Fuel Processing Technology 134 (2015) 456–464

Contents lists available at ScienceDirect

Fuel Processing Technology

journal homepage: www.elsevier.com/locate/fuproc

Density, flash point and heating value variations of corn oil biodiesel–diesel
fuel blends
Mert Gülüm ⁎, Atilla Bilgin
Karadeniz Technical University, Faculty of Engineering, Department of Mechanical Engineering, Trabzon 61080, Turkey

a r t i c l e i n f o a b s t r a c t

Article history: In this study, densities of produced corn oil biodiesel and its blends with commercially available petro-diesel fuel
Received 29 December 2014 have been investigated. The effects of temperature (T) and biodiesel percentage in blend (X) on the densities of
Received in revised form 26 February 2015 blends were examined. The blends (B5, B10, B15, B20, B50 and B75) were prepared on a volume basis and their
Accepted 28 February 2015
densities were measured by following ISO test method at temperatures of 10, 15, 20, 30 and 40 ° C. New one- and
Available online 14 March 2015
two-dimensional equations were fitted to the measurements for identifying of variations of densities with
Keywords:
respect to X and T; and these equations were compared with other equations published in literature. Moreover,
Corn oil biodiesel the qualities of the corn oil biodiesel and its blends were evaluated by determining the other important properties
Diesel such as flash point temperature and higher heating value. In order to predict these properties, some equations
Blend were also evaluated as a function of biodiesel percentage in blend.
Density © 2015 Elsevier B.V. All rights reserved.
Flash point
Heating value

1. Introduction
Rapid depletion of known reserves and rising prices of fossil fuels,
and stringent regulations on exhaust emissions have contributed to
research new and clear renewable alternative fuels. Biodiesel is receiving
increasing attention day by day as an alternative fuel for diesel engines
[1,2]. It can be produced from renewable sources such as virgin or
even used vegetable oils and animal fats via a reaction known as
transesterification [3]. In this reaction, the saturated and unsaturated
fatty acids in oils react with a mono-alcohol (most commonly methanol
or ethanol) in the presence of a catalyst (alkali, acidic or enzyme) to yield
glycerol and mono-alkyl methyl/ethyl esters (i.e. biodiesel) [4,5]. Bio-
diesel has already entered the fuel market and has many advantages
as: (i) it comprises of 10 to 12% oxygen by mass in the molecular struc-
ture, thus improving combustion efficiency and decreasing emissions of
carbon monoxide (CO), unburned or half-burned hydro carbons (HC)
and smoke [6], (ii) it reduces net carbon dioxide (CO2) emissions by
78% on a life-cycle basis when compared to diesel fuel [7], (iii) it is
biodegradable and non-toxic, being beneficial for reservoirs, lakes,
marine life and other environmentally sensitive places [8], (iv) its
flash point temperature is higher than that of petro-diesel fuel (hereafter
referred to as diesel fuel), which makes it less volatile and safer regarding
the storage and transport than diesel fuel [9], (v) it does not contain
sulphur or aromatic compounds and thus it contributes to the reduction
of the diesel engine exhaust emission levels [10,11], (vi) it enhances
⁎ Corresponding author. Tel.: +90 462 377 41 30; fax: +90 462 377 33 36.
E-mail address: gulum@ktu.edu.tr (M. Gülüm).
cetane number, which shortens ignition delay [12], (vii) it can be
produced by using domestic renewable feedstock, reducing country's
dependency on foreign fuel supplies [13], (viii) biodiesel–diesel fuel
blends or even pure biodiesel can be used in diesel engines with small
modifications [14], and (viii) it improves the lubricity, which results in
longer engine component life [15].
These properties of biodiesel make it an ideal fuel for diesel engines.
However, there are some disadvantages of biodiesel such as lower
energy content and volatility, higher viscosity and NOx emissions
compared to diesel fuel [16,17]. Also, the biodiesel produced from oils,
no matter if it is neat vegetable oil or animal fat, is usually more expen-
sive than diesel fuel from 10% to 50%. Therefore, the high cost of biodiesel
is the major obstacle for its commercialization [18].
Density, on the other hand, is one of the key properties since some
other crucial fuel properties such as cetane number and heating value
are related to it. In addition, because the amount of fuel injected to
combustion chamber is measured volumetrically for diesel engines,
the variation of the density directly affects the engine output power
and fuel consumption. Also, density influences the start of injection,
injection pressure, and fuel spray characteristics, so that these affect
the combustion and exhaust emissions [15,19].
As the use of biodiesel becomes more widespread, researchers have
shown a strong interest in modeling the combustion process in the
engine in order to understand the fundamental characteristics of
biodiesel combustion [20]. They often use physical properties of biodie-
sel as input data in their combustion models. However, it may not be
practical at every turn to make measurements of physical properties
of biodiesel or biodiesel–diesel fuel blends for each blending ratio
or temperature in any study. Regression equations as a function of

http://dx.doi.org/10.1016/j.fuproc.2015.02.026
0378-3820/© 2015 Elsevier B.V. All rights reserved.
 M. Gülüm, A. Bilgin / Fuel Processing Technology 134 (2015) 456–464 457

Table 2
Nomenclature and units Fatty acid methyl ester composition of the produced biodiesel.

a, b, d, e Regression constants Fatty acid Mass%

B5, B10, B15, B20, B25, B50, B75 biodiesel–diesel fuel blends Palmitic (C16:0) 15.190
B100 pure corn oil biodiesel Oleic (C18:1) 46.954
Linoleic (C18:2) 34.243
D pure diesel fuel
α -Linolenic acid (C18:3) 1.276
HHV higher heating value (kJ/kg) Arachidic (C20:0) 0.754
Kball coefficient of the viscometer ball (mPa ⋅ s ⋅ cm3/g ⋅ s) Gadoleic acid (C20:1) 0.657
mtotal mass of the pycnometer filled with biodiesel (g) Behenic (C22:0) 0.487
Lignoceric (C24:0) 0.439
R correlation coefficient
Average molecular mass 292.870 g/mola
t falling time of the viscometer ball (s) Typical formula C18.77H35.16O2a
T temperature (° C) a
Calculated from fatty acid distribution.
w1, w2, w3, …, wn uncertainties of independent variables
x1, x2, x3, …, xn independent variables
analysis was performed to determine reliability of the measured density
X volumetric biodiesel percentage in blend (%)
values. Moreover, in order to estimate flash point temperatures and
higher heating values of these blends, some equations were evaluated
Greek symbols
as a function of biodiesel percentage in blend.
ρ density (kg/m3), (g/cm3)

temperature, percentage of blend and the chemical structure have been
generally used to calculate these properties without measurements.
Some studies reporting these equations are summarized as follows.
Tat and Gerpen [21] carried out a study about determination of specific
gravities of soybean oil biodiesel and its blends with No. 1 and No. 2
diesel fuels at 75, 50 and 20% by weight from onset of crystallization
temperature to 100 °C. According to experimental results, linear type
relationship was found between specific gravity and temperature of bio-
diesel–diesel fuels blends. Therefore, the straight line correlation was
well fitted to the experimental data and the lowest regression coefficient
(R2) was computed as 0.9989. Alptekin and Canakci [15] measured den-
sities and kinematic viscosities of biodiesels derived from six different
vegetable oils and its blends with two different diesel fuel (purchased
from Shell Extra Diesel and Petrol Office firms). Linear and Arrhenius
mixing equation were used to predict the densities and viscosities of
the blends, respectively. In the study performed by Benjumea et al. [22]
some basic properties (viscosity, density, heating value, cloud point
temperature, calculated cetane index and T10, T50 and T90 distillation
temperatures) of several palm oil biodiesel–diesel fuel blends were
measured. In order to predict values of these properties, Kay and
Arrhenius equations were evaluated as a function of the volume fraction
of biodiesel. According to the absolute average deviation values (AAD), it
was found that simple Arrhenius equations were suitable for predicting
the basic properties. Tate et al. [23] obtained densities of three commer-
cially available biodiesels at between from 20 to 300 °C by means of a
capacitance type liquid level meter. According to results, densities were
found to be linear behavior with temperature. Moreover, the linear
regression coefficients (R2) were obtained as 0.9911, 0.9872 and 0.9825
for canola methyl ester, soybean methyl ester and fish oil ethyl ester,
respectively
The main objective of the present study is to investigate the effects of
volumetric biodiesel percentage in blend (X) and temperature (T) on
densities of produced corn oil biodiesel (B100) and its blends (B5,
B10, B15, B20, B25, B50 and B75) with commercially available Ultra
Force Euro diesel fuel (D) purchased from Shell. Some new one- and
two-dimensional equations were fitted to the measurements and
compared with other equations published in literature. Uncertainty
2. Experimental methods
2.1. Biodiesel production
In this study, commercially available refined corn oil was used in bio-
diesel production. It was not needed to perform a pretreatment to the
oil because of being refined. Thus, methanol (CH3OH) of 99.80% purity
as alcohol and pure grade sodium hydroxide (NaOH) as catalyst were
used in transesterification reaction. In a master's thesis by Gülüm [24]
and previous study by the authors [25], many reaction parameters,
which influence the transesterification reaction, such as catalyst
concentration, reaction temperature, reaction time and alcohol/oil
molar ratio were varied within the range of 0.25–1.50%, 40–70 °C,
30–120 min and 3:1–12:1, respectively, and the effects of these
parameters on changes of densities and, dynamic and kinematic
viscosities of produced biodiesel were investigated parametrically
to produce corn oil biodiesel having the lowest viscosity. Optimum re-
action parameters were obtained as 0.90% catalyst concentration, 50 °
C reaction temperature, 60 minute reaction time and 9:1 alcohol/oil
molar ratio [24,25]. Transesterification reaction was carried out in a
1 L flat-bottomed flask, equipped with a magnetic stirrer heater,
thermometer and spiral reflux condenser. Isolab pycnometer and top
loading balance with an accuracy of ±0.01 g were used to measure den-
sity. Before starting the reaction, 1.80 g catalyst was dissolved in meth-
anol in a narrow-neck flask to make alcoholic solution of catalyst. In the
flat bottomed flask, this alcoholic solution was added to the 200 g corn
oil that was formerly warmed to about 80 °C in a beaker. These reactants
were mixed for 60 min at 50 °C with stirring speed of 500 rpm by means
of the magnetic stirrer heater. Transesterification reaction was carried
out with the spiral reflux condenser for avoiding loss of alcohol. Also, re-
action temperature was controlled by using thermometer to remain at a
constant temperature during the reaction. At the end of reaction, the
resulting product mixture was transferred to a separating funnel. After
a day, two phases occurred in the separating funnel. The upper phase
consists of methyl esters (biodiesel) while the lower one consists of
glycerol, excess methanol and the remaining catalyst together with
soap. After separation of the two layers by gravity, the upper layer
(biodiesel) was washed with warm distilled water until the water

Table 1
Some fuel specifications of diesel fuel, produced biodiesel and their blends, and corresponding standard values for biodiesel.

Properties Unit D B5 B10 B15 B20 B25 B50 B75 B100 EN14214 ASTM-D6751

Density at 15 ° C kg/m3 832.62 834.25 836.89 838.72 841.37 844.45 855.62 865.99 876.37 860–900 a

Flash Point °C 63 68 69 70 72 75 99 120 169 101≤ 130 ≤

a a
HHV kJ/kg 45,950 45,368 44,984 44,643 44,399 44,205 42,700 41,310 39,930
a
Not specified.
458 M. Gülüm, A. Bilgin / Fuel Processing Technology 134 (2015) 456–464

uncertainty analysis was carried out depending on the sensitivities of

Fig. 1. Changes of density values of fuel blends with respect to biodiesel percentage in
blend for various temperatures.
measurement devices.
In this study, densities were measured in Internal Combustion
Engines Laboratory in the Mechanical Engineering Department at
Karadeniz Technical University. The fatty acid methyl esters of produced
corn oil biodiesel were qualitatively and quantitatively analyzed by gas
chromatography using a Hewlett-Packard HP-6890 Series GC system
fitting with a HP-6890 mass selective detector (1909N-133 innowax
capillary column of 30 m length, 0.25 mm I.D and 0.25 μm film thick-
ness) in Science Research and Application Center at Mustafa Kemal
University. The other properties of the pure fuels and fuel blends such
as flash point temperature (EN ISO 3679) and higher heating value
(DIN 51900-2) were measured in the Prof. Dr. Saadettin GÜNER Fuel
Research and Application Center at Karadeniz Technical University.
These properties, EN 14214 and ASTM D 6751 standard values are
given in Table 1. Also, the fatty acid compositions of the produced
corn oil biodiesel and its calculated average molecular mass and typical
formula are given in Table 2.
2.3. Uncertainty analysis

became clear. Washed biodiesel was heated up to about 100 °C to The results obtained from experimental studies are generally calcu-
remove methyl alcohol and water residuals. lated from measured physical quantities. These quantities have some
uncertainties due to uncertainties of measuring tools and measurement
2.2. Density measurement systems. Therefore, uncertainty analysis should be applied for proving
reliability of the calculated results. In this study, uncertainties of the
The densities of the produced biodiesel and its blends were deter- measured density values were determined by the method proposed
mined by means of Eq. (1) and measurements in accordance with ISO by Kline and McClintock given in [26]. According to this method,
4787 standard: if the result R is a given function of the independent variables
x1, x2, x3, …, xn, and w1, w2, w3, …, wn are the uncertainties of each
mtotal −mpycnometer independent variables, then the uncertainty of the result wR is calculated
ρbiodiesel or blends ¼ ρwater ð1Þ by using the equation:
mwater
"
2
2
2
2 #1=2
∂R ∂R ∂R ∂R
where ρ and m represent density and mass, respectively. In order to wR ¼  w1 þ  w2 þ  w3 þ⋯þ  wn
∂x1 ∂x2 ∂x3 ∂xn
minimize measurement errors, all the measurements were conducted
three times for each sample and the results were averaged. Also ð2Þ

Table 3
For different temperatures, the measured densities, calculated densities from Eq. (3), errors between measured and calculated densities, regression constants and correlation coefficients.

Temp.T (°C) Measured, ρ (kg/m3) Regression constants R

Blend, X (%)

0 5 10 15 20 25 50 75 100 a (10+3) b

10 833.12 834.75 837.39 839.23 841.87 843.91 854.91 865.90 876.90 0.8329 0.4403 0.9999
15 832.62 834.25 836.89 838.72 841.37 843.61 853.99 865.18 876.37 0.8324 0.4383 0.9999
20 831.87 833.49 836.14 837.97 840.61 842.64 853.62 864.61 875.59 0.8316 0.4398 0.9999
30 829.74 831.36 834.00 835.83 838.46 840.49 851.44 862.40 873.35 0.8295 0.4386 0.9999
40 826.95 828.57 831.20 833.01 835.64 837.66 848.58 859.49 870.41 0.8267 0.4371 0.9999

Calculated, ρ (kg/m3)

Blend, X (%)

0 5 10 15 20 25 50 75 100

832.9000 835.1015 837.3030 839.5045 841.7060 843.9075 854.9150 865.9225 876.9300

832.4000 834.5915 836.7830 838.9745 841.1660 843.3575 854.3150 865.2725 876.2300
831.6000 833.7990 835.9980 838.1970 840.3960 842.5950 853.5900 864.5850 875.5800
829.5000 831.6930 833.8860 836.0790 838.2720 840.4650 851.4300 862.3950 873.3600
826.7000 828.8855 831.0710 833.2565 835.4420 837.6275 848.5550 859.4825 870.4100

Absolute errors (%)

Blend, X (%)

0 5 10 15 20 25 50 75 100

0.0264 0.0421 0.0104 0.0327 0.0195 0.0003 0.0006 0.0026 0.0034

0.0264 0.0409 0.0128 0.0303 0.0242 0.0299 0.0381 0.0107 0.0160
0.0325 0.0371 0.0170 0.0271 0.0255 0.0053 0.0035 0.0029 0.0011
0.0289 0.0401 0.0137 0.0298 0.0224 0.0030 0.0012 0.0006 0.0011
0.0302 0.0381 0.0155 0.0296 0.0237 0.0039 0.0029 0.0009 0.0000
 M. Gülüm, A. Bilgin / Fuel Processing Technology 134 (2015) 456–464 459

3. Results and discussion

3.1. One-dimensional lineal models

3.1.1. Effects of biodiesel percentage on density

Fig. 1 demonstrates the variations of densities of fuel blends (B5,
B10, B15, B20, B25, B50 and B75) with respect to biodiesel percentages
(X) for different temperatures (T). In this figure, the points correspond
to measured density values at studied temperatures and biodiesel per-
centages while the lines are plots of curve fit equation. All the fuel
blends follow the similar trend, i.e., densities increase with increasing
in biodiesel percentage for a specific temperature, and density values
are getting closer to each other as temperature decreases for any biodiesel
percentage. The linear model, given in Eq. (3), is considered to be suitable
to characterize these variations because densities of the blends are
directly proportional to biodiesel percentage in the blend:

ρ ¼ ρðXÞ ¼ a þ bX ð3Þ

where ρ is density of the blends in kg/m3, a and b are regression

constants.
Table 3 shows the measured and calculated (from Eq. (3)) densities
of the blends and pure fuels (D and B100), % errors between measured
and calculated values, regression constants and correlation coefficients
(R). The correlation coefficient is a quantitative measure of goodness-
of-fit of the regression equation to the measured data. For a perfect fit,
for example, R becomes 1 which means that the equation explains
100% of the variability of the measured data [27]. The maximum
error was obtained for B5 blend at 10 °C as 0.0421%. The correlation
coefficients were computed as 0.9999 for all blends. These results and
Fig. 2. Variations of density values of pure fuels and fuel blends with respect to tempera-
ture for different regression equations.
the figure show that the linear model properly fits the measured data
qualitatively and quantitatively, and represents almost exactly the
density-biodiesel percentage relationship.

According to Eq. (2), the highest uncertainty for measured densities 3.1.2. Effects of temperature on density
was determined as 0.0346%, which means that the results are highly Fig. 2 shows the effects of temperature on densities of pure fuels
reliable. and biodiesel-diesel fuel blends. As shown in the figure, the densities,

Table 4
For different biodiesel percentages, the measured densities, calculated densities from Eq. (4), errors between measured and calculated densities, regression constants and correlation
coefficients.

Blend X (%) Measured, ρ (kg/m3) Regression constants R

Temp., T (°C)

10 15 20 30 40 a b

0 833.12 832.62 831.87 829.74 826.95 835.7000 −0.2083 0.9881

5 834.75 834.25 833.49 831.36 828.57 837.3000 −0.2086 0.9883
10 837.39 836.89 836.14 834.00 831.20 839.9000 −0.2090 0.9881
15 839.23 838.72 837.97 835.83 833.01 841.8000 −0.2098 0.9881
20 841.87 841.37 840.61 838.46 835.64 844.4000 −0.2103 0.9880
25 843.91 843.61 842.64 840.49 837.66 846.6000 −0.2138 0.9872
50 854.91 853.99 853.62 851.44 848.58 857.3000 −0.2081 0.9870
75 865.90 865.18 864.61 862.40 859.49 868.4000 −0.2135 0.9877
100 876.90 876.37 875.59 873.35 870.41 879.6000 −0.2190 0.9880

Calculated, ρ (kg/m3) Absolute errors (%)

Temp., T (°C) Temp., T (°C)

10 15 20 30 40 10 15 20 30 40

833.6170 832.5755 831.5340 829.4510 827.3680 0.0597 0.0053 0.0404 0.0348 0.0505
835.2140 834.1710 833.1280 831.0420 828.9560 0.0556 0.0095 0.0434 0.0383 0.0466
837.8100 836.7650 835.7200 833.6300 831.5400 0.0502 0.0149 0.0502 0.0444 0.0409
839.7020 838.6530 837.6040 835.5060 833.4080 0.0562 0.0080 0.0437 0.0388 0.0478
842.2970 841.2455 840.1940 838.0910 835.9880 0.0507 0.0148 0.0495 0.0440 0.0416
844.4620 843.3930 842.3240 840.1860 838.0480 0.0654 0.0257 0.0375 0.0362 0.0463
855.2190 854.1785 853.1380 851.0570 848.9760 0.0361 0.0221 0.0565 0.0450 0.0467
866.2650 865.1975 864.1300 861.9950 859.8600 0.0422 0.0020 0.0555 0.0470 0.0430
877.4100 876.3150 875.2200 873.0300 870.8400 0.0582 0.0063 0.0423 0.0366 0.0494
460 M. Gülüm, A. Bilgin / Fuel Processing Technology 134 (2015) 456–464

Table 5
For different biodiesel percentages, the measured densities, calculated densities from Eq. (5), errors between measured and calculated densities, regression constants and correlation
coefficients.

Blend X (%) Measured, ρ (kg/m3) Regression constants R

Temp., T (°C)

10 15 20 30 40 a b (10−2) c (10−3)

0 833.12 832.62 831.87 829.74 826.95 833.7000 −1.3590 −3.8670 0.9999

5 834.75 834.25 833.49 831.36 828.57 835.3000 −1.5110 −3.8440 0.9999
10 837.39 836.89 836.14 834.00 831.20 837.9000 −1.3050 −3.8920 0.9999
15 839.23 838.72 837.97 835.83 833.01 839.8000 −1.2750 −3.9140 0.9999
20 841.87 841.37 840.61 838.46 835.64 842.4000 −1.3070 −3.9190 0.9999
25 843.91 843.61 842.64 840.49 837.66 844.5000 −1.2100 −4.0060 0.9992
50 854.91 853.99 853.62 851.44 848.58 855.4000 −1.8230 −3.7700 0.9982
75 865.90 865.18 864.61 862.40 859.49 866.4000 −1.3170 −3.9800 0.9996
100 876.90 876.37 875.59 873.35 870.41 877.5000 −1.3080 −4.0900 0.9999

Calculated, ρ (kg/m3) Absolute errors (%)

Temp., T (°C) Temp., T (°C)

10 15 20 30 40 10 15 20 30 40
833.1774 832.6261 831.8814 829.8120 826.9692 0.0069 0.0007 0.0014 0.0087 0.0023
834.7645 834.2084 833.4602 831.3871 828.5452 0.0017 0.0050 0.0036 0.0033 0.0030
837.3803 836.8285 836.0822 834.0057 831.1508 0.0012 0.0073 0.0069 0.0007 0.0059
839.2811 838.7281 837.9794 835.8949 833.0276 0.0061 0.0010 0.0011 0.0078 0.0021
841.8774 841.3222 840.5710 838.4808 835.6068 0.0009 0.0057 0.0046 0.0025 0.0040
843.9784 843.4171 842.6556 840.5316 837.6064 0.0081 0.0229 0.0019 0.0049 0.0064
854.8407 854.2783 853.5274 851.4601 848.6388 0.0081 0.0338 0.0108 0.0024 0.0069
865.8703 865.3070 864.5446 862.4229 859.5052 0.0034 0.0147 0.0076 0.0027 0.0018
876.9602 876.3836 875.6024 873.4266 870.4328 0.0069 0.0015 0.0014 0.0088 0.0026

as expected, decrease with increasing temperature and there are similar The quadratic model:
trends for all fuels and the blends in the studied temperature range.
2
According to the distributions of measured data, the following linear, ρ ¼ ρðTÞ ¼ a þ bT þ cT ð5Þ
quadratic and exponential models were tried to represent the variations
of densities with temperature. The exponential model:
The linear model:
bT dT
ρ ¼ ρðTÞ ¼ ae þ ce ð6Þ

ρ ¼ ρðTÞ ¼ a þ bT ð4Þ where T is temperature in ° C and a, b, c and d are regression constants.

Table 6
For different biodiesel percentages, the measured densities, calculated densities from Eq. (6), errors between measured and calculated densities, regression constants and correlation
coefficients.

Blend X (%) Measured, ρ (kg/m3) Regression constants R

Temp., T (°C)

10 15 20 30 40 a b (10−4) c d (10−2)

0 833.12 832.62 831.87 829.74 826.95 854.0000 −5.9540 −20.7800 −2.7420 1.0000
5 834.75 834.25 833.49 831.36 828.57 855.0000 −5.8630 −20.1300 −2.7940 1.0000
10 837.39 836.89 836.14 834.00 831.20 857.7000 −5.8740 −20.2500 −2.8070 1.0000
15 839.23 838.72 837.97 835.83 833.01 865.4000 −6.5200 −26.0000 −2.3680 1.0000
20 841.87 841.37 840.61 838.46 835.64 862.5000 −5.9000 −20.5400 −2.7920 1.0000
25 843.91 843.61 842.64 840.49 837.66 850.7000 −3.7690 −8.0670 −8.1070 0.9997
50 854.91 853.99 853.62 851.44 848.58 −0.8166 502.1000 856.7000 −0.0060 0.9984
75 865.90 865.18 864.61 862.40 859.49 −32.1100 134.4000 898.7000 0.0435 0.9996
100 876.90 876.37 875.59 873.35 870.41 901.3000 −6.2210 −24.2900 −2.5560 1.0000

Calculated, ρ (kg/m3) Absolute errors (%)

Temp., T (°C) Temperatures, T (°C)

10 15 20 30 40 10 15 20 30 40
833.1339 832.6341 831.8827 829.7529 826.9622 0.0017 0.0017 0.0015 0.0016 0.0015
834.7787 834.2754 833.5206 831.3869 828.5980 0.0034 0.0030 0.0037 0.0032 0.0034
837.3827 836.8848 836.1319 833.9942 831.1937 0.0009 0.0006 0.0009 0.0007 0.0008
839.2581 838.7508 837.9968 835.8596 833.0387 0.0033 0.0037 0.0032 0.0035 0.0034
841.8900 841.3886 840.6309 838.4794 835.6600 0.0024 0.0022 0.0025 0.0023 0.0024
843.9136 843.5131 842.7173 840.4266 837.6560 0.0004 0.0115 0.0092 0.0075 0.0005
854.8370 854.1951 853.4435 851.4765 848.5616 0.0085 0.0240 0.0207 0.0043 0.0022
865.8888 865.3011 864.5403 862.4489 859.5053 0.0013 0.0140 0.0081 0.0057 0.0018
876.8990 876.3740 875.5870 873.3524 870.4109 0.0001 0.0005 0.0003 0.0003 0.0001
 M. Gülüm, A. Bilgin / Fuel Processing Technology 134 (2015) 456–464 461

Tables 4–6 list the measured and calculated (from Eqs. (4)–(6)) den- fitted to the experimental density values and compared to each other
sities of the blends and pure fuels, % errors between them, regression by determining the absolute errors and correlation coefficients. The
constants and correlation coefficients. For linear and exponential tried surface models are given in the following:
models, the maximum errors were computed as 0.0654% and 0.0240%, The linear surface model:
respectively. The R values are between 0.9870–0.9883 and 0.9984–
1.0000 for the linear and exponential models, respectively. The ρ ¼ ρðT; XÞ ¼ a þ bT þ cX ð7Þ
maximum error and correlation coefficients of the quadratic model lie
between these values. According to these results, the worst model is The surface model quadratic with respect to T:
linear while the best agreement of the calculated and measured density
2
values is captured by the exponential model for the investigated ρ ¼ ρðT; XÞ ¼ a þ bT þ cX þ dT þ eTX ð8Þ
temperature ranges.
The surface model linear with respect to X and exponential with
3.2. Two-dimensional surface models respect to T:

bT
In this section, two-dimensional surface models were derived to ρ ¼ ρðT; XÞ ¼ ae þ cX ð9Þ
make quick estimates of densities for a given X and a specific T simulta-
neously. As mentioned previously, the relationship between density where ρ is density in kg/m3, a, b, c, d and e are regression constants.
and biodiesel percentage was linear in characteristic, while, linear,
quadratic and exponential models were tried to represent changes of
densities with temperature, which may have nonlinear characteristics. Table 8
In light of this knowledge, new two-dimensional surface models were For different biodiesel percentages and temperatures, the measured densities, calculated
densities from Eq. (8), errors between measured and calculated densities, regression
constants and correlation coefficient.

Table 7 Temp. Blend Measured Calculated Absolute Regression Correlation

For different biodiesel percentages and temperatures, the measured densities, calculated T (°C) X (%) ρ (kg/m3) ρ (kg/m3) errors constants coefficient
densities from Eq. (7), errors between measured and calculated densities, regression con- (%)
stants and correlation coefficient.
10 0 833.12 832.8981 0.0266
Temp.T Blend Measured Calculated Absolute Regression Correlation 5 834.75 835.0979 0.0417
(°C) X (%) ρ (kg/m3) ρ (kg/m3) errors (%) constants coefficient 10 837.39 837.2977 0.0110
15 839.23 839.4975 0.0319
10 0 833.12 833.3890 0.0323
20 841.87 841.6972 0.0205
5 834.75 835.5830 0.0998
25 843.91 843.8970 0.0015
10 837.39 837.7770 0.0462
50 854.91 854.8960 0.0016
15 839.23 839.9710 0.0883
75 865.90 865.8949 0.0006
20 841.87 842.1650 0.0350
100 876.90 876.8938 0.0007
25 843.91 844.3590 0.0532
15 0 832.62 832.3532 0.0320
50 854.91 855.3290 0.0490
5 834.25 834.5508 0.0361 a = 833.4000
75 865.90 866.2990 0.0461
10 836.89 836.7485 0.0169 b = −1.0990 × R = 0.9999
100 876.90 877.2690 0.0421
15 0 832.62 832.3335 0.0344 10−2
5 834.25 834.5275 0.0333 15 838.72 838.9462 0.0270 c = 0.4408
10 836.89 836.7215 0.0201 a = 835.5000 20 841.37 841.1439 0.0269 d = −3.9200 ×
15 838.72 838.9155 0.0233 b = −0.2111 R = 0.9996 10−3
20 841.37 841.1095 0.0310 c = 0.4388 25 843.61 843.3415 0.0318 e = −8.4270 ×
25 843.61 843.3035 0.0363 10−5
50 853.99 854.2735 0.0332 50 853.99 854.3299 0.0398
75 865.18 865.2435 0.0073 75 865.18 865.3183 0.0160
100 876.37 876.2135 0.0179 100 876.37 876.3067 0.0072
20 0 831.87 831.2780 0.0712 20 0 831.87 831.6122 0.0310
5 833.49 833.4720 0.0022 5 833.49 833.8078 0.0381
10 836.14 835.6660 0.0567 10 836.14 836.0033 0.0163
15 837.97 837.8600 0.0131 15 837.97 838.1989 0.0273
20 840.61 840.0540 0.0661 20 840.61 840.3945 0.0256
25 842.64 842.2480 0.0465 25 842.64 842.5901 0.0059
50 853.62 853.2180 0.0471 50 853.62 853.5679 0.0061
75 864.61 864.1880 0.0488 75 864.61 864.5458 0.0074
100 875.59 875.1580 0.0493 100 875.59 875.5237 0.0076
30 0 829.74 829.1670 0.0691 30 0 829.74 829.5423 0.0238
5 831.36 831.3610 0.0001 5 831.36 831.7337 0.0450
10 834.00 833.5550 0.0534 10 834.00 833.9250 0.0090
15 835.83 835.7490 0.0097 15 835.83 836.1164 0.0343
20 838.46 837.9430 0.0617 20 838.46 838.3077 0.0182
25 840.49 840.1370 0.0420 25 840.49 840.4991 0.0011
50 851.44 851.1070 0.0391 50 851.44 851.4559 0.0019
75 862.40 862.0770 0.0375 75 862.40 862.4127 0.0015
100 873.35 873.0470 0.0347 100 873.35 873.3695 0.0022
40 0 826.95 827.0560 0.0128 40 0 826.95 826.6884 0.0316
5 828.57 829.2500 0.0821 5 828.57 828.8755 0.0369
10 831.20 831.4440 0.0294 10 831.20 831.0627 0.0165
15 833.01 833.6380 0.0754 15 833.01 833.2498 0.0288
20 835.64 835.8320 0.0230 20 835.64 835.4370 0.0243
25 837.66 838.0260 0.0437 25 837.66 837.6241 0.0043
50 848.58 848.9960 0.0490 50 848.58 848.5599 0.0024
75 859.49 859.9660 0.0554 75 859.49 859.4956 0.0007
100 870.41 870.9360 0.0604 100 870.41 870.4313 0.0024
462 M. Gülüm, A. Bilgin / Fuel Processing Technology 134 (2015) 456–464

Tables 7–9 show the regression constants, measured densities,

calculated densities from Eqs. (7)–(9), and % errors between them.
The maximum errors and R values of Eqs. (7) and (8) are 0.0998%,
0.0450% and 0.9996, 0.9999, respectively. The maximum error and the
correlation coefficient of Eq. (9) lie between these values. Consequently,
the best fit for variations of densities with X and T simultaneously is
obtained by Eq. (8). Plots of changes of constant density lines for
fuel blends as functions of T and X calculated from these models are
given in Figs. 3–5. According to Eq. (8) by which changes of densities
are well correlated, there exist two characteristic regions where
constant density lines have different gradients, as shown in Fig. 4. At
the higher gradient regions, corresponding to the higher temperature
and lower biodiesel percentage area, less temperature changes are
required for a unit change of density for a given blending percentage,
while higher temperature changes are needed at lower gradient
regions, corresponding to the lower temperature and higher biodiesel
percentage area.

Table 9
For different biodiesel percentages and temperatures, the measured densities, calculated
densities from Eq. (9), errors between measured and calculated densities, regression con-
stants and correlation coefficient.

Temp. Blend Measured Calculated Absolute Regression Correlation

T (°C) X (%) ρ (kg/m3) ρ (kg/m3) errors constants coefficient
(%) Fig. 3. Variations of constant density lines of fuel blends as functions of temperature and
biodiesel percentage simultaneously calculated from Eq. (7).
10 0 833.12 833.3780 0.0310
5 834.75 835.5720 0.0985
10 837.39 837.7660 0.0449
15 839.23 839.9600 0.0870 3.3. Effects of biodiesel percentage on flash point temperature and higher
20 841.87 842.1540 0.0337 heating value
25 843.91 844.3480 0.0519
50 854.91 855.3180 0.0477
75 865.90 866.2880 0.0448 Flash point temperature is the minimum temperature at which
100 876.90 877.2580 0.0408 vapor pressure of the hydrocarbon is sufficient to produce the vapor
15 0 832.62 832.3191 0.0361
5 834.25 834.5131 0.0315 a = 835.5000
10 836.89 836.7071 0.0219 b = −2.5430 × R = 0.9997
10−4
15 838.72 838.9011 0.0216 c = 0.4388
20 841.37 841.0951 0.0327
25 843.61 843.2891 0.0380
50 853.99 854.2591 0.0315
75 865.18 865.2291 0.0057
100 876.37 876.1991 0.0195
20 0 831.87 831.2614 0.0732
5 833.49 833.4554 0.0042
10 836.14 835.6494 0.0587
15 837.97 837.8434 0.0151
20 840.61 840.0374 0.0681
25 842.64 842.2314 0.0485
50 853.62 853.2014 0.0490
75 864.61 864.1714 0.0507
100 875.59 875.1414 0.0512
30 0 829.74 829.1502 0.0711
5 831.36 831.3442 0.0019
10 834.00 833.5382 0.0554
15 835.83 835.7322 0.0117
20 838.46 837.9262 0.0637
25 840.49 840.1202 0.0440
50 851.44 851.0902 0.0411
75 862.40 862.0602 0.0394
100 873.35 873.0302 0.0366
40 0 826.95 827.0444 0.0114
5 828.57 829.2384 0.0807
10 831.20 831.4324 0.0280
15 833.01 833.6264 0.0740
20 835.64 835.8204 0.0216
25 837.66 838.0144 0.0423
50 848.58 848.9844 0.0477
75 859.49 859.9544 0.0540
Fig. 4. Variations of constant density lines of fuel blends as functions of temperature and
100 870.41 870.9244 0.0591
biodiesel percentage simultaneously calculated from Eq. (8).
 M. Gülüm, A. Bilgin / Fuel Processing Technology 134 (2015) 456–464 463

Table 10
For different biodiesel percentages, the measured flash point temperatures, calculated
flash point temperatures from Eq. (10), errors between measured and calculated flash
point temperatures, regression constants and correlation coefficient.

Blend Measured a b c R Calculated Absolute

X (%) flash point flash point errors
temperature temperature (%)
TFP (°C) TFP (°C)

0 63 65.4800 0.1730 0.0084 0.9969 65.4800 3.9365

5 68 66.5550 2.1250
10 69 68.0500 1.3768
15 70 69.9650 0.0500
20 72 72.3000 0.4167
25 75 75.0550 0.0733
50 99 95.1300 3.9091
75 120 125.7050 4.7542
100 169 166.7800 1.3136

Table 10 lists the measured flash point temperatures, calculated

Fig. 5. Variations of constant density lines of fuel blends as functions of temperature and
biodiesel percentage simultaneously calculated from Eq. (9).
needed for spontaneous ignition of the fuel with air with the presence of
an external source, i.e., spark or flame [28]. The higher the flash point
the safer is the fuel regarding the storage, transport and handling [29].
Fig. 6 shows the effects of biodiesel percentage in blend on the flash
point temperatures of the blends. As can be seen in this figure, flash
point temperature quadratically increases with increasing biodiesel
content. Therefore, the quadratic model is used for representing the var-
iation of flash point temperature with respect to biodiesel percentage
as:
2 higher heating values from Eq. (11), errors between them, regression
T FP ¼ T FP ðXÞ ¼ a þ bX þ cX
where X is biodiesel percentage in blend, TFP is flash point temperature
in °C, a, b and c are regression constants.
flash point temperatures from Eq. (10), errors between them, regres-
sion constants and correlation coefficient. The maximum error between
calculated and measured values, and the correlation coefficient were
obtained as 4.7542% and 0.9969, respectively.
Heating value or heat of combustion is the amount of chemical
energy released by the combustion of a unit mass of fuel [30]. Heating
value is of great importance since it also affects the transportation and
storage of fuel, as with flash point temperature. Fuel having higher
heating value can provide longer transportation distance with smaller
storage fuel tank, and also gives higher power output from smaller
engine. The variation of higher heating values of blends with biodiesel
percentage is given in Fig. 7. As was expected, the higher heating values
decrease in direct proportion to biodiesel percentage. For this reason,
HHV values are found to be well correlated by linear model:
HHV ¼ HHVðXÞ ¼ a þ bX ð11Þ
In Eq. (11), X is biodiesel percentage in blend, HHV is in kJ/kg, a and b
are regression constants.
Table 11 shows the measured higher heating values, calculated
ð10Þ
constants and correlation coefficient. The maximum error and the
correlation coefficient were computed as 0.6529% and 0.9979.

Fig. 6. Variations of flash point temperatures of fuel blends with respect to biodiesel Fig. 7. Variations of higher heating values of fuel blends with respect to biodiesel percentage
percentage in blend. in blend.
464 M. Gülüm, A. Bilgin / Fuel Processing Technology 134 (2015) 456–464

Table 11 References
For different biodiesel percentages, the measured higher heating values, calculated higher
heating values from Eq. (11), errors between measured and calculated higher heating [1] M. Alam, J. Song, R. Acharya, A. Boehman, K. Miller, Combustion and emissions
values, regression constants and correlation coefficient. performance of low sulfur, ultra low sulfur and biodiesel blends in a DI diesel
engine, SAE Technical Paper Series, 2004 (2004-01-3024).
Blend Measured a b R Calculated Absolute [2] M. Canakci, A.N. Ozsezen, E. Arcaklioğlu, A. Erdil, Prediction of performance and
X (%) higher higher errors exhaust emissions of a diesel engine fueled with biodiesel produced from waste
heating heating (%) frying palm oil, Expert Syst. Appl. 36 (5) (2009) 9268–9280.
value HHV value HHV [3] M.C. Math, S.P. Kumar, S.V. Chetty, Technologies for biodiesel production from used
(kJ/kg) (kJ/kg) cooking oil—a review, Energy Sustain. Dev. 14 (4) (2010) 339–345.
[4] B.D. Wahlen, R.M. Willis, L.C. Seefeldt, Biodiesel production by simultaneous
0 45,950 4.5650 × 104 −57.9600 0.9979 45,650.0000 0.6529 extraction and conversion of total lipids from microalgae, cyanobacteria, and wild
5 45,368 45,360.2000 0.0172 mixed-cultures, Bioresour. Technol. 102 (3) (2011) 2724–2730.
10 44,984 45,070.4000 0.1921 [5] G. Knothe, Dependence of biodiesel fuel properties on the structure of fatty acid
15 44,643 44,780.6000 0.3082 alkyl esters, Fuel Process. Technol. 86 (10) (2005) 1059–1070.
20 44,399 44,490.8000 0.2068 [6] B.K. Barnwal, M.P. Sharma, Prospects of biodiesel production from vegetable oils in
25 44,205 44,201.0000 0.0090 India, Renew. Sust. Energ. Rev. 9 (4) (2005) 363–378.
50 42,700 42,752.0000 0.1218 [7] Z. Helwani, M.R. Othman, N. Aziz, W.J.N. Fernando, J. Kim, Technologies for produc-
tion of biodiesel focusing on green catalytic techniques: a review, Fuel Process.
75 41,310 41,303.0000 0.0169
Technol. 90 (12) (2009) 1502–1514.
100 39,930 39,854.0000 0.1903
[8] G. Vicente, M. Martı́nez, J. Aracil, Integrated biodiesel production: a comparison of
different homogeneous catalysts systems, Bioresour. Technol. 92 (2004) 297–305.
[9] I. Barabás, A. Todorut, D. Băldean, Performance and emission characteristics of an CI
engine fueled with diesel–biodiesel–bioethanol blends, Fuel 89 (12) (2010)
4 . Conclusions 3827–3832.
[10] M. Diasakou, A. Louloudi, N. Papayannakos, Kinetics of the non-catalytic
In this study, the effects of biodiesel percentage and temperature on transesterification of soybean oil, Fuel 77 (12) (1998) 1297–1302.
[11] S. Kalligeros, F. Zannikos, S. Stournas, E. Lois, G. Anastopoulos, C. Teas, F.
the densities of blends of produced corn oil biodiesel with commercially Sakellaropoulos, An investigation of using biodiesel/marine diesel blends on the
available Ultra Force Euro Diesel fuel were experimentally investigated. performance of a stationary diesel engine, Biomass Bioenergy 24 (2) (2003)
New one- and two-dimensional regression models have been devel- 141–149.
[12] R.O. Dunn, Effect of antioxidants on the oxidative stability of methyl soyate (biodiesel),
oped for estimating densities at various temperatures and biodiesel
Fuel Process. Technol. 86 (10) (2005) 1071–1085.
percentages. Flash point temperatures and higher heating values of [13] A. Demirbas, Importance of biodiesel as transportation fuel, Energy Policy 35 (9)
the corn oil biodiesel–diesel fuel blends were also measured. In order (2007) 4661–4670.
[14] J.V. Gerpen, Biodiesel processing and production, Fuel Process. Technol. 86 (10)
to predict these properties, some equations as a function of biodiesel
(2005) 1097–1107.
percentage in blend were also derived. According to the results, the [15] E. Alptekin, M. Canakci, Determination of the density and the viscosities of
following conclusions can be drawn: biodiesel–diesel fuel blends, Renew. Energy 33 (12) (2008) 2623–2630.
In lineal models, linear and exponential ones better represent [16] M.S. Shehata, Emissions, performance and cylinder pressure of diesel engine fuelled
by biodiesel fuel, Fuel 112 (2013) 513–522.
density–biodiesel percentage and density–temperature relationship, re- [17] J.P. Szybist, A.L. Boehman, J.D. Taylor, R.L. McCormick, Evaluation of formulation
spectively. Correlation coefficients (R) and maximum errors are 0.9999, strategies to eliminate the biodiesel NOx effect, Fuel Process. Technol. 86 (10)
0.0421% and 0.9984–1.0000, 0.0240% for the linear and exponential (2005) 1109–1126.
[18] D.Y.C. Leung, Y. Guo, Transesterification of neat and used frying oil: optimization for
models. biodiesel production, Fuel Process. Technol. 87 (10) (2006) 883–890.
In two-dimensional surface models, Eq. (8) shows the highest de- [19] B. Tesfa, R. Mishra, F. Gu, N. Powles, Prediction models for density and viscosity of
gree of accuracy with correlation coefficient of 0.9999 for representing biodiesel and their effects on fuel supply system in CI engines, Renew. Energy 35
(12) (2010) 2752–2760.
change of density with temperature and biodiesel percentage simulta- [20] W. Yuan, A.C. Hansen, Q. Zhang, Predicting the physical properties of biodiesel for
neously. The maximum error between the measured and calculated combustion modeling, Am. Soc. Agric. Eng. 46 (6) (2003) 1487–1493.
density values is 0.0450% for this model. [21] M.E. Tat, J.H.V. Gerpen, The specific gravity of biodiesel and its blends with diesel
fuel, J. Am. Oil Chem. Soc. 77 (2) (2000) 115–119.
[22] P. Benjumea, J. Agudelo, A. Agudelo, Basic properties of palm oil biodiesel–diesel
In the two-dimensional constant density line plot (Fig. 4) obtained blends, Fuel 87 (10–11) (2008) 2069–2075.
by Eq. (8), there exist two characteristic regions. At the higher gradi- [23] R.E. Tate, K.C. Watts, C.A.W. Allen, K.I. Wilkie, The densities of three biodiesel fuels at
temperatures up to 300 ° C, Fuel 85 (7–8) (2006) 1004–1009.
ent regions, less temperature changes are required for a unit change [24] M. Gülüm, Experimental investigation of the effect of various production parame-
of density for a given blending percentage, while higher temperature ters on the some fuel properties of produced biodiesels from corn and hazelnut
oils(Master's Thesis) Department of Mechanical Engineering, Karadeniz Technical
changes are needed at lower gradient regions. University, Trabzon, TURKEY, 2014.
[25] M. Gülüm, A. Bilgin, The effects of various production parameters on the some fuel
Quadratic and linear models as a function of biodiesel percentage are properties of produced biodiesel from corn oil by using NaOH, Seventh Automotive
Technologies Congress, Bursa (TURKEY), May 26–27 2014, 2014, pp. 1–8.
proposed for estimating flash point temperatures and higher heating [26] J.P. Holman, Experimental Methods for Engineers, seventh ed. McGraw-Hill, New
values of the pure fuels and fuel blends, respectively. Maximum errors York, 2001.
and correlation coefficients (R) are 4.7542%, 0.9969 and 0.6529%, [27] S.C. Chapra, P.R. Canale, Numerical Methods for Engineers with Software and
Programming Applications, third Ed. Mc-Graw-Hill, America, 1998.
0.9979 for the quadratic and linear models. [28] R. Ghanei, G.R. Moradi, R. TaherpourKalantari, E. Arjmandzadeh, Variation of
physical properties during transesterification of sunflower oil to biodiesel as an
approach to predict reaction progress, Fuel Process. Technol. 92 (8) (2011)
1593–1598.
Acknowledgments [29] Z. Al-Hamamre, A. Al-Salaymeh, Physical properties of (jojoba oil + biodiesel),
(jojoba oil + diesel) and (biodiesel + diesel) blends, Fuel 123 (2014) 175–188.
The authors express their gratitude to Karadeniz Technical University [30] A.E. Atabani, A.S. Silitonga, I.A. Badruddin, T.M.I. Mahlia, H.H. Masjuki, S. Mekhilef, A
comprehensive review on biodiesel as an alternative energy resource and its
Scientific Research Projects Fund for financial support received (Project
characteristics, Renew. Sust. Energ. Rev. 16 (4) (2012) 2070–2093.
No: 9745).