Fuel 209 (2017) 224–231

Contents lists available at ScienceDirect

Fuel
journal homepage: www.elsevier.com/locate/fuel

Full Length Article

Novel concepts for onboard determination of fuel quality in plug-in hybrid MARK
cars
⁎
M. Eskinera, , F. Ammera, D. Thena, J. Staufenbiela, M. Rossnera, O. Schrödera, J. Krahla,b,c
a
Coburg University of Applied Sciences and Arts, Friedrich-Streib-Straβe 2, D-96450 Coburg, Germany
b
Ostwestfalen-Lippe University of Applied Sciences, Liebigstraβe 87, D-32657 Lemgo, Germany
c
Fuels Joint Research Group, Germany

A R T I C L E I N F O A B S T R A C T

Keywords: In consideration of the CO2 saving potentials of biofuels, a quota of biodiesel in fossil diesel fuel is established in
Fuel a number of countries all over the world. However, primary investigations on fossil diesel fuel/biodiesel blends
Biofuel showed a certain tendency to form precipitates, which can be considered as oligomers. They result from aging of
Fuel design biodiesel. However, the tendency of fuel blends to form precipitates during long term storage could jeopardize
Fuel aging
the operational safety in plug-in hybrid vehicles (PHEV). In PHEV, the combustion engine is only used when the
Fuel sensors
power supply from the battery is too low. On short distance rides, the combustion engine is not necessary,
Dielectric spectroscopy
Plug-in hybrid vehicle because of frequent battery recharging. Consequently, the storage time of the fuel in the tank is much longer than
usual. For this reason the fuel long-term stability in PHEV is of particular importance. A mixture design based on
a simplex-lattice design fourth-degree was used to obtain variable fuel composition. Aged samples containing
linoeleic acid methyl ester (C18:2) formed precipitates that could sediment at the bottom of the vials. After a
waiting period of 30 and 300 s of sedimentation time, samples of the upper phase of the vial were taken and
investigated by dielectric- and fourier transform infrared spectroscopy (FTIR) to determine the stability of the
suspension. For reasons of comparison size-exclusion chromatography (SEC) was used too. It was shown, that the
stability of the suspension is very much dependent on the fuel mixture. The lowest stability was found in the
range of 15 to 22.5% C18:2. Finally, a fourth-order regression model for the real part of the permittivity was
found, which could estimate the composition of fresh fuels.

1. Introduction fuels or as an addition to fossil fuels [5], also aromatic free fuel are
introduced in the market. Furthermore, fuel degradation must be con-
Variable diesel fuel compositions, e.g. through different biodiesel sidered: Firstly, aged fuels differ in their emission behavior [6]; sec-
qualities, different biofuel mixing rates and different proportion of ondly, fuel degradation, especially blends, showed a certain tendency to
aromatics in fossil fuels pose new challenges for commercial vehicle form precipitates which can be considered as oligomers [6–8]. The
manufacturers. The motor management in vehicles is not designed for cause of undesired fuel instability are unsaturated hydrocarbons: They
such a fuel variability. Diesel fuels for example are highly complex undergo autoxidation forming alkyl and peroxy radicals [9,10] which
mixtures containing linear alkanes (n-alkane), branched alkanes (iso- can react with oxygen to form peroxide or hydroperoxide. In the next
alkane) in a carbon range of C8 to C22, polyaromatic hydrocarbons and reaction step these peroxides and hydroperoxides degrade into low
cycloalkanes [1]. Investigations with gas chromatography show a molecular oxidation products such as aldehydes, alcohols, ketones,
normal distribution of n-alkane in the range of n-C11 to n-C-14 [2]. acids and high molecular oligomers [8,11,12]. Problems due to aged
Additionally, different quotas of biodiesel in fossil diesel fuel is estab- fuels can occur at the fuel pump, injection systems or exhaust after-
lished in several countries all over the world [3]. In Germany for ex- treatment. Fuel pump and injection system are very sensitive to an
ample, a proportion of max. 7% v/v biodiesel is added to fossil diesel increase in viscosity due to oligomerization. The increased viscosity
fuel [4]. Aggravated by the fact that modern fuels like HVO (hydro- leads to poorer nebulization during injection. This can lead to an in-
treated vegetable oil) and GTL (gas-to-liquid) – a mixture of mainly crease of soot or unburned fuel due to an incomplete combustion. En-
branched and unbranched alkanes – are becoming more popular as neat abling factors for fuel aging are oxidative and thermal stress. Since the

⁎
Corresponding author.
E-mail address: [email protected] (M. Eskiner).
URL: http://www.fuels-jrg.de (J. Krahl).

http://dx.doi.org/10.1016/j.fuel.2017.07.102
Received 9 May 2017; Received in revised form 3 July 2017; Accepted 26 July 2017
0016-2361/ © 2017 Elsevier Ltd. All rights reserved.
M. Eskiner et al. Fuel 209 (2017) 224–231

Table 1
Mixture design of several fuels with upper and lower bounds according to the simplex-lattice design fourth-degree.

Sample Nr. Design matrix Response marks Operational matrix

Alkane (X1) Aromatics (X2) C18:2 (X3) Alkane Aromatics C18:2

1 1 0 0 y1 1 0 0
2 0 1 0 y2 0.7 0.3 0
3 0 0 1 y3 0.7 0 0.3
4 0.50 0.50 0 y12 0.85 0.15 0
5 0.5 0 0.5 y13 0.85 0 0.15
6 0 0.5 0.5 y23 0.7 0.15 0.15
7 0.75 0.25 0 y1112 0.925 0.075 0
8 0.25 0.75 0.0 y1222 0.775 0.225 0
9 0.75 0 0.25 y1113 0.925 0 0.075
10 0.25 0 0.75 y1333 0.775 0 0.225
11 0 0.75 0.25 y2223 0.7 0.225 0.075
12 0 0.25 0.75 y2333 0.7 0.075 0.225
13 0.5 0.25 0.25 y1123 0.85 0.075 0.075
14 0.25 0.5 0.25 y1223 0.775 0.15 0.075
15 0.25 0.25 0.50 y1233 0.775 0.075 0.15

introduction of plug-in hybrid vehicles (PHEV), an additional factor εr (ω,T ) = εr′ (ω,T )−iεr″ (ω,T ) (2)
could be added – the long-term stability of fuels. In PHEV, the com-
Losses in the dielectric caused by polarization and conductivity ef-
bustion engine is only used when the power supply from the battery is
fects were described with the imaginary part of Eq. (2). The imaginary
too low. On short distance rides, which are the majority of all driven
part can be subdivided into two terms as shown in Eq. (3). There is a
kilometres (about 80% below 50 km) [13] the combustion engine is not
strong dependence of the losses with the frequency as shown in Eq. (3):
necessary, because of frequent battery recharging. Consequently, the
[17]
storage time of the fuel in the tank is much longer than usual. For this
reason the fuel long-term stability in PHEV is of particular importance. σ
εr″ (ω,T ) = εr″,pol (ω,T ) +
In the long term, an onboard sensor system must be able to de- ωε0 (3)
termine the fuel composition as well as the degradation degree in fuels. εr″,pol (ω,T ) is associated with losses due to polarization, whereas the
Series-production PHEV equipped with diesel engines are already second term is directly proportional to conductivity σ . When frequency
available in the market or planned for the future (Volvo V60, Audi Q7 is low enough, the second part of the term in Eq. (3) is dominating and
e-tron quattro, Peugeot 508 RXH Hybrid4, Range Rover SDV6 Hybrid conductivity is measureable. ε0 is the permittivity constant of free space
Diesel). (8.854∗10−12 F/m).
In these investigations, therefore, we introduce a direct measuring
technique based on dielectric spectroscopy, focusing the in situ de- 3. Methods and materials
termination of degradation products in fossil diesel fuels. Besides a
sensor in lab scale, a smart micro sensor was developed in a laboratory 3.1. Sample preparation
test setup, too. In principle it is suited for mass production and could be
applied e.g. as fuel tank sensor in PHEV. In this case the fuel must be A basic solution of n-Alkanes containing n-undecane, n-dodecane, n-
monitored regarding the most undesired aging products that lead to pentadacane and n-hexadacane (volumetric ratio of 0.98/1/0.79/0.63)
filter blocking and injector fouling. To fulfill the variety of fuels, a was mixed. This basic solution was mixed with Shellsol® T, a neat iso-
mixture experiment based on the simplex-lattice design fourth-degree alkane mixture of branched alkanes (carbon chain length 11 and 12).
was used [14]. The ratio of iso-/n-Alkane was 0.3/0.7 (degree of branching 30%). In
the following these solution is called alkane. As aromatics Shellsol®-A
2. Theoretical background was used, a mixture auf several aromatics (carbon chain length 9).
Representative for biodiesel linoeleic acid methyl ester was used, which
2.1. Theory of dielectric spectroscopy is in many biodiesel feedstocks contained. It has two double bounds
with a bis-allylic position and a carbon chain length of 18. In the fol-
Dielectric parameters were performed by applying an alternating lowing it is called C18:2.
voltage between the electrodes at different frequencies. A change in the Table 1 shows the matrix for the mixture design based on a simplex-
liquid composition (e.g. a fuel during its aging) between electrodes lattice design fourth-degree [14,18–20]. As the resulting mixtures from
leads to a change in its primary dielectric parameters. These changes the design matrix are not all reasonable for real fuel blends, limitations
can be described by several polarization effects like displacement po- have been implemented (alkane: 70 to 100%; aromatics: 0 to 30%;
larization, orientation polarization or interfacial polarization [15]. C18:2: 0 to 30%) thus making the operational matrix. All samples were
The most important polarization effect in the low frequency region accelerated aged by using the Rancimat setup described in DIN EN
can be described by the interfacial polarization. In this case, free charge 15751 (110 °C and 10 L/h air supply, 10 mL sample volume). The aging
carrier in the mixture can migrate to the electrode surface which in- duration was 10 h. To accelerate the aging procedure, no stabilizing
fluences the permittivity. additives were used.
The relative permittivity εr′ (ω,T ) is defined as the quotient of the
capacitance with a dielectric C (ω,T ) and the empty capacity C0 (ω,T ) 3.2. Mathematical model for the determination of permittivity in the simplex
(Eq. (1)), where ω is the angular frequeny and T the temperature: [16]
The response function of the mixture model was explained by using
C (ω,T )
εr′ (ω,T ) = following fourth-degree regression model (Eq. (4)) and its coefficients
C0 (ω,T ) (1)
(Eq. (5)) [21]. Xi is the variable and Xi Xj represents the interaction
Furthermore, the permittivity is a complex valued function (Eq. (2)) between the variables. The coefficient from eq. 5 are the regression

225
M. Eskiner et al. Fuel 209 (2017) 224–231

coefficients of the pure component (βi ) and its interaction. The fourth- samples were shaken. After a waiting period of 30 and 300 s of sedi-
degree regression model was used to predict unknown permittivity in mentation time, samples were taken from the upper phase of the vials
the simplex for fresh fuels for determining the composition of the fuels. and permittivity measurements were done.

ŷ= ∑ βi Xi + ∑ βij Xj Xi + ∑ γij Xi Xj (Xi −Xj )

1⩽i⩽q 1⩽i<j⩽q 1⩽i<j⩽q 3.4. Size-exclusion-chromatography (SEC)
+ ∑ δij Xi Xj (Xi −Xj )2 + ∑ βiijk Xi2 Xj Xk
1⩽i<j⩽q 1⩽i<j<k⩽q Oligomers from aged samples were analyzed via SEC (Agilent 1260
Infinity Quaternary LC System) with refractive index (RI) detection
+ ∑ βijjk Xi Xj2 Xk + ∑ βijkk Xi Xj Xk2
(Agilent 1260 Infinity Refractive Index Detector). (SEC-RI). The se-
1⩽i<j<k⩽q 1⩽i<j<k⩽q
paration was done with 3 × 30 cm PSS SDV (particle size 5 µm por-
+ ∑ βijkl Xi Xj Xk Xl osity; first column 50 Å; second column 100 Å; third column 1000 Å)
1 ⩽ i < j < k < l⩽ q (4) columns.
βi = yi Further parameters of the SEC-RI were:

• Temperature of RI detector and column: 45 °C

βij = 4yij −2yi −2yj

• Injection volume: 100 µL

8
γij = ( 3 )(−yi + 2yiiij −2yijjj + yj )
8
δij = ( 3 )(−yi + 4yiiij −6yij + 4yijjj −yi ) • Concentration: 100 mg analyte per mL tetrahydrofuran (THF)
• Flow rate: 0.5 mL/min
• Mobile phase: Tetrahydrofuran
8
βiijk = 32(3yiijk −yijjk −yijkk ) + ( 3 )(6yi −yj −yk )−16(yij + yik )
16
−( 3 )(5yiiij + 5yiiik −3yijjj −3yikkk −yjjjk −yjkkk )
A calibration curve was performed with polyethylene glycol (PEG)
8
βijjk = 32(3yijjk −yiijk −yijkk ) + ( 3 )(6yj −yi −yk )−16(yij + yjk ) (Agilent PEG calibration kit Part No.: PL2070-0100), with molar masses
16
−( 3 )(5yijjj + 5yjjjk −3yiiij −3yjkkk −yiiik −yikkk ) between 196 Da and 21300 Da. To verify the accuracy of the SEC-RI
8
system, an analytical standard of methyl oleate was measured. With
βijjk = 32(3yijkk −yiijk −yijjk ) + ( 3 )(6yk −yi −yj )−16(yik + yjk ) PEG as calibration, the molar mass of methyl oleate was determined to
16
−( 3 )(5yikkk + 5yikkk −3yiiik −3yjjjk −yiiij −yijjj) about 375 Da, while the real molar mass of methyl oleate is 296.45 Da.
In order to make quantitative statements with regard to the molecule
βijkl = 256yijkl −32(yiijk + yiijl + yiikl + yijjk + yijjl + yjjkl + yijkk + yikkl + yijll size, calibration standards and analytes must be identical. However,
32
+ yjkkl + yikll ) + ( 3 )(yiiij + yiiik + yiiil + yijjj + yjjjk + yjjjl + yikkk since precise oligomer structures are not known so far, an accurate
calibration could not be performed.
+ yjkkk + ykkkl + yilll + yjlll + yklll )
(5) 3.5. Fourier transform infrared spectroscopy (FTIR)

3.3. Laboratory test setup for permittivity measurement FTIR (Nicolet 6700 FT-IR Spectrometer Thermo Scientific with ATR)
examinations on aged samples were performed to determine the sta-
The capacitive sensor is based on a printed circuit board with in- bility of functional groups in the suspension. Here, the samples were
terdigital structure which is used to determine the dielectric parameters shaken too and after a waiting period of 30 and 300 s of sedimentation
(Fig. 1). Similar sensor principles were also mentioned in [22,23]. To time, samples were taken from the upper phase of the vials and the FTIR
avoid interaction with fuels, the copper surface was coated with gold. signal was measured.
The surface of the sensitive area can be covered with fuel. Since the
permittivity has a strong dependency on temperature, a constant tem- 4. Results and discussion
perature control of the dielectric is necessary. For this reason, a peltier
element, which is fixed below the printed circuit board, a PT100 sensor 4.1. Permittivity of fresh fuels
and a cooler (for cooling down the peltier element) are connected to a
temperature control unit (Belektronig HAT control). The sensor is Fig. 2 shows the real part of the permittivity for the fresh fuel
connected to an impedance analyzer (Keysight E4990A) that detects samples obtained by the fuel mixture design presented in Table 1.
capacitance and dissipation factor. Calibration was performed using Considering a constant amount of alkanes (e.g. 0.70; sample number 2,
cyclohexane as reference liquid. Detailed information on the procedure 11, 6, 12, 3), there is an increase in the real part of the permittivity from
can be found in [24–26]. εr′ = 2.13( ± 0.02) up to εr′ = 2.40( ± 0.02) due to the higher amount of
Permittivity measurements of fresh fuels in different mixtures as C18:2-Ester. The higher permittivity can be explained by the higher
shown in Table 1 were performed at 25 °C ( ± 0.1 °C). In order to be polarity of the C18:2-Ester: On the one hand side, there are highly
able to assess the stability of the suspension in the aged fuels, these electronegative oxygens atoms in the ester function attached to the
carbon atom, which pull the electrons of the bonding towards to the
oxygen atom and creates polarity. On the other hand side, delocalized
electrons in the double bounds between the carbon atoms in the ester
molecule leads to an increase of the polarity and permittivity. Con-
sidering a constant amount of C18:2 (e.g. sample number 9, 13, 14 and
11) it is obvious, that a higher amount of the aromatics leads to a slight
increase of the permittivity from εr′ = 2.12( ± 0.02) to εr′ = 2.20( ± 0.02) .
In this case as well delocalized electrons in the aromatics are mainly
responsible for the marginally higher polarity. All in all, the behavior of
the permittivity for the fresh fuels in the simplex was expected. It has to
be mentioned that a standard fossil diesel fuel with biodiesel proportion
in the range of 0 and 30% (v/v) has an permittivity in the range of 2.1
Fig. 1. Printed circuit board with interdigital structure.
to 2.5 (room temperature), [27,28].

226
M. Eskiner et al. Fuel 209 (2017) 224–231

Fig. 2. Real part of permittivity before aging for a ternary

fuel mixture containing alkane/aromatics/C18:2 in dif-
0.002.13 2.02 ferent ratios.
0.30
2 2.07
2.12
0.05
2.16
0.25
2.21
2.20 2.10
2.26
0.10 11 8
2.30
0.20

Ar
2.35
8:2

om
2.40
0.152.27
C1

ati
2.17 2.08
0.15

cs
6 14 4

0.20
0.10
2.32 2.24 2.15 2.05
0.25 12 15 13 7
0.05

0.302.40 2.30 2.22 2.12 2.02

0.00
3 10 5 9 1
0.70 0.75 0.80 0.85 0.90 0.95 1.00
Alkane

4.2. The formation of oligomers in fuels via SEC concentration of 7.5% of C18:2 (sample number 11, 14, 13 and 9) the
total area under the graph is between 77 and 124, the total area in-
Fig. 3 shows the SEC measurements on the example of sample creases to 217 to 273 for a concentration of 22.5% of C18:2 (sample
number 15, which contains the mixture of alkane, aromatics and C18:2 number 12 and 10). A multiple measurement for determination of the
in the volumetric ratio of 0.775/0.075/0.15. The molar mass ranges total area was performed for sample number 15 so that the error was
between 350 and 425 g/mol are those from the C18:2-Ester. Below estimated to be ± 30 for all other samples.
350 g/mol the mixture of alkane and aromatics are visible. The de-
crease of the C18:2 peak during aging is associated with an increase of 4.3. Stability of oxidation products in the mixture
the oligomer range (> 425 g/mol) due to its reaction to high-molecular
oxidation products. The results indicate that in reference to their molar Carbonyl bonds (CO) from esters and oxidation products like alde-
masses there were no preferable oligomers produced during the aging hydes, ketones, carboxylic acids can stretch in the FTIR-spectra from
process. Hence, it is difficult to detect a valid peak maximum with in- 1715 to 1746 cm−1 [6,8]. The lower the transmission degree, the more
creasing aging time. The highest detectable molar mass was about carbonyl bonds are present in the mixture. The values for the degree of
1500 g/mol. For comparative purposes the total area under the graphs CO-transmission can be found in Table 2. It must be added that no CO
for all aged sample were mathematically integrated. From Fig. 4 it stretch is observable for sample number 1, 2, 4, 7 and 8, because no
becomes clear that a higher concentration of C18:2-esters leads to an C18:2 were present and no oxidation products in these samples were
increase of the amount of oligomers in the mixtures. While at a constant formed. Therefore transmission is almost 100% for these samples. After
an idle period of 30 s, the transmission degree varies from 62.79%
( ± 1.05%) (Sample 3) to 94.15% ( ± 0.68%) (Sample 13). This scat-
Fresh fuel (Sample Nr. 15)
C18:2-ester tering is caused by the oxidation products and different proportion of
Aged fuel (Sample Nr. 15)
10 C18:2. It is also striking that the standard deviation after 30 s is rela-
tively high (highest value ± 5.81% for sample 10) compared to the
8 standard deviation with a duration of 300 s after shaking (highest
value ± 1.87% for sample 9). The main reason for this behavior is, that
rel W(log M) [-]

Alkanes and
6 aromatics there is a very unstable suspension due to the oxidation products. 300 s
after the sample were shaken, some oxidation products could sediment
4 at the bottom of the vials. As a result, in the mixture, only those oxi-
Formation of oligomers dation products remained on the upper phase, which could largely be
kept in suspension. Assuming that functional groups like CO were part
2
of oligomers, the differences in the value of the transmission degree
after a duration time of 30 and 300 s after shaking, should give in-
0
formation about the stability of precipitate in the fuels (see Table 2,
400 800 1200 1600 2000
column 6). The results from Table 2 are illustrated in the ternary dia-
gram (see Fig. 5). Despite the absolute standard deviation is consider-
Relative molar mass [g/mol]
able high, it is obvious that the degree of sedimentation for mixtures in
Fig. 3. Molar mass distribution for fresh and aged fuel on the example of sample number the range of about 0.15% of C18:2 seems to be at a highest level, which
15 – mixture of alkane, aromatics and C18:2 in the volumetric ratio of 0.8/0.1/0.1.
means that sedimentation in the region is very probably. This is in

227
M. Eskiner et al. Fuel 209 (2017) 224–231

Fig. 4. Illustration of the integrated total area under the

graphs obtained from the SEC measurement for a ternary
0.00 -6 0 fuel mixture containing alkane/aromatics/C18:2 in dif-
0.30 ferent ratios.
2 37
75
0.05
113
0.25
150
124 -0
187
0.10 11 8
225
0.20

Ar
263
8:2

om
300
0.15176
C1

ati
77 -0
0.15

cs
6 14 4

0.20
0.10
217 151 109 -3
0.25 12 15 13 7
0.05

0.30289 273 153 117 -1

0.00
3 10 5 9 1
0.70 0.75 0.80 0.85 0.90 0.95 1.00
Alkane

contrast to the SEC measurements shown in Fig. 4, where with higher these mixtures. In addition, it cannot be ruled out that volatile oxida-
C18:2 concentration the amount of oligomers is increasing too. Con- tion products could evaporate due to the used Rancimat setup. This
sequently, the stability of the suspension should decrease with higher would, however, affect all mixtures. For all other samples an increase in
concentration of C18:2, because more oligomerization occurs. Such an permittivity is observable due to oxidation products. Since the aging
antagonistic effect is already known and described in the literature products are predominantly polar and the permittivity correlates with
[6,8]. the polarity, a statement about the stability of the suspension by means
Again, insoluble precipitated in all aged samples were emulsified by of permittivity should be adequate. Therefore in column 5 (Table 3) the
shaking. After a duration of 30 and 300 s, permittivity measurements differences of the permittivity signal for all aged samples after shaking
were performed, similar to the procedure before. In Table 3 all per- them, were calculated after a duration time of 30 and 300 s. The higher
mittivity data are listed. Data for the fresh samples (column 3) are al- the value is, the higher high molecular oxidation products could settle
ready shown in Fig. 2. It can be assumed that under the given aging down on the bottom of the vials, which means a lower stability of the
conditions no polar oxidation products are formed in the mixture 1, 2, suspension. However, one cannot exclude that some of the low mole-
4, 7 and 8, which could be caused by the C18:2 oxidation, because these cular oxidation products were attracted by the high molecular ones, due
samples are without C18:2 proportion. Nevertheless, oxidation of the to interactions between them (e.g. hydrogen-bonding). Despite a de-
remaining admixture components cannot be ruled out in these mix- viation of data shown in Fig. 5 and Fig. 6 are singly recognizable, which
tures. However, the permittivity results show, that changes in the can be attributed to the standard deviation and only the detection of
permittivity under the given aging conditions are not measureable in CO-stretching, the data illustrated in Fig. 6 show at least a similar

Table 2
Carbonyl (CO)-Transmission at 1735 cm−1 via FTIR.

Sample nr. CO30 sec [Transmission Δ CO30 sec [Transmission CO300 sec [Transmission Δ CO300 sec [Transmission CO30sec - CO300sec Total Δ [Transmission
%] %] %] %] [Transmission%] %]

1 100 – 100 – 0 –
2 100 – 100 – 0 –
3 62.79 ± 1.05 63.11 ± 0.21 −0.32 ± 1.07
4 100 – 100 – 0 –
5 88.11 ± 4.09 93.00 ± 0.25 −4.89 ± 4.09
6 83.14 ± 2.89 92.40 ± 0.56 −9.26 ± 2.94
7 100 – 100 – 0 –
8 100 – 100 – 0 –
9 91.83 ± 3.07 96.17 ± 0.70 −4.34 ± 3.15
10 84.42 ± 5.81 91.54 ± 1.26 −7.12 ± 5.95
11 89.70 ± 1.82 92.93 ± 1.87 −3.23 ± 2.60
12 86.68 ± 0.49 88.20 ± 0.81 −1.52 ± 0.94
13 94.15 ± 0.68 96.49 ± 0.14 −2.33 ± 0.69
14 89.55 ± 3.39 94.51 ± 1.11 −4.95 ± 3.57
15 83.39 ± 4.71 92.92 ± 0.56 −9.53 ± 4.74

228
M. Eskiner et al. Fuel 209 (2017) 224–231

0.000.00 -1.650
Fig. 5. Illustration of differences in the transmission degree
+/-0.00 at 1735 cm−1 for all aged samples after the idle period of 30
0.30 and 300 s after shaking for a ternary fuel mixture containing
2 -2.694
alkane/aromatics/C18:2 in different ratios.
-3.737
0.05
-4.781
0.25
-5.825
-3.23 +/-2.60 0.00 +/-0.00
-6.869
0.10 11 8
-7.912
0.20

Ar
-8.956
8:2

om
-10.00
0.15-9.26
C1

ati
-4.95 0.00
+/-2.94 +/-3.57 +/-0.00
0.15 [%]

cs
6 14 4

0.20
0.10
-1.52 +/-0.94 -9.53 +/-4.74 -2.33 +/-0.69 0.00 +/-0.00

0.25 12 15 13 7
0.05

0.30-0.32 -7.12 -4.89 -4.34 0.00

+/-1.07 +/-5.95 +/-4.09 +/-3.15 +/-0.00
0.00
3 10 5 9 1
0.70 0.75 0.80 0.85 0.90 0.95 1.00
Alkane

antagonistic effect. Especially samples 6, 12 und 15 show the highest y ̂ = 2.02X1 + 2.13X2 + 2.40X3 + 0.01X1 X2 + 0.02X1 X3 + 0.02X2 X3
increase in permittivity differences, which are therefore to be assessed
+ 0.00X1 X2 (X1−X2 )−0.01X1 X3 (X1−X3) + 0.05X2 X3 (X2 −X3)
as particularly critical. However, whether the low precipitation ten-
dency of sample 3 is due to exclusively polarity effects or the absence of −0.01X1 X2 (X1−X2 )2−0.09X1 X3 (X1−X3)2−0.24X2 X3 (X2 −X3)2
interactions with aromatics is responsible for this, is not apparent from + 0.04X12 X2 X3−0.41X1 X22 X3−0.01X1 X2 X32 (6)
the data. At least the literature suggests that there may be oxidation
pathways of aromatics [29–31]. To validate the equation, a control point was added in the simplex,
which has the mixture ratio of 0.33/0.33/0.33 in the design matrix,
0.8/0.8/0.1 respectively in the operational matrix. The calculated value
4.4. Model validation for the four degree model for determining the fuel for the permittivity from Eq. (5) is εr′ = 2.16, whereas the measured
composition value is εr′ = 2.18. Results from the response function from equation 5
are within the error limits of the measurement ( ± 0.02).
Based on the data from the real part of the permittivity (see Fig. 2). Currently, the method based on the fourth-degree equation works
The coefficients from equation 5 were calculated and fitted into Eq. (4). only for fresh samples. Work is ongoing to find an adequate response
The fourth-order regression model for the real part of the permittivity function which fully describes also aged fuels in the simplex. If a
for fresh fuels has the form: polynomial of a higher degree can be determined for the values ob-
tained from the aged fuels in the simplex, a method could be provided
which would allow the prediction of precipitation tendency of unknown

Table 3
Permittivity values (100 kHz) for all aged samples - After a waiting period of 30 and 300 s of sedimentation time, samples were taken from the vials and permittivity measurements were
performed.

Sample nr. εr′,0h εr′,10h,30sec εr′,10h,300sec Δεr′ = εr′,10h,30sec−εr′,10h,300sec

Δεr′ = ± 0.02 Δεr′ = ± 0.03 Δεr′ = ± 0.03 Δεr′ = ± 0.04

1 2.02 2.04 2.04 0.00

2 2.13 2.11 2.12 0.00
3 2.40 3.28 2.83 0.45
4 2.08 2.07 2.07 0.00
5 2.22 5.26 4.91 0.35
6 2.27 4.86 2.45 2.41
7 2.05 2.05 2.05 0.00
8 2.10 2.09 2.09 0.00
9 2.12 3.42 2.24 1.18
10 2.30 3.41 2.36 1.05
11 2.20 3.00 2.53 0.47
12 2.32 5.53 2.53 3.00
13 2.15 3.81 2.26 1.55
14 2.17 2.88 2.31 0.57
15 2.24 5.69 2.65 3.04

229
M. Eskiner et al. Fuel 209 (2017) 224–231

Fig. 6. Illustration of differences in the permittivity all aged

samples after a waiting period of 30 and 300 s after shaking
0.00-0.00 0.00 for a ternary fuel mixture containing alkane/aromatics/
0.30 C18:2 in different ratios.
2 0.39
0.78
0.05
1.16
0.25
1.55
0.47 -0.00
1.94
0.10 11 8
2.32
0.20

Ar
2.71
8:2

om
3.10
0.152.41
C1

ati
0.57 -0.00
0.15

c
6 14 4

s
0.20
0.10
3.00 3.04 1.55 -0.00
0.25 12 15 13 7
0.05

0.300.45 1.05 0.35 1.18 -0.00

0.00
3 10 5 9 1
0.70 0.75 0.80 0.85 0.90 0.95 1.00
Alkane

mixtures, if they are within the simplex. References

[1] Mollenhauer K, Tschöke H. Handbuch dieselmotoren. 3rd ed. Berlin, Heidelberg:

5. Conclusion Springer; 2007.
[2] Nollet LML. Chromatographic analysis of the environment. 3rd ed. Boca Raton:
CRC/Taylor & Franciss; 2006.
In the future, parameters such as long-term stability will become an [3] Naumann K, Oehmichen K, Zeymer M, Müller-Langer F, Scheftelowitz M, Adler P,
important factor, especially if PHEV capture new markets. In this work, et al. DBFZ Report Nr. 11. Monitoring Biokraftstoffsektor. 3rd ed. Leipzig; 2016.
[4] DIN EN 590:2014–04. Kraftstoffe für Kraftfahrzeuge – Dieselkraftstoff –
we have shown that it is possible to adapt a sensor system to variable Anforderungen und Prüfverfahren; 2014.
fuel composition by using fuel model substances and a mixture ex- [5] Götz K, Zickmann S, Fey B, Bünger J, Stapf W, Fan Z, et al. Abschlussbericht zum
periment based on the simplex-lattice design. Such a mixture experi- Projektvorhaben Diesel R33. 1st ed. Göttingen: Cuvillier; 2016.
[6] Schmidt L. Wechselwirkungen zwischen Kraftstoffkomponenten in biodieselba-
ment is necessary, because a high degree of variability regarding to the
sierten Mischkraftstoffen unter besonderer Berücksichtigung der Alterungsprodukte
fuel composition all over the world is to be expected. For example, it von Fettsäuremethylestern. 1st ed. Göttingen: Cuvillier; 2014.
can be assumed that aromatic free and regenerative fuels like HVO will [7] Christensen E, McCormick RL. Long-term storage stability of biodiesel and biodiesel
blends. Fuel Process Technol 2014;128:339–48.
become increasingly important. Such fuels are added to fossil fuels in
[8] Fang HL, McCormick RL. Spectroscopic Study of Biodiesel Degradation Pathways.
order to improve the carbon footprint. The formation of oxidation In: Powertrain & Fluid Systems Conference and Exhibition. SAE International400
products in fuels, tends to form undesired precipitates and therefore Commonwealth Drive, Warrendale, PA, United States; 2006.
their determination enables a fuel quality control. Therefore a look at [9] Bartz WJ. Additive für Schmierstoffe. Renningen-Malmsheim: Expert-Verl; 1994.
[10] Mortier RM, Fox MF, Orszulik ST. Chemistry and technology of lubricants. 3rd ed.
the ability of the different mixtures, obtained from the simplex-lattice Dordrecht, New York: Springer; 2010.
design fourth degree, to determine the stability of the suspension is [11] deMan JM, Tie F, deMan L. Formation of short chain volatile organic acids in the
useful. It has been shown by determining the permittivity that the automated AOM method. J Am Oil Chem Soc 1987;64(7):993–6.
[12] Belitz H-D, Grosch W, Schieberle P. Lehrbuch der Lebensmittelchemie. 6th ed.
stability of the suspension is strongly dependent on the concentration of Berlin, Heidelberg: Springer; 2008.
alkanes/aromatics and C18:2 (as a substitute for biodiesel). An antag- [13] Wallentowitz H, Freialdenhoven A, Olschewski I. Strategien zur Elektrifizierung des
onistic effect was found in the range of 15 to 22.5% C18:2. For purposes Antriebstranges: Technologien, Märkte und Implikationen. Wiesbaden: Vieweg
+ Teubner Verlag/GWV Fachverlage GmbH Wiesbaden; 2010.
of comparison, the FTIR signal was measured, which correlates with the [14] Scheffé H. Experiments with mixtures. J Roy Stat Soc B 1958;20(2):344–60.
sensor signal. The small size of the sensor system is an option for on- [15] Hippel ARV, editor. Dielectric materials and applications Boston: Artech House;
board control in PHEV. Works are ongoing to expand this method by a 1995.
[16] Crastan V. Elektrische Energieversorgung 2: Energiewirtschaft und Klimaschutz
further sensor system based on a spectroscopic method (fluorescence
Elektrizitätswirtschaft, Liberalisierung Kraftwerktechnik und alternative
and near-infrared spectroscopy). This is to achieve a redundancy and to Stromversorgung, chemische Energiespeicherung. 3rd ed. Heidelberg, New York:
identify varying fuel compositions more specifically. This strategy Springer; 2012.
[17] Romano SD, Sorichetti PA. Dielectric spectroscopy in biodiesel production and
prevents engine damages since fuel degradation is detectable at an early
characterization. London: Springer; 2011.
stage and the combustion engine can be started before precipitation [18] Cornell JA. How to run mixture experiments for product control. Milwaukee: ASQC;
occur. It is also a promising approach to apply this method for gasoline 1990.
powered PHEV: Here for example the vapour pressure during long-term [19] Cornell JA. A primer on experiments with mixtures. Hoboken, N.J.: Wiley; 2011.
[20] Cornell JA. Experiments with mixtures: designs, models and the analysis of mixture
storage of varying fuel mixtures (e.g. ethanol content) could be de- data. New Jersey: John Wiley & Sons; 2011.
termined. [21] Lazić Ž. Design of experiments in chemical engineering: a practical guide.
Weinheim: Wiley-VCH; 2004.
[22] Golio JM. The RF and microwave handbook. Boca Raton, Forida: CRC Press; 2001.

230
M. Eskiner et al. Fuel 209 (2017) 224–231

[23] Zoolfakar A, Mahzan N, Manut A, Idros M. Design, fabrication and characterization [27] Corach J, Sorichetti PA, Romano SD. Permittivity of diesel fossil fuel and blends
of pH transducer using printed circuit board (PCB). In: 2010 International with biodiesel in the full range from 0% to 100%: application to biodiesel content
Conference on Electronic Devices, Systems & Applications: ICEDSA proceedings estimation. Fuel 2017:367–73.
Best Western Premier Sri Pacific Hotel, Kuala Lumpur, 12–13 April 2010. [28] Gonzalez Prieto L, Sorichetti PA, Romano SD. Electric properties of biodiesel in the
Piscataway: IEEE; 2010, p. 206–211. range from 20 Hz to 20 MHz. Comparison with diesel fossil fuel. Int J Hydrogen
[24] Eskiner M, Bär F, Krahl J, Rossner M, Munack A, editors. Onboard determination of Energy 2008;13:3531–7.
fuel stability in plug-in hybrid diesel cars by dielectric relaxation spectroscopy [29] Altarawneh M, Dlugogorski BZ, Kennedy EM, Mackie JC. Theoretical investigation
Germany: AMA Service GmbH; 2015. into the low-temperature oxidation of ethylbenzene. Proc Combust Inst
[25] Eskiner M, Bär F, Rossner M, Munack A, Krahl J. Determining the aging degree of 2013;34(1):315–23.
domestic heating oil blended with biodiesel by means of dielectric spectroscopy. [30] Ye L, Xing L, Yuan W, Li Y, Zhang L, Qi F. Predictive kinetics on the formation and
Fuel 2015;143:327–33. decomposition of ethylbenzene. Proc Combust Inst 2016.
[26] Eskiner M, Krahl J. Forschung für zukünftige Mobilität: 7. Biokraftstoffsymposium [31] Yuan W, Li Y, Pengloan G, Togbé C, Dagaut P, Qi F. A comprehensive experimental
am 26. und 27. Februar 2015. Kraftstoffqualitätssensoren für Plug-in and kinetic modeling study of ethylbenzene combustion. Combust Flame
Hybridfahrzeuge. 1st ed. Göttingen: Cuvillier; 2015. p. 2016. 2016;166:255–65.

231