James 2020
James 2020
a r t i c l e i n f o a b s t r a c t
Article history: Accurate calculations of properties such as enthalpy, entropy, and fugacity are crucial for
Received 24 September 2019 chemical process design. These properties are calculated from equations of state in com-
Revised 20 December 2019
monly used process design software such as CHEMCAD [1], and software-based calcula-
Accepted 12 March 2020
tions of properties have been routine for decades. It is important to be able to rely on
Available online 21 April 2020
the accuracy of these calculations. Furthermore, students of thermodynamics and process
Keywords: engineers must be able to easily reproduce the calculations from the process simulator.
Lee-kesler In this study, we attempted to validate the compressibility factors, enthalpy, entropy, and
Enthalpy fugacity coefficients from CHEMCAD by comparing the results of the process simulator to
Entropy results from a published solution algorithm [2]. Our results show good consistency for
Fugacity most molecules. However, we observe large deviations for nitric oxide, water and hydro-
Compressibility gen, and somewhat smaller deviation for a few other molecules. This report presents a
Process detailed comparison of the thermodynamic properties of 48 molecules at two different
CHEMCAD
states to demonstrate that our calculation is correct and that the deviations for hydrogen
and water are real.
Published by Elsevier B.V.
Specifications Table
∗
Corresponding author.
E-mail address: [email protected] (A. Biaglow).
https://doi.org/10.1016/j.cdc.2020.100371
2405-8300/Published by Elsevier B.V.
2 C. James, M. Armstrong and G. Washington et al. / Chemical Data Collections 27 (2020) 100371
1. Rationale
Accurate calculation of properties such as enthalpy, entropy, and fugacity are crucial for chemical process design. These
properties can be calculated from equations of state in process design software such as CHEMCAD [1], and software-based
calculations of properties have been routine for decades. Because of the importance of thermodynamic properties in process
design, engineers and students must be able to easily reproduce the data from the process simulator. In this study, we
attempted to confirm the compressibility, enthalpy, entropy, and fugacity coefficients calculated by CHEMCAD by comparing
the results of the process simulator to results from a published solution algorithm [2].
Our procedure was to first calculate the properties at two defined states as well as changes between two states using
the Lee-Kesler method [2]. State 1 in our study is defined in terms of reduced temperature and pressure as Tr =0.90 Pr =0.10,
and state 2 is defined as Tr =2.2 and Pr =0.8, where Tr =T/Tc and Pr =P/Pc . We selected these two states carefully so as to
give relatively large property changes while at the same time staying well away from any phase transitions. All of our
calculations were conducted using the algorithm published in Reference 1 and implemented by us in Mathematica [3].
We then implemented the same method in CHEMCAD. Property data were then arranged in a tabular format for direct
comparison. We analyzed the same corresponding states and state changes for all 48 molecules in Table C.1 of Introduction
to Chemical Engineering Thermodynamics [4], a commonly used text.
Our results show consistency for most molecules in our study, with large deviations for nitric oxide, water and hydro-
gen, and smaller but still significant deviations for a few other molecules, namely dinitrogen tetroxide, nitrogen dioxide,
and sulfur dioxide. We repeated the calculations for 48 molecules to demonstrate that the observed deviations are real
and reproducible. These results are important because of the ubiquitous nature of water and hydrogen in process design
simulations, so it is important to understand how they arise.
2. Procedure
The procedure used here is published in Reference 1 and briefly outlined here. Eqs. (1.1) to (1.5) appear in Reference 1
and are repeated here. We first briefly describe our algorithm and then we discuss how the simulation is implemented in
CHEMCAD.
Equation of State – Modified Benedict-Webb-Rubin
P V
B C D c4 γ γ
Z=
r r
=1+ + 2 + 5 + 3 2 β+ exp − (1.1)
Tr Vr Vr Vr Tr Vr Vr2 Vr2
ω
Z = Zsimple + · Zre f erence − Zsimple (1.2)
ωre f erence
Fugacity Coefficient
f B C D
ln = Z − 1 − ln (Z ) + + + +E (1.3)
P Vr 2Vr2 5Vr5
Residual Enthalpy and Entropy
R H − H ig b2 + 2 b3 /Tr + 3 b4 Tr2 c2 − 3 c3 /Tr2 d2
H = = Tr Z − 1 − − 2
+ + 3E
RTC TrVr 2TrVr 5TrVr5
S − Sig P
b1 + b3 /Tr2 + 2 b4 Tr3 c1 − 2 c3 /Tr3 d1
SR = + ln = ln (Z ) − + − + 2E (1.4)
TC P ig Vr 2Vr2 5Vr5
b2 b b c c d c4 γ γ
where B = b1 − − 32 − 43 , C = c1 − 2 + 33 , D = d1 + 2 , and E = β + 1 − β + 1 + 2 exp − 2
Tr Tr Tr Tr Tr Tr 2Tr3 γ Vr Vr
Enthalpy and Entropy
T
2 2
c/T e/T
H= H of + 0.001 · a+ b· + d· dT + H R
298.15 Sinh[c/T ] Cosh[e/T ]
2 2 P (1.5)
T 1 c/T e/T
S = Sof + 0.001 · 298.15 a+ b· + d· dT − ln + SR
T Sinh[c/T ] Cosh[e/T ] 1.0
In Eq. (1.5), the standard enthalpy and entropy of formation and the constants a-e for ideal gas heat capacity calculations
are specific to the molecule being studied. For consistency, all of these constants were taken from CHEMCAD, which uses
the DIPPr data set [5], as described below and shown in Fig. 1.
The first step in the procedure is to solve the equation of state (1.1) for reduced volume (Vr ) given reduced temperature,
pressure, and acentric factor (Tr , Pr , and ω). This is done once for the simple fluid (defined as ω=0), and once for the ref-
erence fluid (ωreference =0.3978). Calculation of Vr then allows calculation of Z for the simple and reference fluids. Second,
using Eq. (1.2), the compressibility factors (Z) for the simple and reference fluid results are combined in a Pitzer-type corre-
lation (1.2) using the acentric factors for the reference compound and the compound being analyzed (methane, ethane, etc.)
Third, the resulting Z values are used to calculate the fugacity coefficient (f/P), residual enthalpy (HR ) and residual entropy
(SR ) from Eqs. (1.3) and (1.4) for the simple and reference fluids and again combined with corresponding Pitzer-type cor-
relations similar to Eq. (1.2). Finally, absolute enthalpy and entropy are calculated from the residuals, the ideal gas values,
and the standard enthalpy and entropy of formation using Eqs. (1.5). “Absolute” in this context refers to gas-phase forma-
tion properties at 298.15 K and 1 bar, and is analogous to CHEMCAD stream properties, as opposed to differences between
streams.
The Lee-Kesler computations were also performed in CHEMCAD. The first step is to draw the process in the simulator.
Referring to Fig. 1, a simple process using a compressor (A) and heat exchanger (B) converts the fluid from T1 and P1 to
T2 and P2 . CHEMCAD requires specification of actual temperatures and pressures as opposed to the reduced values. The
stream box to the left of the process flow diagram is set to display the relevant properties from the streams. Next, we set
the desired thermodynamic model. The enthalpy model is chosen in “Thermodynamic settings (C),” and was set to Lee-
Kesler with ideal gas heat capacity constants from DIPPr. These settings are each indicated with red arrows in the figure.
The critical temperature, pressure, standard enthalpy of formation, and standard Gibbs energy of formation used in our
Mathematica algorithm are found under “Basic Data” (D). Enthalpy of formation and Gibbs energy of formation are used to
calculate the entropy of formation used in Eq. (1.5). The heat capacity coefficients are found in “Library Heat Capacity Data”
(E). All relevant settings and constants are indicated with red arrows in the figure.
Our results are summarized in Tables 1 through 5. Tables 1 to 4 are comparisons of absolute enthalpies, absolute en-
tropies, fugacity coefficients, and compressibility factors, respectively. Table 5 is a comparison of the change in enthalpy
and entropy from State 1 to State 2. The tables also show a comparison of our calculations with the results obtained from
CHEMCAD. All comparisons are at State 1 and State 2, where State 1 is at a reduced temperature and reduced pressure of
Tr1 = 0.9 and Pr1 = 0.10 and State 2 is at Tr2 = 2.2, Pr2 = 0.8.
Significant differences in the results are highlighted in bold red font in the tables. We used a comparison of percentage
errors to determine these outliers. The values highlighted in red show more than ten standard deviations difference between
the percentage error for that data point and the average percentage error of the non-outliers. For example, for the H1
values for nitric oxide in Table 1, the calculated enthalpy from Mathematica is 85,419 J/mol and the CHEMCAD value is
85,813 J/mol, giving a difference of 394 J/mol and an average of the two values of 85,616 J/mol. The difference expressed as
a percent deviation from the average is 100(394/85,616), or 0.46%. The average of all deviations in H1 over the entire table,
not including highlighted outliers, is 0.0159% with a standard deviation of 0.0351%, and 0.46% is greater than 10 standard
deviations from the average. That is to say, in this example, 0.46 > 0.0159+10 × 0.0351, or 0.46>0.37), which justifies the
indication of this point as an outlier. All other bold red values in Tables 1-5 were similarly analyzed.
Using the criterion described above, discrepancies in the data occur for nitric oxide, nitrogen dioxide, dinitrogen tetroxide,
and water in Table 1. In Table 2, we see discrepancies for ethylbenzene, formaldehyde, hydrogen, hydrogen chloride, nitric
oxide, sulfur dioxide, and water. Of these, the most significant deviation appears to be for water. In Table 5, for enthalpy
and entropy differences, we see discrepancies for hydrogen, nitric oxide, dinitrogen tetroxide, and water. The discrepancies
for water and hydrogen are explained below.
From the thermodynamics documentation in the help menu, “Steam tables are used to determine the enthalpy of water
(if water is the only component and SRK enthalpy is selected).” It appears CHEMCAD also makes this choice when Lee-Kesler
was selected for enthalpy. We also repeated the calculation using the steam table equations cited by CHEMCAD [7] and ob-
tained the same result. Interestingly, when water is “cloned” in CHEMCAD and the cloned water is used in the simulation,
the values are much closer to the Mathematica calculation. For H1 for cloned water, we observe −232,913 J/mol in CHEM-
CAD, versus −232,898 J/mol in Mathematica, while H2 becomes −198,044 J/mol in CHEMCAD versus −198,036 J/mol in
Mathematica. For entropy values S1 and S2 , the values for cloned water values are −47.84 and −28.53 J/mol•K, respec-
tively, while we calculate −48.15 and −28.69 J/mol•K. The ability to use the cloning feature to circumvent the use of
steam tables does not appear to be documented in the software and was made apparent to us during the review of this
manuscript.
Hydrogen is also noteworthy with a large deviation in entropy, as shown in Table 2. This deviation is due to the way
CHEMCAD handles the ideal gas heat capacity integral (eq. (1.5)). The software uses upper and lower bounds on the func-
tion to prevent the return of unreasonably high values from the integral. This is a very reasonable and prudent programming
approach but also causes the large discrepancy seen for this molecule. It would be useful if the company provided documen-
4 C. James, M. Armstrong and G. Washington et al. / Chemical Data Collections 27 (2020) 100371
Fig. 1. Screenshots of the implementation of the simulation methods in CHEMCAD. Methane data published with permission of the American Institute
of Chemical Engineering (AIChE®) and its thermo-physical property research consortium, the Design Institute for Physical Properties (DIPPR®) using the
DIPPR Evaluated Database ©.
C. James, M. Armstrong and G. Washington et al. / Chemical Data Collections 27 (2020) 100371 5
Table 1
Comparison of calculated Lee-Kesler enthalpies with CHEMCAD Lee-Kesler enthalpies at reduced temperatures and
pressures of (Tr1 = 0.9, Pr1 = 0.10) and (Tr2 = 2.2, Pr2 = 0.8). The subscripts MMA and CC designate Mathematica
and CHEMCAD, respectively. Enthalpy units are J/mol.
tation of this feature. Furthermore, the library heat capacity data shown in Fig. 1(E) shows that CHEMCAD indicates a lower
bound for the ideal gas heat capacity polynomial, and the lower bound for hydrogen is 250 K. The absolute temperature
at which we report the discrepancy is 29.94 K, which is well below 250 K and seems at first to be extreme. However, the
reduced temperature of 0.9 at 29.94 K is not unreasonable as a process condition, and corresponding states theory indicates
that thermodynamic properties should be consistent with the other molecules in this study at the same reduced conditions.
At the time of writing of this paper, CHEMCAD was in the process of revising the software to mitigate this issue.
We took the following additional steps to validate our work. First, each author independently repeated the calculations
for the highlighted molecules. Second, we and repeated the entire study using the Peng-Robinson equation of state [6] in
Mathematica and CHEMCAD, with similar discrepancies. Third, we applied exactly the same algorithm to all 48 molecules
to minimize the chance of data entry error. This also demonstrates that our algorithm is in good agreement with CHEMCAD
6 C. James, M. Armstrong and G. Washington et al. / Chemical Data Collections 27 (2020) 100371
Table 2
Comparison of calculated Lee-Kesler entropies with CHEMCAD Lee-Kesler entropies at reduced temperatures and
pressures of (Tr1 = 0.9, Pr1 = 0.10) and (Tr2 = 2.2, Pr2 = 0.8). The subscripts MMA and CC designate Mathematica
and CHEMCAD, respectively. Entropy units are J/mol•K.
for the vast majority of the molecules in this study, providing evidence that the algorithm is most probably correct. Fourth,
further validation was obtained using a third-party online Lee-Kesler calculator [8]. In all 48 cases, the online calculator was
in perfect agreement with our results but differed from CHEMCAD in the molecules discussed above.
In summary, CHEMCAD is a state-of-the-art process simulator and is widely used in the engineering community. Engi-
neers should understand what the simulator is doing and how different selections affect the underlying math and results.
As engineers, we are also challenged to understand why an outlier may still be a valid result. To explore how the simulator
works and understand the outliers, it is interesting and informative to check the simulation calculations, and this paper
provides a performance report for users of the software who might rightfully want such an assessment.
C. James, M. Armstrong and G. Washington et al. / Chemical Data Collections 27 (2020) 100371 7
Table 3
Comparison of calculated Lee-Kesler fugacity coefficients (f/P) with CHEMCAD Lee-Kesler fugacity coefficients at
reduced temperatures and pressures of (Tr1 = 0.9, Pr1 = 0.10) and (Tr2 = 2.2, Pr2 = 0.8). The subscripts MMA and
CC designate Mathematica and CHEMCAD, respectively.
Table 4
Comparison of calculated Lee-Kesler compressibility factors with CHEMCAD Lee-Kesler compressibility factors at
reduced temperatures and pressures of (Tr1 = 0.9, Pr1 = 0.10) and (Tr2 = 2.2, Pr2 = 0.8). The subscripts MMA and
CC designate Mathematica and CHEMCAD, respectively.
Table 5
Comparison of differences in Lee-Kesler enthalpy and entropy for change from reduced temperatures and pressures
of (Tr1 = 0.9, Pr1 = 0.10) to (Tr2 = 2.2, Pr2 = 0.8), calculated versus CHEMCAD, in J/mol and J/mol•K, respectively.
The subscripts MMA and CC designate Mathematica and CHEMCAD, respectively.
Declarations of Interest
None.
References
[1] CHEMCAD version 7.1 by chemstations, inc., https://www.chemstations.com/. (Last accessed March 24, 2020), 2020.
[2] B.I. Lee, M.G. Kesler, A generalized thermodynamic correlation based on three-parameter corresponding states, AIChE J. 21 (3) (1975) 510–527.
[3] Mathematica version 11.3 by wolfram research, https://www.wolfram.com/mathematica/. (Last accessed March 24, 2020), 2020.
[4] J.M. Smith, H.C. van Ness, M.M. Abbott, Introduction to Chemical Engineering Thermodynamics, 7th Edition, Mc-Graw-Hill, New York, New York, 2005.
[5] Design Institute for Physical Properties (DIPPr), American Institute of Chemical Engineers, https://www.aiche.org/dippr. (Last accessed March 24, 2020),
2020.
[6] D.-.Y. Peng, D.B. Robinson, A new two-constant equation of state, Ind. Eng. Chem. Fundam. 15 (1) (1976) 510–527.
[7] The international association for the properties of water and steam, steam table version IAPWS R7-97 (2012), Lucenre, Switzeland, August 2007, http://
www.iapws.org/relguide/IF97-Rev.html.
[8] H, S, and f/P were verified with the Lee-Kesler calculator by J.M. Haile, Macatea Productions, http://www.macatea.com/wshop/lk/. (Last accessed January
3, 2019), 2019.