Sensors and Actuators A 168 (2011) 22–29

Contents lists available at ScienceDirect

Sensors and Actuators A: Physical

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

Application of dielectric spectroscopy for engine lubricating oil degradation

monitoring
L. Guan ∗ , X.L. Feng, G. Xiong, J.A. Xie
Oil Application and Management Department, Logistical Engineering University, Shapingba District, Chongqing 401311, China

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

Article history: In this paper, Dielectric Spectroscopy (DS) was employed to analyze the oxidation degradation process of
Received 29 September 2010 engine lubricating oil qualitatively and quantitatively compared with Fourier Transform Infrared Spec-
Received in revised form 15 March 2011 troscopy (FTIR). It was found that both DS and FTIR can directly obtain the degradation features from
Accepted 15 March 2011
the spectral data. With the combination of DS and multivariate calibration (Partial Least Square PLS),
Available online 23 March 2011
three main oil monitoring properties including Oxidation Duration (OD), Total Acid Number (TAN) and
Insoluble Content (IC) can be determined quantitatively and accurately. It was proved that operating
Keywords:
temperature had more influence on DS data than excitation amplitude. The results in the article show
Engine lubricating oil
Oil monitoring
that DS can be developed into an effective oil monitoring/analysis method.
Fourier transform infrared spectroscopy © 2011 Elsevier B.V. All rights reserved.
(FTIR)
Dielectric spectroscopy (DS)
Multivariate calibration

1. Introduction [6,7], atomic spectroscopy [8,9], X-ray fluorescence spectroscopy

Lubricating oil plays a key role in internal-combustion engines.
It consists of complex mixtures of hydrocarbons and is a combi-
nation of base oils and additives [1]. Engine lubricants are used
to reduce the frictions of the mobile components and to keep the
different elements clean, being able to work as detergents and
dispersant agents. The engine lubricating oil ageing process is a
very complex process during which degradation of the base oil
and depletion of its additives take place simultaneously. Oxida-
tive high temperature degradation and contamination by water,
ethylene glycol, fuel, soot, and wear metals are the main factors.
Currently, the main methods for determining engine lubricating
oil condition are the routine physical & chemical tests to assess
the properties including kinematic viscosity, Total Acid Number
(TAN), Total Base Number (TBN) and Insoluble Content (IC), which
are always time-consuming, laborious and require specific equip-
ment for the determination of each property of interest. In this
context, kinds of efficient alternative methods have been devel-
oped. Ferrography [2–5] is currently the prevalent and effective
method to evaluate the wear ferromagnetic particles. Analytical
and direct reading (DR) ferrography are the two main types. As for
determination of wear metals such as Na, Mg, Al, Ca, Ti, V, Cr, Cu,
Zn, Mo, Ag, and Cd. Laser-induced breakdown spectroscopy (LIBS)
∗ Corresponding author. Tel.: +86 023 86730900.
E-mail address: gl 200122@163.com (L. Guan).
[10] are already introduced respectively. Visible spectrophotomet-
ric detection in association with flow injection analysis (FIA-visible
spectrometry) is proposed and employed to determinate the insol-
uble content [11]. In particular, middle infrared spectroscopy (MIR)
has been largely applied [1,11–17], which offers several advantages
for this type of application, such as nondestructive nature.
Recently, much attention has been paid to dielectric and elec-
trochemical impedance properties of industrial lubricants. Lvovich
and Schmiechowski [18] have discussed the relationship between
chemical composition of lubricants and their electrochemical
properties obtained by means of electrochemical impedance spec-
troscopy (EIS). The non-linear impedance analysis of industrial
lubricants has also been performed by higher harmonic non-linear
electrochemical impedance spectroscopy (NLEIS) [19]. EIS and
NLEIS are able to offer the opportunities to characterize, evalu-
ate and provide insights into chemical composition, changes and
mechanisms of lubricants. Wang and Lee [20,21] have used a.c.
impedance technique to detect glycol contamination in engine oil.
And a new technique to detect minor antifreeze in engine oil by
measuring the changes of engine oil resistance was introduced too.
EIS has been used to estimate soot and diesel contamination in
engine oil simultaneously [22]. Oil condition sensors [23–27] based
on dielectric constant and conductivity measurement of the engine
lubricating oil have been designed and fabricated. These sensors
can detect the relative variation of lubricant degradation. Wang
[28] has established a good correlation between TAN and this kind
of sensor’s output.

0924-4247/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.sna.2011.03.033
 L. Guan et al. / Sensors and Actuators A 168 (2011) 22–29
Fig. 1. Schematic diagram of sample preparation device.
Dielectric and electrochemical impedance analysis methods are
relatively fast, simple, inexpensive and free from the difficulties
associated with the current standard testing methods. Moreover,
dielectric and electrochemical properties are related with the
chemical composition and physical structures of engine lubricating
oil.
Dielectric spectroscopy (DS) is an analytical technique on
the interaction between dielectric material and electromagnetic
energy in the radio-frequency and microwave range, which is a
powerful structural detection technique for dielectric materials.
What DS studies is the dependence of materials’ dielectric prop-
erties on wavelength or frequency. The difference between DS
and EIS: the main interest of DS is on the intrinsic electric mate-
rial properties. The complex permittivity ε*(ω) or conductivity
*(ω) spectra can be easily evaluated from Z*(ω) with the help of
sample dimensions; the focus of EIS is mostly on the properties
of electrode/material interfaces and the materials under test are
often electrolytes or ion conductors. Petroleum products including
engine lubricating oil are all typical dielectric materials. So DS is
more suitable than EIS for engine lubricating oil analysis.
At present, DS technique enables researchers to make sound
contributions to contemporary problems in modern physics. DS
has been employed to quality sensing application of agricultural
product [29,30]. Our previous publications have investigated the
correlations between DS data and petroleum products’ compo-
sition and quality indexes by means of multivariate calibration,
which include classification of virgin engine lubricating oils by SAE
and source [31] and determination of clean gasoline octane num-
bers [32]. It is already proved that DS is a practical and effective
analysis method to obtain rich composition and structure infor-
mation of complex mixture systems. Especially, with the help of
chemometrics multivariate calibration methods the direct relation-
ships between DS data and quality properties can be established
efficiently, which is more effective than the interpretation of EIS
results by means of complex equivalent circuit (EC) models. The
Table 1
Information on three virgin engine lubricating oils.
Name API SAE
Set one CD 15 W/40
Set two CD-SE 50
Set three SJ 15 W/40
23
new analysis idea is explained in our previous publication too
[32].
The monitoring sensor for oil condition, especially for engine
lubricating oil should be easy to be cleaned and is able to extract
sufficient dielectric information. Interdigitated comb capacitor sen-
sor is a good and practical selection, which has been applied for
oil analysis widely [27,31–33]. So the measuring sensor for engine
lubricating oil in this article is a type of interdigitated comb capac-
itor sensor too.
In this article, DS is employed to examine the oxidative degra-
dation of engine lubricating oil and to determine the Oxidation
Duration (OD), Total Acid Number (TAN) and Insoluble Content (IC)
properties. All the DS analysis results are discussed compared with
the Fourier Transform Infrared spectroscopy (FTIR).
2. Experimental
2.1. Samples
All the samples were prepared with the device illustrated in
Fig. 1. The main function of the device is to oxidize the virgin sam-
ples under given oxidation conditions. The air flow generated by
the air pump and the copper wire catalyst are used to accelerate
the oxidization process.
According to the sample preparation device, three types of vir-
gin engine lubricating oils were employed to produce three series
of samples with different degrees of degradation. The three virgin
engine lubricating oils are listed in Table 1.
For each series of samples, three types of properties including
Oxidation Duration (OD), Total Acid Number (TAN) and Insoluble
Content (IC) were recorded. OD values were recorded according
to sampling time. TAN and IC properties were respectively deter-
mined by TAN&TBN Analyzer and Insoluble Content Analyzer made
by Beijing China Invent Instrument Technology Ltd. Company. The
samples with three properties are shown in Tables 2–4.
Company Oxidation temperature (◦ C)
KunLun (China) 150
Shell 150
Dalian Petroleum (China) 150
24 L. Guan et al. / Sensors and Actuators A 168 (2011) 22–29

Table 2
Sample list generated by set one.

No. Oxidation TAN Insoluble content

during (min) (mgKOH/g) (mg/g)

1-00 0 0.87 0.00

1-01 666 1.44 0.02
1-02 1501 1.45 0.05
1-03 2282 1.25 0.11
1-04 3075 1.09 0.22
1-05 3907 0.91 0.39
1-06 4756 1.26 0.59
1-07 5525 1.83 0.76
1-08 6375 1.96 1.02
1-09 7215 2.81 1.42
1-10 8033 5.43 1.61
1-11 8945 5.60 2.81

Table 3
Sample list generated by set two.

No. Oxidation TAN Insoluble content

during (min) (mgKOH/g) (mg/g)

2-00 0 1.10 0.02

2-01 668 1.00 0.17
2-02 1485 0.85 0.35
2-03 1947 0.76 0.58
2-04 2667 0.72 0.68
2-05 3474 0.79 0.81
2-06 3618 0.70 1.13
2-07 4303 0.94 1.32
2-08 5079 0.79 1.54
2-09 5890 1.01 1.87
2-10 6733 1.18 2.21
2-11 7540 1.09 2.68
2-12 8342 1.36 3.24
2-13 9150 1.27 3.48
2-14 9869 0.99 3.87 Fig. 2. Construction of sensor (a) schematic. Construction of sensor (b) photograph.
2-15 10659 1.23 4.08
2-16 10963 1.13 4.29
2.3. DS apparatus and method

A Dielectric Spectroscopy Analyzer for Petroleum (DSAP)

Table 4 instrument made by Logistical Engineering University (LEU) was
Sample list generated by set three.
employed to obtain the DS data of samples [31,32]. During DS
No. Oxidation TAN Insoluble content measurement, DSAP generates a. c. signals with desired frequen-
during (min) (mgKOH/g) (mg/g) cies and amplitudes which are applied to the samples under
3-00 0 1.34 0.00 test. Oil-composition-dependent current, which is similar to oil-
3-01 772 0.94 0.05 condition-dependent current [23], is amplified, filtered, rectified,
3-02 1539 0.55 0.14 converted into a d. c. voltage output and put into a 12 bits A/D
3-03 2328 0.43 0.26
converter. So the values DSAP measures, called ‘response signals’,
3-04 3103 0.56 0.37
3-05 3445 0.74 0.43 are non-dimensional numbers and range from 0 to 4096. The main
3-06 3889 0.68 0.58 parameters of DSAP are:
3-07 4072 1.00 0.65
3-08 4693 0.60 1.02
Waveform generated: 1. sine wave, 2. square wave, 3. triangle
3-09 4971 1.10 1.28
3-10 6180 1.11 3.60 wave.
3-11 6660 1.05 7.56 Amplitude range: −10 V to +10 V.
Frequency range: 50 kHz–16 MHz.
Temperature held: room temperature to 100 ± 1 ◦ C.
Analysis time for one sample: about 3 min.

2.2. FTIR apparatus and method
The FTIR instrument used in this work was Perkin Elmer 1725X
FT-IR spectrometer. A pair of well-polished KBr windows with
0.05 mm spacer creating a fixed thin film was used to acquire the
sample spectral data. During spectra measurement, a blank KBr
window spectra was collected and used as reference for sample
absorbance calculation. And then the sample was injected into the
blank KBr windows and the sample spectra were collected in the
range of 400 to 4000 cm−1 at 4 cm−1 resolution.
For this study, the sine waveform was selected; amplitudes were
4 V, 6 V, and 8 V; frequencies ranged from 50 kHz to 16 MHz and
frequency interval was 200 kHz; measurement temperatures were
40 ± 1 ◦ C and 100 ± 1 ◦ C.
A type of interdigitated comb capacitor sensor is fabricated and
employed in this article. The construction of the sensor is shown
in Fig. 2. The sensor is fabricated by printed circuit board (PCB)
technique.
The comb electrodes of the sensor are made of gold plated cop-
per layers with 35 ␮m height h. The copper finger width and spacing
 L. Guan et al. / Sensors and Actuators A 168 (2011) 22–29 25

Table 5
Peak area results of set one.

No. S1 S2 S3

1-00 22.193 248.666 128.894

1-01 17.331 228.165 135.949
1-02 18.897 230.387 158.754
1-03 19.300 271.496 178.981
1-04 21.300 311.430 204.315
1-05 25.781 385.124 210.927
1-06 35.395 496.285 231.570
1-07 54.494 659.748 237.350
1-08 86.910 827.334 251.910
1-09 105.635 861.001 251.832
1-10 163.407 1167.237 281.753
1-11 226.322 1177.521 319.109

Table 6
Peak areas results of set two.

No. S1 S2 S3

2-00 0.000 0.000 248.462

2-01 −1.826 59.469 208.507
2-02 25.437 141.699 236.571
2-03 12.173 250.713 273.927
2-04 15.530 366.493 280.454
2-05 30.444 354.302 266.599
2-06 55.156 770.127 303.360
2-07 55.968 845.631 317.584
2-08 96.491 1146.900 307.353
2-09 115.090 1241.389 380.286
2-10 155.654 1352.902 392.526
2-11 188.805 1576.000 420.048
2-12 177.432 1389.495 378.501
2-13 255.886 1757.598 494.598
2-14 219.195 1468.520 423.234
2-15 294.524 1804.537 525.734
2-16 288.930 1711.515 551.087

Table 7
Peak areas results of set three.

No. S1 S2 S3

3-00 9.043 170.631 131.805

3-01 13.187 197.463 132.567
3-02 12.904 222.201 110.505
3-03 16.111 281.947 109.664
3-04 20.263 291.729 98.938
3-05 22.478 368.709 116.547
3-06 27.519 422.198 127.855
3-07 26.342 366.387 115.760
3-08 38.515 455.334 133.959
Fig. 3. MIR spectra (a) set one. MIR spectra (b) set two. MIR spectra (c) set three. 3-09 44.867 500.059 148.623
3-10 97.849 838.778 237.963
3-11 124.464 908.065 401.006
are both 200um. The substrate of the sensor is a type of polyte-
trafluoroethylene (Teflon or PTFE or F4) whose dielectric constant
is 2.65.
presence of carboxyl matter namely the deeply oxidized prod-
3. Results and discussion ucts.
The absorption peak areas centered at 1774 cm−1 , 1713 cm−1
3.1. FTIR measurements and 1604 cm−1 for three series of samples, which were labeled by
S1, S2 and S3, respectively, were calculated by Thermo Scientific
The MIR spectra of three series of samples are shown in Fig. 3. Omnic software. The results are shown in Tables 5–7.
The features in the three series of MIR spectra can be con- Nine multiple linear regression (MLR) models were, respec-
cluded that absorption bands centered at 1774 cm−1 , 1713 cm−1 tively, built to examine the relationships between the peak areas of
and 1604 cm−1 have the increasing trends as for absorption peak three series of samples and those three properties. In these models,
intensity with the oxidation of the lubricating oil, which can be the predictor variables were peak areas and the response variables
explained as below: were the three properties including OD, TAN and IC. The programs
The band feature centered at 1604 cm−1 is due to the pres- were written in Matlab2008a and based on the ‘regress’ function
ence of saponified matter namely the initial oxidized products; in Matlab Statistics Toolbox. For each series of samples, three MLR
the band feature centered at 1713 cm−1 is due to the presence models were, respectively, constructed according to OD, TAN and
of carbonyl matter namely the intermediate oxidized prod- IC properties. During modeling leave-one-out full cross validation
ucts; the band feature centered at 1774 cm−1 is due to the was used to obtain the predicted results. The predicted correlation
26 L. Guan et al. / Sensors and Actuators A 168 (2011) 22–29

Table 8 ferent from the peak area or height changing feature in MIR
Predicted results based on MIR peak areas in correlation coefficient.
spectra.
Property Set one Set two Set three (2) Similar characteristics can be obtained under different testing
Oxidation duration (min) 0.9905 0.9723 0.5796 conditions. For example, at 40 ◦ C and 100 ◦ C testing tempera-
TAN (mgKOH/g) 0.8670 0.4918 0.1138 tures and at 4Vpp, 6Vpp and 8Vpp excitation signal amplitudes
Insoluble content (mg/g) 0.9385 0.9718 0.9131 have almost the same testing results.
(3) Characteristic band peak features cannot be seen from the spec-
tra data of DS, which is different from MIR and somewhat
coefficients between the predicted and the measured were listed
similar with NIR. It can be difficult to assign specific features
in Table 8.
to specific chemical components.
As shown in Table 8, poor correlation coefficients for TAN were
obtained for three series. For set one and set two, the correlation
coefficients for OD and IC were relatively acceptable. For set three, 3.2.2. Quantitative calculation based on DS data
only IC property had good predicted results. As DS is based on the interaction of an external field with
the electric dipole moment of the sample, it is difficult to assign
3.2. DS measurements specific features to specific chemical components of the sam-
ple like MIR analysis. However characteristic ingredient and
3.2.1. DS spectra features analysis structure information of the sample can be obtained from DS
DS spectra of all three series of samples were collected and data. In order to find out the relationship between DS data
shown in Fig. 4. From Fig. 4 we can conclude the main features and usual chemical and physical properties such as TAN and
of DS spectra: insolubles, multivariate (multiple frequencies) calibration tech-
niques (e.g. principal components analysis PCA, partial least
(1) The output DS spectra of DSAP have the shift trend to the lower squares PLS, or artificial neural networks) are often employed
frequency range as oxidation duration increases, which is dif- [31,32].

Fig. 4. DS spectra (a) set one at 40 ◦ C. DS spectra (b) set one at 100 ◦ C. DS spectra (c) set two at 40 ◦ C. DS spectra (d) set two at 100 ◦ C. DS spectra (e) set three at 40 ◦ C. DS
spectra (f) set three at 100 ◦ C.
 L. Guan et al. / Sensors and Actuators A 168 (2011) 22–29 27

Table 9
Predictive results of sample set one based on DS data.

Property 4Vpp 6Vpp 8Vpp

◦ ◦ ◦ ◦
40 C 100 C 40 C 100 C 40 ◦ C 100 ◦ C

Oxidation duration (min) 0.8560 0.9623 0.9263 0.9906 0.9889 0.9896

TAN (mgKOH/g) 0.9614 0.9923 0.9078 0.9926 0.9355 0.9963
Insoluble content (mg/g) 0.7884 0.9814 0.7330 0.9846 0.9151 0.9762

Table 10
Predictive results of sample set two based on DS data.

Property 4Vpp 6Vpp 8Vpp

40 ◦ C 100 ◦ C 40 ◦ C 100 ◦ C 40 ◦ C 100 ◦ C

Oxidation duration (min) 0.9905 0.9860 0.9763 0.9857 0.9954 0.9912

TAN (mgKOH/g) 0.9613 0.9764 0.9722 0.9806 0.9818 0.9785
Insoluble content (mg/g) 0.9944 0.9890 0.9473 0.9858 0.9954 0.9893

Table 11
Predictive results of sample set three based on DS data.

Property 4Vpp 6Vpp 8Vpp

40 ◦ C 100 ◦ C 40 ◦ C 100 ◦ C 40 ◦ C 100 ◦ C

Oxidation duration (min) 0.8571 0.9557 0.7853 0.9592 0.8101 0.9797

TAN (mgKOH/g) 0.6859 0.8652 0.5527 0.8516 0.6857 0.8826
Insoluble content (mg/g) 0.7976 0.9971 0.7605 0.9977 0.9534 0.9988

Table 12
Predicted results of all samples based on transformed DS data.

Property 4Vpp 6Vpp 8Vpp

◦ ◦ ◦ ◦
40 C 100 C 40 C 100 C 40 ◦ C 100 ◦ C

Oxidation duration (min) 0.9038 0.9908 0.9734 0.9733 0.9524 0.9849

TAN (mgKOH/g) 0.9729 0.9911 0.9544 0.9933 0.9808 0.9986
Insoluble content (mg/g) 0.8371 0.9265 0.9031 0.9867 0.8635 0.9866

In this article partial least-squares (PLS) regression is used to

build the relationship between DS data and OD, TAN and IC proper-
ties. Three series of samples including set one, set two and set three
were respectively calculated by means of PLS. The target properties
are the OD, TAN and IC. As the number size of samples was relatively
few, leave-one-out method (full cross validation) was adopted. All
the programs were written and run in Matlab2008a. The calculation
results of correlation coefficients are listed in Tables 9–11.
The obvious conclusion can be obtained from Tables 9–11: more
accurate results can be obtained at 100 ◦ C than at 40 ◦ C while the
excitation amplitude has less influence.
To summarize, the respective predicted results of three series
of samples indicated that DS data contained the compositional
and structural information, which accounted for the degradation
of engine lubricating oil. With PLS, TAN and IC properties can be
determined accurately based on the DS data.
The results above were obtained by respective calculations. So
one PLS model had to be constructed for each series of samples.
However, it is necessary and practical to build an overall PLS model
for different series of samples. During oxidation-induced degrada-
tion of lubricating oil, kinds of polar oxidation and nitration were
newly generated. The presence of these polar products can be seen
and interpreted from the changes (shifts) in DS spectra. Before con-
struction of the overall PLS model, all three series of DS spectra
were transformed by subtracting the corresponding virgin sample
spectra, respectively. And the overall PLS multivariate calibration
model was built based on the transformed DS data. The transformed
spectra of the three series of samples were shown in Fig. 5.
According to the overall PLS model, the predicted results are
shown Table 12. Fig. 5. The transformed spectra of three series (a) at 40 ◦ C. The transformed spectra
of three series (b) at 100 ◦ C.
28 L. Guan et al. / Sensors and Actuators A 168 (2011) 22–29
Fig. 6. Regression line for oxidation duration (a). Regression line for TAN (b). Regres-
sion line for insoluble content (c).
From results in Table 12, more accurate results can be obtained
by transformed DS data with the overall PLS model. And higher
operating at 100 ◦ C was more enhancing than at 40 ◦ C, which shown
that operating temperature was an important parameter to control.
In Fig. 6 shows the regression lines at 100 ◦ C operating temper-
ature and 8Vpp amplitude.
From Fig. 6, good performance can be seen except for one sam-
ple’s predicted value of IC property as shown in Fig. 6(f). However,
we found that the sample which had the largest residual value
(1.965 mg/g) had the largest IC value (7.56 mg/g).
4. Conclusions
Based on the described engine lubricating oil oxidation device,
three series of samples with different oxidation degrees were pro-
duced. MIR and DS were employed to discover the degradation
characteristics of the samples. The oxidation features of the sam-
ples can be directly seen from both MIR and DS spectra. Based on
the characteristic peak areas calculated based on MIR spectra data,
OD and IC can get good predicted performance but TAN failed. We
have demonstrated that DS is suitable to monitor the degradation of
engine lubricating oil. OD, TAN and IC properties can be determined
simultaneously quantitatively and accurately.
DS, as the most effective method to extract the dielectric char-
acteristic from the dielectric material, can be developed into an
efficient oil monitoring/analysis method. With the chemometrics
multivariate calibration such as PLS, qualitative and quantitative
analysis models can be built to monitor the degradation degrees
and to determine the main oil monitoring properties of engine
lubricating oil such as OD, TAN and IC. We believe that the remain-
ing useful life of engine lubricating oil can be predicted based on
on-line or in situ DS data.
References
[1] A. Borin, R.J. Poppi, Application of mid infrared spectroscopy and iPLS for the
quantification of contanminants in lubricating oil, Vib. Spectrosc. 37 (2005)
27–32.
[2] Y. Liu, Z. Liu, S.Z. Wen, Y.B. Xie, Motion analysis on the particles in a magnetic
field detector, Tribol. Int. 33 (2000) 837–843.
[3] B.J. Roylance, Ferrography then and now, Tribol. Int. 38 (2005) 857–862.
[4] H.L. Xiao, The development of ferrography in China-some personal refflections,
Tribol. Int. 38 (2005) 904–907.
[5] V. Macian, R. Payri, B. Tormos, L. Montoro, Applying analytical ferrography as a
technique to detect failures in diesel engine fuel injection systems, Wear 260
(2006) 562–566.
[6] P. Yaroshchyk, R.J.S. Morrison, D. Body, B.L. Chadwick, Quantitative determina-
tion of wear metals in egine oils using laser-induced breakdown spectroscopy:
a comparison between liquid jets and static liquids, Spectrochim. Acta Part B
60 (2005) 986–992.
[7] P. Yaroshchyk, R.J.S. Morrison, D. Body, B.L. Chadwick, Quantitative determina-
tion of wear metals in engine oils using LIBS: the use of paper substrates and
a comparison between single-and double-pulse LIBS, Spectrochim. Acta Part B
60 (2005) 1482–1485.
[8] E.J. Ekanem, J.A. Lori, S.A. Thomas, The determination of wear metals in used
lubricating oils by flame atomic absorption spectrometry using sulphanilic acid
as ashing agent, Talanta 44 (1997) 2103–2121, 2108.
[9] Z. Yang, X. Hou, B.T. Jones, Determination of wear metals in engine oil by mild
acid digestion and energy dispersive X-ray fluorescence spectrometry using
solid phase extraction disks, Talanta 59 (2003) 673–680.
[10] M. Pouzar, T. Cernohorsky, A. Krejcova, Determination of metals in lubricating
oils by X-ray fluorescence spectrometry, Talanta 54 (2001) 829–835.
[11] M. Knochen, A. Sixto, G. Pignalosa, S. Domenech, S. Garrigues, M. de la Guardia,
Determination of insolubles in diesel lubricating oil by FIA-visible spectrome-
try, Talanta 64 (2004) 1359–1363.
[12] ASTM E-2412-0, Standard Practice for Condition Monitoring of Used Lubri-
cants by Trend Analysis Using Fourier Transform Infrared (FT-IR) Spectrometry.
Standard Volume.03.06 (2005).
[13] A.R. Caneca, M. Fernanda Pimentel, Roberto Kawakami Harrop Galvao, C.E.
da Matta, F.R. de Carvalho, I.M. Raimundo@@Jr., C. Pasquini, J.J.R. Rohwedder,
Assessment of infrared spectroscopy and multivariate techniques for monitor-
ing the service condition of diesel-engine lubricating oils, Talanta 70 (2006)
344–352.
[14] H.F. Yuan, X.L. Chu, H.R. Li, Y.P. Xu, Determination of multi-properties of residual
oils using mid-infrared attenuated total reflection spectroscopy, Fuel 85 (2006)
1720–1728.
[15] M.A. Al-Ghouti, L. Al-Atoum, Virgin and recycled engine oil differentiation: a
spectroscopic study, J. Environ. Manage. 90 (2009) 187–195.
[16] N. Gracia, S. Thomas, P. Bazin, L. Duponchel, F. Thibault-Starzyk, O. Lerasle,
Combination of mid-infrared spectroscopy and chemometric factorization
tools to study the oxidation of lubricating base oils, Catal. Today 155 (2009)
255–260.
[17] Y. Felkel, N. Dorr, F. Glatz, K. Varmuza, Determination of the total acid number
(TAN) of used gas engine oils by IR and chemometrics applying a com-
bined strategy for variable selection, Chemom. Intell. Lab. Syst. 101 (2010)
14–22.
[18] V.F. Lvovich, M.F. Smiechowski, Impedance characterization of industrial lubri-
cants, Electrochim. Acta 51 (2006) 1487–1496.
[19] V.F. Lvovich, M.F. Smiechowski, Non-linear impedance analysis of industrial
lubricants, Electrochim. Acta 53 (2008) 7375–7385.
[20] S.S. Wang, H.-S. Lee, The application of a.c. impedance technique for
detecting glycol contamination in engine oil, Sens. Actuators B 40 (1997)
193–197.
[21] S.S. Wang, Y. Lin, A new technique for detecting antifreeze in engine oil during
early stage of leakage, Sens. Actuators B 96 (2003) 157–164.
[22] H.P. Christian Ulrich, H. Sundgren, F. Bjorefors, C. Krantz-Rulcker, Simultaneous
estimation of soot and diesel contamination in engine oil using electrochemical
impedance spectroscopy, Sens. Actuators B 127 (2007) 613–618.
[23] H.-S. Lee, S.S. Wang, In situ monitoring of high-temperature degraded engine
oil condition with microsensors, Sens. Actuators B 20 (1994) 49–54.
[24] S.S. Wang, H.-S. Lee, D.J. Smolenski, The development of in-situ electrochemical
oil condition sensors, Sens. Actuators B 17 (1994) 179–185.
[25] S.S. Wang, H.-S. Lee, An electrochemical sensor for distinguishing two-stroke-
engine oils, Sens. Actuators B 40 (1997) 199–203.
 L. Guan et al. / Sensors and Actuators A 168 (2011) 22–29
[26] S. Wang, Road tests of oil condition sensor and sensing technique, Sens. Actu-
ators B: Chem. 73 (2001) 106–111.
[27] S. Raadnui, S. Kleesuwan, Low-cost condition monitoring sensor for used oil
analysis, Wear 259 (2005) 1502–1506.
[28] S.S. Wang, Engine oil condition sensor: method for establishing correlation with
total acid number, Sens. Actuators B 86 (2002) 122–126.
[29] S.O. Nelson, Dielectric spectroscopy for agricultural applications, Proc. IEEE
Instr. Meas. Technol. Conf. 1 (2004) 752–754.
[30] S.O. Nelson, Dielectric spectroscopy of fresh fruits and vegetables, Proc. IEEE
Instr. Meas. Technol. Conf. 1 (2005) 360–364.
[31] L. Guan, Feng.F X.L., G. Xiong, Engine lubricating oil classification by SAE
grade and source based on dielectric spectroscopy data, Chim. Acta 628 (2008)
117–120.
[32] L. Guan, X.L. Feng, Z.C. Li, G.M. Lin, Determination of octane num-
bers for clean gasoline using dielectric spectroscopy, Fuel 88 (2009)
1453–1459.
[33] V.F. Lvovich, C.C. Liu, M.F. Smiechowski, Optimaization and fabrication of planar
interdigitated impedance sensors for highly resistive non-aqueous industrial
fluids, Sens. Actuators B 119 (2006) 490–496.
29
Biographies
L. Guan Ph.D. (Engineering), Logistical Engineering University (LEU), Chongqing,
China, 2009. He is now the lecturer of LEU. Testing and analysis technologies, devel-
opment of analytical instruments for fuel and lubricating oil especially based on
dielectric spectroscopy technology are the current study topics.
X.L. Feng Ph.D. (Engineering), Logistical Engineering University (LEU), Chongqing,
China, 2004. He is now the professor of LEU. Chemometrics, testing and analysis
technologies and development of analytical instruments for fuel and lubricating oil
are the current fields of interest. Near infrared spectroscopy and dielectric spec-
troscopy is his hot study topics nowadays.
G. Xiong Master, Logistical Engineering University (LEU), Chongqing, China, 2003.
He is now the engineer of LEU. Electronics engineering and development of analysis
instrument is his major fields of interest.
J.A. Xie Master, Logistical Engineering University (LEU), Chongqing, China, 2008.
Chemometrics is his current study field of interest.