.] Wang, Z.-F., Wang, L.-S., Fan, T.-B., 2007. Densities and Viscosities of Ternary
Mixtures of Heptane, Octane, Nonane, and Hexyl Benzene from 293.15 K to
313.15 K. J. Chem. Eng. Data. 52, 1866-1871.
[39.] Fan, T.-B., Wang, L.-S., 2006. A viscosity model based on PengRobinson
equation of state for light hydrocarbon liquids and gases. Fluid Phase Equilib.
247, 5969.
[40.] Wang, X. Wu, J., Liu, Z., 2007. Viscosity modeling of several HFC refrigerants
using the friction theory. Fluid Phase Equilib., 262, 251263.
[41.] Khosharay, S., 2014. Suggestion of mixing rule for parameters of PR model for
light liquid hydrocarbon mixtures. Korean J. Chem. Eng., 31, 1246-1252.
[42.] Guo, X.Q., Wang, L.S., Rong, S.X., Guo, T.M., 1997. Viscosity model based on
equations of state for hydrocarbon liquids and gases. Fluid Phase Equilib. 139,
405421.
[43.] Guo, X.-Q., Sun, C.-Y., Rong, S.-X., Chen, G.-J., Guo, T.-M., 2001. Equation of
state analog correlations for the viscosity and thermal conductivity of
hydrocarbons and reservoir fluids. J. Petrol. Sci. Eng. 30, 1527.
[44.] Patel, N.C., Teja, A.S., 1982. A New Cubic Equation of State for Fluids and
Fluids Mixtures. Chem. Eng. Sci. 77, 463-473.
[45.] DIPPR-801, Project 801, 2006. Evaluated Process Design Data (Public Release
Documentation). Design Institute for Physical Properties (DIPPR), American
Institute of Chemical Engineers, AIChE.
190
f (P,T ,v) = 0
Starting from this equation it is possible to calculate the volume of a specific fluid, knowing
pressure and temperature.
However, by applying fundamental thermodynamic relations to the basic EOS form, the
departures from ideality of all thermodynamic properties can be derived as functions of
temperature and volume.
The first attempt to describe fluids behavior is the famous equation of Boyle and GayLussac, but that equation, and some attempt later, consider only the idea gases, and not the
volume occupied by molecules. The first that made the turning-point of history was Van
der Waals in 1873 [1].
After the Van der Waals equation lots of equations have been proposed. The book of
Vukalovich and Novikov [2] contains a list of 150 EoS published since 1944.
There are many types of EOS, but following the classification of Valderrama [3], we can
split the Equation of state in three parts:
Virial-type EoS
van der Waals EoS
Molecular-based EoS.
In particular the second type could be divided into subparts, named cubic, and non cubic
equation of state. In this thesis we consider this last type of equation of state.
The van der Waals EoS was the first equation able to represent vapor-liquid coexistence:
The parameter b is related to the size of each molecule. The volume that the molecules have
to move around in is not just the volume of the container V, but is reduced to (V - b).
191
The parameter a is related to intermolecular attractive force between the molecules. The
effect of the intermolecular attractive force is to reduce the pressure for a given volume and
temperature.
The parameters a and b, called equation of state constant are related to critical properties
of the fluid. The Van der Waals equation can be regarded as the sum of an attractive term,
and a hard sphere, repulsive term. In this way there are two contribute at the same equation
that together form the goodness of this equation of state.
There have been many modification of this equation. These are ranged from simple
modifications, with two or three parameters, to complicated forms with lots of parameters.
For example Benedict-Webb-Rubin (BWR) equation of state contains eight adjustable
parameters:
The BWR equation is one of the most powerful equation, and it is able to predict the critical
point behavior, but there are many disadvantages, such as the difficulty to extend its
applicability to the mixtures, the difficulty to find the 8 parameters, and so on.
All the parameters are fluid dependent, and often it is not so simple to extend these
equation to mixture, although they are precise in PVT representation.
5.1.1
In Table 5.1 are reported the most important modification of the attractive term of Van der
Waals equation, written in term of compressibility factor:
Where
is the compressibility factor, V is the volume, P is the pressure, and R is
the molar universal gas constant.
Table 5.1 Most important modification of attractive term
Equation
Redlich-Kwong (RK)
1949
Soave (SRK) 1972
Peng-Robinson (PR)
1976
Attractive term
(5.5)
(5.6)
(5.7)
192
Fuller (1976)
(5.8)
Heyen (1980)
Sandler (1994)
Schmidt-Wenzel (1980)
(5.9)
(5.10)
Harmens-Knapp (1980)
(5.11)
Kubic (1982)
(5.12)
(5.13)
(5.14)
Stryjek-Vera (PRSV)
(1986a)
Yu and Lu (1987)
(5.15)
(5.17)
(5.16)
Redlich-Kwong [4] proposed the first equation widely accepted for the engineering
applications, and as Prausnitz wrote: Readlich great contribution was to revive the spirit of
van der Waals [5].
This equation was built considering with two assumptions in the low and high density
boundary.
In the low density the equation was constrained to give a reasonable second virial
coefficient
Where K is a constant. At the high density the authors noticed that , the reduced volume,
at infinite pressure could be approximated to 0.26.
The Soave modification [6]of this equation is based on the generalize function,
introduced in 1964 by Wilson [7, 23 Valderrama]
The Soave modification had been proposed to improve correlations of vapor pressure of
fluids, and this modification became popular, and widely accepted. Since is a function of
acentric factor, it is most predictive for nonpolar compounds with acentric factor not
exceeding about 0.6
193
where
and
With
And
that is an adjustable parameter calculated by the same authors for the most
important compounds.
The Peng Robinson equation has got large errors all temperatures for nonpolar compounds,
and the errors increase at low temperature for all compounds. The PRSV equation
194
overcame this problem, and the vapor pressure for nonpolar, polar or associating
compounds may be reduced down the 1.5 KPa.
The two authors, Stryjek, and Vera proposed in the same year [8] a new improvement os
the PRSV, called PRSVII.
This equation required three adjustable parameter for pure compounds, and the errors in the
calculation for pure compound vapor pressure are dramatically reduced for reduced
temperature below 0.7.
Equation
Reiss et al. (1959)
Repulsive term
(5.27)
Guggenheim (1965)
(5.28)
Carnahan-Starling
(1969) (CS)
Scott (1971)
(5.29)
Boublik (1981)
(5.31)
(5.30)
Where
is the packing fraction defined in terms of covolume (b).
Perhaps the most important of this expression is the CS equation of state.
The final Carnahan-Starling [9-10] equation is as following, similar to RK equation:
This equation has been modified by De Santis et al. [11] obtaining this equation
The results show that this equation is in satisfactory agreement with experimental data over
a wide range of states both for volume and pressure representations.
195
For this reasons we choose this equation for our calculations, and test it near critical point
to understand what is the prediction of it for the critical properties.
Compound
n-Butane
n-Pentane
Propane
Propene
R14
R32
Tc [K]
425.13
469.7
369.89
364.21
227.51
351.26
Pc[KPa]
3796.0
3370.0
4251.2
4555.0
3750.0
5782.0
0.201
0.251
0.1521
0.146
0.1785
0.2769
Zc
0.274
0.270
0.276
0.281
0.283
0.244
From Figure 5.1 to Figure 5.6 there are the graphical comparison between the VLE
calculations for between equations PR, CSD, PRSVI, and from table
196
197
198
199
Table 5.4. AAD% butane comparison between equations PR, CSD, PRSVI
T [K]
AAD
%
VLiqui
d PR
140
1.84
150
2.31
160
2.74
170
3.15
180
3.52
190
3.85
200
4.16
210
4.43
220
4.66
230
4.85
240
5.00
250
5.11
260
5.16
270
5.17
280
5.11
290
4.99
300
4.79
310
4.50
320
4.12
330
3.61
340
2.97
350
2.15
360
1.12
370
0.19
380
1.85
390
4.02
400
6.91
410 10.96
420 17.00
Mean 4.49
AAD%
VLiquid
PRSVI
AAD%
VLiqui
d CSD
AAD
%
VVap
or PR
AAD%
VVapor
PRSVI
1.95
2.41
2.85
3.25
3.62
3.95
4.25
4.52
4.74
4.93
5.07
5.16
5.21
5.19
5.12
4.98
4.76
4.46
4.05
3.52
2.85
2.00
0.95
0.39
2.08
4.27
7.18
11.22
17.18
4.56
0.89
0.95
0.97
0.97
0.93
0.87
0.80
0.71
0.61
0.51
0.40
0.29
0.19
0.09
0.00
0.07
0.12
0.15
0.16
0.13
0.08
0.03
0.16
0.35
0.56
0.77
0.84
0.34
3.41
0.56
23.04
18.12
14.10
10.84
8.21
6.11
4.43
3.09
2.03
1.18
0.52
0.01
0.41
0.71
0.92
1.04
1.07
1.03
0.91
0.71
0.44
0.09
0.31
0.76
1.22
1.69
2.07
2.22
1.35
3.75
5.32
2.88
1.16
0.01
0.77
1.24
1.50
1.62
1.64
1.60
1.52
1.42
1.31
1.17
1.03
0.86
0.67
0.44
0.19
0.11
0.44
0.81
1.22
1.64
2.05
2.43
2.71
2.70
1.60
1.45
AAD% AAD%
VVapo Pressur
r CSD
e PR
9.15
7.12
5.55
4.34
3.39
2.64
2.03
1.54
1.14
0.81
0.54
0.34
0.19
0.09
0.04
0.04
0.09
0.17
0.27
0.38
0.48
0.53
0.48
0.27
0.23
1.21
3.08
6.91
17.69
2.44
29.93
22.13
16.42
12.17
8.97
6.55
4.71
3.30
2.24
1.43
0.82
0.37
0.05
0.18
0.32
0.39
0.40
0.37
0.29
0.18
0.03
0.13
0.30
0.47
0.62
0.74
0.78
0.69
0.36
3.98
AAD%
Pressure
PRSVI
AAD%
Pressure
CSD
5.62
2.97
1.18
0.01
0.74
1.18
1.40
1.47
1.43
1.33
1.18
1.01
0.82
0.62
0.42
0.22
0.03
0.17
0.36
0.55
0.73
0.89
1.03
1.15
1.23
1.24
1.16
0.93
0.44
1.09
10.07
7.66
5.88
4.55
3.52
2.73
2.11
1.62
1.23
0.91
0.65
0.44
0.28
0.17
0.08
0.04
0.02
0.03
0.07
0.11
0.17
0.23
0.27
0.29
0.27
0.17
0.04
0.41
1.06
1.56
200
Table 5.5. AAD% pentane comparison between equations PR, CSD, PRSVI
T[K]
AAD%
VLiquid
PR
AAD%
VLiquid
PRSVI
AAD%
VLiquid
CSD
150
160
170
180
190
200
210
220
230
240
250
260
270
280
290
300
310
320
330
340
350
360
370
380
390
400
410
420
430
440
450
460
Mean
1.21
0.67
0.17
0.29
0.72
1.12
1.48
1.81
2.11
2.36
2.59
2.77
2.92
3.03
3.10
3.11
3.08
2.99
2.83
2.60
2.29
1.89
1.38
0.74
0.05
1.03
2.25
3.76
5.69
8.19
11.62
16.76
3.02
1.10
0.56
0.06
0.41
0.83
1.23
1.58
1.92
2.20
2.45
2.67
2.85
2.99
3.09
3.14
3.14
3.09
2.98
2.80
2.55
2.22
1.80
1.26
0.59
0.22
1.23
2.48
4.02
5.97
8.48
11.92
17.03
3.09
0.76
0.82
0.87
0.89
0.90
0.89
0.87
0.83
0.79
0.74
0.68
0.62
0.55
0.49
0.42
0.36
0.30
0.25
0.22
0.18
0.17
0.16
0.17
0.20
0.23
0.27
0.32
0.33
0.28
0.03
0.65
2.85
0.57
AAD
AAD
AAD
AAD
AAD
%
%
%
%
%
VVapo
Pressu
VVapo
VVapo Pressu
r
re
r PR
r CSD re PR
PRSVI
PRSVI
31.08
25.33
20.48
16.43
13.08
10.31
8.04
6.18
4.65
3.41
2.39
1.55
0.87
0.32
0.11
0.45
0.70
0.86
0.95
0.97
0.92
0.80
0.64
0.42
0.16
0.12
0.42
0.71
0.96
1.12
1.06
0.47
4.87
10.12
6.69
4.15
2.31
1.01
0.12
0.48
0.86
1.08
1.19
1.23
1.22
1.18
1.11
1.03
0.93
0.81
0.68
0.52
0.34
0.14
0.09
0.33
0.60
0.87
1.15
1.42
1.65
1.82
1.87
1.66
0.87
1.55
15.88
12.26
9.44
7.26
5.57
4.25
3.23
2.42
1.78
1.27
0.87
0.54
0.28
0.09
0.05
0.14
0.18
0.18
0.15
0.08
0.01
0.07
0.12
0.13
0.06
0.15
0.56
1.29
2.52
4.58
8.26
16.05
3.12
45.09
33.91
25.75
19.67
15.05
11.52
8.78
6.64
4.97
3.66
2.63
1.82
1.19
0.70
0.33
0.06
0.12
0.24
0.29
0.30
0.26
0.18
0.08
0.05
0.20
0.35
0.49
0.63
0.73
0.77
0.72
0.51
5.87
11.26
7.16
4.33
2.37
1.03
0.14
0.44
0.79
0.97
1.04
1.02
0.95
0.84
0.71
0.56
0.40
0.23
0.06
0.11
0.28
0.45
0.62
0.78
0.93
1.07
1.19
1.27
1.33
1.32
1.24
1.04
0.68
1.46
AAD%
Pressure
CSD
18.87
13.97
10.43
7.83
5.90
4.46
3.36
2.52
1.87
1.37
0.98
0.67
0.43
0.24
0.11
0.01
0.04
0.07
0.07
0.04
0.01
0.06
0.13
0.19
0.24
0.27
0.27
0.22
0.09
0.13
0.48
1.04
2.39
201
Table 5.6. AAD% propane comparison between equations PR, CSD, PRSVI
90
100
110
120
130
140
150
160
170
180
190
200
210
220
230
240
250
260
270
280
290
300
310
320
330
340
350
360
Mea
n
AAD%
AAD%
VVapo
VVapo
r
r CSD
PRSVI
AAD%
Pressure
PR
AAD%
AAD%
Pressur
Pressur
e
e CSD
PRSVI
AAD%
VLiquid
PR
AAD%
VLiquid
PRSVI
AAD%
VLiquid
CSD
AAD%
VVapor
PR
2.01
2.62
3.19
3.71
4.19
4.63
5.03
5.38
5.70
5.97
6.20
6.38
6.51
6.57
6.57
6.50
6.35
6.10
5.74
5.25
4.61
3.77
2.69
1.29
0.53
2.96
6.32
11.30
2.10
2.72
3.29
3.81
4.30
4.73
5.13
5.49
5.80
6.07
6.29
6.46
6.57
6.63
6.61
6.53
6.35
6.09
5.70
5.19
4.52
3.65
2.54
1.12
0.73
3.18
6.55
11.51
2.64
2.10
1.64
1.27
0.95
0.70
0.49
0.33
0.20
0.11
0.05
0.00
0.02
0.03
0.03
0.03
0.03
0.02
0.02
0.03
0.04
0.05
0.07
0.06
0.01
0.21
0.86
3.05
48.10
39.11
31.29
24.74
19.33
14.94
11.40
8.57
6.32
4.54
3.13
2.02
1.16
0.49
0.01
0.36
0.58
0.68
0.66
0.54
0.31
0.01
0.43
0.93
1.50
2.09
2.61
2.75
23.74
17.22
12.10
8.26
5.43
3.39
1.94
0.93
0.25
0.20
0.49
0.67
0.76
0.80
0.78
0.71
0.60
0.43
0.21
0.06
0.39
0.78
1.21
1.70
2.23
2.75
3.14
3.10
1.68
1.90
1.95
1.88
1.73
1.54
1.34
1.15
0.97
0.83
0.70
0.58
0.47
0.35
0.20
0.03
0.16
0.38
0.61
0.83
1.02
1.14
1.12
0.89
0.27
1.03
3.74
10.15
92.66
64.22
45.55
32.87
23.97
17.56
12.87
9.40
6.80
4.85
3.37
2.27
1.44
0.83
0.38
0.08
0.11
0.22
0.25
0.21
0.13
0.01
0.13
0.28
0.43
0.54
0.58
0.47
31.14
20.80
13.76
9.01
5.75
3.51
1.99
0.97
0.31
0.10
0.34
0.44
0.47
0.44
0.36
0.26
0.13
0.01
0.16
0.32
0.48
0.63
0.77
0.88
0.96
0.97
0.89
0.63
1.66
1.87
1.91
1.84
1.69
1.51
1.32
1.12
0.94
0.78
0.65
0.54
0.44
0.35
0.27
0.18
0.10
0.01
0.07
0.15
0.22
0.27
0.29
0.24
0.13
0.09
0.46
1.06
4.93
4.99
0.54
8.16
3.37
1.38
11.52
3.45
0.72
90
0.66
0.79
1.54
AAD%
VVapor
PR
AAD%
VVapor
PRSVI
AAD%
VVapor
CSD
57.97
28.31
47.39
137.93
39.48
90.09
202
100
110
120
130
140
150
160
170
180
190
200
210
220
230
240
250
260
270
280
290
300
310
320
330
340
350
360
Mea
n
AAD%
VVapor
PR
AAD%
VVapor
PRSVI
AAD%
VVapor
CSD
1.37
2.02
2.61
3.15
3.65
4.09
4.48
4.83
5.13
5.38
5.57
5.70
5.77
5.77
5.69
5.52
5.25
4.85
4.31
3.59
2.65
1.42
0.17
2.29
5.17
9.31
16.17
1.51
2.17
2.76
3.31
3.80
4.24
4.63
4.97
5.26
5.50
5.68
5.80
5.85
5.83
5.72
5.53
5.22
4.79
4.21
3.46
2.47
1.21
0.43
2.58
5.49
9.64
16.40
1.10
0.75
0.47
0.25
0.08
0.04
0.13
0.17
0.20
0.19
0.17
0.14
0.10
0.05
0.00
0.05
0.08
0.11
0.12
0.11
0.08
0.03
0.01
0.00
0.22
1.14
5.73
47.31
37.88
29.86
23.22
17.83
13.52
10.11
7.42
5.31
3.64
2.34
1.33
0.55
0.02
0.41
0.65
0.75
0.73
0.59
0.34
0.01
0.45
0.98
1.58
2.20
2.72
2.40
19.34
12.63
7.79
4.40
2.10
0.60
0.33
0.87
1.16
1.28
1.30
1.25
1.16
1.03
0.85
0.64
0.39
0.09
0.24
0.62
1.04
1.51
2.01
2.53
3.02
3.34
2.69
38.05
30.19
23.72
18.49
14.30
10.97
8.32
6.22
4.57
3.25
2.24
1.46
0.89
0.51
0.29
0.20
0.22
0.32
0.46
0.60
0.69
0.67
0.40
0.30
1.86
5.44
16.95
89.80
60.99
42.58
30.25
21.71
15.65
11.28
8.08
5.71
3.95
2.64
1.67
0.95
0.45
0.10
0.12
0.23
0.25
0.21
0.11
0.02
0.17
0.33
0.45
0.53
0.51
0.27
23.97
14.46
8.45
4.60
2.15
0.62
0.29
0.79
1.02
1.07
1.01
0.89
0.72
0.53
0.32
0.11
0.11
0.32
0.53
0.72
0.88
1.01
1.10
1.12
1.04
0.83
0.37
61.42
43.23
31.10
22.69
16.69
12.32
9.08
6.66
4.81
3.40
2.33
1.51
0.91
0.47
0.18
0.00
0.08
0.08
0.04
0.05
0.14
0.24
0.30
0.29
0.18
0.09
0.64
4.52
4.62
0.47
9.72
3.66
8.53
15.60
3.88
11.04
130
135
140
145
9.32
9.39
9.43
9.42
9.35
9.42
9.45
9.43
0.06
0.06
0.04
0.02
1.20
0.79
0.48
0.25
AAD%
VVapor
PRSVI
0.23
0.15
0.11
0.11
0.35
0.62
0.56
0.57
1.14
0.70
0.36
0.09
0.18
0.07
0.01
0.05
0.00
0.15
0.06
0.21
203
150
155
160
165
170
175
180
185
190
195
200
205
210
215
220
Mea
n
AAD%
VVapor
PRSVI
9.36
9.23
9.04
8.78
8.42
7.97
7.40
6.68
5.80
4.71
3.36
1.66
0.50
3.34
7.22
9.36
9.23
9.03
8.75
8.38
7.91
7.33
6.60
5.70
4.60
3.23
1.52
0.66
3.50
7.37
0.01
0.04
0.08
0.11
0.15
0.19
0.23
0.27
0.32
0.39
0.50
0.68
1.02
1.70
3.33
0.12
0.05
0.06
0.14
0.28
0.49
0.75
1.07
1.44
1.87
2.35
2.86
3.39
3.85
3.99
0.16
0.25
0.38
0.56
0.78
1.03
1.33
1.66
2.04
2.46
2.91
3.38
3.85
4.24
4.27
0.66
0.81
1.02
1.26
1.54
1.82
2.09
2.31
2.47
2.48
2.30
1.79
0.73
1.36
5.80
0.10
0.23
0.30
0.32
0.30
0.25
0.17
0.08
0.03
0.13
0.23
0.31
0.34
0.31
0.16
0.06
0.04
0.00
0.07
0.15
0.24
0.34
0.43
0.52
0.60
0.66
0.68
0.65
0.54
0.30
0.30
0.35
0.35
0.32
0.27
0.20
0.12
0.05
0.01
0.01
0.02
0.13
0.32
0.64
1.16
6.90
6.89
0.48
1.34
1.57
1.61
0.29
0.29
0.25
150
160
170
180
190
200
210
220
230
240
250
260
270
280
290
300
17.10
16.43
15.81
15.25
14.75
14.31
13.95
13.67
13.47
13.38
13.40
13.57
13.91
14.46
15.28
16.45
17.21
16.54
15.92
15.35
14.85
14.40
14.03
13.74
13.53
13.41
13.42
13.57
13.88
14.41
15.20
16.33
0.71
0.51
0.34
0.21
0.09
0.01
0.07
0.13
0.17
0.22
0.26
0.30
0.35
0.39
0.44
0.46
4.96
4.63
4.22
3.83
3.50
3.26
3.11
3.04
3.05
3.11
3.21
3.32
3.45
3.58
3.72
3.85
AAD%
VVapor
PRSVI
5.56
3.77
2.41
1.35
0.48
0.26
0.91
1.50
2.03
2.51
2.94
3.31
3.64
3.92
4.16
4.35
6.44
4.38
2.82
1.62
0.65
0.16
0.86
1.49
2.08
2.64
3.21
3.80
4.46
5.24
6.21
7.49
4.66
4.29
3.80
3.26
2.72
2.20
1.73
1.31
0.92
0.58
0.27
0.01
0.26
0.49
0.68
0.84
5.96
4.07
2.73
1.80
1.15
0.69
0.37
0.16
0.04
0.02
0.02
0.02
0.09
0.20
0.33
0.45
6.95
4.73
3.16
2.07
1.31
0.79
0.43
0.20
0.07
0.00
0.00
0.03
0.08
0.14
0.19
0.21
204
310
320
330
340
350
Mean
18.08
20.38
23.66
28.66
37.99
17.33
17.94
20.20
23.46
28.44
37.80
17.32
0.41
0.18
0.56
2.87
14.91
1.12
AAD%
VVapor
PRSVI
4.01
4.21
4.55
5.33
9.67
4.08
4.52
4.71
5.00
5.67
9.84
3.47
9.28
11.94
16.30
24.91
59.45
8.35
0.95
1.00
0.95
0.73
0.14
1.51
0.58
0.67
0.69
0.58
0.12
0.99
0.17
0.05
0.19
0.65
1.52
1.09
and
This coefficients have been determined by evaluating the a, and b in the CSD equation
(5.33), and by fitting them to the functions of the equations (5.34), and (5.35). The values
of a, and b were determined starting from the saturation vapor pressure and saturation
liquid and vapor volumes.
According to the thermodynamic principles, equilibrium is reached, considering the
constraints on the system and at given pressure and temperature, when the Gibbs energy is
lowest. It can be shown that this minimum leads to the statement that the chemical potential
is identical in all phases, for any component i. For two phases, in terms of isofugacity
condition, and , can be written as:
205
where , ,
are some weights for the objective function J,
is the vapor volume.
Log(a) = Log(a0 ) + a1 t + a2 t 2 + a3 t 3
Then
Exp(Log(a)) = Exp(Log(a0 ) + a1 t + a2 t 2 + a3 t 3 )
a = a0 Exp(a1 t + a2 t 2 + a3 t 3 )
For b we add a parameter in the form
In Figure 5.7 there are the two different regressions methods for the propane for the a
coefficient. The right one is the new logarithmic regression equation(5.38), and the left one
is the classical exponential regression (5.34). Both seem to fit the data very well, but if we
consider the absolute average deviations, they change significantly: in fact, the first one is
0.00325057%, and the second one is 0.190158%.
206
Figure 5.7. Comparison between logarithmic regression (5.38), and exponential one (5.34)
For the coefficient b, the difference between the new regression with 4 parameters (5.40),
and the classical one with 3 parameters, (5.35) is specified in Figure 5.8
0.13
0.13
0.12
0.12
0.11
Experimental
0.10
4 parameters
0.09
0.11
Experimental
0.10
3 parameters
0.09
0.08
0.08
100
150
200
250
300
100
150
200
250
300
Figure 5.8. Regression with 4 parameters (5.40), and with 3 parameters (5.35)
In the same way the AAD% for the new regression is quite lower, in fact for the 4
parameters regression the AAD% is 0.0496266, and for the 3 parameters regression is
0.220608.
The second improvement to CSD equation of state is the multi-objective optimization
(MOO).
In general, the optimization can be single (SOO)- or multi-objective. The attempt to
optimize a project or system in which a single objective is present usually means applying
methods to the gradient in which the algorithms look for the minimum or maximum of a
function according to the established objective. One of the methods for managing multiobjective optimization consists in summing all of the objectives (adequately weighted) into
a unique function in order to reduce the problem to a single-objective optimization
problem. This is the method used for the classical CSD equation as states in equation
(5.37). This method, however, has the disadvantage that such weights, which should be
furnished a priori by the user, influence the solution. In addition, if the objectives are all
different, it could be difficult or even useless to try to formulate a single-objective function.
Multi-objective optimization techniques sidestep these problems by keeping the objectives
separate within the optimization process. The role of the optimization algorithm consists in
identifying solutions that are found on the trade-off curve, better known as the Pareto
frontier. A common characteristic of such solutions is the fact that by improving one
objective, at least one of the other objectives worsens.
207
Pareto optimisation can be viewed as a concept that formalizes the trade-off between a
given set of mutually contradicting objectives. A solution is Pareto-optimal when it is not
possible to improve one objective without deteriorating at least one of the others. A set of
Pareto-optimal solutions constitutes the Pareto frontier.
The Pareto frontier is defined as the best solutions that are not strictly dominated by other
solutions. (Solution A dominates a solution B if A behaves better than B, all things
considered.) Pareto optimization or Pareto efficiency is a concept introduced by the Italian
engineer Vilfredo Pareto that is widely applied in economics, game theory, engineering,
and social sciences. Pareto optimization can be viewed as a concept that formalizes the
trade-off between a given set of mutually contradicting objectives. A solution is Paretooptimal when it is not possible to bring Pareto improvements to the system, i.e., it is not
possible to improve one objective without worsening at least one of the others. A set of
Pareto-optimal solutions constitutes the Pareto frontier and movement along the frontier
represents a trade-off between the objectives.
In general, the set of solutions along the Pareto frontier can be composed of an infinite
number of points, but a restricted set of objectives is generally chosen for optimization.
Choosing the best solution is not always trivial due to the trade-offs involved, and various
techniques have been developed to select a number of representative points from along the
frontier. One of the most common strategies, and the one used here, is the fast nondominated sorting algorithm NSGA-II [13].
In the later years the importance of MOO inside the research of chemical engineering is
increasing [14]. Following the study of Bhaskar [15] there was 30 journal publication using
MOO in chemical engineering. On the other hand, Gade Pandu [14] shows as from the year
2000 to 2007 the number increases reaching more that 130 journal publications.
For these reasons the MOO was applied also for the optimization of the coefficients of CSD
equation of state.
As a first step we modified the objective function (5.37) as follow:
In this way we split the volumes from the pressure avoiding to sum the entire residuals.
The method used for the minimization of the multi-objective function (5.41) is a non
dominated sorted algorithm named NSGAII [13]. In figures Figure 5.9 there is an example
of the results of the minimization of both pressure and volumes separately. As we can see
there are two box containing the information for pressure and for volumes.
208
For each temperature, of each fluid, a Pareto frontier was formed after 100.000 runs. An
example of results are shown in Figure 5.10. The result is a set of the couple a and b
magnitudes selected from the Pareto frontier that are optimal for the quantity desired,
pressure and volumes.
After the minimization, the results for each compound are shown from Figure 5.11 to
Figure 5.16 that states the new approach is improving the prediction od CSD equation
especially near the critical region. In fact, this is one of the deficiency of the CSD equation,
partially overcame with this modifications.
From Table 5.10 to Table 5.15 there are listed the deviation for each temperature between
the new method and the classical method. In this way the improving of this new technique
is more clear.
209
Figure 5.11. Comparison between new method and classical method for n-butane
Figure 5.12. Comparison between new method and classical method for n-pentane
210
Figure 5.13. Comparison between new method and classical method for propane
Figure 5.14. Comparison between new method and classical method for propene
211
Figure 5.15. Comparison between new method and classical method for R14
Figure 5.16. Comparison between new method and classical method for R32
Table 5.10. Comparison between new method and classical method for n-butane
AAD%
AAD%
P
Classica P New
l
P residuals
Classical
(MPa)
P
residuals
New
(MPa)
AAD%
VLiquid
Classical
AAD%
VLiquid
New
AAD%
AAD%
V Vapor
V Vapor
Classica
New
l
140
10.070 1.8725
-0.0002
0.0000
0.8904
0.6481
9.1470
1.9105
212
AAD%
AAD%
P
Classica P New
l
P residuals
Classical
(MPa)
P
residuals
New
(MPa)
AAD%
VLiquid
Classical
AAD%
VLiquid
New
150
160
170
180
190
200
210
220
230
240
250
260
270
280
290
300
310
320
330
340
350
360
370
380
390
400
410
2
7.6640
5.8841
4.5460
3.5226
2.7344
2.1143
1.6231
1.2281
0.9066
0.6509
0.4445
0.2839
0.1666
0.0828
0.0395
0.0213
0.0338
0.0658
0.1136
0.1710
0.2277
0.2717
0.2919
0.2677
0.1704
0.0351
0.4094
1.0619
0.3548
0.2292
0.0449
0.0001
1.2350
0.0002
0.9652
0.0000
0.2755
1.0409
0.9649
0.8040
0.6404
0.4938
0.3588
0.2498
0.1634
0.1026
0.0676
0.0541
0.0557
0.0670
0.0732
0.0510
0.0290
0.2222
-0.0007
-0.0020
-0.0053
-0.0118
-0.0232
-0.0410
-0.0656
-0.0958
-0.1279
-0.1568
-0.1741
-0.1731
-0.1524
-0.1099
-0.0740
-0.0549
-0.1171
-0.3002
-0.6706
-1.2859
-2.1501
-3.1805
-4.1866
-4.6583
-3.5650
0.8756
12.1104
0.0001
0.0001
-0.0003
-0.0002
0.0000
-0.0239
0.0000
-0.0753
0.0000
-0.0664
-0.4076
-0.5884
-0.7355
-0.8506
-0.9251
-0.9242
-0.8652
-0.7454
-0.6058
-0.5087
-0.5108
-0.6523
-0.9612
-1.2730
-1.0670
0.7231
6.5716
0.9503
0.9738
0.9654
0.9305
0.8731
0.7976
0.7096
0.6106
0.5064
0.3977
0.2911
0.1868
0.0902
0.0046
0.0671
0.1240
0.1549
0.1591
0.1344
0.0762
0.0261
0.1622
0.3479
0.5631
0.7659
0.8444
0.3399
0.2766
0.0232
0.2564
0.4290
0.5464
0.6137
0.6381
0.6224
0.5742
0.4960
0.3962
0.2775
0.1478
0.0132
0.1201
0.2476
0.3536
0.4322
0.4747
0.4687
0.3934
0.2455
0.0094
0.3784
0.8628
1.4252
1.7571
420
1.0650 0.6367
37.1646
22.2187
3.4098
0.4433
0.9920
0.6113
0.5639
0.4680
AAD%
AAD%
V Vapor
V Vapor
Classica
New
l
7.1201
5.5528
4.3394
3.3900
2.6388
2.0323
1.5423
1.1369
0.8071
0.5438
0.3373
0.1856
0.0906
0.0430
0.0406
0.0889
0.1673
0.2703
0.3816
0.4800
0.5289
0.4839
0.2696
0.2305
1.2107
3.0806
6.9133
17.689
0
2.4394
1.0716
0.3611
0.2187
0.0302
0.0258
1.1797
0.0622
0.8778
0.1005
0.1657
0.9366
0.8762
0.7433
0.6193
0.5128
0.4385
0.3838
0.3505
0.3299
0.3084
0.2588
0.1482
0.0829
0.5272
1.3389
2.8510
5.9547
14.960
0
1.2974
Table 5.11. Comparison between new method and classical method for n-pentane
150
AAD%
P
P residuals
P
residuals
AAD%
Classical
Classica P New
New
(MPa)
l
(MPa)
AAD%
VLiquid
Classical
AAD%
VLiquid
New
AAD% V
Vapor
Classical
AAD% V
Vapor
New
18.873
0.7555
0.0859
15.8760
1.5145
1.490
-0.0001
0.0000
213
160
170
AAD%
P
P residuals
P
residuals
AAD%
Classical
Classica P New
New
(MPa)
l
(MPa)
5
13.965
9
10.427
8
180
7.8310
190
5.9030
200
4.4575
210
3.3593
220
2.5181
230
1.8709
240
1.3699
250
0.9752
260
0.6673
270
0.4274
280
0.2418
290
0.1063
300
0.0126
310
0.0416
320
0.0715
330
0.0667
340
0.0403
350
0.0064
360
0.0621
7
1.215
0
0.869
2
0.525
7
0.217
7
0.037
5
0.228
1
0.355
8
0.124
9
0.629
7
0.231
8
0.000
0
0.313
9
0.235
2
0.154
4
0.080
1
0.021
3
0.028
4
0.051
7
0.055
7
0.037
4
0.004
AAD%
VLiquid
Classical
AAD%
VLiquid
New
AAD% V
Vapor
Classical
AAD% V
Vapor
New
-0.0002
0.0000
0.8224
0.1203
12.2567
1.2274
-0.0007
0.0001
0.8675
0.1413
9.4417
0.8785
-0.0018
0.0001
0.8932
0.1511
7.2600
0.5312
-0.0043
0.0002
0.9028
0.1522
5.5663
0.2269
-0.0088
-0.0001
0.8894
0.1373
4.2537
0.0214
-0.0162
-0.0011
0.8731
0.1258
3.2272
0.2019
-0.0270
-0.0038
0.8306
0.0942
2.4228
0.3175
-0.0412
-0.0027
0.7879
0.0678
1.7839
0.0667
-0.0577
-0.0265
0.7418
0.0429
1.2739
0.5418
-0.0740
-0.0176
0.6804
0.0076
0.8662
0.1237
-0.0865
0.0000
0.6209
0.0219
0.5368
0.1373
-0.0904
-0.0664
0.5528
0.0561
0.2793
0.1586
-0.0801
-0.0778
0.4854
0.0863
0.0874
0.0720
-0.0532
-0.0772
0.4195
0.1120
0.0519
0.0139
-0.0092
-0.0586
0.3574
0.1313
0.1403
0.0842
0.0433
-0.0222
0.3027
0.1408
0.1809
0.1314
0.1032
0.0410
0.2527
0.1437
0.1813
0.1540
0.1306
0.1011
0.2152
0.1323
0.1475
0.1510
0.1048
0.1449
0.1847
0.1124
0.0832
0.1201
-0.0218
0.1270
0.1677
0.0775
0.0062
0.0741
-0.2708
0.0193
0.1599
0.0321
0.0696
0.0227
214
AAD%
P
P residuals
P
residuals
AAD%
Classical
Classica P New
New
(MPa)
l
(MPa)
370
0.1263
380
0.1883
390
0.2400
400
0.2714
410
0.2663
420
0.2174
430
0.0933
440
0.1264
450
0.4817
460
1.0373
Mea
n
2.3858
4
0.045
7
0.104
5
0.166
0
0.221
5
0.255
9
0.262
8
0.211
2
0.080
7
0.168
9
0.603
3
0.282
2
AAD%
VLiquid
Classical
AAD%
VLiquid
New
AAD% V
Vapor
Classical
AAD% V
Vapor
New
-0.6971
-0.2522
0.1710
0.0336
0.1231
0.0168
-1.2975
-0.7205
0.1966
0.1156
0.1317
0.0253
-2.0411
-1.4119
0.2299
0.2074
0.0566
0.0330
-2.8187
-2.3004
0.2741
0.3133
0.1537
0.2056
-3.3436
-3.2133
0.3153
0.4208
0.5636
0.5529
-3.2715
-3.9548
0.3318
0.5109
1.2945
1.1904
-1.6698
-3.7800
0.2754
0.5399
2.5151
2.2781
2.6739
-1.7067
0.0350
0.4048
4.5812
4.1559
11.9645
4.1949
0.6534
0.1425
8.2635
7.5612
30.1152
17.516
5
2.8453
2.1237
16.0493
14.8851
0.9110
0.1391
0.5654
0.2183
3.1164
1.1774
Table 5.12. Comparison between new method and classical method for propane
90
100
110
120
130
140
150
160
170
180
AAD% P AAD% P
Classical
New
1.6564
1.8677
1.9122
1.8411
1.6934
1.5131
1.3154
1.1194
0.9378
0.7816
0.5763
2.0562
2.6718
2.7587
2.5331
2.1528
1.7074
1.2639
0.8647
0.5368
P
residuals
Classical
(MPa)
P
residuals
New
(MPa)
AAD%
VLiquid
Classical
AAD%
VLiquid
New
0.0000
0.0000
0.0000
0.0001
0.0003
0.0012
0.0037
0.0095
0.0207
0.0396
0.0000
0.0000
0.0000
0.0001
0.0004
0.0017
0.0048
0.0107
0.0191
0.0272
2.6369
2.0962
1.6442
1.2666
0.9550
0.6992
0.4921
0.3295
0.2047
0.1120
1.7012
0.8278
0.1249
0.4256
0.8365
1.1223
1.2955
1.3654
1.3446
1.2453
AAD% V AAD% V
Vapor
Vapor
Classical
New
1.6846
1.9031
1.9502
1.8753
1.7261
1.5379
1.3405
1.1492
0.9747
0.8255
0.5799
2.0991
2.7458
2.8367
2.6025
2.2015
1.7442
1.2961
0.8975
0.5723
215
190
200
210
220
230
240
250
260
270
280
290
300
310
320
330
340
350
360
Mean
AAD% P AAD% P
Classical
New
0.6507
0.5351
0.4368
0.3498
0.2678
0.1846
0.0977
0.0113
0.0737
0.1546
0.2245
0.2719
0.2872
0.2420
0.1331
0.0904
0.4585
1.0591
0.7202
0.2873
0.1069
0.0063
0.0647
0.0826
0.0753
0.0549
0.0275
0.0026
0.0116
0.0104
0.0063
0.0410
0.0773
0.1236
0.1396
0.1044
0.0643
0.6574
P
residuals
Classical
(MPa)
P
residuals
New
(MPa)
AAD%
VLiquid
Classical
AAD%
VLiquid
New
AAD% V AAD% V
Vapor
Vapor
Classical
New
0.0686
0.1081
0.1574
0.2119
0.2592
0.2731
0.2129
0.0350
-0.3174
-0.8992
-1.7267
-2.7128
-3.6544
-3.8691
-2.6394
2.1984
13.5323
37.6455
1.3914
0.0303
0.0216
-0.0023
-0.0392
-0.0799
-0.1115
-0.1198
-0.0855
-0.0111
0.0675
0.0803
-0.0629
-0.5216
-1.2355
-2.4503
-3.3938
-3.0805
2.2862
-0.3087
0.0477
0.0048
0.0197
0.0320
0.0350
0.0319
0.0260
0.0220
0.0212
0.0259
0.0368
0.0525
0.0676
0.0645
0.0071
0.2125
0.8601
3.0455
0.5375
1.0777
0.8565
0.5929
0.3020
0.0027
0.3062
0.5911
0.8362
1.0201
1.1153
1.0891
0.9008
0.4980
0.1852
1.2416
2.7839
4.9836
7.8409
1.3040
0.7000
0.5849
0.4696
0.3465
0.2046
0.0319
0.1622
0.3791
0.6071
0.8286
1.0164
1.1356
1.1198
0.8854
0.2740
1.0302
3.7370
10.1483
1.3796
0.3247
0.1417
0.0097
0.0837
0.1555
0.2246
0.2896
0.3612
0.4398
0.5230
0.6047
0.6788
0.7211
0.7122
0.5997
0.2643
0.5679
2.8545
0.9690
Table 5.13. Comparison between new method and classical method for propene
90
100
110
120
130
140
150
160
170
180
190
200
AAD% P AAD% P
Classical
New
P
residuals
Classical
(MPa)
P
residuals
New
(MPa)
AAD%
VLiquid
Classical
AAD%
VLiquid
New
AAD% V AAD% V
Vapor
Vapor
Classical
New
90.0930
61.4184
43.2344
31.0988
22.6882
16.6892
12.3237
9.0846
6.6562
4.8102
3.4013
2.3270
0.0000
0.0000
-0.0002
-0.0014
-0.0061
-0.0195
-0.0506
-0.1096
-0.2042
-0.3333
-0.4824
-0.6226
0.0000
0.0000
0.0000
0.0000
0.0000
-0.0007
-0.0041
-0.0145
-0.0388
-0.0837
-0.1507
-0.2303
1.5363
1.1020
0.7511
0.4705
0.2498
0.0820
0.0423
0.1254
0.1746
0.1960
0.1935
0.1735
0.8012
0.3398
0.0222
0.2979
0.4979
0.6298
0.7039
0.7236
0.6979
0.6346
0.5396
0.4218
47.3939
38.0500
30.1852
23.7227
18.4918
14.3015
10.9655
8.3204
6.2236
4.5659
3.2522
2.2371
4.6328
3.1326
1.8490
0.8003
0.0125
0.6000
0.9903
1.1998
1.2652
1.2077
1.0623
0.8608
4.8583
3.2325
1.8827
0.8057
0.0108
0.5935
0.9699
1.1707
1.2218
1.1570
1.0000
0.8029
216
AAD% P AAD% P
Classical
New
210
1.5128
220
0.9056
230
0.4715
240
0.1796
250
0.0019
260
0.0793
270
0.0844
280
0.0367
290
0.0469
300
0.1439
310
0.2353
320
0.2956
330
0.2893
340
0.1833
350
0.0879
360
0.6411
Mean 11.0364
0.6297
0.3926
0.1749
0.0087
0.1509
0.2409
0.2768
0.2629
0.2002
0.0972
0.0400
0.1978
0.3523
0.4813
0.5300
0.3902
0.7871
P
residuals
Classical
(MPa)
P
residuals
New
(MPa)
AAD%
VLiquid
Classical
AAD%
VLiquid
New
AAD% V AAD% V
Vapor
Vapor
Classical
New
-0.7126 -0.2966
-0.7084 -0.3071
-0.5828 -0.2162
-0.3362 0.0163
-0.0052 0.4121
0.3061
0.9305
0.4480
1.4699
0.2615
1.8743
-0.4396 1.8759
-1.7387 1.1753
-3.6086 -0.6129
-5.6720 -3.7961
-6.8590 -8.3531
-5.3096 -13.9459
3.0842 -18.5940
27.0477 -16.4615
0.1195 -1.9769
0.1394
0.0955
0.0471
0.0023
0.0476
0.0841
0.1094
0.1185
0.1090
0.0790
0.0339
0.0079
0.0046
0.2152
1.1399
5.7271
0.4663
0.2878
0.1451
0.0034
0.1291
0.2410
0.3201
0.3545
0.3268
0.2179
0.0031
0.3455
0.8604
1.5750
2.5025
3.4732
2.5484
0.7016
1.4607
0.8911
0.5060
0.2886
0.1971
0.2188
0.3171
0.4568
0.6027
0.6945
0.6683
0.4042
0.2988
1.8624
5.4375
16.9534
8.5346
0.5837
0.3735
0.2000
0.0864
0.0238
0.0268
0.0828
0.1796
0.3051
0.4260
0.5129
0.4963
0.2592
0.4548
2.3785
9.6592
1.2055
Table 5.14. Comparison between new method and classical method for R14
P
AAD% P AAD% P residuals
Classical New
Classical
(MPa)
P
residuals
New
(MPa)
AAD%
VLiquid
Classical
AAD%
VLiquid
New
AAD% V AAD% V
Vapor
Vapor
Classical New
130
135
140
145
150
155
160
165
170
175
180
0.0000
0.1496
0.0601
0.2092
0.3048
0.3520
0.3515
0.3225
0.2672
0.1958
0.1209
0.0504
0.0147
0.0449
0.2066
0.5837
1.2651
2.3194
3.8360
5.8250
8.2493
10.9940
0.0567
0.0577
0.0450
0.0223
0.0070
0.0396
0.0767
0.1133
0.1525
0.1905
0.2302
0.4616
0.1855
0.0309
0.0158
0.0317
0.1579
0.3505
0.5887
0.8568
1.1286
1.3793
0.3473
0.6204
0.5576
0.5673
0.6558
0.8081
1.0155
1.2646
1.5395
1.8210
2.0902
0.1658
0.0314
0.0644
0.2054
0.4134
0.6540
0.8944
1.1251
1.3222
1.4720
1.5640
0.0000
-0.0701
0.0419
0.2104
0.4304
0.6809
0.9114
1.0995
1.1772
1.0973
0.8500
0.1735
0.4322
0.5513
0.5710
0.5417
0.4837
0.4183
0.3587
0.3142
0.2912
0.2990
217
P
AAD% P AAD% P residuals
Classical
Classical New
(MPa)
P
residuals
New
(MPa)
AAD%
VLiquid
Classical
AAD%
VLiquid
New
AAD% V AAD% V
Vapor
Vapor
Classical New
185
190
195
200
205
210
215
220
Mean
0.0528
0.0051
0.0121
0.0192
0.1252
0.3186
0.6433
1.1563
0.2456
13.8052
16.2836
17.7751
17.4359
14.2837
6.7882
-6.2454
-25.5927
4.6275
0.2733
0.3245
0.3930
0.4999
0.6801
1.0161
1.7018
3.3293
0.4847
1.5757
1.6775
1.6334
1.3797
0.8248
0.1538
1.7364
4.1589
0.9646
2.3144
2.4667
2.4842
2.3005
1.7897
0.7266
1.3650
5.8033
1.6072
1.5855
1.5271
1.3753
1.1232
0.7725
0.3105
0.2431
0.8525
0.8264
0.4598
0.0542
-0.1559
0.2974
2.3155
6.9667
16.5242
34.7146
3.5582
0.3377
0.4222
0.5437
0.7129
0.9197
1.1323
1.2493
0.8842
0.5598
Table 5.15. Comparison between new method and classical method for R32
150
160
170
180
190
200
210
220
230
240
250
260
270
280
290
300
310
320
330
340
350
AAD% P AAD% P
Classical
New
6.9523
4.7287
3.1644
2.0749
1.3130
0.7867
0.4305
0.1994
0.0665
0.0031
0.0036
0.0264
0.0775
0.1376
0.1871
0.2052
0.1729
0.0525
0.1921
0.6451
1.5226
3.2571
1.4796
0.2070
0.0004
0.0153
1.4465
1.4996
0.0000
1.2270
0.9892
0.7373
0.4955
0.2847
0.1238
0.0199
0.0260
0.0149
0.0283
0.0739
0.0365
0.3134
P
residuals
Classical
(MPa)
P
residuals
New
(MPa)
AAD%
VLiquid
Classical
AAD%
VLiquid
New
AAD% V AAD% V
Vapor
Vapor
Classical
New
-0.0226
-0.0492
-0.0903
-0.1427
-0.1957
-0.2324
-0.2339
-0.1870
-0.1020
-0.0075
0.0131
-0.1378
-0.5692
-1.3853
-2.5267
-3.6416
-3.9645
-1.5319
7.0482
29.4258
85.7409
0.0106
0.0154
0.0059
0.0000
-0.0023
-0.4274
-0.8149
0.0000
-1.8829
-2.3706
-2.6518
-2.5841
-2.0898
-1.2467
-0.2689
0.4612
0.3413
-0.8248
-2.7116
-1.6652
17.6503
0.7092
0.5069
0.3411
0.2057
0.0936
0.0053
0.0660
0.1256
0.1736
0.2160
0.2571
0.3000
0.3467
0.3940
0.4391
0.4594
0.4106
0.1752
0.5614
2.8672
14.9129
0.8648
0.1119
0.4529
0.8476
1.0926
1.2015
1.1970
1.1023
0.9357
0.7268
0.5050
0.3035
0.1605
0.1170
0.2269
0.5432
1.1276
2.0247
3.1933
4.0957
2.7032
6.4417
4.3796
2.8219
1.6150
0.6468
0.1633
0.8638
1.4924
2.0798
2.6432
3.2058
3.7974
4.4566
5.2370
6.2145
7.4932
9.2826
11.9423
16.3008
24.9109
59.4509
3.4338
1.6492
0.4644
0.4275
0.6440
0.5032
0.2270
1.6872
0.8551
1.5819
2.3939
3.2796
4.2389
5.2852
6.4476
7.7688
9.3676
11.4499
14.5326
20.3333
45.8377
218
Mean
AAD% P AAD% P
Classical
New
1.0925
0.5846
P
residuals
Classical
(MPa)
P
residuals
New
(MPa)
AAD%
VLiquid
Classical
AAD%
VLiquid
New
5.1051
-0.0503
1.1222
1.1206
AAD% V AAD% V
Vapor
Vapor
Classical
New
8.3543
6.7814
The improvement is greater near the critical region roughly for pressure and volumes. Some
other analysis is required for R14. In fact the improvement for this compound there are
some unexpected values.
In conclusion for our results demonstrate that the introduction of two more parameters, and
the multi-objective optimization improve the behavior of the CSD equation. In particular its
main advantage is near the critical point a region particularly difficult to represent for many
equations of state.
219
5.6 Bibliography
[1.]
[2.]
[3.]
[4.]
[5.]
[6.]
[7.]
[8.]
[9.]
[10.]
[11.]
[12.]
[13.]
[14.]
[15.]
220
Concluding Remarks
The above chapters, states my course of study in these years I had spent inside the
Dipartimento di Ingegneria Industriale e Scienze Matematiche of Universit Politecnica di
Ancona.
The intense activity carried out in my thesis doesnt close the arguments of research
contained, but it is only the beginning. Every issues contained in this work, just offered a
chance to develop new ways which are waiting to be explored. For each study I finished,
for every paper published, lots occurs to be started. For example we are at the beginning
with viscosity. Were searching an equation that fits data in a better way. Another problem
is to calculate the thermo physical properties crossing themselves, i.e. calculate surface
tension with the viscosity and thermal conductivity, and vice versa.
Regarding equation of state, the modification of Carnahan-Starling-de Santis equation of
state, is only a start up towards new equations. Calculation of Vapor liquid equilibrium
properties must to be known as the previous authors built their equations and subsequently
to build a new one. Mixture would represent another challenge. How these new
modifications perform with mixture?
At the end, if I look back, I remember the first time I joined the department. I was one of
the very few mathematicians (perhaps only 2 with me) among a lot of engineers, and some
physicians. The research carried out in this precious time, demonstrate that the difference
between some professional figures, disappears in front of a problem. The different approach
to solve problems is the keystone to resolve them and overcome all the apparent
differences. So Ive lost my presumption to be a mathematician and perhaps, now Im now
a little bit a physicians or engineer, hoping that I was able to give people close to me, a
little bit of the mathematician that once I was.
221