Fuel 78 (1999) 1319–1326

www.elsevier.com/locate/fuel

Predicting the viscosity of biodiesel fuels from their fatty acid ester
composition
C.A.W. Allen a, K.C. Watts a,*, R.G. Ackman b, M.J. Pegg c
a
Department of Biological Engineering, DalTech, Dalhousie University, P.O. Box 1000, Halifax, Nova Scotia, Canada, B3J 2X4
b
Department of Food Science, DalTech, Dalhousie University, P.O. Box 1000, Halifax, Nova Scotia, Canada, B3J 2X4
c
Department of Chemical Engineering, DalTech, Dalhousie University, P.O. Box 1000, Halifax, Nova Scotia, Canada, B3J 2X4
Received 25 June 1998; received in revised form 17 December 1998

Abstract
Viscosity is one of the most significant properties to affect the utilization of biodiesel fuels. This paper presents a method, which has been
verified experimentally, for predicting the viscosities of biodiesel fuels from the knowledge of their fatty acid composition. The applicability
of a logarithmic mixture equation was verified using controlled mixtures of standard fatty acid esters and natural biodiesels. Several binary,
ternary and quaternary mixtures of fatty acid ethyl ester (FAEE) gas chromatography (GC) standards were formulated. Their viscosities were
predicted from their component values and were within ^ 3.7% of their measured values. The fatty acid compositions of six typical oils were
simulated by mixing fatty acid methyl ester (FAME) standards in appropriate amounts. Viscosities of these mixtures were also predicted
within ^ 2.1% of their measured values. Five biodiesel types were produced from natural oils and the logarithmic equation was applied to
predict their viscosities. An average prediction error of ^ 3% was obtained for these samples. The viscosities of fifteen biodiesel types were
then predicted based on their fatty acid composition as published in the literature and were found to vary as much as 100% This is most likely
a principal contributing factor to the variation in performance of some biodiesel fuel types. The viscosity of biodiesel fuels reduce
considerably with increase in unsaturation. Contamination with small amounts of glycerides significantly affects the viscosity of biodiesel
fuels. q 1999 Elsevier Science Ltd. All rights reserved.
Keywords: Biodiesel; Properties; Viscosity

1. Introduction in a diesel engine, is significantly affected by the fuel’s

viscosity.
Biodiesel fuels are generally classified as fatty acid There exist in the literature several models for predicting
methyl esters (FAMEs) which are derived from the alkali- the viscosity of fatty acids. Reid et al. [2] presented several
catalysed transesterification of fats and oils with methanol, group contribution methods for computing the viscosity of
although other alcohols can be used. The applicability of individual components but most of these models gave wide
this fuel type as a replacement or subsidy for petroleum ranging errors and thus were considered unreliable. Fisher
based diesel fuels has been investigated by many authors. [3] described a method based on limiting properties to
Graboski and McCormick [1] provided a recent extensive predict several properties of n-fatty acids whereby their
review of the utilisation of biodiesel fuels in compression variation was attributed to the increments in the methylene
ignition engines. The general conclusion from the literature group of the homologous series. This method worked fairly
is that, in terms of power, wear, efficiency and emissions, well for properties like heat capacity, heat of vaporization,
biodiesel fuels are a viable alternative. There are, however, boiling point, molal volume and density, but unfortunately
some variations in the results reported by some authors. One the limiting property for viscosity was approximately zero,
of the contributing factors to the variation in the reported rendering this method inaccurate for viscosity predictions.
results can be the viscosity of the fuel. This is because the Noureddini et al [4], Swern [5] and Markley [6] presented
atomization process, which is the initial stage of combustion empirical models and data for predicting the viscosity of
individual fatty acids and vegetable oils. The above methods
* Corresponding author. Tel: 1 1-902-494-3269; fax: 1 1-902-423-
were applicable to individual fatty acid components but not
2423. to acyl esters or mixtures. This paper presents a method for
E-mail address: [email protected] (K.C. Watts) predicting the viscosity of mixtures of fatty acid esters of
0016-2361/99/$ - see front matter q 1999 Elsevier Science Ltd. All rights reserved.
PII: S0016-236 1(99)00059-9
1320 C.A.W. Allen et al. / Fuel 78 (1999) 1319–1326

which biodiesel is comprised. The importance of this work basis, varying proportions of 8:0, 10:0, 12:0 and 18:1
is that if it is possible to predict the viscosity of any given FAMEs.
vegetable oil source from its components (and hence its • MEs were mixed on a mass basis in appropriate propor-
atomization), there may be no need for expensive testing tions to simulate canola, coconut, palm, peanut, rapeseed
programmes to determine the functioning of the oil in an and soybean oil MEs.
engine. • Biodiesel fuels from natural canola, coconut, palm,
peanut and soya oils were produced in a batch transester-
ification unit described by Allen and Watts [10]. These
2. Mixture model ME fuels were also heated to 1008C under nitrogen and
then filtered through a “Reeve Angel” grade 202 coarse
The Grunberg–Nissan [7] equation has been reported by paper filter to remove any sediments in the oil that were
Monnery et al. [8] and Irving [9] to be the most suitable carried over from the oil extraction process. At 1008C, all
equation for computing the viscosity of liquid mixtures. of the oil’s fatty acids were fluid with low viscosity,
This equation was developed primarily for binary mixtures making it easier for them to pass through the coarse filter.
and works best with non-associated liquids. However, good
results were obtained by Irving [9] for some associated The purity of the biodiesel fuels produced by the transes-
fluids with errors less than 5–10%. This equation has also terification process was verified using a thin-layer chroma-
been extended beyond the binary mixture level with some tography (TLC) on Chromarods-SIII (silica gel) with flame
success ([9]). The Grunberg–Nissan equation is ionization of the components (TLC-FID) by scanning in an
Iatroscan Mark III ([11]). The fatty acid composition of the
X
n XnX biodiesel fuels was measured on a Perkin–Elmer Gas Chro-
ln mm ˆ xi ln mi 1 xi xj Gij …1†
i i±j matograph, model 8420 as described by Ackman [12].

where mm is the mean viscosity of mixture (Pa s), m the 3.2. Viscosity measurement
viscosity of pure ith component (Pa s), xi and xj the mole
fractions of the ith and jth components, Gij the interaction A Paar model AMV 200 micro viscometer operating on a
parameter (Pa s) n the number of components. rolling-ball principle was used to measure the viscosity of
Biodiesel fuels are non-associated liquids that are all samples at the test temperature. Only one replicate for
comprised of mixtures of fatty acid esters whose chemical each sample was required as the instrument provided consis-
structures are similar (aliphatic chains). Due to this similar- tently repeatable results.
ity, the components in a mixture should not interact with
each other and thus should behave in a similar manner as an 3.3. Error analysis
individual component. It was therefore assumed that the
interaction parameter in Eq. (1) would be small and thus The maximum allowable error for the predicted results
could be neglected. Also, the mass fraction was used in was derived through a sensitivity analysis of a generalized
preference to the mole fraction in Eq. (1) to conform with atomization model taken from the literature. This model,
the mass unit that is implicit in the units for viscosity used in which was developed by Msipa et al. [13], computes an
this study. With these two modifications, Eq. (2) was used to atomization characteristic Ka which gives an overall view
predict the viscosity of biodiesel fuels based on their fatty of the atomization quality. It includes the surface tension,
acid composition. viscosity and density of the fuel. Eqs. (3)–(5) summarise
X
n this model and further details can be found in Ref. [13].
ln mm ˆ yi ln mi …2† " #1=3
iˆ1 rf Wef
Ka ˆ …3†
rg Ref
where yi is the mass fraction.
where Ka is the atomization characteristic, Wef the Weber
number for fuel, Ref the Reynolds number for fuel, r the
3. Materials and methods
densities for fuel [r f] and gas atmosphere [r g] (kg/m 3).
3.1. Sample preparation The Reynolds and Weber numbers are given by
V 0 d0
Both FAME and ethyl ester (EE) standards were acquired Ref ˆ …4†
vf
from Sigma Aldrich Chemical Supplies company. EEs were
used for binary, ternary and quaternary mixtures since it was
more economical to use these standards. The following rf V02 d0
Wef ˆ …5†
samples were prepared for different tests, as described later: sf
• Twenty EE samples were produced by mixing, on a mass where s f is the surface tension of the fuel (N/m), vf the
 C.A.W. Allen et al. / Fuel 78 (1999) 1319–1326 1321

Table 1 was converted to an effective shear rate per unit rolling

Mean value of properties used for sensitivity analysis distance and is given in Eq. (8).
Property Mean value 4d
g_ L ˆ …8†
Surface tension of fuel 0.0292 N/m …D2 2 d2 †t
Viscosity of fuel 4.18 mPa s where g_ L is the equivalent shear rate per unit length (1/
Density of fuel 871.4 kg/m 3
Nozzle diameter 0.4 mm
(m s)) and t the rolling time of ball (s).

4. Results and discussion

kinematic viscosity of the fuel (m 2/s) and V0 the velocity of
the fuel jet (m/s).
4.1. Allowable error
Using the conventional 95% confidence, it was decided to
vary the viscosity to a maximum of ^ 5% from its mean Fig. 1 shows a graphical representation of the results of
value to determine if there was any significant change in the the sensitivity analysis. A ^ 5% variation in the viscosity
atomization characteristic. The mean value of all properties, resulted in a 3% variation in Ka from its maximum value.
taken from Allen et al. [14], are given in Table 1. This 3% variation in Ka over the entire ^ 5% (10% varia-
tion from maximum viscosity) range of viscosities used can
3.4. Rheological properties
be considered negligible. Therefore, the maximum allow-
To ensure a correct comparison of the viscosities of able error for a prediction equation to estimate the viscos-
different biodiesel types, a check was made to determine ities of the ester mixtures was set at 5%.
whether these fuels were significantly non-Newtonian at
4.2. Rheological properties of experimentally produced
the test temperature of 408C. All of the biodiesel ME fuels
biodiesel fuels
produced were subjected to controlled, variable shear stress
viscosity measurements on the rolling-ball viscometer and For every biodiesel fuel type that was produced, a plot of
the effective shear rate was determined as a function of the the equivalent shear stress versus the equivalent shear rate
rolling time. The effective shear rate and the effective shear per unit length was made (Fig. 2) and linear regression lines
stress for a rolling-ball viscometer [15] are given in Eqs. (6) were then fitted to these points. For each fuel type, all the
and (7). data points fell on their regression line which passed
Mean Velocity of Ball 4dV through a point very close to the origin (Fig. 2). This very
g_ E ˆ ˆ …6† small shear stress (yield stress) at zero flow indicates a slight
Mean Gap …area based† …D 2 d2 †
2
“Bingham-plastic” behaviour which is more pronounced for
tE ˆ d…Dr†gsin…u† …7† peanut, palm and coconut oils. However, this effect is small
and these fuels can, therefore, be considered to behave in a
where g_ E is the equivalent shear rate (1/s), t E the equivalent Newtonian manner at 408C which means that the higher
shear stress (Pa), D the capillary diameter (m), d the ball saturates present in some of the oils did not significantly
diameter (m), V the velocity of the rolling ball (m/s), Dr the affect the nature of the fuel’s viscosity. Also, during the
difference in density between ball and fluid (kg/m 3), u the utilization of these vegetable oil ME fuels in engines, the
angle of inclination of capillary (8) and g the acceleration shear stresses encountered are much higher than the yield
due to gravity (9.81 m/s 2). stresses observed. Therefore, these fuels may be considered
The viscometer measured the rolling time of the ball over to behave in a Newtonian manner in actual engine opera-
a fixed distance. This distance was not known for the instru- tions.
ment used in these tests, therefore, the effective shear rate
4.3. Viscosity of individual ester components

The measured viscosities of individual FAMEs and EEs

had a complex pattern with respect to the saturation of the
fatty acids and their chain length. For saturated fatty acid
esters, the measured data at 408C indicated that the viscosity
increased with carbon number (CN) in a curvilinear trend
rather than a linear one. Good correlation was found
between the measured data and a second-order polynomial
function over the CN range 8:0–18:0. Eqs. (9) and (10) give
the fitted polynomials for saturated MEs and EEs.
Fig. 1. Variation of normalized atomization characteristic (Kai/Kamax) with
5% variation of viscosity for constant surface tension and density: (B) Ka mME2sat ˆ 1:05E 2 4M 2 2 0:0242M 1 2:15 s ˆ 0:0145
and (K) % variation. …9†
1322 C.A.W. Allen et al. / Fuel 78 (1999) 1319–1326

Fig. 3. Viscosity trend lines for methyl ester and ethyl ester GC standards at
408C: (A) EEs and (W) MEs.
Fig. 2. Rheograms for vegetable oil methyl ester biodiesel fuels at 408C
plotted on a linear scale: (W ) canola; (X) coconut; (K) diesel; ( 1 ) palm; 18:0 became unsaturated to 18:1. A completely different
(A) peanut; (B) soya. curve was observed as the degree of unsaturation progressed
from 18:1 to 18:3 (Fig. 3). As the number of double bonds
mEE2sat ˆ 1:16E 2 4M 2 2 0:0264M 1 2:28 s ˆ 0:0182 increased, there was a non-linear decrease in viscosity, with
…10† a 21% difference between 18:0 and 18:1 (based on 18:0), an
18% difference between 18:1 and 18:2 (based on 18:1), and
where m is the viscosity (mPa s), M the molecular weight (g/ a 13% difference between 18:2 and 18:3 (based on 18:2).
mol) and s the standard error. Although this phenomenon was not verified at other CN
Fig. 3 shows the data points and their trend curves. The due to the cost and availability of these individual
viscosities of the saturated EEs were only slightly higher components, there is no a priori reason why the same result
than those for MEs, an average of 5.4% higher over the CN would not occur at other CN since their chemical structures
range 8:0–18:0. are similar.
The viscosity trend for unsaturated esters at 408C showed The trend for unsaturated C18 esters also correlated well
a sharp deviation from the trend of the saturated esters when with second-order polynomial functions given in Eqs. (11)

Table 2
Viscosities of binary, ternary and quaternary mixtures of fatty acid ethyl ester GC standards at 258C

Sample # Fatty acid ethyl ester Measured viscosity (mPa s) Predicted viscosity (mPa s) Per cent error

Mass fraction

18:1 12:0 10:0 8:0

1 1.00 – – – 5.50 – –
2 – 1.00 – – 2.88 – –
3 – – 1.00 – 1.99 – –
4 – – 1.00 1.37 – –
5 – – 0.52 0.48 1.67 1.66 0.60
6 – 0.50 0.50 1.91 1.98 2 3.66
7 – 0.50 0.50 2.39 2.39 0.00
8 – – 0.25 0.75 1.51 1.5 0.66
9 – – 0.75 0.25 1.82 1.81 0.55
10 – 0.25 – 0.75 1.62 1.65 2 1.85
11 – 0.75 – 0.25 2.32 2.39 2 3.02
12 – 0.25 0.75 – 2.19 2.19 0.00
13 – 0.74 0.26 – 2.62 2.62 0.00
14 – 0.34 0.33 0.33 1.94 1.99 2 2.58
15 – 0.25 0.25 0.51 1.77 1.81 2 2.26
16 – 0.19 0.54 0.27 1.91 1.93 2 1.05
17 – 0.50 0.25 0.25 2.12 2.18 2 2.83
18 0.25 0.25 0.25 0.25 2.51 2.57 2 2.39
19 0.52 0.16 0.16 0.16 3.31 3.38 2 2.11
20 0.16 0.51 0.16 0.16 2.62 2.67 2 1.91
21 0.16 0.16 0.51 0.16 2.32 2.35 2 1.29
22 0.16 0.16 0.16 0.51 2.01 2.06 2 2.49
 C.A.W. Allen et al. / Fuel 78 (1999) 1319–1326 1323

Table 3
Viscosities and mass fraction of simulated vegetable oil ME at 408C (using measured viscosities of pure components)

Fatty acid methyl ester Viscosity (mPa s) Simulated vegetable oil methyl ester-component mass fraction

Coconut Palm Rapeseed Peanut#1 Soybean Canola Peanut#2 a

8:0 0.99 0.078 0.001 0.000 0.000 0.000 0.000 0.000

10:0 1.40 0.070 0.001 0.000 0.000 0.000 0.000 0.000
12:0 1.95 0.466 0.009 0.000 0.000 0.000 0.000 0.000
14:0 2.69 0.182 0.013 0.000 0.000 0.002 0.002 0.000
16:0 3.60 0.092 0.441 0.029 0.106 0.102 0.037 0.104
0.75 4.74 0.028 0.051 0.025 0.010 0.046 0.024 0.033
(0.054) b
18:1 3.73 0.068 0.384 0.128 0.499 0.222 0.600 0.529
18:2 3.05 0.017 0.095 0.116 0.334 0.546 0.218 0.327
18:3 2.65 0.000 0.004 0.085 0.005 0.082 0.113 0.005
20:0 – 0.000 0.000 0.000 0.013 0.000 0.000 0.000
(0.000) b
22:0 – 0.000 0.000 0.000 0.031 0.000 0.000 0.000
(0.000) b
22:1 5.91 0.000 0.000 0.617 0.002 0.000 0.005 0.002
Measured viscosity (mPa s) 2.15 3.59 4.7 3.57 3.26 3.45 3.51
Predicted viscosity (mPa s) 2.19 3.59 4.72 3.51 3.27 3.45 3.50
Error (mPa s) 20.04 0 20.03 0.06 20.01 0 0.01
Per cent error 22.04 20.08 20.57 1.49 20.24 0.11 0.31
a
Peanut#2 has no 22:0 and 20:0.
b
Mass fraction after being combined with 18:0.

and (12) for MEs and EEs, respectively. ester standards. Initially, the model was tested using binary,
ternary and quaternary mixtures of 8:0, 10:0, 12:0 and 18:1
mME2unsat2C18 ˆ 0:153 NDB 2 1:15 NDB 1 4:73
2
EE standards at 258C. The mixing ratios and results are
…11†
summarized in Table 2 where it can be seen that the errors
s ˆ 0:0112
in predicting the viscosities are all less than 3.7%, showing
that the assumption of a negligible interaction parameter,
mEE2unsat2C18 ˆ 0:147 NDB2 2 1:09 NDB 1 4:82 and that the use of the measured mass fraction of the control
…12† samples provide acceptable results. It was thus decided to
s ˆ 0:000 extend the verification procedure to more complex mixtures
where NDB is the number of double bonds in the C18 chain. of MEs that simulated biodiesel fuels.
The ME standards that were mixed on a mass basis to
4.4. Measurement and prediction of the viscosity of mixtures simulate rapeseed, canola, peanut, soybean, palm and coco-
of GC standards nut oil ME fuels were used in this phase. Peanut oil has
greater than 4% of 20:0 and 22:0 fatty acids combined
The logarithmic equation (Eq. 2) for predicting the whose viscosities are not measurable at the test temperature
viscosities of multi-component mixtures was tested by of 408C, since these individual components are solid at this
comparing measured and predicted viscosities of fatty acid temperature. For comparison purposes, it was decided to

Table 4
Viscosities of individual fatty acid methyl ester

FAME Viscosity taken from Swern [5] Viscosity of GC standard Per cent difference

8:0 1.01 1.01 0.00

10:0 1.47 1.4 4.76
12:0 1.98 1.95 1.52
14:0 2.81 2.69 4.27
16:0 3.76 3.6 4.26
18:0 4.88 4.74 2.87
18:1 3.87 3.73 3.62
18:2 3.17 3.05 3.79
18:3 2.84 2.65 6.69
22:1 6.27 5.91 5.74
1324 C.A.W. Allen et al. / Fuel 78 (1999) 1319–1326

Table 5
Viscosities of methyl ester biodiesel fuels at 408C (using viscosities of pure components taken from Swern [5] and combined mass fractions)

Fatty acid methyl ester Viscosity (mPa s) Methyl ester component mass fraction

Coconut Peanut Soya Palm Canola

24:0 – 0.000 0.035 0.000 0.000 0.000

– (0.000) a (0.000) a – –
22:1 6.27 0.000 0.000 0.000 0.000 0.000
22:0 – 0.000 0.024 0.000 0.000 0.001
– (0.000) a (0.000) a – (0.000) a
20:1 – 0.000 0.014 0.016 0.001 0.021
– (0.000) b (0.000) b (0.000) b (0.000) b
20:0 – 0.002 0.013 0.007 0.003 0.012
(0.000)* (0.000) a (0.000) a (0.000) a
8:3 2.84 0.000 0.010 0.096 0.002 0.112
18:2 3.17 0.014 0.301 0.199 0.080 0.213
18:1 3.87 0.055 0.466 0.600 0.373 0.574
– (0.485) b (0.623) b (0.377) b (0.599) b
18:0 4.88 0.019 0.027 0.017 0.040 0.020
– (0.099) (0.023) (0.042) (0.033)
16:1 – 0.000 0.004 0.008 0.003 0.004 (0.000) b
– (0.000) b (0.000) b (0.000) b (0.000) b
16:0 3.76 0.073 0.105 0.058 0.481 0.042
14:0 2.81 0.171 0.000 0.000 0.013 0.000
12:0 1.98 0.533 0.000 0.000 0.004 0.000
10:0 1.47 0.060 0.000 0.000 0.000 0.000
8:0 1.01 0.075 0.000 0.000 0.000 0.000
Measured viscosity (mPa s) 2.32 3.77 3.67 3.87 3.7
Predicted viscosity (mPa s) 2.19 3.71 3.62 3.76 3.61
Error (mPa s) 0.13 0.06 0.05 0.11 0.09
Per cent error 5.72 1.71 1.25 2.76 2.50
a
Mass fraction after being combined with 18:0.
b
Mass fraction after being combined with 18:1.

produce two mixtures to simulate peanut oil, one including

the 20:0 and 22:0 components (peanut#1) and the other
without the 20:0 and 22:0 components (peanut#2). In the
peanut#1 sample, the 20:0 and 22:0 were assumed to have Table 6
the same viscosity as the 18:0 and so the mass fractions of Biodiesel fuel purity and composition obtained by TLD-FID analysis
(FFA ˆ free fatty acids)
the three components were combined for calculation
purposes. All other simulations required less than 1% of Oil type Component type Composition (%)
these fatty acids which were thus considered negligible.
Soya ME 99.76
Table 3 shows the predicted and measured viscosities along
TG&FFA* 0.00
with the mass fractions of the simulated vegetable oil FAMEs. Others 0.24
The errors in predicting the viscosities of all the simulated oils Total 100.00
were 2% or less, with the highest absolute error being Palm ME 98.21
0.06 mPa s. The peanut#1 oil sample with the mass fractions TG 0.00
Others 1.79
of its 20:0 and 22:0 components combined together with 18:0,
Total 100.00
had a 1.5% error compared with a 0.3% error for the peanut#2 Peanut ME 99.81
sample with those two components excluded from the mixture. TG&FFA* 0.00
Although the error obtained by lumping the three components Others 0.19
together is higher than the sample without 20:0 and 22:0, this Total 100.00
Coconut ME 95.01
error is well within the acceptable limit previously defined
TG&FFA* 0.00
and is comparable with the errors obtained for the other Others 4.99
simulated oils. As all the errors were within the defined Total 100.00
acceptable limit, it was concluded that, for practical Canola ME 100.00
purposes, the logarithmic equation (Eq. 2) was applicable TG&FFA* 0.00
Others 0.00
for predicting the viscosities of MEs and EEs of multi-
Total 100.00
component fatty acid ester mixtures used as biodiesel fuels.
 C.A.W. Allen et al. / Fuel 78 (1999) 1319–1326 1325

4.5. Measurement and prediction of the viscosity of five

Erucic 22:1
biodiesel types

50.9
0.2

0.0
0.5

0.0

0.0
0.0

0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
As Eq. (2) was demonstrated to predict viscosities well,
knowing the fatty acid ester composition, it was possible to
estimate the viscosity of five biodiesel fuels prepared in the
Linolenic 18:3

laboratory. The FAME composition of the fuels were deter-

mined by GC analysis and determined to be the same as in
0.5
8.6
11.1
0.6
0.0

0.3
1.1

0.1
0.6
8.3
0.1
0.5
0.5
0.6
1.1
Ackman. When the viscosities of the component FAME as
determined above were used to predict the viscosity of the
biodiesel fuels, the error on viscosity (compared to the
Linoleic 18:2

measured values) were low by 7.1, 8.1, 7.0, 6.0 and 6.3%
for canola, coconut, palm, peanut and soya oil, respectively.
This caused the authors to question the composition of the
32.9
13.1
21.1
9.5

56.8

71.4
54.1

54.2
1.7

9.5
75.5

4.3

2.7
3.1
8.9
standards especially because there were some inexplicable
anomalies in the surface tension of some of the standards
compared to literature values, as reported by Allen [16]
Oleic 18:1

which included a sharp drop for 16:0 EE from an otherwise

47.1
21.9
60.2
75.3

29.1
39.0
14.4
16.0
22.5
88.2
18.5
48.9
27.8
47.1
6.9

smooth trend of saturated esters, and very erratic surface

tension data for the unsaturated esters. Thus another source
of viscosity for the component ME was sought. Since Swern
[5] has a complete listing of viscosities for 8:0 to 18:0, 18:1
Stearic 18:0

to 18:3 and 22:1 at 408C, it was decided to compare viscos-

ities developed by Swern with the viscosities measured in
19.2
8.9
2.8

11.4
15.8
3.1
3.6
2.9
3.1
4.9
3.3
5.9
4.7
4.4
3.1

this study. A comparison of these two sets of data showed

that the values measured in this study were 1.6–6.7% lower
than that reported by Swern [5] (Table 4). No anomalous
Palmitic 16:0

trends were noted in his data, and thus more confidence was
obtained in the absolute value of his viscosities. The biodie-
10.4

11.0

43.9

10.3

22.9
25.2
2.7

33.4
25.5
3.9

9.2
9.9

6.6
6.0

3.0

sel fuels produced contained minor quantities of some

FAMEs whose viscosities were not available. To approxi-
mate for these components, their mass fractions were
Myristic 14:0

combined together with a component that had a comparable

CN and saturation. For example, the mass fractions of
higher saturates 20:0, 22:0 and 24:0, which occurred in
11.6
0.0
0.0

18.3
0.1
0.0

0.0
1.3
0.1
0.1
0.1
0.0
0.8
3.3

1.4

minor quantities, were combined together with 18:0, and

the mass fractions of the unsaturates 16:1 and 20:1 were
Lauric 12:0

combined together with 18:1.

The viscosities predicted using the component viscosities
46.7
0.0
0.0
0.0
0.0

0.0
0.9
0.0
0.0
0.0
0.0
0.0
0.1
3.6
0.1

taken from Swern [5], and with some of the mass fractions
Fatty acid composition of 15 fats and oils (From Ref. [15])

combined together were all within 2.8% of the measured

viscosities, except for coconut oil which had a 5.7% error
Capric 10:0

(Table 5).
The results of the TLC-FID analysis, given in Table 6,
showed that all of the oils contained 98% or more MEs
Fatty acid composition

0.0
0.0
0.0
0.0
6.0
0.0
0.1
0.0
0.0
0.0
0.0
0.0
0.1
3.0
0.1

except for coconut oil which contained approximately

95% MEs with the remaining 5% being one or a combina-
Caprylic 8:0

tion of mono-, di-glycerides, sterols or polar lipids. Since

mono- and di-glycerides were not available to the research-
ers, small amounts of triglycerides (TG) were systematically
0.0
0.0
0.0
0.0
8.3
0.0
0.1
0.0
0.0
0.0
0.0
0.0
0.0
5.5
0.0

added to ME to give an indication of the effect of glycerides

on viscosity. Canola oil, which was readily available, was
used in these tests. TG in the form of canola oil in mass
Beef Tallow
Cottonseed
Sunflower
Safflower
Rapeseed

fractions of 1, 2, 4 and 6% were respectively added to canola

Butterfat
Soybean
Coconut
Oil type
Table 7

Canola

Sunola
Peanut

oil ME. It was found that 1% TG in the ME resulted in a

Olive

Palm
Corn

Lard

1.5% increase in viscosity; 2% gave a 3.8% increase; 4%

1326 C.A.W. Allen et al. / Fuel 78 (1999) 1319–1326

Table 8 Acknowledgements
Predicted viscosities for 15 biodiesel fuel types (408C)

Oil type Viscosity (mPa s) This project was supported by the Natural Science and
Engineering Research Council, Ottawa, grant to K.C. Watts.
Predicted Lower prediction limit Upper prediction limit

Peanut 3.69 3.58 3.79

Rapeseed 4.72 4.59 4.85
Canola 3.61 3.51 3.72 References
Olive 3.81 3.70 3.92
Coconut 2.25 2.19 2.32 [1] Graboski MS, McCormick RL. Combustion of fat and vegetable oil
Corn 3.46 3.36 3.56 derived fuels in diesel engines. Progress in Energy and Combustion
Palm 3.74 3.64 3.85 Science 1998;24(1):125–164.
Safflower 3.35 3.25 3.44 [2] Reid RC, Prausnitz JM, Poling BE. The properties of gases and
Sunflower 3.39 3.29 3.48 liquids. New York: McGraw-Hill, 1987.
Soybean 3.41 3.32 3.51 [3] Fisher CH. Evaluating and predicting n-fatty acid properties. Journal
Sunola 3.87 3.76 3.98 of the American Oil Chemists Society 1988;65(10):1647–1651.
Cottonseed 3.46 3.36 3.56 [4] Noureddini H, Teoh BC, Clemens LD. Viscosities of vegetable oils
Beef Tallow 3.94 3.83 4.05 and fatty acids. Journal of the American Oil Chemists Society
Butterfat 3.31 3.22 3.41 1992;66(12):1189–1191.
Lard 3.74 3.63 3.84 [5] Swern D. Bailey’s industrial oil and fat products, 1. New York: Wiley,
1979.
[6] Markley KS. Fatty acids—their chemistry, properties, production,
gave a 7.8% increase; and 6% resulted in an 11.8% increase and uses. New York: Interscience Publishing Inc, 1960.
[7] Grunberg L, Nissan AH. Mixture law for viscosity. Nature
in the viscosity of the mixture. Assuming that mono- and di- 1949;164:799–800.
glycerides may have similar effects, these results serve to [8] Monnery WD, Svrcek WY, Mehrotra AK. Viscosity: a critical review
explain why the measured viscosity of coconut oil was 5.8% of practical predictive and correlative methods. Canadian Journal of
higher than its predicted viscosity compared with an average Chemical Engineering 1995;73:3–40.
2.1% error for the other oils. This also indicates a possible [9] Irving JB. Viscosity of binary liquid mixtures: a survey of mixture
equations, NEL Report 630, National Engineering Laboratory,
source of variation between different worker’s results. Department of Industry, London, England, 1977.
[10] Allen CAW, Watts KC. A batch type transesterification unit for
4.6. Predicted viscosities of 15 biodiesel types biodiesel fuels, Technical paper 96-404, Canadian Society of Agri-
cultural Engineers, 1996.
The results obtained above showed that the errors [11] Ackman RG, McLeod CA, Banerjee AK. An overview of analysis by
obtained for predicting the viscosities of mixtures of MEs Chromarod-Iatroscan TLC-FID. Journal of Planar Chromatography
were within the maximum allowable error previously 1990;3:450–490.
defined. It was thus decided to use Eq. (2) to predict the [12] Ackman RG. Application of gas–liquid chromatography to lipid
separation and analysis: qualitative and quantitative analysis. In:
viscosities of the MEs of typical oils. Perkin EG, editor. Analysis of fats, oils and lipoproteins, Champaign,
The fatty acid composition of 15 fat and oils are given by IL: American Oil Chemists’ Society, 1991.
Ackman [17]. To predict the viscosity of the MEs of these [13] Msipa C, Goering C, Karcher T. Vegetable oil atomization in a DI
oils using Eq. (2), some minor quantities of their fatty acids diesel engine. Transactions of ASAE 1983;26(6):1669–1672.
had to be combined together in the manner previously [14] Allen CAW, Watts KC, Ackman RG. Properties of methyl esters of
interesterified triacylglycerols Proceeding of the Third Liquid Fuel
described. Table 7 gives a listing of the oils/fats and the Conference. Liquid fuels and industrial products from renewable
fatty acid composition used in the analyses. resources. St. Joseph, MI: ASAE, 1996 pp. 73-82.
The predicted viscosities of the MEs of the 15 oils are [15] Briscoe BJ, Luckham PF, Ren SR. An assessment of a rolling-ball
summarized in Table 8 along with the upper and lower viscometer for studying non-Newtonian fluids. Colloids and Surfaces
prediction limits, based on the average error of the model. 1992;62:153–162.
[16] Allen CAW. Prediction of biodiesel fuel atomization characteristics
It can be seen that most of the oils had similar viscosities based on measured properties, Ph.D thesis, DalTech, Dalhousie
ranging from 3.31 to 3.94 mPa s. However, rapeseed oil ME University, Nova Scotia, Canada, 1998.
had the highest predicted viscosity (4.72 mPa s) while coco- [17] Ackman RG. Fatty acids in newer fats and oils. In: Hui YH, editor.
nut oil ME had the lowest (2.25 mPa s). This 100% differ- Bailey’s industrial oil and fat products, 1. New York: Wiley, 1996.
ence in viscosity range is significant in the utilization of
biodiesel fuels.