Journal of Petroleum Science and Engineering 57 (2007) 166 – 180

www.elsevier.com/locate/petrol

Classification and characterisation of crude

oils based on distillation properties
Peter Behrenbruch ⁎, Thivanka Dedigama
Australian School of Petroleum, University of Adelaide, Australia 5005
Accepted 31 October 2005

Abstract

True boiling point (TBP) distillation is a widely used batch distillation process for the characterisation of crude oils, traditionally
mainly for marketing and refining purposes. The TBP curve is obtained by plotting the cumulative mass or volume distillation
fraction with increasing temperature. The shape of these curves is dependent on the volatility of components in a given crude oil.
As such, these curves give a “footprint” of the composition of crude oils.
A new method of characterising crude oils based on the shape of TBP distillation curves is proposed. A gamma distribution is used to
characterise the TBP distillation curve, and the parameters of the fitted distribution are used as characterisation parameters. The proposed
method is found to describe experimental data very well with just two parameters, and as such offers a very practical approach in terms of
classifying crude oils. Ranges of values for the characterisation parameters for different types of crude oil are identified for a large set of
TBP data. The characterisation parameters can be correlated with a number of crude oil properties. They can also be used to predict the
petroleum fractions that can be produced from a given crude oil and hence correlated to the value of that crude.
As an alternative, it is shown how crude oil cut fractions may be classified with the aid of a ternary diagram, and the link
between this approach and the characterisation parameters introduced above is demonstrated.
© 2006 Elsevier B.V. All rights reserved.

Keywords: True boiling point distillation; Crude oil; Characterisation; Classification; Gamma distribution; Ternary diagram

1. Introduction to as crude oil, natural gas and solid hydrocarbons such

Petroleum has an organic origin based on the
accumulation of plant and animal matter and the action
of heat and pressure over a long period of time on this
biological material. It is a complex mixture of hydro-
carbons that naturally occurs in reservoir rocks.
Petroleum encompasses liquid hydrocarbons referred
⁎ Corresponding author. Tel.: +61 8 8303 8020; fax: +61 8 8303 8011.
E-mail address: peter.behrenbruch@adelaide.edu.au
(P. Behrenbruch).
0920-4105/$ - see front matter © 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.petrol.2005.10.016
as tars. The chemical constituents of petroleum and
techniques available to separate and analyse the various
compounds are discussed in Waples (1985). Major
groups of compounds found in petroleum are saturated
hydrocarbons, including straight chained, branched and
cyclic hydrocarb ons, simple aromatic hydrocarbons,
small sulphur bearing compounds, resins and very large
aromatic asphaltene compounds.
The varied composition of petroleum has led to the
existence of a number of characterisation systems. Each
of these characterisation systems is important for deci-
sion making at some stage of exploration, production
 P. Behrenbruch, T. Dedigama / Journal of Petroleum Science and Engineering 57 (2007) 166–180 167

Table 1
List of TBP distillation data (source: Chevron, 2005)
Country Location Crude oil Date Whole crude properties
of
Gravity Specific Characterization Sulfur Pour point
assay
(°API) gravity factor (K factor) (wt.%) (°C)
1 Angola Africa Kuito 2000 19.2 0.9390 11.6 0.68 − 32
2 Angola Africa Cabinda 1996 32.8 0.8610 12.3 0.13 16
3 Angola Africa Nemba 2001 39.3 0.8282 12.2 0.19 −4
4 Angola Africa Palanca 2001 38.4 0.8329 12.1 0.17 10
5 Chad Africa Doba Blend 2002 20.5 0.9308 12.0 0.16 −4
6 Congo Africa Kitina 2002 37.7 0.8365 12.2 0.09 7
7 Congo Africa N'kossa 2002 46.8 0.7936 12.4 0.03 − 21
8 Nigeria Africa Bonny Light 2000 35.8 0.8459 11.9 0.14 − 18
9 Nigeria Africa Escravos 2000 34.4 0.8528 11.7 0.15 7
10 Nigeria Africa Pennington 1995 35.0 0.8497 11.7 0.08 10
11 Australia Australia Barrow Island 1987 38.0 0.8346 11.7 0.05 − 54
12 Australia Australia Cossack 1998 48.1 0.7877 12.0 0.04 − 18
13 Australia Australia NWS condensate 1997 61.2 0.7344 12.3 0.01 − 51
14 Australia Australia Thevenard Island 2001 41.3 0.8189 NA 0.02 NA
15 Indonesia SE Asia Duri 1994 20.8 0.9293 12.0 0.20 10
16 Indonesia SE Asia Minas 1997 35.3 0.8482 12.6 0.09 38
17 PNG SE Asia Kutubu Light 1999 45.1 0.8014 12.2 0.04 −1
18 China South China Sea Nanhai Light 2000 40.1 0.8246 12.4 0.06 32
19 Thailand South China Sea Benchamas 2002 43.0 0.8108 12.7 0.04 32
20 Thailand South China Sea Tantawan 2000 43.3 0.8094 12.5 0.05 16
21 Canada Canada Hibernia 2000 35.9 0.8454 11.6 0.34 13
22 Chile Sth America Rincon 1999 36.0 0.8447 11.8 0.37 −4
23 Argentina Sth America Medanito 1994 35.2 0.8486 12.2 0.41 −1
24 Venezuela Sth America Hamaca 2003 26.0 0.8987 12.1 1.55 − 40

and refining of crude oil. These systems can be broadly Geologists and geochemists use geological para-
grouped as follows. meters to identify and characterise crude oil, for crude
oil — source rock correlation and to measure the degree
(a) Geochemical parameters of evolution (Tissot and Welte, 1984). A range of
(b) Whole crude (live oil) properties geochemical parameters is available for the character-
(c) Assay properties isation of crude oil. Live crude properties are of interest

Fig. 1. TBP distillation curves for a selection of samples listed in Table 1. Numbers correspond to Sample No in Table 1.
168 P. Behrenbruch, T. Dedigama / Journal of Petroleum Science and Engineering 57 (2007) 166–180

Table 2
Detailed TBP distillation data for Griff in crude oil (source: BHP Billiton, 2005)
Cumulative Temperature Temperature Mid Specific API Cumulative Cumulative Temperature Weight Weight
volume (°F) (°C) volume gravity gravity weight weight difference fraction fraction
percentage percentage percentage fraction yact ΔT (°C) difference Δyact density
4.0 66.2 19 2.0 0.5738 115.1 3.1 0.031 19 0.031 0.0016
11.0 140 60 7.5 0.6351 91.3 9.0 0.090 41 0.059 0.0014
14.2 158 70 12.6 0.6715 79.2 11.8 0.118 10 0.028 0.0028
16.5 176 80 15.4 0.6888 73.9 13.9 0.139 10 0.021 0.0021
20.0 194 90 18.3 0.7002 70.6 17.1 0.171 10 0.032 0.0032
26.3 212 100 23.2 0.7129 67.0 23.0 0.230 10 0.059 0.0059
30.3 230 110 28.3 0.7230 64.2 26.9 0.269 10 0.039 0.0039
36.0 248 120 33.2 0.7271 63.1 32.4 0.324 10 0.055 0.0055
40.9 266 130 38.5 0.7325 61.7 37.1 0.371 10 0.047 0.0047
44.2 284 140 42.6 0.7390 60.0 40.3 0.403 10 0.032 0.0032
49.3 302 150 46.8 0.7434 58.8 45.3 0.453 10 0.050 0.0050
52.8 320 160 51.1 0.7523 56.6 48.8 0.488 10 0.035 0.0035
54.6 329 165 53.7 0.7567 55.5 50.6 0.506 5 0.018 0.0036
56.8 338 170 55.7 0.7590 54.9 52.8 0.528 5 0.022 0.0044
60.2 356 180 58.5 0.7645 53.6 56.2 0.562 10 0.034 0.0034
63.6 374 190 61.9 0.7714 51.9 59.7 0.597 10 0.035 0.0035
67.3 392 200 65.4 0.7866 48.4 63.5 0.635 10 0.038 0.0038
70.9 410 210 69.1 0.7935 46.8 67.3 0.673 10 0.038 0.0038
74.5 428 220 72.7 0.7995 45.5 71.1 0.711 10 0.038 0.0038
75.7 446 230 75.1 0.8050 44.3 72.4 0.724 10 0.013 0.0013
76.9 464 240 76.3 0.8127 42.6 73.7 0.737 10 0.013 0.0013
79.6 482 250 78.3 0.8204 41.0 76.6 0.766 10 0.029 0.0029
81.7 500 260 80.6 0.8272 39.6 78.9 0.789 10 0.023 0.0023
83.2 518 270 82.4 0.8339 38.2 80.6 0.806 10 0.017 0.0017
84.8 536 280 84 0.8380 37.4 82.4 0.824 10 0.018 0.0018
86.6 554 290 85.7 0.8419 36.6 84.4 0.844 10 0.020 0.0020
88.5 572 300 87.5 0.8465 35.7 86.6 0.866 10 0.022 0.0022
89.5 590 310 89 0.8513 34.7 87.7 0.877 10 0.011 0.0011
90.4 608 320 89.9 0.8556 33.9 88.8 0.888 10 0.011 0.0011
91.4 626 330 90.9 0.8595 33.1 89.9 0.899 10 0.011 0.0011
92.6 644 340 92 0.8635 32.4 91.3 0.913 10 0.014 0.0014
93.5 662 350 93.1 0.8706 31.0 92.3 0.923 10 0.010 0.0010
94.4 680 360 94 0.8818 29.0 93.4 0.934 10 0.011 0.0011
95.3 734 390 94.9 0.8790 29.5 94.5 0.945 30 0.011 0.0004
96.2 752 400 95.8 0.8794 29.4 95.6 0.956 10 0.011 0.0011
96.8 770 410 96.5 0.8797 29.4 96.3 0.963 10 0.007 0.0007
97.7 806 430 97.3 0.8831 28.7 97.3 0.973 20 0.010 0.0005
98.4 842 450 98.1 0.8928 27.0 98.1 0.981 20 0.008 0.0004
99.0 896 480 98.7 0.9085 24.3 98.8 0.988 30 0.007 0.0002
99.3 932 500 99.2 0.9273 21.1 99.1 0.991 20 0.003 0.0002
99.6 986 530 99.5 0.9536 16.9 99.5 0.995 30 0.004 0.0001
100.0 986.0+ 530.0+ 99.8 0.9702 14.3 100.0 1.000 0.005
1.000

to reservoir engineers, who use these properties to
understand fluid flow in the reservoir and to predict
changes in reservoir conditions with time, during
production. Crude oil assay properties are of most in-
terest to petroleum refiners and crude oil marketers,
who need to know the quantities of distillate fractions
that a particular crude will produce during distillation.
They also need to know the physical and chemical
properties of the various fractions. Finally, the compo-
sition of crude oil is closely linked to its price.
TBP distillation is one of the most common experi-
mental procedures employed in ascertaining assay proper-
ties of a crude oil. Many researchers have investigated the
use of TBP distillation data as a basis for characterisation of
crude oils. Miquel et al. (1992) recognised that TBP distil-
lation data were the most commonly available information
regarding the volatile behaviour of hydrocarbon mixtures.
They proposed a characterisation system based on opti-
mally selecting pseudo-fractions from the TBP distillation
curve. Haynes and Matthews (1991) proposed a continuous
 P. Behrenbruch, T. Dedigama / Journal of Petroleum Science and Engineering 57 (2007) 166–180
Fig. 2. TBP distillation curves for Griffin crude oil.
approach to use TBP distillation data for heavy end char-
acterisation of live crude oils in vapour–liquid equilibrium
calculations. Specific cut fractions (part of the TBP range),
used in the generation of petroleum products directly influ-
ence pricing of crude oils. The other major use of these data
is in deciding refinery processes needed to refine a given
crude oil. TBP distillation can also be used as a method to
isolate a specified fraction from a crude oil for testing.
The method for performing a TBP distillation exper-
iment is described in ASTM D 2892 (American Society of
Testing and Materials, 1999a) and ASTM D 5236
(American Society of Testing and Materials, 1999b) and
the procedure can be used to analyse hydrocarbon mixtures,
including crude oils, condensates and petroleum fractions.
However, the method cannot be used to analyse liquefied
petroleum gasses, very light naphtha fractions and fractions
with initial boiling points (IBP) greater than 400 °C. The
TBP experiment is performed by distilling a sample of
crude oil or petroleum fraction in a standardised fraction-
ating column that is subject to specified operating
conditions. Distillation is carried out at atmospheric pres-
sure from the IBP to about 210 °C vapour temperature, and
169
at reduced pressure beyond this point. Distillation under a
partial vacuum avoids cracking of the more complex
components at elevated temperatures. Samples of distillate
are collected at specified temperature cut points. Mass and
density of each fraction are measured, and the distillation
yield by mass is calculated. Volumetric yield can be
estimated with mass and density data. Vapour temperatures
measured at reduced pressure are translated to Atmospheric
Equivalent Temperature (AET). Distillation can usually be
continued up to an AET of approximately 400 °C. Final
results of such experiments are TBP curves in mass and/or
volume yield versus boiling temperature expressed in AET.
The shape of a TPB curve is dependent on the type and
quantities of hydrocarbon compounds that make up the
mixture being analysed. As such, this curve uniquely des-
cribes a given crude oil in terms of its chemical makeup. As
an indicator of a crude oil's compositional makeup, a TBP
distillation curve has the added advantage that it is exten-
sively performed for marketing purposes on all crude oils.
These data are also easily accessed, as crude oil marketing
companies publish these data in the public domain. Some
TBP distillation data and other assay properties published
on the Internet by Chevron (2005) are listed in Table 1. The
list covers a range of crude oil types from gas condensates
to heavier oils, containing samples from across the world,
and constitutes one of two major data sets used in the
current study. Fig. 1 shows the TBP curve shapes for a
selection of crude oils from this set of data. Three marker
crude oils (Brent, West Texas Intermediate (WTI) and
Tapis) are also shown for comparison. The following
detailed assay reports for Australian crude oils published on
the BHP Billiton web site (BHP Billiton, 2005) have also
been used in the analysis.
• Griffin Crude Oil
• Gippsland Crude Oil
• Cossack Crude Oil
• Laminaria Crude Oil
• North West Shelf Condensate
Typical detailed TBP distillation data for Griffin crude
oil are shown in Table 2 as an example. The corresponding
TBP distillation (cumulative weight fraction distilled versus
temperature) curve is shown in Fig. 2(a). This curve can be
converted in to a weight fraction density curve, as shown in
Fig. 2(b). The weight fraction density can be calculated as
follows.
Weight fraction density of the ith fraction ¼ Dyact =DT
where,
Δyact = yact,I − yact,i−1
ΔT = TI − Ti−1
170 P. Behrenbruch, T. Dedigama / Journal of Petroleum Science and Engineering 57 (2007) 166–180
yact,I Cumulative weight fraction of the ith
fraction
Ti Boiling temperature of ith fraction
The scatter seen on the experimental differential
curve could be attributed to two factors. The first is due
to the lumping of compounds with close boiling points.
The other is due to fluctuations in conditions during the
experiment. The variation is highlighted in the weight
fraction density curve, as this is a differential form of the
cumulative weight fraction distribution, which is some-
what smoother by comparison.
A mathematical characterisation of TBP distillation
curves would be useful for a number of applications.
These include physical property predictions, heavy end
characterisation for EOS modelling and crude oil
valuation. In addition, the comparison of shapes of
different TBP curves would allow for the identification
of groupings of oils based on composition.
The following sections detail a method of mathe-
matically characterising a TBP distillation curve using
the two-parameter gamma distribution. The fitting
methodology used is described and results for a range
of crude oil types have been presented, demonstrating
the flexibility of the proposed methodology. A
correlation between the fitting parameters and API
Fig. 3. Probability density functions and cumulative probability distributions for the gamma distribution with varying values of α and β.
gravity is also shown. A new classification system
based on temperature cut fractions from TBP
distillation data, using a ternary diagram is introduced
as an alternative to the PNA (Paraffins, Naphthenes
and Aromatics) ternary diagram (Tissot and Welte,
1984).
2. TBP curve characterisation
2.1. Gamma distributions
The gamma distribution has been widely used as a
mathematical function to characterise hydrocarbon
mixtures. The main focus of this work has been to
describe the molecular weight distribution with the
objective of predicting phase behaviour of hydrocarbon
mixtures. Whitson (1983) proposed a continuous ap-
proach for modelling the molecular weight distribution
of a hydrocarbon mixture using the three-parameter
gamma distribution for Equation of State (EOS) model-
ling. Cotterman et al. (1985) also used the three-parame-
ter gamma distribution to describe the molecular weight
distribution of hydrocarbon mixtures for process design
applications.
The probability distribution function (PDF) for a
gamma distribution can be described using the
 P. Behrenbruch, T. Dedigama / Journal of Petroleum Science and Engineering 57 (2007) 166–180 171

Table 3
Gamma distribution fitting for Griffin crude oil TBP distillation data
Experimental Predicted
Temperature Cumulative weight Cumulative weight Weight fraction Cumulative weight Weight fraction (yact − (yact −
T (°C) percentage fraction yact density percentage ypred density ypred)2 yact)2
19 3.1 0.031 0.0016 0.005 0.0007 0.00070 0.39017
60 9.0 0.090 0.0014 0.081 0.0030 0.00008 0.31994
70 11.8 0.118 0.0028 0.114 0.0035 0.00002 0.28905
80 13.9 0.139 0.0021 0.150 0.0038 0.00012 0.26691
90 17.1 0.171 0.0032 0.189 0.0041 0.00034 0.23487
100 23.0 0.230 0.0059 0.231 0.0042 0.00000 0.18116
110 26.9 0.269 0.0039 0.274 0.0043 0.00002 0.14949
120 32.4 0.324 0.0055 0.317 0.0044 0.00004 0.10998
130 37.1 0.371 0.0047 0.361 0.0043 0.00010 0.08102
140 40.3 0.403 0.0032 0.404 0.0043 0.00000 0.06382
150 45.3 0.453 0.0050 0.446 0.0042 0.00004 0.04106
160 48.8 0.488 0.0035 0.487 0.0040 0.00000 0.02810
165 50.6 0.506 0.0036 0.507 0.0039 0.00000 0.02239
170 52.8 0.528 0.0044 0.527 0.0038 0.00000 0.01629
180 56.2 0.562 0.0034 0.564 0.0037 0.00000 0.00877
190 59.7 0.597 0.0035 0.600 0.0035 0.00001 0.00344
200 63.5 0.635 0.0038 0.633 0.0033 0.00000 0.00043
210 67.3 0.673 0.0038 0.665 0.0031 0.00006 0.00030
220 71.1 0.711 0.0038 0.695 0.0029 0.00027 0.00307
230 72.4 0.724 0.0013 0.722 0.0027 0.00000 0.00467
240 73.7 0.737 0.0013 0.748 0.0025 0.00011 0.00662
250 76.6 0.766 0.0029 0.771 0.0023 0.00003 0.01218
260 78.9 0.789 0.0023 0.793 0.0021 0.00002 0.01779
270 80.6 0.806 0.0017 0.813 0.0019 0.00005 0.02261
280 82.4 0.824 0.0018 0.831 0.0018 0.00005 0.02835
290 84.4 0.844 0.0020 0.848 0.0016 0.00002 0.03548
300 86.6 0.866 0.0022 0.863 0.0015 0.00001 0.04425
310 87.7 0.877 0.0011 0.877 0.0013 0.00000 0.04900
320 88.8 0.888 0.0011 0.890 0.0012 0.00000 0.05399
330 89.9 0.899 0.0011 0.901 0.0011 0.00001 0.05923
340 91.3 0.913 0.0014 0.912 0.0010 0.00000 0.06624
350 92.3 0.923 0.0010 0.921 0.0009 0.00000 0.07148
360 93.4 0.934 0.0011 0.930 0.0008 0.00002 0.07749
390 94.5 0.945 0.0004 0.950 0.0006 0.00003 0.08373
400 95.6 0.956 0.0011 0.956 0.0005 0.00000 0.09022
410 96.3 0.963 0.0007 0.961 0.0005 0.00001 0.09447
430 97.3 0.973 0.0005 0.969 0.0004 0.00002 0.10072
450 98.1 0.981 0.0004 0.976 0.0003 0.00003 0.10586
480 98.8 0.988 0.0002 0.983 0.0002 0.00002 0.11047
500 99.1 0.991 0.0002 0.987 0.0002 0.00002 0.11247
530 99.5 0.995 0.0001 0.991 0.0001 0.00002 0.11517
530+ 100.0 1.000
0.00228 3.57276

following two-parameter form (Hastings and Peacock, β Scale parameter (β N 0)

1975). Γ(α) The gamma function with parameter α, given
xa−1 ex=b by,
pð x Þ ¼ ð1Þ Z l
ba CðaÞ CðaÞ ¼ e−u ua−1 du ð2Þ
0
where,
The cumulative probability function is given by,
Z x
x Variable (0 ≤ x b ∞) PðX V xÞ ¼ pðxÞdx ð3Þ
α Shape parameter (α N 0) 0
172 P. Behrenbruch, T. Dedigama / Journal of Petroleum Science and Engineering 57 (2007) 166–180

Probability density functions and cumulative proba-

bility distributions for different values of parameters α
and β are shown in Fig. 3, indicating the diversity of
shapes that can be accommodated. From the definition of
the two-parameter gamma distribution, the mean, stan-
dard deviation and mode of the distribution are as follows.

Mean αβpffiffiffi
Standard deviation β a
Mode β(α − 1)

2.2. Methodology

The two parameters (α and β) of the gamma dis-

tribution described previously are varied to obtain the best
fit with experimental data. The data in Table 2 for the
detailed assay of Griffin crude oil are used as an example.
The predefined function GAMMADIST in MS Excel is
used to generate the cumulative weight fraction data and
weight fraction density data. The goodness of fit is
measured using the coefficient of determination (R2) and
the Root Mean Square Error (RMSE). The Solver plug-in

Fig. 5. Gamma distribution fitted TBP distillation curves for Cossack

crude oil.

from MS Excel is used to tune α and β to the optimum

values while minimising RMSE.
The application of the methodology described above
for the Griffin crude oil (Table 2 and Fig. 2(a) and (b)) is
shown in Table 3. The fitted values of α and β are 2.90 and
63.48 respectively, and the cumulative weight fraction
and weight fraction density curve are shown in Fig. 4(a)
and (b), respectively. The RMSE and R2 are calculated
using Eqs. (4) and (5), respectively.
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
uX n
u
u ðyact −ypred Þ2
t
RMSE ¼ i¼1
ð4Þ
n
X
n
ðyact −ypred Þ2
i¼1
R2 ¼ 1− P
n ð5Þ
ðyact −P
yact Þ2
i¼1

where,

Fig. 4. Gamma distribution fitted TBP distillation curves for Griffin crude yact Actual (experimental) value
oil. ypred Predicted value from model
 P. Behrenbruch, T. Dedigama / Journal of Petroleum Science and Engineering 57 (2007) 166–180 173

ȳ act Mean of actual values

n Number of points

For the above example,

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
0:00228
RMSE ¼ ¼ 0:0075
41

0:00228
R2 ¼ 1− ¼ 0:9994
3:57276

These values, together with a visual comparison of the

fitted curve to the actual data, indicate that a good fit has
been achieved. RMSE was selected as the parameter for
optimisation (i.e., minimisation of error) using the Solver
as it showed a better convergence when compared to the
results obtained by using R2 (i.e., aiming for an R2 value
of 1.0). This fitting is independent of the initial values
used for α and β. The fitted curve shows a good visual
agreement with experimental data points and R2 values of
0.99 or better in all cases.

Fig. 7. Gamma distribution fitted TBP distillation curves for Laminaria

Crude Oil.

2.3. Results

The same methodology described above has been ap-

Fig. 6. Gamma distribution fitted TBP distillation curves for Gippsland
crude oil.
plied to other detailed samples. Fitted distributions
obtained for four other detailed TBP distillation curves
are shown in Figs. 5 6 7 and 8. As can be seen from these
graphs, the methodology described has the ability to fit
experimental TPB curves very well for a range of oil types.
The partial TBP distillation curves listed in Table 1
consist of weight fraction data collected at eight tem-
perature cut points ranging from 80 °C to 570 °C. Other
basic assay properties such as API gravity, sulphur content
and UOP K factor are also available. The UOP K factor (or
Watson characterisation factor) is defined as follows
(Nelson, 1958)
p
3
ffiffiffiffiffi
TB
K¼ ð6Þ
S
where,
TB Average molal boiling temperature in degrees
Rankine
S Specific gravity at 60 °F
174 P. Behrenbruch, T. Dedigama / Journal of Petroleum Science and Engineering 57 (2007) 166–180
The value of the K factor allows the characterisa-
tion of the base of a crude oil as paraffin base,
naphthene base or intermediate base. The ranges of K
factor values corresponding to each base are listed in
Table 4. The same methodology described above for
detailed assay results was used to fit a gamma
distribution to each of these curves. The results are
shown in Table 5. As with the detailed assay results,
good fitting can be observed in all cases. These
results also demonstrate the flexibility of the new
methodology. As mentioned previously and shown in
Fig. 1, the 24 samples in Table 1 cover a wide range
of crude oil types. Represented are oils with API
gravity ranging from 19.2 (for Kuito crude oil, a
heavy oil from Angola) to 61.2 (for North West Shelf
Condensate, a gas condensate). In addition, the list
contains crude oils with a wide distribution of total
sulphur content (0.0 to 1.55 wt.%), UOP K factor
(11.6 to 12.7) and pour point (− 54 °C to 38 °C). For
this diverse group of oils, the methodology is able to
fit gamma distributions with the coefficient of
Fig. 8. Gamma distribution fitted TBP distillation curves for North
West Shelf Condensate.
Table 4
UOP K factor values for paraffin, naphthene and intermediate base
crude oils (from Nelson, 1958)
UOP K factor Crude oil base
12.9–12.2 Paraffinic base
12.2–11.5 Intermediate base
11.5–10.5 Naphthene base
determination (R2 ) being better than 0.99 in each
case. A visual comparison of the fitted and actual
curves confirms the excellent fit.
An attempt was made to correlate the gamma dis-
tribution characterisation parameters (α and β) with
API gravity. Various functional forms were tested
against the available data. The form that gave the
best approximation to the experimental data was as
follows.
APIGravity ¼ aðbða−1Þ þ bÞc ð7Þ
where,
α and β Gamma distribution characterisation
parameters
a, b and c Constants
The functional form of Eq. (7) was selected as the
term β(α − 1) corresponds to the modal temperature of
the fitted distribution. As such the expectation is that
there will be a relationship between the modal tem-
perature and API gravity. The constants a, b and c
were tuned to get the best possible fit to the
experimental data by using the Solver plug-in from
MS Excel. The results for prediction of API gravity
using Eq. (7) are compared to experimental values for
all 24 samples in Fig. 9. Values for constants a, b and
c, and the coefficient of determination (R2 ) and RMSE
for these data are shown on the graph. Although there
is reasonable agreement between experimental and
predicted values, the degree of scatter on the graph is
high, with indication of a secondary trend. This aspect
was further investigated in order to determine the
reasons for this scatter, in terms of other available
crude oil properties. The scatter observed on Fig. 9
appeared to be related to the UOP K factor and
outlying points have been annotated on that figure.
As evident from Fig. 9, samples with a UOP K
factor greater than 12.2 cluster as a separate trend,
distinct from the main trend. This group corresponds
to crude oils having a paraffin base. Therefore, an
 P. Behrenbruch, T. Dedigama / Journal of Petroleum Science and Engineering 57 (2007) 166–180 175

Table 5
Gamma distribution fitting results for the samples listed in Table 1
Crude oil Whole crude properties Gamma distribution
fitting parameters
Gravity (°API) Characterization factor (K factor) Sulfur (wt.%) Pour point (°C) Alpha Beta R2
1 Kuito 19.2 11.6 0.68 − 32 4.21 109.9 1.000
2 Cabinda 32.8 12.3 0.13 16 2.42 182.2 0.996
3 Nemba 39.3 12.2 0.19 −4 2.22 147.2 0.997
4 Palanca 38.4 12.1 0.17 10 2.54 136.6 0.996
5 Doba Blend 20.5 12.0 0.16 −4 6.09 88.7 0.999
6 Kitina 37.7 12.2 0.09 7 2.18 170.1 0.997
7 N'kossa 46.8 12.4 0.03 − 21 1.95 139.3 0.997
8 Bonny Light 35.8 11.9 0.14 − 18 3.30 93.4 0.994
9 Escravos 34.4 11.7 0.15 7 3.20 100.5 0.996
10 Pennington 35.0 11.7 0.08 10 4.56 66.0 0.998
11 Barrow Island 38.0 11.7 0.05 − 54 3.67 69.0 0.999
12 Cossack 48.1 12.0 0.04 − 18 2.23 103.7 0.998
13 NWS condensate 61.2 12.3 0.01 − 51 2.42 54.7 1.000
14 Thevenard Island 41.3 NA 0.02 NA 4.87 50.6 0.998
15 Duri 20.8 12.0 0.20 10 4.07 142.0 0.998
16 Minas 35.3 12.6 0.09 38 3.32 131.6 0.998
17 Kutubu Light 45.1 12.2 0.04 −1 2.15 112.9 0.995
18 Nanhai Light 40.1 12.4 0.06 32 3.54 97.9 0.994
19 Benchamas 43.0 12.7 0.04 32 2.43 144.2 0.994
20 Tantawan 43.3 12.5 0.05 16 2.81 111.7 0.999
21 Hibernia 35.9 11.6 0.34 13 2.37 154.4 0.998
22 Rincon 36.0 11.8 0.37 −4 2.77 128.0 1.000
23 Medanito 35.2 12.2 0.41 −1 2.49 146.0 0.999
24 Hamaca 26.0 12.1 1.55 − 40 3.04 139.2 0.996

attempt was made to predict the API gravity of crude constants a, b and c to match the experimental data.
oils with a paraffin base (K factor greater than 12.2) Results from this API gravity prediction for paraffin
using the functional form of Eq. (7) and tuning the base crude oils are compared to experimental values

Fig. 9. Predicted versus experimental values for API gravity using Eq.
(7) for the 24 samples listed in Table 1. The UOP K factor for the Fig. 10. Predicted versus experimental values for API gravity using Eq.
outlying samples has been annotated. (7) for paraffin base crude oils (UOP K factor N 12.2).
176 P. Behrenbruch, T. Dedigama / Journal of Petroleum Science and Engineering 57 (2007) 166–180
Fig. 11. Predicted versus experimental values for API gravity using Eq.
(7) for non-paraffin base crude oils (UOP K factor b 12.2). The UOP K
factor for the outlying samples has been annotated.
in Fig. 10. It can be seen that there is very good
agreement between the predicted and experimental
values. An attempt was then made to predict API
Fig. 12. Paraffins, naphthenes and aromatics (PNA) composition classification ternary diagram (from Tissot and Welte, 1984, p. 440).
gravity for the non-paraffin base crude oils (K factor
less than 12.2) using the same functional form and
tuning a, b and c to get the best match with expe-
rimental data. Results are shown in Fig. 11, indicating
that, with the exception of three scattered point, there
is also very good agreement between the predicted
and experimental API gravities.
3. Cut fraction characterisation
3.1. Methodology
One possible framework for classification of crude
oils, based on assay properties is given in Tissot and
Welte (1984). They propose a classification based on
relative quantities of paraffins, naphthenes and
aromatics (PNA) contained in a particular crude oil,
displayed on a ternary diagram. They have applied
this technique to a large number of samples and
show how this ternary diagram can be used for oil–
oil and oil–source correlations. An example of their
classification technique is shown in Fig. 12. One
draw back of this PNA framework is that data for
plotting the ternary diagram are not readily available.
 P. Behrenbruch, T. Dedigama / Journal of Petroleum Science and Engineering 57 (2007) 166–180 177

Fig. 13. Cut fraction temperature definitions used by three companies based on data published on their respective web pages.

An assay report would normally only contain the resulting ternary plot for various types of crude oils
PNA break down for the lighter fraction (usually available, ranging from gas condensates to heavy
boiling at less than 200 °C) of the crude oil. As such, oils. Table 6 shows the three cut fractions that were
it is necessary to perform a separate analysis to selected for the purpose of drawing the new ternary
populate this type of ternary plot. The latter is plot.
perhaps one reason that this terminology appears to Weight fractions are read by interpolating cumu-
be used relatively infrequent. lative weight fractions distilled at 200 °C and 350 °C
For the abovementioned reasons, a new ternary from the distillation curve. The three weight
plot based on assay properties is proposed, utilising
three boiling temperature cut fractions from TBP
distillation. It should be noted that boiling temper-
ature ranges for petroleum fractions vary regionally
and also among companies. Definitions used by
three companies are shown in Fig. 13, based on data
published on the respective company web sites. This
figure indicates that, although companies use slightly
different definitions for individual cut fractions,
there is some agreement between them. Three cut
fractions were selected to correspond to distinct
refinery products. The cuts fractions were also
picked to have a good distribution of points on the

Table 6
Boiling temperature cut fractions selected for plotting the proposed cut
fraction ternary diagram
Cut Fraction No. Temperature range Main distillation products
1 IBP to 200 °C Naphtha
2 200 °C to 350 °C Kerosene and Gas Oil
Fig. 14. Composition of temperature cut fractions (modified after
3 350 °C+ Atmospheric Residue
Hunt, 1996, p. 45).
178 P. Behrenbruch, T. Dedigama / Journal of Petroleum Science and Engineering 57 (2007) 166–180

Fig. 15. Cut fraction ternary diagram showing the 24 samples listed in Table 1.

fractions, adding up to 1.0, are then calculated as Fig. 14 shows the link between cut fractions
follows. selected above and chemical composition for a naphthenic
crude oil (Hunt, 1996). It can be seen that there is some
IBP to 200 °C fraction W200 degree of correspondence between the selected fractions
200 °C to 350 °C fraction W350 − W200 and composition elements. The IBP to 200 °C fraction is
350 °C + fraction 1 − W350 richer in normal and iso-paraffin, while the 350 °C +
fraction is richer in heavy aromatics and nitrogen, sulphur
where, and oxygen (NSO) compounds.

W200 Cumulative weight fraction distilled at 3.2. Results

200 °C in
W350 Cumulative weight fraction distilled at The proposed cut fraction ternary diagram, with the
350 °C samples from Table 1 plotted, is shown in Fig. 15. It can

Fig. 16. Relationship between gamma distribution characterisation parameters and cut fraction ternary diagram.
 P. Behrenbruch, T. Dedigama / Journal of Petroleum Science and Engineering 57 (2007) 166–180 179

Fig. 17. Constant API gravity lines for paraffin base crude oils (UOP K factor N 12.2).

be seen from the plot that there is a good distribution of
points on this ternary diagram, considering the varied
fluid types plotted.
The location of a crude oil on the proposed ternary
diagram is governed by the relative quantities of each
fraction produced, being a function of the shape of the
TBP distillation curve. As such, there is a unique rela-
tionship between the two-parameter gamma distribution
characterisation system described previously and this
ternary diagram. Fig. 16 shows constant α and constant
β lines superimposed on the ternary plot. It can be seen
that a given pair of α and β lines have only one
intersection. A given TPB curve shape (as defined by its
α and β values) will be uniquely located on the ternary
plot.
It has also been possible to identify the distribution
of API gravity on the new cut fraction ternary
diagram. As described previously, a good correlation
has been obtained between the gamma distribution
characterisation parameters (α and β) and API gravity
for paraffin and non-paraffin base crude oils. This
relationship, based on the functional form of Eq. (6),
was used to calculate constant API gravity lines on the
ternary diagram.

Fig. 18. Constant API gravity lines for non-paraffin base crude oils (UOP K factor b 12.2).
180 P. Behrenbruch, T. Dedigama / Journal of Petroleum Science and Engineering 57 (2007) 166–180
Eq. (6) can be rearranged as follows.
n o
b ¼ ðG=aÞ1=c −b =ða−1Þ
where,
α and β gamma distribution characterisation
parameters
G API gravity
a, b and c Constants
This formulation allows the β value corresponding to
a given α value to be calculated for a specified API
gravity. As shown previously, a given set of α and β
values results is a unique point on the ternary diagram.
As such, a constant API gravity locus on the ternary
diagram can be established by varying α and calculating
β, using Eq. (8). Fig. 17 shows constant API gravity
lines for the paraffin base crude oils from Table 1 (as
defined by having a UOP K factor N 12.2). It can be seen
that the data points match well with the predicted
constant API gravity curves. Constant API gravity lines
for non-paraffin base crude oils (as defined by having a
UOP K factor b 12.2) are shown in Fig. 18, indicating
good agreement between the data points and predicted
curves. A comparison of the two sets of constant API
gravity curves is shown in Figs. 17 and 18, indicating
that the two sets are significantly different. It can be
concluded that the distribution of crude oils according to
API gravity on the cut fraction ternary diagram is
dependent on their base.
4. Summary and conclusions
A system for the characterisation of crude oils,
based on TBP distillation curve data using a two-
parameter form of the gamma distribution from sta-
tistics, has been described. The new methodology has
been tested against a broad selection of crude oil
types and is able to fit the cumulative TBP distillation
curve with a high degree of correlation in all cases.
The gamma distribution characterisation parameters
(α and β) have been related to API gravity using a
single functional form, having different constants
depending on the base of the crude oil as defined by
the UOP K factor.
A cut fraction ternary diagram for crude oils has also
been introduced as an alternative to the traditional PNA
diagram. It is possible to display a wide range of crude oil
types on the cut fraction ternary diagram. This ternary
diagram has the advantage that the data required to
ð8Þ populate it is easily available from crude oil assay reports.
The distribution of crude oils according to their API gravity
on the cut fraction ternary diagram has also been described.
It has been shown that this distribution is a function of the
base of a crude oil as defined by the UOP K factor. The
relationship developed suggests that API gravity is a
function of the modal boiling temperature of a crude oil.
Acknowledgments
The authors wish to gratefully acknowledge the
support of the sponsors: BHP Billiton, Chevron, Santos
and Woodside Energy, for their financial support, field
data and permission to publish the results presented.
References
American Society of Testing and Materials, 1999a. ASTM D 2892:
Distillation of Crude Petroleum (15-Theoretical Plate Column).
Annual Handbook of ASTM Standards, vol. 05.01.
American Society of Testing and Materials, 1999b. ASTM D 5236:
Distillation of Heavy Hydrocarbon Mixtures (Vacuum Potstill
Method). Annual Handbook of ASTM Standards, vol. 05.01.
BHP Billiton, 2005. BHP Billiton (online). Available from: http://
globaloil.bhpbilliton.com/ (accessed on 4 April 2005).
Chevron, 2005. Chevron (online)Available from: http://www.chevron.
com/crudemarketing/ (accessed on 4 April 2005).
Cotterman, R.L., Bender, R., Prausnitz, J.M., 1985. Phase equilibria
for mixtures containing very many components—development
and application of continuous thermodynamics for chemical
process design. Ind. Eng. Chem. Process Des. Dev. 24, 194–203.
Hastings, N.A.J., Peacock, J.B., 1975. Statistical Distribution. Butter-
worth and Co (Publishers) Ltd., London.
Haynes, H.W., Matthews, M.A., 1991. Continuous-mixture vapour–
liquid equilibria computations based on true boiling point
distillation. Ind. Eng. Chem. Res. 30, 1911–1915.
Hunt, J.M., 1996. Petroleum Geochemistry and Geology, 2nd edition.
W.H. Freeman and Company, USA.
Miquel, J., Hernandez, J., Castells, F., 1992. A new method for
petroleum fractions and crude oil characterization. SPE Reserv.
Eng. 265–270 (May).
Nelson, W.L., 1958. Petroleum Refinery Engineering, 4th edition.
McGraw-Hill Book Company Inc, New York.
Tissot, B.P., Welte, D.H., 1984. Petroleum Formation and Occurrence,
2nd edition. Springer-Verlag, Berlin Germany.
Waples, D.W., 1985. Geochemistry in Petroleum Exploration.
D. Reidel Publishing Company, Dordrecht, Holland.
Whitson, C.H., 1983. Characterizing hydrocarbon plus fraction. Soc.
Pet. Eng. J. 683–694 (August).