Engineering Practice

Selecting the Right Thermodynamic Models

for Process Simulation
Software that enables complex processes to be simulated continues to evolve for a
wide range of thermodynamic conditions. Selection of the most appropriate models
plays a crucial role in adequately representing real-life conditions
Atul B. Choudhari
Tata Consulting Engineers Ltd.

ne of the first and most

O
important decisions to
be made when model­ Select an
equation of
Are there polar

ling a chemical process state (EOS) model

*-©- components present?

system using a computer simulation

package is the selection of an ap­
propriate method for simulating ther­ Are the operating
modynamic properties. Thermody­ Select a modified
EOS model
conditions near the
critical region?
namic-property methods enable the
mathematical calculation of pres­
sure, temperature, composition and
specific volume for multi-component Are there any light gases
or supercritical components present
and multi-phase chemical systems. (such as H , CO2, N2)?
2
These calculation methods are also
used as a basis for determining the
state functions of a system, such as Select an activity coefficient model with Select an activity coefficient model
Henry's law (if available) for the liquid for the liquid phase and an EOS for
enthalpy and entropy and hence, phase, and an EOS for the vapor phase the vapor phase
Gibbs free energy. This article pro­
FIGURE 1. This decision tree can aid in the selection of the most appropriate thermodynamic method
vides an overview of the primary
thermodynamic-property meth­
ods, and discusses the basis that junction with either the ideal gas the following:
underlies each one, the strengths equation or an EOS, to accurately • Steam tables
and weaknesses of each, and appli­ incorporate vapor-phase behavior. • API Sour-Water Method
cability of each to various chemical This is sometimes referred to as a • Glycol
systems, to assist in the selection dual-model approach. • Amines
decision. 3. Henry’s Law. Henry’s Law is gen­ Commercial simulation pack­
In general, thermodynamic-prop­ erally employed for environmental ages often use terminology such
erty methods can be categorized applications. It is applied along with as ‘‘thermodynamic system” or
according to the following four op­ liquid-activity-coefficient methods, “thermodynamic package" in the
tions (each of which is discussed in such as non-random two-liquid input data section. Selection of the
greater detail below): (NRTL) methods. Henry’s Law deter­ most appropriate thermodynamic
1. Equations of state (EOS). These mines the amount of a supercritical method is not just a decision on the
are primarily suited to describing va­ component (non-condensable) gas thermodynamic-property method.
por-phase behavior and predicting in the liquid phase. Instead, it involves selection of all
the vapor-liquid equilibrium (VLE) be­ 4. Special models. For specific appli­ the relevant equations and methods
tween the vapor phase and an ideal, cations, a specially developed ther­ that the simulation will use to predict
saturated liquid phase. modynamic package (for instance, vapor-liquid equilibrium (VLE), vapor-
2. Liquid-activity-coefficient meth­ having binary interaction param­ liquid-liquid equilibrium (VLLE), state
ods. These have been developed eters that are specially developed functions (such as enthalpy and
specifically to model non-ideal liquid- for the components associated with entropy), physical properties and
phase behavior. These methods are these systems) is generally transport properties.
more empirical in nature than EOS available in most of the commer­
and therefore require more data cial simulation packages. The types Equations of state (EOS)
about the system and its compo­ of systems that are often included Any equation that attempts to re­
nents. They must be used in con­ in these special models include late the pressure, temperature and

50 CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM DECEMBER 2019

 TABLE 1. EQUATION OF STATE (EOS) APPLICABILITY GUIDE
Classification Low High Liquid Polar Comments
pressure pressure phase molecules
Ideal gas equation Yes No No No Strictly applicable to ideal gas systems

Generalized correla­ Yes Yes No Yes The compressibility factor and accentnc factor attempt to account for non-ideality. These
tions (such as Lee- correlations are of a graphical form and are easy to use. The basic premise is that all gas­
Kesler-Plocker) eous components with similar accentricity behave similarly at the same reduced tempera­
ture (TJ and reduced pressure (PJ
Virial equations Yes No No No Though these equations have reasonable accuracy for non-polar gas mixtures at low to
medium pressures, they have limited applicability, as they have not been developed to the
extent of the more popular cubic EOS
Cubic EOS (such as Yes <1,000 Yes No (unless Cubic EOS (cubic with respect to specific volume) are the most popular EOS used in the
Peng Robinson (PR), bar for specifically oil-and-gas and petrochemical industries. They are relatively simple to apply and are able
SRK, RK and others) PR,and modified) to predict the behavior of saturated liquid-phase components, in addition to vapor-phase
<350 bar behavior. They are all susceptible to errors in prediction at or very near the critical point,
for SRK but are considered to be more accurate than activity coefficient methods in that region.
However, by themselves they cannot predict within the two-phase (V-L) region of a pure
component
More complex EOS Yes Yes Yes Yes Such EOS contain many adjustable parameters, which can extend their applicability,
(such as Benedict- though often only for a limited range of components. Their applicability to describing
Webb-Rubin; BWRII liquid-phase behavior is equation-specific

specific (molar) volume of a compo­
nent (or mixture of components) in a
single phase can be termed an EOS.
The ideal gas law represents the
simplest EOS.
PV = nRT
Where:
P = The pressure of the gas
V = The volume of the gas
n = The number of moles
T = The absolute temperature
R = The ideal gas constant
This equation makes no allowance
for intermolecular forces, and as­
sumes individual molecules do not
occupy any volume relative to the
total volume. Due to these simplify­
ing assumptions, this EOS is only
applicable to simple, spherical mole­
cules in the gas or vapor-phase that
are present at relatively low pressure contain binary interaction param­
and high temperature. eters, which have been determined
In order to extend the applicabil­ from experimental data and actual
ity of the ideal gas equation to real plant operating data. Descriptions
gas systems and include liquid- of the applicability of various EOS
phase pressure-vapor-temperature to various chemical systems can be
(PVT) behavior, a number of pa­ found in online or in published oper­
rameters and modifications are ating manuals such as those found
introduced below: in Refs. 1 and 2 that accompany
• The compressibility factor most commercial computer simula­
• Acentric factor tion packages. The general applica­
• Van der Waal a and b correction bility of EOS methods is summarized
factor in Table 1.
• The binary interaction parameter However, EOS typically have a
Proprietary modifications to vari­ number of limitations, which re­
ous EOS can help to extend their strict their application to relatively
applicability to moderately non-ideal simple and primarily non-polar com­
mixtures of components commonly ponents, such as hydrocarbons
encountered in oil-and-gas and pet­ and light gases. The main
rochemical processes. For example, limitations include the following:
mixtures of glycol, alcohol, water • Complexities introduced by the
and hydrocarbons can be described mathematical methods that are
by some modified EOS. These EOS used to determine the fugacity co­

TABLE 2. LIQUID ACTIVITY COEFFICIENT METHOD APPLICABILITY GUIDE

Method Binary Different-sized Non-ideal/ Phase Comments
parameters molecules Polar binary splitting
mixtures
Ideal solution None No No No Only applicable to similarly sized molecules and components consisting of simi­
lar molecular species
Regular solution None Yes Yes Yes Allows pure component data to be used to predict activity coefficients. Non-ad-
justable parameters attempt to account for differences in polarity and molecular
size. Applicable to typical non-polar molecules such as hydrocarbons
Two-suffix Margules One No No Yes Simplistic. Assumes all molecules are of a similar size and only considers inter­
actions between two molecules (hence two-suffix) throughout a mixture
Two-constant Mar- Two No No Yes Assumes similarly sized molecules, but assumes that intermolecular interac­
gules tions between pairs ot molecules (ot different molecular species) are not equal
Three-suffix Mar­ Two No Yes Yes By extending the number ot molecules involved in interactions to three, the
gules range ot applicability is extended. Most widely used of the Margules methods
Van Laar Two Yes Yes Yes Does not assume similar sized molecular components, and has been extended
to non-ideal mixtures. Wide usage and simole to apply
Wilson Two Yes Yes No The main drawback is its inability to predict LLE (for instance, liquid-liquid im-
miscibility)
NRTL Three -Five Yes Yes Yes Widely applicable. Good forVLE. LLE and VILE
UNI-OUAC Two Yes Yes Yes Parameters not as dependent on temperature as NRTL parameters. Its form
makes it more applicable to solutions containing small or large molecules (such
as polymers)

chemical ENGINEERING WWW.CHEMENGONLINE.COM DECEMBER 2019

 TABLE 3. EOS VERSUS ACTIVITY COEFFICIENT METHODS
EOS model
Suitable for ideal and non-polar liquids
Consistent property prediction in “critical" region
Applicable for vapor and liquid phase
Parameters can be easily extrapolated with
temperature
efficient from an EOS
• Mixing rules that are used to ex­
tend the use of EOS to mixtures
are somewhat arbitrary, and there­
fore reduce the validity of the
results
• EOS are best suited to describing
vapor-phase behavior. However,
in combination with the advanced
mixing rules, EOS can (sometimes)
be used to model the liquid-phase
behavior for simple and non-po-
lar liquid mixtures. Examples for
sample systems are mixtures of
propane, ethane and butane and
mixtures of benzene and toluene
• EOS become inaccurate at higher
pressure and lower temperature,
and fail to predict condensation of
a gas to a liquid
relations, such as those by Chao-
Activity coefficient models Seader (CS) and Grayson-Streed
Suitable for defininq hiqhlv non-ideal liquids (GS), can also be used to better
Inconsistent in predicting behavior near the
predict the fugacity coefficient for a
"critical" region
range of components in the liquid
Represent the liquid phase only. Additionally, the gas
phase must still be defined using an EOS model phase. Thus, when used in con­
Binary interaction parameters depend on junction with an EOS to predict the
temperature vapor-phase fugacity coefficient,
VLE data can be predicted for sys­
Using EOS to predict fugacity tems containing light hydrocarbons
coefficients. EOS can be used to and hydrogen. In fact, it is gener­
mathematically determine the va­ ally recommended [7] to use the GS
por-phase fugacity coefficient and, method for systems involving heavy
though often with less accuracy, the hydrocarbons and hydrogen, such
liquid-phase fugacity coefficient. This as hydrotreating processes. It is also
is done by relating the fugacity coef­ recommended for use in the simula­
ficient to measurable properties of tion of topping units and vacuum ap­
pressure, temperature and specific plications handling heavy ends.
volume through the use of residual
functions. Residual functions use the Liquid-activity coefficient
thermodynamic properties of Gibbs methods
free energy, enthalpy and entropy to Non-ideal behavior in the liquid
mathematically describe deviations phase is evidenced by the following:
from ideal gas behavior, and can • Strong departures from Raoult's
therefore be used to evaluate the fu­ law (which relates ideal-gas phase
gacity coefficient of a component in behavior and ideal-liquid phase
a phase. behavior), due to polarity or hydro­
For predicting liquid-phase fugac­ gen bond formation
ity coefficients, semi-empirical cor­ • The formation of azeotropes

Ultrapen PTBT1, PTBT2, PTBT3

Smartphqne/Tablet Compatible Tester

Compatible with
900 S enes
Android and iOS Devices

Multi-Parameter
Monitor/Controller
2018 onductivity.i Resistivity. TDS
ORP, Salinity, ISE, Flow
ressurey with Touch Screen LCD
►moral

w.ww.myronlfcom
a* GiBluetoothj m CBM PAlSiy/
• AppStore 7.603 38:202<1
s

For details visit adlinks.chemengonline.com/73861-29

CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM DECEMBER 2019

 TABLE 4. RECOMMENDED THERMODYNAMIC METHODS FOR VARIOUS APPLICATIONS
Type of system
Glycol dehydration
Refinery crude tower (low pressure)
Refinery crude tower (high pressure)
Reformers and hydrotreaters
Sour water
Cryoqenic qas-processinq. natural qas
Air separation
Light hydrocarbons, petrochemicals
Aromatics
Ethylene towers
Reservoir systems
Steam systems
HF Alkylation
Aromatics plus non-aromatics
Hydrocarbon systems where water solubility in hy­
drocarbon is important
Non-ionic, chemical applications
Amine system
• The formation of two liquid phases
(via liquid-liquid extraction, LLE)
The liquid-activity coefficient is es­
sentially a correction factor that at­
tempts to quantify deviations from
ideal solution behavior. It is a strong
function of composition and to a
lesser extent, temperature (thus ac­
counting for interactions between
molecules), but a much weaker
function of pressure. Since the activ­
ity coefficients of the components in
a mixture are related to the compo­
sition, they are not independent of
each other.
There are two principal ap­
proaches for determining activity co­
efficients for liquid mixtures:
1. Excess Gibbs free energy
methods
2. Local compositional models
By estimating activity coefficients
using semi-empirical or semi-theo­
retical equations and fitting adjust­
able binary interaction parameters to
experimental data, VLE behavior can
be predicted.
Excess Gibbs free-energy meth­
ods use excess Gibbs free energy
functions, which are analogous to
residual Gibbs free-energy functions
in that they mathematically (and
thermodynamically) relate devia­
tions from ideal solution behavior to
functions involving the activity coef­
ficients of the components in the liq­
uid phase. Examples of activity coef­
ficient methods that use the excess
Gibbs free-energy methodology,
include the regular solution model,
various forms of the Margules equa­
Recommended property method
Peng Robinson (PR), specialized package
Barun K10 (BK10), Grayson-Streed (GSKGrayson-
Streed Erbar(GSE). SRK/PR
GS/GSE, SRK/PR
GS/SRK/PR
PR, sour PR, or special packages
SRK/PR. Benedict-Webb-Rubm Sterlinq (BWRS)
SRK/PR
SRK/PR, Soave Redlich Kwong Kabadi Oanner
(SRKKD) for hiqli pressure
Ideal for low pressure, SRK/PR for pressure greater
than 2 atm
Lee Kesler Plocker (LKP), BWRS
PR/SRK
Steam tables
NRTL
NRTL/UNIQUAC
SRKKD
Wilson. NRTL, UNIQUAC
EOS-based special packages
tion, and the Van Laar equation.
Local compositional models are
semi-empirical models based on the
concept that intermolecular forces
will cause non-random arrangement
of molecules in the liquid phase.
These models incorporate param­
eters related to energies of interac­
tion and distribution tendencies of
molecules, and this increases their
general application and reliability
while also increasing the need for
binary interaction data and experi­
mental data. Examples of activity
coefficient methods that use local
compositional models include NRTL
methods, Wilson, universal quasi­
chemical (UNIQUAC) methods
and more.
A word of caution when applying
liquid-activity-coefficient methods is
warranted here: Due to their empiri­
cal nature and reliance on fitted pa­
rameters, liquid-activity-coefficient
methods should be applied with
caution, giving heed to the following
criteria:
• Are components in the system
common and well understood?
• Are operating conditions relatively
moderate (that is, is the system at
low pressure, since fitted param­
eter data available in commercial
software packages are usually
based on the assumption of ideal
gas conditions in equilibrium with
the liquid phase)?
• Are experimental data available
and valid for the applicable range
of operating conditions, either for
regression to tune binary interac­
CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM DECEMBER 2019
tion parameters, or to verify the
predictions of the model?
If these criteria cannot be satisfied
with the available experimental data,
then the list of recommendations for
selecting an appropriate thermody­
namic property method should be
consulted to ascertain whether there
might be an appropriately modified
EOS to describe the system. Many
modified EOS are available to de­
scribe a variety of common ideal
and partially non-ideal systems. In
general, EOS are less cumbersome
to use because of their ability to de­
scribe both vapor-phase and ideal
liquid-phase behavior (unless modi­
fied to account for non-ideal behav­
ior), and because they are less reli­
ant on parameters that are fitted for
a specific range of conditions. In any
case, there is no substitute for ap­
propriate experimental data for tun­
ing and checking model predictions.
Group contribution methods
When experimental data are not
available, various methods can be
used to predict activity coefficients
based on the theory that functional
groups within a molecule of any
component contribute uniquely to
the activity coefficient for that com­
ponent. These methods are approxi­
mate because they rely on the fol­
lowing two simplifying assumptions:
• That the contribution of a func­
tional group within any molecule is
the same, and
• That the contribution of each func­
tional group is independent of the
contribution of any other group
within a molecule
Examples of these group contri­
bution methods include UNIFAC
and ASOG. In order to apply the
UNIFAC method, all components
must be condensable and electro­
lyte solutions cannot be modeled.
Given the assumptions associated
with these methods, in the absence
of experimental data for both the fit­
ting of adjustable parameters and
the comparison with model predic­
tions, such methods should be used
with extreme caution. From the au­
thor’s experience, however, it can be
stated that the data obtained from
the use of UNIFAC model is not as
accurate as the standard fitted val­
ues obtained from the various simu­
lation packages.
A summary showing the general
applicability of the liquid activity coefficient methods is
given in Table 2. Table 3 lists key parameters that dif­
FLOWROX
ferentiate EOS and activity coefficient models, allowing Proven Performance
for a comparison.

Selecting an applicable thermodynamic

property method
The thermodynamic property decision tree for selecting
Flowrox Filtration
the thermodynamic property route is shown in Figure 1.
Table 4 provides general recommendations for the
Solutions
selection of a thermodynamic system for modeling
and simulation. These recommendations are generic in Filter service:
nature and are based on the author’s experience and
• Original quality and performance
available literature [7,2]. Readers are cautioned to verify
the applicability based on the operating conditions in­ with lower costs
volved, and are strongly recommended to validate the
simulation model. The recommended selection is in­ Smart Filtration Digital Service:
tended to serve as initial guide to model the system. • Real-time insight into the filtration process
Further, the availability of binary interaction parameters
& troubleshooting
and applicable temperature and pressure range must
be verified before the selection. Readers shall note
that the developers of specific simulation packages Flowrox Filter Press:
have several times modified the typical thermodynamic • Fully automatic operation
methods and have presented customized solutions in • High quality
their software products. Readers should also take pre­ • Superb performance
caution to check the suitability of a particular thermody­
namic method with respect to the reference and user
manuals of respective simulation software.

Concluding remarks
Commercially available simulation packages contain
databases of both pure component and binary interac­
tion parameter data. Binary interaction data can also be ..
extended to multi-component systems. When used in
conjunction with the mathematical methods pressure,
temperature and phase composition can be predicted.
The predictive accuracy of the thermodynamic model­
ing methods is dependent on both (a) the suitability of
Flowrox Ceramic Disc Filter:
selected mathematical method(s) to describe behavior • Highly efficient filtration
in the chemical system, and (b) the quality of the com­ • Lowest energy consumption
ponent data (pure and binary interaction) available for • Reliable operation
use in the model. ■
Edited by Suzanne Shelley

References
1 PRO/ll with Provision Reference Manual - Thermodynamic Property Methods Applica­
tion Guidelines.
2. HYSYS User Manual: Thermodynamics: Appendix A - Property Methods and Calcula­
tions.
3. Reid. R.C. Prausnitz. J.M.. and Shenvood. T.K.. ‘Properties of Gases and Liquids; 3rd.
Ed.,McGraw-Hi, 1977.
4.Smith J.M., Van Ness H.C.. Abbott M. M.. "Introduction to Chemical Engineering Ther­
modynamics.* 7th Ed.. McGraw-Hill. 2005.
0 Complete filtration solutions to
• maximize production
Author • reduce downtime
Atul B. Choudhari is technology principal at Tata Consulting • lower total costs!
Engineers Ltd, (Empire Tower, Cloud City campus, Thane-Bela-
pur Road. Airoli, Navi Mumbai 400708, India; Phone:+91 -22-
61148620; Email: achoudhari@tce.co.in). In his 26 years of
professional experience. Choudhari has been involved with CONTACT US!
flowsheet simulations, process optimization, basic and detailed www.flowrox.com
888-356-9797
engineering for petrochemicals, petroleum refinery and hydro­
carbon processes. He has authored various technical papers sales-us@flowrox.com ©oo
in the field ot process optimization. Choudhari holds a bache­
lor's degree in chemical engineering trom Marathwada
University (Aurangabad). For details visit adlinks.chemengonline.com/73861-17
55
CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM DECEMBER 2019