Fuel 85 (2006) 748754

www.fuelfirst.com

Enhanced method for predicting the properties of light petroleum fractions

Tareq A. Albahri *
Chemical Engineering Department, Kuwait University, P.O. Box 5969, Safat 13060, Kuwait
Received 10 March 2005; received in revised form 4 September 2005; accepted 5 September 2005
Available online 28 September 2005

Abstract
A molecularly explicit property prediction mode (MEPP) is developed for predicting the properties of light petroleum fractions from ASTM
D86 distillation data. The method uses a molecularly-explicit characterization model (MECM) based on optimization to define the molecular
composition of the petroleum fuel from which more than 20 thermodynamic and transport properties are calculated. The model is tested using a set
of 68 pure components to characterize and predict the properties for 30 different petroleum naphtha samples and is found to be in excellent
agreement with the current conventional prediction methods based on global properties. The method is suitable for incorporation into the new
generation of molecularly-explicit models for simulating the kinetics and dynamics of petroleum refining processes and has the flexibility to be
tailored to use any set of pure components as desired.
q 2005 Elsevier Ltd. All rights reserved.
Keywords: Characterization; Property estimation; ASTM D86 distillation

1. Introduction
Accurate characterization of petroleum fuels is an important
step in the application of kinetic and thermodynamic
calculations for the design, operation, and simulation of
petroleum refining processes. An insufficient description of
heavier hydrocarbons (e.g. pentane and heavier; CC
5 ) reduces
the accuracy of predictions. Ideally, fuels properties are
determined experimentally in the laboratory on actual fluid
samples taken from the process under study. Because of the
expense of the experimental determination of such data, there
is interest in their accurate prediction.
In order to speed up the execution of real-time simulation, a
series of simplified correlations has been proposed for the
evaluation of physical properties of petroleum fractions during
the last five decades [1]. Several charts and correlations were
developed many of which require as input parameters the fuels
bulk or global properties such as the average boiling point, the
specific gravity and some characterization factors. However,
wide boiling range fractions are mixtures of a large number of
hydrocarbon compounds the type of which varies along the
distillation curve, therefore, a single value for boiling point or
* Tel.: C965 664 0663; fax: C965 483 9498.
E-mail address: albahri@kuc01.kuniv.edu.kw.
URL: http://www.albahri.info.

0016-2361/$ - see front matter q 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.fuel.2005.09.005

specific gravity does not characterize the fraction very well.

Moreover, as many existing correlations are based on properties of pure compounds, errors in predicted values from the
correlations increase significantly when the methods are
applied to mixtures.
Among the effort to enhance the characterization of
petroleum fractions, Riazi developed an advanced method
based on a two-parameter distribution model to predict the
complete property distributions for a CC
7 petroleum fraction
using either the pseudo-component or the single carbon
number (SCN) groups approach [2,3]. Although very useful,
these methods are far from being considered molecularly
explicit, and has not proven predictive ability for other
properties.
It is possible to indirectly predict the properties of petroleum
fuels using sophisticated analytical equipment such as GC, IR,
and NMR. Sasano [4] for example developed a method for
measuring the research octane number of gasoline by GC. The
property value for the mixture was calculated using a
correlation the input parameter of which was the GC-measured
pure component concentrations. However, these extensive
experimental programs are complex, expensive and time
consuming, thus, did not find wide acceptance in daily refinery
operation.
Recognizing that property measurements in petroleum feedstocks would be difficult, we sought to define and characterize
these feedstocks in a simple mathematical way that would
allow for a molecular property representation of a given

T.A. Albahri / Fuel 85 (2006) 748754

feedstock. Since, the ASTM D86 distillation test has found

wide acceptance in refinery daily operation on a routine basis
we sought utilizing these traditional definitions of complex
petroleum feedstocks for achieving that objective. Hence, we
present a novel technique for predicting the thermo-physical
and transport properties of light petroleum fractions, in terms
of molecular components, based on the knowledge of their
global properties that are easily measurable in the laboratory
such as ASTM D86 distillation. It is worth mentioning that
there is no method in the literature that predicts the various
properties of petroleum fractions from the ASTM distillation
test alone.
2. Method
2.1. Model development
The proposed model is based on the concept that the global
properties of a petroleum fraction must be equal to those
calculated from the pure components comprising that
petroleum fraction. When both bulk and pure component
properties are available, the composition of a limited set of
pure components in the petroleum fraction may be predicted.
In our previous work on the simulation of light petroleum
fractions we developed a molecularly explicit characterization
model (MECM) that can determine the optimal molecular
distribution in light petroleum fractions in terms of pure
components [5,6]. The concentration of a pre-selected set of
representative true-components is calculated using the global
as well as the internal and structural properties of the petroleum
fraction. Upon further experimentation in the present work we
found that it not essential for all the properties of the petroleum
fuel to be optimized against those from the pure components.
In fact only the ASTM D86 Distillation, the PNA content and
the RVP were sufficient to provide a feasible solution. The
predicted composition may then be used to predict the other
global properties of the petroleum fuel using appropriate
mixing rules as simplified in Fig. 1.
The standard input global-properties for the model are the
petroleum fractions distillation data (either ASTM D86, TBP,
SimDist, etc.), the RVP, and the PNA content. The internally
calculated global properties are shown in Table 1. These
properties are determined using well established methods in the
literature or from methods developed specifically for this
purpose [5]. The boiling point distribution, RVP and PNA are
also calculated from pure components data. The two methods
are contrasted and the difference is minimized using an
optimization algorithm. The model output is a computationally
generated explicit atomic detail of the petroleum feedstock.
RVP
ASTMD 86 or
TBPDistillation
PNA

MECM
Model

749

Experimental values of the RVP and PNA content are

always desirable. However, if not supplied as input, they are
calculated by the MECM model using generalized correlations
available in the literature, making the ASTM D86 distillation
or the true boiling point (TBP) the only experimental data
required as input. The availability of other input properties
such as, for example, the API gravity is an additional benefit to
improve the model predictions but is not essential.
2.2. Methodology
The chemical logic diagram for this system is depicted in
Fig. 2 which illustrates the methodology used to develop
the molecularly explicit property prediction (MEPP) model and
the procedure followed to analyze the simulation problem. The
petroleum fraction input to the model is characterized in terms
of bulk properties (RVP, PNA, and ASTM D86 distillation).
The MECM model is used to generate a molecular ensemble
that retains the qualitative features expected to mimic the
petroleum fraction. The predicted molecular distributions are
aggregated in the MEPP model to calculate the global
properties (API, MW, viscosity, etc.). The model performance
in simulating the actual petroleum fraction is compared to
products from generalized correlations which use the contemporary method of feed characterization. The theoretical
background of the MECM model and the methods used for
determination of the global properties, concentration of light
components, aggregating the molecular ensemble, model
solution using optimization, and molecular distribution of the
molecular ensemble are explained in details by Albahri [5].
Here, we take it one step further by incorporating the MECM
model in the MEPP model to make it viable for property
prediction purposes.
The distribution of the molecular ensemble in the MEPP
model is determined in terms of volume fractions by
minimizing the following objective function modified from
that of the MECM model [5],
!2 

n
X
100
100 2
0
0
SZ
Tbj KTbj W0
C PNAKPNA W1
Tbj
PNA
jZ1
(1)
where j is the index number of the molecule, and n is the total
number of molecules. PNA is the paraffin, naphthene, and
aromatic content for the petroleum fraction determined
experimentally or calculated from experimental using generalized correlations with the global properties as input
parameters. PNA 0 is the paraffin, naphthene, and aromatic
content for the petroleum fraction calculated from aggregating

C3 Detailed
composition
of 68
predefined
C11 molecules

Aggregation
Model

API
MW
RI
H/C
Viscosity
SurfaceTension
Otherproperties

Fig. 1. Simplified schematic representation of the proposed model.

750

T.A. Albahri / Fuel 85 (2006) 748754

Table 1
Comparison of global properties of light reformate calculated from generalized correlations and from aggregation of pure components using MEPP model

Specific gravity
Vapor pressure @ 37.8 8C, bar
Reid vapor pressure, bar
Cubic average boiling point, 8C
Mean average boiling point, 8C
Volume average boiling point, 8C
Weight average boiling point, 8C
Mole average boiling point, 8C
Watson characterization factor
Molecular weight, gm/mole
Refractive index
Hydrogen content, wt frac
Kinematic viscosity @ 98.9 8C, mm2/s
Kinematic viscosity @ 37.8 8C, mm2/s
Surface tension @ 25 8C, dynes/cm
Aniline point, 8C
Critical temperature, 8C
Pseudo critical temperature, 8C
Pseudo critical pressure, bars
Critical compressibility factor
Paraffins content, mole%
Naphthenes content, mole%
Aromatics content, mole%
Acentric factor
Freezing point, 8C
Research octane number
Motor octane number
Heat of vaporization @ NBPb, J(abs)/gm
Net heat of combustion @ 25 8C, J(abs)/gm
Isobaric liquid heat capacity @ 15.6 8C, J(abs)/gm K
Isobaric vapor heat capacity @ 15.6 8C, J(abs)/gm K
Liquid thermal conductivity @ 25 8C, J(abs)/s m2 K/m
a
b

Experimental or predicted
from experimental

MECM simulationa

% Error (deviation)

0.7143
0.7033
0.655
87.7
84.4
89.3
90.3
81.6
12.09
93.5
1.3973
0.1458
0.32
0.48
20.44
37.8
265.6
254.4
31.03
0.2773
71.50
17.19
11.3
0.284
K102.3
72.3
70
322.08
44,245
2.437
1.882
0.1099

0.7083
0.7033
0.655
82.6
81.1
84.1
87.1
79.6
12.18
93.1
1.3960
0.1507
0.31
0.49
19.48
57.1
261.3
254.8
31.23
0.2664
70.88
17.43
11.7
0.285
K104.7
73.2
71.7
323.06
44,047
2.039
1.604
0.1161

K0.84
0.00
0.00
K1.44
K0.94
K1.45
K0.92
K0.57
0.77
K0.42
K0.09
3.34
K3.56
1.70
K4.70
6.21
K1.51
0.14
0.67
K3.91
K0.87
1.43
3.34
0.32
K1.44
1.23
2.37
0.31
K0.11
K16.38
K14.79
K9.08

From aggregation of pure components.

At 1 bar and the mean average boiling points [1].

pure components in the molecular ensemble using mixing

rules. Tbj0 is the boiling point of pure component j and Tbj is the
boiling point value on petroleum fractions TBP curve
corresponding to component j. W0 is the weighting factor for
the boiling points and W1 is the weighting factor for the PNA
fractional composition valued 1 and 25, respectively. S is the
objective function to be minimized.
The objective function consists of two parts. The first
compares the boiling point of the pure component to the boiling
point on the TBP curve of the petroleum fraction. The second
part of the objective function compares the PNA fractional
composition of the petroleum fraction and those from
aggregation of the molecular ensemble.
The distribution of the molecular ensemble outputted from
the optimization module combined with pure component data
is converted into global properties (e.g. molecular weight and
H/C content, etc.) using mixing rules [5]. The ultimate output
of the program is the global properties of light petroleum
fractions from the knowledge of their conventional laboratory
analysis such as RVP, and PNA content and ASTM D86
distillation, with the latter being the minimum input required.
To verify the validity and accuracy of the MEPP model
the simulation outcome in terms of average global properties

of the petroleum fraction are compared to those calculated

from other methods in the literature using generalized
correlations.
3. Results and discussion
Experimental data on 30 petroleum naphtha cuts were
obtained from the literature [7]. The ASTM D86 distillation
data for each was used to compute molecular distribution
results from which the transport and thermodynamic properties
were calculated. Using the MEPP methods and correlations to
determine the petroleum fractions global properties, the PNA
compositions and physical properties of one of the samples
(petroleum naphtha reformate fraction) are estimated and
tabulated in Table 1. The same properties estimated from
aggregation of the molecular ensemble for the MEPP simulated
petroleum fraction are also shown in the same table. The
percent deviations of these estimated properties are also shown.
The small deviations indicate how the MEPP estimated
properties are representative of the undefined petroleum
mixture.
Detailed comparison for some selected properties for the 30
naphtha samples are shown in the summary presented in

T.A. Albahri / Fuel 85 (2006) 748754

751

START

Preseleced set of
pure components

ASTMD86
DistillationData
+ PNA+RVP

Pure Component
Data from Literature
or Predicted

Initial Estimation
of composition

Convert
to
TBPcurve

Mixing
Rules

Global Properties of Light

Petroleum Fractions from
generalized correlations

END

Comparison&
ErrorAnalysis

Global
Properties

Optimization
Model

Objective
Function

Optimal Composition
of the Simulated
Petroleum Fraction

Mixing
Rules

Fig. 2. Chemical logic diagram for the methodology used to develop the property model.

Table 2 and the bar plots are also shown in Fig. 3. The fit
between the experimental and stimulation results as shown by
the average percentage errors and correlation coefficients is
quite good. The average absolute deviation for all the
properties in Table 2 is about 3.5% and the average correlation
coefficient is 0.96.

The RVP and PNA fractional composition results

demonstrate an almost exact match to within statistical
errors. This was to be expected in that the input distribution
was directly derived from the experimental distribution.
From all the properties investigated, the true vapor pressure
at 37.8 8C, the specific (API) gravity, molecular weight,

Table 2
Error analysis for some of the properties investigated.
No.

Property

Value range

Av. % error

Correlation coefficient

1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.

API gravity
Cubic average boiling point, 8C
Mean average boiling point, 8C
Volume average boiling point, 8C
Molar average boiling point, 8C
Weight average boiling point, 8C
Watson characterization factor
Molecular weight, gm/mole
Refractive index
Hydrogen content, wt fraction
Motor octane number
Kinematic viscosity at 98.9 8C, mm2/s
Kinematic viscosity at 37.8 8C, mm2/s
Surface Tension of liquid at 25 8C, dynes/cm
Aniline point, 8C
True critical temperature, 8C
Pseudocritical temperature, 8C
Pseudocritical pressure, bar
Heat of vaporization at NBP, J(abs)/gm
Net heat of combustion at 25 8C, J(abs)/gm
Freezing, 8C
Acentric factor
Critical compressibility factor
Flash point, 8C
Liquid thermal conductivity at 25 8C, J(abs)/s m2 K\m

3594
34134
32133
35134
30134
35134
10.6113.06
68.2110.8
1.35461.4764
0.10630.1716
6097
0.250.43
0.320.7
14.5127.96
2065
196338
191338
26.235.9
311.6363.7
43,99645,362
(K119)(K72)
0.2440.312
0.27500.2805
(K74)(23)
0.10200.1211

2.14
1.34
0.99
1.34
0.83
1.07
0.80
2.06
0.21
2.57
4.15
4.04
5.41
2.67
1.74
0.93
0.80
2.22
1.17
0.18
5.38
3.13
0.25
5.16
1.78

0.990
0.995
0.995
0.995
0.995
0.996
0.970
0.990
0.993
0.965
0.850
0.960
0.972
0.995
0.830
0.991
0.989
0.900
0.948
0.960

0.946
0.832
0.924
0.976

752

T.A. Albahri / Fuel 85 (2006) 748754

(b) 40.0

(a) 140

35.0

120

30.0

100

25.0

80

20.0

60

API
MeABP
MW
MON

40
20

Surface Tension

10.0

Pc (pseudo)

5.0

0
0
(c)

Kw

15.0

20 40 60 80 100 120 140

(d)

0.8
0.7

15

25

35

1.6
1.55

0.6
1.5

0.5

1.45

0.4
H content
Viscosity 37.8C
Viscosity 98.9C
Acentric factor
Zc
Thermal cond.

0.3
0.2
0.1
0
0
(e)

0.2

0.4

0.6

1.4
RI
?Hc/30,000

1.35
1.3
1.3

0.8
(f)

1.4

1.5

1.6

400

40
350

20
0

300

20
250

40
60
80
100
100

50

Tc (true)
Tc (pseudo)
?Hv

200

Flash Point

50

150
150

200

250

300

350

400

Fig. 3. Bar plots (x-axis is property value calculated from generalized correlations, y-axis is property value calculated by MEPP model).

surface tension of liquid at 25 8C, Watson characterization

factor, refractive index, hydrogen content, kinematic viscosity at 37.8 and 98.9 8C, true critical temperature,
pseudocritical temperature, pseudocritical pressure, acentric
factor, liquid thermal conductivity at 25 8C, flash point, net
heat of combustion at 25 8C, heat of vaporization at the
normal boiling point, and the mean, cubic, weight, molar,
and volume average boiling points correlated very well as
evident by the average percentage errors and correlation
coefficients that were in the upper nineties. This indicates
how good the feed modeling and characterization by the
proposed technique are.
Less accurate were the predications for the aniline point,
motor octane number, and critical compressibility factor with
correlation coefficients ranging between 0.8 and 0.9 followed
by the isobaric liquid heat capacity at 15.6 8C with correlation
coefficient of 0.69 and an average error of 10%. No correlation
was evident for the research octane number, isobaric vapor heat
capacity at 15.6 8C, cloud point or freezing point although

the later is predicted with an average percentage error of about

5.38% which is not too bad.
Differences between the conventional and simulated results
may be caused by inconsistencies in the data provide by the
analytical tests, approximations made during construction of
the MEPP model including the choice of the molecular
ensemble and inherent errors in the various correlations used.
Although, the MEPP model predictions for the aniline point
and critical compressibility factor do not show as good
correlation as the other properties in Table 2, the average
percentage error for the two properties is quite acceptable. The
model predictions for MON shows promise for further
improvement with using interaction parameters since this
property does not mix linearly with composition because of the
synergistic and antagonistic effect the molecules possess when
they are present together. It is important to note that because
sufficient experimental data do not exist, the method has been
tested for about 25 physical properties of 30 petroleum naphtha
fractions. More is required to develop interaction parameters to

predict the properties that do not mix conventionally such as

octane number.
3.1. Simulation time
The steady state MEPP simulation of the petroleum naphtha
systems was run successfully on Pentium IV personal computer
running at 3.0 GHz. The CPU time of the simulation to
characterize the petroleum naphtha feed into molecular
ensemble and recalculate its global properties by aggregating
the molecular ensemble was in the order of 10 s. The CPU time
required to predict the properties of the feed was proportional to
the number of molecules simulated. For this reason, it was
efficient to simulate the fewest molecules possible that
accurately described the petroleum naphtha fraction. Increasing
the number of molecules, results in a more accurate estimate of
the fraction bulk properties and the average property is a better
estimate of the true simulated property. A sample size of 68
molecules offers an accurate estimate of the feedstock properties
and allows for the future incorporation of reaction kinetic
models at a reasonable computation expense.
3.2. Model advantages
The MEPP model proves to be a powerful tool for
simulating and predicting the properties of petroleum fuels.
The procedure for predicting the fluid properties is simple and
straightforward. It requires limited information from readily
available lab analysis and simple analytical characterizations to
describe the feedstock. The computerized procedure can be
combined with or incorporated in petroleum process simulation
packages thus enhancing their information content.
Moreover, this model provides a foundation for developing
molecular-based property relationships and incorporating wellestablished correlations to estimate mixture properties, which
is an essential need of future process models; the capability of
predicting physical and performance properties of undefined
multicomponent hydrocarbon mixtures.
The clear advantage of the model is its ability to
complement the molecularly-explicit kinetic models of
petroleum refining processing making use of the vast
information available in the literature on pure components.
Another advantage of the model is its flexibility where the
number and type of model components in the molecular
ensemble may be tailored as needed by the user to
accommodate a specific need with the limitations being only
the computation power available and the intended use of the
simulation.
Furthermore, the developed computerized procedure can be
incorporated as software in the ASTM D86 distillation
apparatus hardware to provide estimation of more than 25
properties of petroleum fractions using one single laboratory
test. This leads to large savings in terms of energy, time and
cost since one ASTM distillation test can replace the test
equipment needed to predict all of these properties.
The model can also be used to probe existing correlations
for petroleum fractions based on global properties as shown for

Heat of combustion from MEPP model

(J/kg) .

T.A. Albahri / Fuel 85 (2006) 748754

753

46000

45000

44000
Gorenkov et al.
API method

43000
43000
44000
45000
46000
Heat of Combustion from generalized
correlations (J/kg)

Fig. 4. Bar plot comparison for the net heat of combustion using the methods of
API and Gorenkov et al.

the net heat of combustion in Fig. 4. The figure shows the bar
plot for using both the API method [1] and the method of
Gorenkov et al. [8]. It is evident that the latter method provides
better correlation with the molecular ensemble and can be
considered more accurate than the API correlation.
4. Conclusion
4.1. Summary
It was an objective of this work to develop a model for
predicting the properties of undefined multicomponent
mixtures and light petroleum fractions like naphtha and
gasoline from ASTM distillation data alone using a characterization method that is more suitable for incorporation in the
molecularly-explicit simulation models than the current
methods and to enhance the prediction performance of
chemical process simulation packages.
The developed technique combines routine analytical tests
and a molecularly-explicit modeling approach to provide
quantitative insight to the petroleum fractions structure which
permitted easy accounting of molecules and enabled the direct
estimation of the thermodynamic and transport properties
thereof. The method was used to predict the properties of 30
petroleum naphtha samples with good agreement with the
experimental data. The said method can calculate the properties of light petroleum fractions with good accuracy when at
least one bulk property is available (namely ASTM D86 or
TBP distillation temperatures). However, these properties can
be predicted more accurately, if in addition RVP and PNA are
available. Although the technique is shown to predict properties of petroleum naphtha as an example, the method in
principle is applicable to any petroleum fraction.
This work further demonstrates that the complex nature of
petroleum fuels may be modeled by a limited set of
representative molecules using non-linear-regression optimization models. Considering the difficulty and complexity of
accounting for the thousands of compounds in petroleum fuels
and the limitations of the conventional pseudo-component

754

T.A. Albahri / Fuel 85 (2006) 748754

technique, the proposed method can be an effective alternative.

It was found that through proper characterization of feed
composition the MEPP model can provide accurate predictions
of the various properties.
The developed methodology is a useful tool in representing
a broad range of different petroleum feedstocks provided
relatively simple set of experiments are performed to
characterize the attributes of the feed. This molecularlyexplicit simulation of feedstock structure and properties is
direct input for molecular reaction models which can
ultimately be used to map out the changing molecular
population with respect to various processing conditions.
4.2. Future outlook
Work is currently in progress to develop mixing rules with
interaction parameters to predict other properties of light
petroleum fractions including motor and research octane
numbers, true critical pressure, and liquid and vapor enthalpies.
Effort is also being made to improve the existing prediction
capacity for the critical compressibility factor, aniline point,
and isobaric liquid and vapor heat capacities and to predict
additional properties for heavier fractions such as cetane
number, cloud point, pour point, freezing point, and smoke
point. Another future aspect is to enhance the property
prediction capabilities of the MEPP model as a function of
temperature and pressure. One more aspect includes prediction
of the properties of heavier petroleum fractions such as

kerosene and diesel as a pretext for the future molecular-level

characterization and modeling of crude oil.
Acknowledgements
This work was supported by Kuwait University, Research
Grant No. EC04/01. It has been partially presented at the
American chemical society 228th national meeting [9].
References
[1] Technical API, API. Technical data book on petroleum refining. In:
Daubert, Danner, editors.. Washington, DC: API; 1989.
[2] Riazi MR, Daubert TE. Improved characterization of wide boiling range
undefined petroleum fractions. Ind Eng Chem Res 1987;26:62932.
[3] Riazi MR. A continuous model for C7C fraction characterization of
petroleum fluids. Ind Eng Chem Res 1997;36:4299307.
[4] Sasano Y. Measuring Research Octane Number of Gasoline by Gas
Chromatograph. Patent JP 09-318613; 1997.
[5] Albahri TA. Molecularly-Explicit Characterization Model for Light
Petroleum Fractions. Ind. Eng. Chem. Res. 2005; under review.
[6] Albahri TA. Simulation of light petroleum fractions. ACS Fuel Chem Div
Prepr 2004;49(1):3278.
[7] Healy WC, Maassen CW, Peterson RT. A new approach to blending
octanes. Am Pet Inst Div Refining 1959;39:13292.
[8] Gorenkov AF, Lifanova TA, Klyuiko IG, Kupreev AI. Determination of
heat of combustion of jet fuels by calculation. Chem Technol Fuels and
Oils 1984;20:2624.
[9] Albahri TA. Enhanced method for predicting the properties of petroleum
fractions. ACS Fuel Chem Div Prepr 2004;49(2):9256.