ARCHIVES OF ELECTRICAL ENGINEERING VOL. 65(4), pp.

733-744 (2016)

DOI 10.1515/aee-2016-0051

Fault diagnostics of DC motor using acoustic signals

and MSAF-RATIO30-EXPANDED

ADAM GLOWACZ
AGH University of Science and Technology
Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering
Department of Automatics and Biomedical Engineering
A. Mickiewicza 30, 30-059 Kraków, Poland
e-mail: adglow@agh.edu.pl
(Received: 11.01.2016, revised: 25.08.2016)

Abstract: An early fault diagnostic method of Direct Current motors was presented in
this article. The proposed method used acoustic signals of a motor. A method of feature
extraction called MSAF-RATIO30-EXPANDED (method of selection of amplitudes of
frequencies – ratio 30% of maximum of amplitude – expanded) was presented and im-
plemented. An analysis of proposed method was carried out for early fault states of a real
DC motor. Four following states of the DC motor were measured and analyzed: the
healthy DC motor, DC motor with 3 shorted rotor coils, DC motor with 6 shorted rotor
coils, DC motor with a broken coil. Measured states were caused by natural degradation
of the DC motor. The obtained results of analysis were good. The presented early fault
diagnostic method can be used for protection of DC motors.
Key words: acoustic signal, analysis, diagnostic, DC motor, machine

1. Introduction

DC motors are easy to control and a speed of a DC motor can be controlled. DC motors are
used in propulsion of electric vehicles (for example trams), hoists and elevators. Small DC
motors are used in toys and various tools such as: printers, hard disks and CD/DVD ROM
drives. Early faults of motors are caused by natural degradation of machines. These early
faults may damage a motor. Damaged motors generate economic losses caused by breakdowns
in production lines. The damaged motor should be replaced by another one. Methods of fault
diagnostics are developed to prevent economic losses. In the literature diagnostic methods
have been developed for the detection and localization of faults of a DC motor, induction
motor [1, 2] and synchronous motor [3]. A new methodology to assist the preventive main-
tenance of induction motor was presented in literature [4]. A distributed system for induction
motors was also implemented [5]. An on-line condition monitoring system for induction mo-
tors was also described [6]. Analyses of vibrations [7-11] and acoustic signals [12-18] were
discussed in literature. Analyses of thermal images of faulty states of electrical machines and

Unauthenticated
Download Date | 11/30/16 1:32 PM
734 A. Glowacz Arch. Elect. Eng.

equipment were also presented [19-22]. Techniques of material diagnostic were also deve-
loped [23]. The most of rotating electrical machines generate acoustic signals. It is a reason to
analyze this type of signal for the DC motor. Information about windings of electrical ma-
chines can be found in following literature [24]. In this article the author presented a technique
of recognition of acoustic signals of the DC motor (Fig. 1). This technique is based on original
method of feature extraction called MSAF-RATIO30-EXPANDED (method of selection of
amplitudes of frequencies – ratio 30% of maximum of amplitude – expanded).

Fig. 1. The analyzed DC motor and computer with microphone

2. Technique of fault diagnostics of DC motor

Technique of fault diagnostics of a DC motor is based on processing of an acoustic signal.

It is a very difficult task. The proposed technique uses 2 processes: pattern creation and iden-
tification.

Fig. 2. Block diagram of proposed technique of fault diagnostics of the DC motor

Unauthenticated
Download Date | 11/30/16 1:32 PM
Vol. 65 (2016) Fault diagnostics of DC motor using acoustic signals and MSAF-RATIO30-EXPANDED735

The pattern creation process was used to obtain patterns. The identification process com-
pared a test sample with patterns (processed a training sample). Both processes were imple-
mented for proposed technique (Fig. 2). The implemented technique used a computer with
a microphone. The analyzed DC motor was a source of acoustic signals.
The pattern creation process consisted of following steps: recording of acoustic signals,
soundtrack splitting, normalization of amplitude, feature extraction with the use of MSAF-
RATIO30-EXPANDED, patterns creation. The identification process consisted of the same
steps except the patterns creation. Last step of the identification process is classification of
a test sample. In the classification step, two methods can be chosen – Nearest Neighbour and
Support Vector Machine (SVM). The recording of acoustic signals was based on the sound
card of the computer and condenser microphones such as: OLYMPUS TP-7, ZALMAN
ZM-MIC1. These microphones were chosen because of their low costs. The soundtrack was
recorded on the computer with following parameters of audio format: WAVE PCM, the num-
ber of channels – mono, sampling rate – 44.1 kHz, 16-bit depth. Next soundtrack was split
into smaller audio files (5-second samples). Normalization of amplitude was applied in the
range of [–1, 1]. It divided each point by the maximum value of the sample. The next FFT
spectrum was calculated for each sample (16384 values, because Hamming window was
32768) [25-28]. The calculated FFT spectra were used by MSAF-RATIO30-EXPANDED.
This method was discussed in Section 2.1. At the end of the signal processing, the classi-
fication step was performed.

2.1. MSAF-RATIO30-EXPANDED
MSAF-RATIO30-EXPANDED is a feature extraction method. The author decided to
show a block diagram of the proposed method (Fig. 3).

Fig. 3. Block diagram of MSAF-RATIO30-EXPANDED

Unauthenticated
Download Date | 11/30/16 1:32 PM
736 A. Glowacz Arch. Elect. Eng.

The steps of the proposed MSAF-RATIO30-EXPANDED method were following:

1) Calculate the frequency spectrum of an acoustic signal for each state of the DC motor. The
calculated frequency spectrum of an acoustic signal of the healthy DC motor was a vector
hdc = [hdc1, hdc2, ..., hdc16384]. The calculated frequency spectrum of an acoustic signal of
the DC motor with 3 shorted rotor coils was a vector sdc = [sdc1, sdc2, ..., sdc16384]. The
calculated frequency spectrum of an acoustic signal of the DC motor with 6 shorted rotor
coils was a vector gdc = [gdc1, gdc2, ..., gdc16384]. The calculated frequency spectrum of an
acoustic signal of the DC motor with a broken coil was a vector bdc = [bdc1, bdc2, ...,
bdc16384].
2) Calculate differences between frequency spectra of states of the DC motor: |hdc-sdc|,
|hdc-gdc|, |hdc-bdc|, |sdc-gdc|, |sdc-bdc|, |gdc-bdc|.
3) Calculate a ratio R for each amplitude of frequency. The ratio was defined as:

R = (100%)Xy/Xmax, (1)

where Xy – amplitude of frequency with index y, Xmax – maximum amplitude of frequency

in the sample, R = 30% for MSAF-RATIO30-EXPANDED, Xy, Xmax were based on dif-
ferences between frequency spectra of training samples.
4) Select amplitudes of frequencies, which have the ratio R greater than (30%)Xmax. Some-
times the calculated differences |hdc-sdc|, |hdc-gdc|, |hdc-bdc|, |sdc-gdc|, |sdc-bdc|,
|gdc-bdc| may have maximum values at different frequencies. In this case, selection of
proper frequencies may be difficult. Let's discuss the example MSAF-RATIO30-EXPANDED
when following frequencies are selected 80, 120, 160, 200, 240, 280 Hz for |hdc-sdc|; 80,
130, 170, 210, 250, 290 Hz for |hdc-gdc|; 80, 160, 180, 190, 210 Hz for |hdc bdc|, 80, 160,
240, 320 Hz for |sdc-gdc|, 80, 110, 160, 220, 310 Hz for |sdc-bdc|, 160, 210, 230, 340 Hz
for |gdc-bdc|. It can be noticed that frequencies 80 Hz and 160 Hz are the best (5 common
frequencies). For this reason, the parameter CF was introduced. This parameter expanded
a normal MSAF-RATIO30 method.
5) Set the parameter CF. This parameter was defined as:

CF = (the number of required common frequencies)/(the number of differences). (2)

The parameter CF was responsible for common frequencies. In the example above the
parameter CF was equal to 0.83, then 5 of 6 frequencies were required ((5/6) > 0.83) to
make decision about selection of common frequencies. Then frequencies 80 Hz and
160 Hz were selected. 1 of 6 frequencies was required ((1/6) > 0.16) for the parameter CF
equaled to 0.16. The higher value of the parameter CF the better for analysis.
6) Form a feature vector.
Differences of frequency spectra of acoustic signals of the DC motor |hdc-sdc|, |hdc-gdc|,
|hdc-bdc|, |sdc-gdc|, |sdc-bdc|, |gdc-bdc| were presented in Figs. 4-9.
The author used 3 training sets to select the best frequencies for analysis. Each of them had
4 training samples. There were 18 differences of frequencies spectra (6 for each training set).
Common frequencies of 4 acoustic signals of the DC motor were selected depending on the
parameter CF and training sets (Table 1).

Unauthenticated
Download Date | 11/30/16 1:32 PM
Vol. 65 (2016) Fault diagnostics of DC motor using acoustic signals and MSAF-RATIO30-EXPANDED737

Fig. 4. The difference between frequency spectra of acoustic signal of healthy DC motor and acoustic
signal of the DC motor with 3 shorted rotor coils (|hdc-sdc|) – rotor speed of 500 rpm

Fig. 5. The difference between frequency spectra of acoustic signal of healthy DC motor and acoustic
signal of the DC motor with 6 shorted rotor coils (|hdc-gdc|) – rotor speed of 500 rpm

Fig. 6. The difference between frequency spectra of acoustic signal of healthy DC motor and acoustic
signal of the DC motor with broken coil (|hdc-bdc|) – rotor speed of 500 rpm

Fig. 7. The difference between frequency spectra of acoustic signal of DC motor with 3 shorted rotor
coils and acoustic signal of the DC motor with 6 shorted rotor coils (|sdc-gdc|) – rotor speed of 500 rpm

Unauthenticated
Download Date | 11/30/16 1:32 PM
738 A. Glowacz Arch. Elect. Eng.

Fig. 8. The difference between frequency spectra of acoustic signal of DC motor with 3 shorted rotor
coils and acoustic signal of the DC motor with broken coil (|sdc-bdc|) – rotor speed of 500 rpm

Fig. 9. The difference between frequency spectra of acoustic signal of DC motor with 6 shorted rotor
coils and acoustic signal of the DC motor with broken coil (|gdc-bdc|) – rotor speed of 500 rpm

Table 1. Common frequencies of 4 acoustic signals of the DC motor

depending on the parameter CF and training sets
CF = 0.98 (6 common frequencies) Frequency [Hz]
Common frequencies –
CF = 0.83 (5 common frequencies) Frequency [Hz]
Common frequencies 293, 371, 880, 895, 1049

Amplitudes of frequencies 293, 371, 880, 895, 1049 Hz (hdcv = [hdc219, hdc277, hdc657,
hdc665, hdc783], sdcv = [sdc219, sdc277, sdc657, sdc665 sdc783], gdcv = [gdc219, gdc277, gdc657, gdc665
gdc783], bdcv = [bdc219, bdc277, bdc657, bdc665 bdc783] formed feature vectors for CF = 0.83. The
calculated feature vectors hdcv, sdcv, gdcv, bdcv were processed by the Nearest Neighbour
classifier and Support Vector Machine.

2.2. Nearest Neighbour classifier

The classification was the last step of recognition of an acoustic signal of the DC motor.
The problem of classification was already discussed in literature [29-45]. Neural networks
were described in many scientific articles [37-42]. However the author decided to use the
Nearest Neighbour classifier [2, 36, 43, 44] and Support Vector Machine [45], because they
had high efficiency of recognition of similar problems. The Nearest Neighbour classifier was

Unauthenticated
Download Date | 11/30/16 1:32 PM
Vol. 65 (2016) Fault diagnostics of DC motor using acoustic signals and MSAF-RATIO30-EXPANDED739

very good to classify a high dimensional feature vector. The author used the Manhattan dis-
tance for the Nearest Neighbour classifier. Of course other distance functions such as: Jac-
quard, Euclidean [36] and Minkowski distances were also proper for recognition of an acous-
tic signal of the DC motor. The author decided to use the Manhattan distance, because the
results of recognition for other mentioned distance functions were similar. The Manhattan dis-
tance M was expressed as:
5
M (hdcv, sdcv ) = ∑| (hdcv - sdcv ) | ,
i i (3)
i =1

where feature vectors hdcv = [hdc219, hdc277, hdc657, hdc665, hdc783] and sdcv = [sdc219, sdc277,
sdc657, sdc665 sdc783]. The calculations were performed for 4 feature vectors hdcv, sdcv, gdcv,
bdcv.

2.3. Support Vector Machine

A support vector machine (SVM) was a very well-known method of classification. It con-
structed a hyperplane. SVM classified feature vectors by finding the best hyperplane that sepa-
rated all vectors between the classes. The mentioned hyperplane had the largest distance be-
tween training vectors (support vectors) of any class. The decision function was:

svm = ∑ w f (sv , x) + q ,
i
i i (4)

where wi were the weights, f was a kernel function, svi were the support vectors, x was a fea-
ture vector, q was the bias.
The advantages of SVM was: it used support vectors in the decision function (memory
efficient), SVM used different Kernel functions for the decision function, classifier has high
efficiency in high dimensional spaces. More about SVM was available in the literature [45].

3. Analysis of acoustic signals of the DC motor

The author used 4 acoustic signals of the DC motor: a healthy DC motor, a DC motor with
3 shorted rotor coils, a DC motor with 6 shorted rotor coils, a DC motor with broken coil. The
DC motor had a rotor speed of 500 rpm. Motor power was P = 13 kW (Fig. 1). Parameters of
the DC motor were depended on states:
– healthy DC motor: Uav = 42.8 V, Iac = 56.7 A, Uev = 159.3 V, Iec = 2.5 A,
– DC motor with 3 shorted rotor coils: Uav = 42.6 V, Iac = 58 A, Uev = 160.9 V, Iec = 2.5 A,
Icsc = 49 A (Fig. 10),
– DC motor with 6 shorted rotor coils: Uav = 42 V, Iac = 57.5 A, Uev = 162 V, Iec = 2.5 A,
Icsc = 106 A (Fig. 11),
– DC motor with broken coil: Uav = 42.7 V, Iac = 58.25 A, Uev = 163.4 V, Icsc = 2.5 A
(Fig. 12),

Unauthenticated
Download Date | 11/30/16 1:32 PM
740 A. Glowacz Arch. Elect. Eng.

where Iec is the excitation current of the DC motor, Iac is the armature current of the DC mo-
tor, Icsc is the current of the short-circuit, Uev is the excitation voltage of the DC motor, Uav is
the armature voltage of the DC motor.
Three and six loops of rotor coils were shorted out by resistance RscDC = 0.5 Ω. The author
used 12 training samples (3 training sets) and 80 test samples for the analysis of acoustic
signals of the DC motor. Each sample had duration of 5 seconds (220500 values, window was
32768, so the author calculated FFT from 5 windows). Training and test samples were proces-
sed. Next amplitudes of frequencies 293, 371, 880, 895, 1049 Hz were obtained and classified.
Efficiency of recognition of an acoustic signal was calculated by the following formula:
Np
EF = 100% , (5)
Na

where: EF is the efficiency of recognition of acoustic signal. Np is the number of properly

recognized test samples, Na is the number of all test samples.
The results of recognition of acoustic signals with the use of MSAF-RATIO30-EXPANDED,
Nearest Neighbour were presented in Table 2. The value of EF was from 85% to 100%.

Fig. 10. Rotor windings of the

DC motor with 3 shorted coils

Fig. 11. Rotor windings of the

DC motor with 6 shorted coils

Fig. 12. Rotor windings of the

DC motor with broken coil

Unauthenticated
Download Date | 11/30/16 1:32 PM
Vol. 65 (2016) Fault diagnostics of DC motor using acoustic signals and MSAF-RATIO30-EXPANDED741

Table. 2. Results of recognition of acoustic signals of the DC motor using

MSAF-RATIO30-EXPANDED and Nearest Neighbour classifier
State of DC motor EF [%]
Healthy DC motor 100
Motor with 3 shorted rotor coils 90
Motor with 6 shorted rotor coils 100
Motor with broken coil 85

The results of recognition of acoustic signals with the use of MSAF-RATIO30-EXPANDED,

SVM were presented in Table 3. The value of EF was from 80% to 95%.

Table 3. Results of recognition of acoustic signals of the DC motor

using MSAF-RATIO30-EXPANDED and SVM
State of the DC motor EF [%]
Healthy DC motor 95
Motor with 3 shorted rotor coils 90
Motor with 6 shorted rotor coils 80
Motor with broken coil 80

4. Conclusions

In this paper the author proposed the early fault diagnostic method of the DC motor. The
proposed method used acoustic signals of the DC motor. A method of feature extraction called
MSAF-RATIO30-EXPANDED was presented and analyzed. The analysis was conducted out
for the healthy DC motor and DC motor with early faults (a rotor speed of 500 rpm). Efficien-
cy of recognition of acoustic signals of the DC motor was better for the Nearest Neighbour
classifier than SVM. It was from 85% to 100%. The proposed early fault diagnostic method
was not expensive (price of computer – 300$). The presented early fault diagnostic method
can be used for protection of DC motors. It can found application for monitoring electrical
motors in mining and fuel industry. The further analyses should be performed for different
types, sizes and parameters of DC motors. The next researches will be carried out for other
diagnostic signals such as: thermal radiation, vibration, currents and voltages.

Acknowledgements
This work has been financed under AGH researcher grant in 2016 (Adam Glowacz).

References
[1] Glowacz W., Glowacz Z., Diagnostics of separately excited DC motor based on analysis and reco-
gnition of signals using FFT and Bayes classifier, Archives of Electrical Engineering, vol. 64,
no. 1, pp. 29-35 (2015).

Unauthenticated
Download Date | 11/30/16 1:32 PM
742 A. Glowacz Arch. Elect. Eng.

[2] Glowacz A., Glowacz Z., Diagnostics of stator faults of the single-phase induction motor using
thermal images, MoASoS and selected classifiers, Measurement, vol. 93, pp. 86-93 (2016).
[3] Glowacz A., Recognition of Acoustic Signals of Loaded Synchronous Motor Using FFT, MSAF-5
and LSVM, Archives of Acoustics, vol. 40, no. 2, pp. 197-203 (2015).
[4] Gonzalez-Cordoba J.L., Granados-Lieberman D., Osornio-Rios R.A. et al., Methodology for over-
heating identification on induction motors under voltage unbalance conditions in industrial pro-
cesses, Journal of Scientific & Industrial Research, vol. 75, no. 2, pp. 100-107 (2016).
[5] Wegiel T., Sulowicz M., Borkowski D., A distributed system of signal acquisition for induction
motors diagnostic, IEEE International Symposium on Diagnostics for Electric Machines, Power
Electronics & Drives, Cracow, POLAND, pp. 261-265 (2007).
[6] Irfan M., Saad N., Ibrahim R., Asirvadam V.S., An on-line condition monitoring system for induc-
tion motors via instantaneous power analysis, Journal of Mechanical Science and Technology, vol.
29, no. 4, pp. 1483-1492 (2015).
[7] Krolczyk G.M., Krolczyk J.B., Legutko S., Hunjet A., Effect of the disc processing technology on
the vibration level of the chipper during operations, Tehnicki Vjesnik-Technical Gazette, vol. 21,
no. 2, pp. 447-450 (2014).
[8] Carletti E., Miccoli G., Pedrielli F., Parise G., Vibroacoustic measurements and simulations
applied to external gear pumps. An integrated simplified approach, Archives of Acoustics, vol. 41,
no. 2, pp. 285-296 (2016).
[9] Li Z.X., Jiang Y., Hu C., Peng Z., Recent progress on decoupling diagnosis of hybrid failures in
gear transmission systems using vibration sensor signal: A review, Measurement, vol. 90, pp. 4-19
(2016).
[10] Perun G., Stanik Z., Evaluation of state of rolling bearings mounted in vehicles with use of
vibration signals, Archives of Metallurgy and Materials, vol. 60, no. 3, pp. 1679-1683 (2015).
[11] Jozwik J., Mika D., Diagnostics of workpiece surface condition based on cutting tool vibrations
during machining, Advances in Science and Technology Research Journal, vol. 9, no. 26, pp. 57-65
(2015).
[12] Lara R., Jimenez-Romero R., Perez-Hidalgo F., Redel-Macias M.D., Influence of constructive pa-
rameters and power signals on sound quality and airborne noise radiated by inverter-fed induction
motors, Measurement, vol. 73, pp. 503-514 (2015).
[13] Caesarendra W., Kosasih B., Tieu A.K. et al., Acoustic emission-based condition monitoring me-
thods: Review and application for low speed slew bearing, Mechanical Systems and Signal Proces-
sing, vol. 72-73, pp. 134-159 (2016).
[14] Figlus T., Liscak S., Wilk A., Aazarz B., Condition monitoring of engine timing system by using
wavelet packet decomposition of a acoustic signal, Journal of Mechanical Science and Technology,
vol. 28, no. 5, pp. 1663-1671 (2014).
[15] Zhang X., Feng N.Z., Wang Y., Shen Y., Acoustic emission detection of rail defect based on wa-
velet transform and Shannon entropy, Journal of Sound and Vibration, vol. 339, pp. 419-432
(2015).
[16] Pleban D., Definition and measure of the sound quality of the machine, Archives of Acoustics.
vol. 39, no. 1, pp. 17-23 (2014).
[17] Jena D.P., Panigrahi S.N., Automatic gear and bearing fault localization using vibration and
acoustic signals, Applied Acoustics, vol. 98, pp. 20-33 (2015).
[18] Jozwik J., Identification and monitoring of noise sources of CNC machine tools by acoustic Holo-
graphy methods, Advances in Science and Technology-Research Journal, vol. 10, no. 30, pp.
127-137 (2016).
[19] Bedkowski B., Madej J., The innovative design concept of thermal model for the calculation of the
electromagnetic circuit of rotating electrical machines, Eksploatacja i Niezawodnosc – Mainte-
nance and Reliability, vol. 17, no. 4, pp. 481-486 (2015).
[20] Glowacz A., Glowacz A., Glowacz Z., Recognition of thermal images of direct current motor with
application of area perimeter vector and Bayes classifier, Measurement Science Review, vol. 15,
no. 3, pp. 119-126 (2015).

Unauthenticated
Download Date | 11/30/16 1:32 PM
Vol. 65 (2016) Fault diagnostics of DC motor using acoustic signals and MSAF-RATIO30-EXPANDED743

[21] Sebok M., Gutten M., Kucera M., Diagnostics of electric equipments by means of thermovision,
Przeglad Elektrotechniczny, vol. 87, no. 10, pp. 313-317 (2011).
[22] Smalcerz A., Aspects of application of industrial robots in metallurgical processes, Archives of
Metallurgy and Materials, vol. 58, no. 1, pp. 203-209 (2013).
[23] Krolczyk J. B., Gapinski B., Krolczyk G. M., Samardzic I., Maruda R. W., Soucek K., Legutko S.,
Nieslony P., Javadi Y., Stas L., Topographic inspection as a method of weld joint diagnostic, The-
nicki Vjesnik-Technical Gazette, vol. 23, no. 1, pp. 301-306 (2016).
[24] Demenko A., Description of electrical machine windings in the finite element space, COMPEL-
The international journal for computation and mathematics in electrical and electronic engineering,
vol. 27, no. 4, pp. 711-719 (2008).
[25] Glowacz A., Glowacz Z., Diagnostics of induction motor based on analysis of acoustic signals with
application of FFT and classifier based on words, Archives of Metallurgy and Materials, vol. 55,
no. 3, pp. 707-712 (2010).
[26] Stepien K., Research on a surface texture analysis by digital signal processing methods. Tehnicki
Vjesnik-Technical Gazette, vol. 21, no. 3, pp. 485-493 (2014).
[27] Duspara M., Sabo K., Stoic A., Acoustic emission as tool wear monitoring, Tehnicki Vjesnik-Tech-
nical Gazette, vol. 21, no. 5, pp. 1097-1101 (2014).
[28] Michalak M., Sikora M., Sobczyk J., Analysis of the longwall conveyor chain based on a harmonic ana-
lysis, Eksploatacja i Niezawodnosc – Maintenance and Reliability, vol. 15, no. 4, pp. 332-336 (2013).
[29] Valis D., Pietrucha-Urbanik K., Utilization of diffusion processes and fuzzy logic for vulnerability
assessment, Eksploatacja i Niezawodnosc – Maintenance and Reliability, vol. 16, no. 1, pp. 48-55
(2014).
[30] Gorny Z., Kluska-Nawarecka S., Wilk-Kolodziejczyk D., Regulski K., Methodology for the con-
struction of a rule-based knowledge base enabling the selection of appropriate bronze heat treat-
ment parameters using rough sets, Archives of Metallurgy and Materials, vol. 60, no. 1, pp.
309-312 (2015).
[31] Kundegorski M., Jackson P.J.B., Ziolko B., Two-Microphone dereverberation for automatic speech
recognition of Polish, Archives of Acoustics, vol. 39, no. 3, pp. 411-420 (2014).
[32] Valis D., Zak L., Pokora O., System condition estimation based on selected tribodiagnostic data,
Quality and Reliability Engineering International, vol. 32, no. 2, pp. 635-645 (2016).
[33] Hachaj T., Pattern classification methods for analysis and visualization of brain perfusion CT
maps, Computational Intelligence Paradigms in Advanced Pattern Classification, Book Series:
Studies in Computational Intelligence, vol. 386, pp. 145-170 (2012).
[34] Jaworek-Korjakowska J., Kleczek P., Automatic classification of specific melanocytic lesions using
artificial intelligence, BioMed Research International. Article Number: 8934242 (2016).
[35] Deptula A., Kunderman D., Osinski P., Radziwanowska U., Wlostowski R., Acoustic diagnostics
applications in the study of technical condition of combustion engine, Archives of Acoustics,
vol. 41, no. 2, pp. 345-350 (2016).
[36] Orzechowski P., Boryczko K., Parallel approach for visual clustering of protein databases, Com-
puting and Informatics, vol. 29, no. 6, pp. 1221-1231 (2010).
[37] Jun S., Kochan O., Investigations of thermocouple drift irregularity impact on error of their inho-
mogeneity correction, Measurement Science Review, vol. 14, no. 1, pp. 29-34 (2014).
[38] Roj J., Cichy A., Method of measurement of capacitance and dielectric loss factor using artificial
neural networks, Measurement Science Review, vol. 15, no. 3, pp. 127-131 (2015).
[39] Zhang Y.G., Yang J.Y., Wang K.C., Wang Z.P., Wind power prediction considering nonlinear
atmospheric disturbances, Energies, vol. 8, no. 1, pp. 475-489 (2015).
[40] Jamroz D., Niedoba T., Application of multidimensional data visualization by means of self-orga-
nizing Kohonen maps to evaluate classification possibilities of various coal types, Archives of
Mining Sciences, vol. 60, no. 1, pp. 39-50 (2015).
[41] Jun S., Kochan O., Kochan V., Wang CZ., Development and investigation of the method for com-
pensating thermoelectric inhomogeneity error, International Journal of Thermophysics, vol. 37,
no. 1, (2016).

Unauthenticated
Download Date | 11/30/16 1:32 PM
744 A. Glowacz Arch. Elect. Eng.

[42] Panek D., Skalski A., Gajda J., Tadeusiewicz R., Acoustic analys is assessment in speech pathology
detection, International Journal of Applied Mathematics and Computer Science, vol. 25, no. 3,
pp. 631-643 (2015).
[43] Marzec M., Koprowski R., Wrobel Z., Methods of face localization in thermograms, Biocyber-
netics and Biomedical Engineering, vol. 35, no. 2, pp. 138-146 (2015).
[44] Jiang Y., Li Z.X., Zhang C., Hu C., Peng Z., On the bi-dimensional variational decomposition ap-
plied to nonstationary vibration signals for rolling bearing crack detection in coal cutters, Measu-
rement Science and Technology, vol. 27, no. 6, Article Number: 065103 (2016).
[45] Hwang D.H., Youn Y.W., Sun J.H et al., Support vector machine based bearing fault diagnosis for
induction motors using vibration signals, Journal of Electrical Engineering & Technology, vol. 10,
no. 4, pp. 1558-1565 (2015).

Unauthenticated
Download Date | 11/30/16 1:32 PM