International Journal For Technological Research In Engineering

Volume 6, Issue 7, March-2019 ISSN (Online): 2347 - 4718

FAULT IDENTIFICATION IN ELECTRICAL MACHINES USING

THERMAL IMAGE PROCESSING
Mari Selvam B1, Siva Bala Singh K2, Harish R3, Ms. K.S. Krishna Veni4, M.E.
1,2,3
UG students, 4Assistant Professor,
Department of Electrical and Electronics Engineering
Mepco Schlenk Engineering College (Autonomous), Sivakasi, Tamilnadu, India

Abstract: Recently, the induction motors has attracted more once external fault is occurred it will leads to internal faults
attention in industries as well as in commercial use. The within very short interval of time. There are lot of techniques
design and construction of induction motors are very simple available to detect the fault in the induction motor. Some of
compared to DC and synchronous machines. Whatever the them are motor current signature analysis, vibration analysis
design these machines has the tendency to fail during the and thermographic analysis. Among them motor current
occurrence of fault . This will leads to large economic and signature analysis is the most commonly used method to
safety risk to the place where it is installed. In case of a detect fault in induction motors [1-2]. However motor
fault in machine many protective devices are required to current signature analysis has some drawbacks. One of the
detect and diagnose them. But by using Thermal Image main drawback is that it does not detect the fault if it is
Processing Technique any type of failure can be identified present in the kinematic chain [2]. Similarly vibration
in the earlier stage thereby reducing the maintenance cost. analysis has some drawbacks. The major issue in the
By using infrared camera, thermal images of the machines vibration analysis is it is an invasive method of detecting the
under normal and fault conditions can be captured. This fault since it requires two or more accelerometers to monitor
images are processed by the MATLAB software. The fault the condition of the motor under test at every instant of time
in the machine can be detected by means of identifying [1-4]. In industry point of view this method has only few
hotspots in the image of the machine under running applications. Thermographic analysis is a non-invasive
condition. If any fault is identified by this online technique to detect failure in the machines under test.
monitoring technique then the concern people will be Therefore there is no need to interfering the system to
notified. Therefore fault in a machine can be detected by analyze its operating conditions. In general any objects
non- contact means and prior identification of the fault in having temperature above 0K emits infrared radiations [4-9].
the machine leads to great use of this method to the The intensity level of such infrared radiations represents the
identification of the fault in the industrial machineries in temperature of the various points on the system on which the
earlier stage itself. measurements was done [2-4, 8-10]. The infrared camera is
Index Terms: fault detection, infrared camera, image used to capture the images of the machine under various
segmentation, thresholding. operating conditions. This captured images are analyzed
using MATLAB to identify the hot spot. Thus by using this
I. INTRODUCTION method fault is identified in earlier stage than any other
Industrial machinery mainly use Induction motors for their method of detecting the faults in the electrical machinery [1-
applications because of their low initial cost, low 4]. In electrical machines major fault is occur then necessary
maintenance cost, high starting torque and durability [1]. action have been taken during the minor stages of the faults.
However these motors are very prone to failures and hence For example phase to earth fault may converted into the
making the detection of the fault is a major problem for phase to phase fault if the machine is not isolated in the
industries [1-3]. Fault in the electrical machines may be of earlier stage [7-10]. By using the thermographic analysis this
many types. Fault in the electrical machines primarily problem is completely eliminated. Once the hotspots are
classified into two broad categories. They are internal faults detected then the respective authorities will be notified for
and external faults. Internal faults refers to failures in the further actions through e-mail.
stator or rotor windings which may arises due to any one of
the following reasons. The most common failures are II. THEORETICAL BACKGROUND
malfunction of the stator or rotor winding insulation, inter The infrared cameras are devices that is used to capture the
turn faults in the windings, improper winding connections, radiated heat off the machines under study. This infrared
any broken windings, high stress on the insulating materials cameras does not require any source of radiation for
due to excessive current flow in the windings etc.., Similarly capturing the image, it utilizes the heat radiated from the
external faults arises due to any of the following reasons. The target objects. This method is a non-invasive method of
most common external fault is bearing faults, fault due to detecting the fault in industrial machines. The spectrum of
supply voltage fluctuation, malfunction of the starting infrared camera fall into the infrared region and the main
devices, power quality issues in the supply system, sudden advantage of this method is safety and harmless to the
increase or decrease of large load on the system etc... The environment. Based on the emission of heat from the target
common tendency of failures in the electrical machines are objects infrared camera produce images of the target [7-8].

www.ijtre.com Copyright 2019.All rights reserved. 5125

International Journal For Technological Research In Engineering
Volume 6, Issue 7, March-2019 ISSN (Online): 2347 - 4718

Thermal cameras are calibrated based on some industrial the machines under test. This is done by segmentation of
standards before taking the images of the objects under study images by applying global thresholding. The threshold value
and this is maintained throughout the experiment [5-10]. If for red color region separation is set as per the formula. The
low quality infrared cameras are used to capture the images region of interest is the entire machine for complete analysis
of the machines it is necessary to use temperature sensors in of the machines. In this proposed method we are using the
order to ensure exact thermal radiation from the machines. In rectangular region to select the particular area for analysis
case of high quality infrared cameras are used then there is no from the thermal image hence our approach is a user friendly
need for temperature sensors [4-8]. The effectiveness of this method. Surrounding temperature also influence the thermal
method largely depends upon the quality of the cameras used. image of the machines which can be nullified by proper
adjustment in the settings of the infrared cameras.
According to Stefan-Boltzmann law every object at
temperature T, will radiate energy which is proportional to
the fourth power of the temperature T[1-3].The following
mathematical expression will give the relationship between
temperature and the energy emitted i.e.,
4
W=KeT (1)
where
e - Emissivity of the object
K- Stefan-Boltzmann constant
T-Temperature of the object in Kelvin
W-power emitted in W/m² Fig. 1. Block diagram of proposed method.
Emissivity of the material is the measure of how well the
surface of the material emitting energy as thermal radiation. IV. IMPLEMENTATION
Emissivity constant is considered as 1 for a perfect reflector. Figure 2 shows the flow chart of the proposed method.
But in practice there is no perfect reflector. Similarly
emissivity constant is taken as 0 for a white body. The value
of emissivity ranges from 0-1. The approximate value of
Stefan- Boltzmann constant is 5.67w/m²/K4. In thermal
image the color represents the temperature of the object .If
the image is red in color temperature of the machine is very
dangerous [1-2]. Once the fault is occurred in the machine
especially in the rotor or the stator windings the stress
imposed on the insulating material used in the machine
increase rapidly. In this scenario the machine is strictly
isolated from the mains to avoid large maintenance cost. In
this case the current flowing through the winding increases
and leads to formation of hot spots. The temperature of the
hot spots exceed the normal temperature of the machine by
very large amount. This high temperature gradually spread
over the insulating materials and leads to malfunction of the
materials. By utilizing this temperature the fault is identified
in the earlier stage and the machine is isolated from the mains
for repair as soon as possible. If any abnormalities are
identified then the respective authorities will be notified by
sending an e-mail from the MATLAB.
Fig. 2. Flow chart of the proposed approach. The
III. PROPOSED METHOD experimental set up is done on the 3 phase,
In electrical machines the power loss will happen in the form 3.75 KVA, 415V, 1440 rpm slip ring induction motor. Slip
ring induction motors are employed in places where high
of heat dissipation through the windings and it is mainly due
to the presence of winding resistance. If fault is occurred then starting torque is required and this can be achieved by adding
external resistant to the rotor circuit through the slip rings.
it indirectly means that more heat dissipation happens
We are purposefully introducing a minor fault on the stator
through the windings. Figure 1 shows the block diagram
and capture the thermal images. The successive steps
representation of the proposed method. The heat dissipation
involved in the processing thermal image is clearly shown in
is captured by the thermal cameras. These images are
preprocessed initially. The preprocessing method involves the flow chart shown above. This thermal radiation is mainly
region of interest selection and separation of red region. We due to 3I²R loss. Where I represents the fault current flowing
are focusing only on red region because red color in thermal through the windings. Figure 3 (a) shows the thermal image
image indicates the corresponding high temperature areas in of the machine under normal operating condition. Fig. 2 (b)

www.ijtre.com Copyright 2019.All rights reserved. 5126

International Journal For Technological Research In Engineering
Volume 6, Issue 7, March-2019 ISSN (Online): 2347 - 4718

depicts the thermal image of the machine during stator fault rather than the distributed data. Figure 4 depicts the
condition. histogram of the thermal image shown in fig. 2 (a).

(a) (b)
Fig. 2. (a) Thermal image of the machine under normal
condition, (b) Under stator fault condition.
Figure 3 depicts the thermal image of the faulted bearing. Fig. 4. Histogram of the thermal image
Once the bearing is failed then mechanical balance between The next step is to extract the color channel. In this paper we
the machine and the load will be lost. extract the red color channel from the region of interest. As
already mentioned red color is the direct measurement of the
temperature. Figure 5 represents the red color channel
extraction from the thermal image of stator faulted machine.
Similarly for thermal image corresponds to the different fault
also the red channel is extracted to make the analysis easier.
Figure 6 shows the histogram of the red channel of the
image.

Fig. 3. Thermal image of faulted bearing.

The next step is to choose the region of interest for further
analysis. In our paper we considered the entire machine as a
region of interest for complete analysis of the machine under
study in order to improve the efficiency of the proposed
approach.

The next step is segmentation of the selected region. There

are lot of techniques available to segment the image. The Fig. 4. Red channel extracted from the image
most commonly used method is applying watershed
algorithm on the selected region. The watershed algorithm
divides the image by foreground and background.
Segmentation provides the binary image from the given
grayscale image. In this paper global thresholding method is
used to perform segmentation. There is another one method
of thresholding is available known as local thresholding.
However local thresholding provides less accurate results
than the global thresholding method. Threshold value can be
computed by the following relationship.
Threshold = 0.5 (Max (I) + Min (I))(2)
where
Max (I) -Maximum value of intensity.
Min (I) -Minimum value of intensity.
These two values are obtained from the histogram of the
thermal image. Histogram is nothing but a chart that draws
the intensity distribution of grayscale image. Histogram Fig. 5. Histogram of the red channel
differ from the bar chart by plotting the continuous data The next step is to calculate the intensity value of every pixel

www.ijtre.com Copyright 2019.All rights reserved. 5127

International Journal For Technological Research In Engineering
Volume 6, Issue 7, March-2019 ISSN (Online): 2347 - 4718

in the segmented image. Intensity level of the thermal image M. Loccufier, R. Van de Walle, and S. Van Hoecke,
is directly proportional to the temperature of the machine. “Thermal image based fault diagnosis for rotating
The intensity values are calculated on the grayscale image by machinery,” Infrared Physics & Technology, vol.
using standard function available in the MATLAB. The 73, pp. 78–87, 2015.
intensity values are compared to the critical value of intensity [3] Harishkumar, V. Razick Mohammed, B.
which corresponds to the normal operating temperature of the Mohammed Mujtaba “Detection of hot spots by
machine under test. For this comparison if constructs in the thermal imaging to protect power equipments ’’
matlab function is used. The measured intensity values are Department of Electrical and Electronics
denoted as F. If the value exceed the critical intensity value at Engineering C. Abdul Hakeem College of
one single pixel or group of pixel this result is send to the Engineering & Technology, sep2012.
concerned authorities via e-mail. The clients will see the [4] Zhiqiang Huo, Yu Zhang, Richard Sath, Lei Shu
operating condition of the machine by login to the gmail account “Self-adaptive Fault Diagnosis of Roller Bearings
and it helps to take respective steps in order to safeguard the
using Infrared Thermal Images” Guangdong
machine against all kinds of internal as well as external faults.
University of Petrochemical Technology,
Using sendmail function available in the matlab the result is
send to the respective authorities for immediate remedy. To
Maoming, China Vol.3,2006.
ensure proper function of the sendmail function the internet [5] P. Tavner, “Review of condition monitoring of
connectivity should be maintained properly. rotating electrical ma- chines,” IET Electric Power
Applications, vol. 2, no. 4, pp. 215–247, 2008.
TABLE I. Results [6] R. Das, “Determining the locations of faults in
INTENSITY VALUE distribution systems”, Ph.D.Dissertation, University
CONDITION of Saskatchewan, Saskatoon, Canada, 1998.
(for stator faults only) [7] D. Rossetti, Y. Zhang, S. Squartini, and S.
Collura, “Classification of bearing faults through
Machine is operating time-frequency analysis and image processing,” in
F<230 under normal operating Mechatronics-Mechatronika (ME), 2016 17th
condition. International Conference on. IEEE, 2016, pp. 1–7.
Stator winding insulation [8] A. Glowacz and Z. Glowacz, “Diagnostics of stator
230 ≤ F ≤ 250 temperature exceeds the faults of the single- phase induction motor using
normal value. thermal images, moasos and selected classifiers,”
Measurement, vol. 93, pp. 86–93, 2016.
Machine is under severe
[9] O. Munoz-Ornelas, D.A. Elvira-Ortiz, R.A.
F>250 threat so disconnect the
Osornio-Rios, R.J. Romero-Troncoso, L.A.
motor from the mains.
Morales-Herandez, “Methodology for thermal
Analysis of Induction Motors with Infrared
V. CONCLUSION Thermography Considering Camera Location”, in
Our paper aim is to develop a non-invasive online monitoring Proc. IEEE Ind. Electron. Soc. IECON; 2016.
system for detecting faults in electrical machines. This can be [10] A.S.N. Huda and S. Taib, “Suitable features
achieved with a help of thermal images captured by means of selection for monitoring thermal condition of
infrared camera. The proposed approach can be used for the electrical equipment using infrared thermography”,
fault diagnosis in industrial machinery by identifying the in Infrared Phys. Techn., vol.61, pp. 184- 191,
high temperature area in the machinery and distinguish the 2013.
abnormal operating condition from the normal operating [11] M.E. Sieracki, S.E. Reichenbach, K.L. Webb,
condition very accurately. The region of interest is segmented Evaluation of automated threshold selection
by using global thresholding concept which can provide methods for accurately sizing microscopic
better result. The segmented image is further processed to get fluorescent cells by image analysis, Applied and
the intensity of each pixel. This intensity value is compared Environmental Microbiology, 55 (1989) 2762-2772.
with the normal intensity value and the result is send to the [12] Okuda et al.,“Development of inspection systems
concerned authorities via e-mail succesfully. This proposed for distribution power apparatuses self-running on a
approach is further applied to any electrical apparatus. This overhead ground wire of pole”, IJEE technical
proposed method provide information not only the area report, PE93-56, 199.
where the fault is happen it also provide the effect of this
fault on the other part of the machines.

REFERENCES
[1] M. Younus and B.-S. Yang, “Intelligent fault
diagnosis of rotating machinery using infrared
thermal image,” Expert Systems with Applica-
tions, vol. 39, no. 2, pp. 2082–2091, 2012.
[2] O. Janssens, R. Schulz, V. Slavkovikj, K. Stockman,

www.ijtre.com Copyright 2019.All rights reserved. 5128