ABSTRACT

There is a constant increase in the Internal heat of electrical devices due to variety of reasons
ranging from incorrect loading, contact defects ,defects in insulation and protection, problems with
the relays or terminal blocks, etc until final failure which can lead to unpredicted outages and
possible harm. Thus an earlier knowledge is required for prior prevention of thermal defects in
electrical equipment to avoid failure. These days a quick, reliable, non-contact and cost effective
infrared thermographic inspection system is being used widely for determinig where and how
internal and external defects occur in equipment. The system gathers data for analysing thermal
condition through the careful analysis of thermal images of electric equipments. The images are
taken by infrared camera without interrupting the running operation of system. Both precautionery
and predictive maintenance programs are chosen for defect diagnosis. The number of data points
and sequence of inspection of thermal images depends on various factors such as thermal
condition, cost of a failure, type of production and age of equipment, etc. The objectives of this
article are to discuss about precautionary and predictive maintenance, thermal defects in electrical
equipment and also a smart thermal defect identification system for accelerating the predictive
defect diagnosis technique. The intelligent system proposes the use of artificial neural network and
statistical features to determine the existence of defect within equipment by monitoring its thermal
condition. For the working of the neural network, the optimum features were chosen as the inputs,
using discrimant analysis. The performance of the neural network was compared with the
performances results of discriminant analysis classifier and it showed that discriminant analysis
classifier outperformed with an accuracy of 82.40%.

INTRODUCTION

All objects with a temperature greater than absolute zero emit infrared radiation and cause a rise in
temperature. The naked eye is unable see the infrared energy because it is emitted as heat energy.
Infrared thermography is a technique for transforming invisible heat to a visual image which shows
the heat iamge of the object surface. The colour of the object surface in the heat image depends on
the surface temperature. An infrared camera also known as thermal imager is used to capture the
heat image and measures temperature gradient of the object surface [1,2]. For several years,
thermographic inspection has become an crucial tool for precautionary and predictive maintenance
of surface defect in various materials due to its non-contact,non-invasiveness, safety and relatively
cheaper approach. Thus number of studies were executed to prove thermography as a useful
technique in various practical applications [3e11]. Heat energy is an crucial factor in electrical
equipment for improving operational reliability. Electrical current traverses through a resistive
component and produces heat. Thermal energy produced from an electrical component is directly
proportional to the square of the current passing through it times the resistance (I 2 R Loss). Thus a
rise in electrical resistance causes in an overall increase in heat. With progress in time the condition
of the electrical component goes on deteriorating due to various reasons such as loose connection,
corrosion and insulators crack As a component deteriorates, its resistance increases and also
produces more heat. The increase in heat energy can lead the failure of electrical equipment and
also fire may break out. By using thermography for inspecting the equipment under loading
condition, the defective component can be spotted and classified by its level of severity [12].
Infrared thermography detects the heat which is produced in the electrical equipment. The thermal
profiles of electrical devices and connectors are captured using an infrared camera. Thermal profile
which is also called thermogram consists of a thermal image and a temperature scale. The different
colours of temperature scale represent the different temperature regions of the equipment. Heating
components are generally detected as red or brighter coloured areas in an infrared image. A circuit
breaker (CB) thermogram determined in the current study is shown in Fig. 1. In Fig. 1A, the
brighter area (R phase) shows the higher resistance and results overheating of that phase. The most
probable cause of this high resistance could be a bad or dirty connection. Fig. 1B shows the thermal
profile across the line L0 showing the variation of temperature in the equipment surface. Thus
errors in components are identified and then recorded. The maintenance programs are planned and
executed before breakdown of the electric equipment. Maintenance cost of electrical equipment
plays an crucial role for reducing the total operating cost of electrical power system. The
maintenance cost can be increased due to the lack of information about the maintenance scheduling
and type of maintenance needed for the electric equipment. Usually the operating performance of
electrical system is not monitored periodically which can play an crucial role in early prevention of
a fault and for increasing the usage of equipment. Instead, the schedule for maintenance is decided
by using the failure history of equipment or according to the prior experiences of maintenance
personnel. Infrared thermography is used to keep an eye on the performance of equipment which
provides a complete information on the operating status of equipment for increasing production,
reducing or removing unnecessary repairs, avoiding eventual failure and reducing the maintenance
cost of equipment. Usually, applying infrared thermography technology, the electrical thermal
defect and degree of defect are identified by monitoring its Delta T (DT) criteria [13]. The method
is known as qualitative based temperature measurement system [14]. The Delta T criteria of a
component is defined as a rise in temperature value above the temperature of a reference value,
which is usually the ambient temperature, temperature of a similar component under the same
condition or the maximum allowable temperature of the component [15]. Several set standards for
DT value are found such as Electrical Testing Association (NETA) [15], National Fire Protection
Association (NFPA) e NFPA 70-B [16] and InterNational American Society for Testing &
Materials (ASTM) e E [17], etc. Sometimes thermographers prefer to group thermal condition
based on their own inspection experiences instead of using the available standards. Artificial neural
networks (ANNs) are more powerful, robust and adaptive mathematical model for pattern
identification and classification [18,19]. Nowadays ANNs are being used to solve difficult non-
linear practical world problems. Several articles about the application of artificial neural networks
(ANN) for electrical fault diagnosis system using infrared thermography have been put forth in the
past few years. In these articles, different features were utilized as inputs of ANN. Almeida et al.
[20] used a neuro-fuzzy approach for fault detecting of lighting arrester while the inputs of the ANN
were thermographic and identification variables. The system was capapble of resulting index of
wrong diagnosis lower than 10% for classifying thermal condition into normal, suspicious and
faulty classes. Shafi’i and Hamzah [21] put forth the application of multilayered perceptron
network. The RGB colour scale data and temperature data were utilized to classify internal faults
into four different classes namely low, intermediate, medium and high. The result obtained 99.38%
of accuracy. Rahmani et al. [22] developed an smart system using support vector machine (SVM) as
a classifier and 22 image features of Zernike moments for study of ground substation errors. The
diagnosis was used on 20 thermal images and was able to obtain 68.42% of accuracy. Smedberg
[23] and Wretman [24] proposed an intelligent classification system based on ANN for detecting
three-phase fuses and different forms of connections defects. The four input parameters of ANN
were needed which are absolute maximum temperature, relative maximum temperature, mean
temperature difference compared to the other regions of the image and histogram distance to the
other regions of the image. The test error rate using all four different feature parameters as input of
ANN was 9.5% and the error rate using only the histogram distance as input was 31.2%. The set of
data was 74 infrared images. One of the disadvantages of working with a small dataset is that the
reliability of the results can become questionable. This paper will study the application of infrared
thermography technology as predictive/preventive maintenance program to detect the presence of
thermal errors in electrical equipment. The technique uses multilayered perceptron network,
statistical features and discriminant analysis classifier to characterize the thermal status of hotspot
into ‘defect’ and ‘no defect’ groups. Discriminant analysis classifier was utilized to select suitable
features and also to produce the linear regression model of each group using suitable features.
2. Preventive/predictive maintenance

Typically, for maintaining the equipment, two different approaches,

either run-to-failure or preventive maintenance are used.
The run-to-failure approach is easy and straightforward, where
there is no outflow of money on maintenance before
eventual breakdown of equipment. This method waits for equipment
failure before any maintenance procedure is taken. Thus the
method is more costly. Thermography based diagnosis system
allows the preventive/predictive maintenance for early prevention
of equipment failure without stopping the running operation
and saves money.
According to the historical data in USA [25], proper use of
preventive/predictive maintenance will cutdown about 33% to 50% of
maintenance cost that are wasted by most of the manufacturing
and production factories and also introduced the following advantages
from various aspects:

1. Cutdown in unscheduled downtime or system ultimate failure

2. Rise in the proper utilization of manpower
3. More production capacity of the plant
4. Decrease in the maintenance expenditure of equipment
5. Increased life time of equipment

2.1. Preventive maintenance

Maintenance of electrical devices according to the statistical

or historical data of operating capacity, prior failure history,
mean-time-to-failure (MTTF) instead of tracking device performance
is referred as precautionary maintenance [25]. precautionary
maintenance program schedules repairing and rebuilding activities
for device. Suppose, an electrical equipment normally runs 10
months before needed any sort of maintenance or repairing. While using preventive methods,
the equipment will be discarded from usage and rebuilt after 10
months of use. In this example, the time table of repairing of
equipment is already decided which is determined by maintenance
personnel. Now, if there is no requriement to rebuild after 10 months, then,
there would be wastage of labour and material. Again, if the equipment
fails before 10 months it is required to fix problem
after failure which is usually more costly than schedule maintenance.
This is a just random decision which is not supported by
any kind of actual data.

2.2. Predictive maintenance

Maintenance of electrical devices based on the direct monitoring
of the actual operating condition, daily collecting
measurements, efficiency of the equipment, heat distribution and other indicators
instead of relying on statistical or historical data is termed as predictive
maintenance. The program schedules all maintenance activities
based on actual data and if necessary, repairs before the
system breakdown.
Prior detection of imminent equipment failures and avoiding
them before crucial condition considerably reduces downtime, cost
for maintenance, maximize usage.

3. Defect in electrical equipment

Electrical equipments such as transformers in substations, circuit breaker, fuse

and capacitor, etc can cause defects when its internal operating temperature
reaches at an abnormal level. Most of the anomalies are due to various reasons such as contact
problems, unbalance current distributions,
cracks in insulators, defective relays or terminal blocks,
etc. Some of the examples of the error in electrical system which
can be detected by thermographic technology are mentioned here.

3.1. Transformer
Heat is generated due to the losses in transformer resistance and temperature
increases within the ferrite core and copper windings. The transformer is
cooled by either oil or air and the operating temperature is
higher than surrounding. The cooling system helps to maintain the rise
of temperature within desired limits so as to avoid the breakdown. The thermographic
inspection is used for detecting for thermal defects diagnosis of oil-filled
transformers instead of dry transformers.usually the coil
temperatures of the dry transformers are much greater than
ambient, it is difficult to differentiate the internal defect using thermography
before it becomes serious. Therefore, other monitoring
tools including in built thermal gauge and Dial gauge can
perform more accurate measurement for the dry transformer. In the
oil-filled transformers, thermographic inspection helps in detecting
the defects present at the primary and secondary bushing connections,
cooling fins, cooling fans and internal bushing connections,
etc. The problems can be detected visually by using following methods:

-> Primary and secondary bushing connections.

Over heating can be caused by loose or dirty connections, overloaded connections or when not
balanced . Thermographic inspections
detects hotspots at the defected phase connection compared to
other phases.

-> Cooling fins

When dealing with an oil-filled transformer, cooling mainly depends on
the circulation of oil through the transformer. The oil gets heated up as it carries the heat generated
in the core and windings. Thus cooling fins will be detected to be
hotter compared to the inactive cooling fins. From the thermal
picture , if it is observed that one or more fins has lower
temperature or cooller compared to other warm fins, then it can be
concluded that the circulation of oil is being limited to those fins,
probably due to the lower oil level.

-> Cooling fans

In higher rating transformer, where more heat will be generated compared to other transformers,
cooling fans are used to boost the cooling
rate. In general, the fan in action is normally warmer and thus
thermograph will show whether it is operating or not.

-> Internal bushing connections

If there is any error in the connections or if its dirty or loose,
it will have higher temperature compared to ambient temperature as detected by the thermograph.

3.2. Defect in circuit breaker

A circuit breaker (CB) always detects the heat generated by the
current. If it finds the sudden variation of heat, it instantly opens the circuit
automatically and protects the electrical equipment from
damage. Thermographic inspection has the advantages to find the
defects by analysis the surface temperature variation of circuit
breaker and can prevent these defecrs earlier.
3.3. Fuse connection
Fuses are used to protect the equipment against overloading condition.
Overloading condition in a system can lead to the fuse
connection heating and moreover, the fuse clips can be overheated
due to the loose and dirty connections.
3.4. Defect in contact
Electrical connection switches, breakers and other reclosing
devices make use of contacts. If these contacts become loose,oxidized or dirty this would lead to an
increase in resistance thus leading to defect to be occurred in the contacts.

3.5. Insulation problem

Insulation or protection defects can cause short circuit current between two
conductors wire. Overcurrent generates overheating and allows the circuit
breaker or fuse to open automatically . Poor insulation can lead to increase in resistance thus cause
overheating.

3.6. Utility connection box break elbows

In residential regions with many underground connection and
switching locations are in utility boxes. Usually break
elbows provide a point of disconnection and if these break elbow
connections are not correctly installed or connection resistance
rises, then defects can occur.

3.7. Surge arrestors problem

Problems due to surge protection and lightning arrestors seeping
to ground and current tracking over protection can also be detected
using thermography. However, finding such problems requires the
capture of minute temperature differences which are often too
difficult to be detected.

4. Intelligent defect identification system

4.1. picture acquisition and Delta T criteria analysis

The infrared camera manufactured by Fluke Ti25 was used to

capture the thermogram of device. Its range of temperature
measurement was 20 #Ce350 #C. The thermal picturer consists of a
160 # 120 focal plane array with uncooled microbolometer detector and
functions in the infrared spectral band of 7.5 mme14 mm. The thermal
lens capture pictures of 320 # 240 pixels while the ordinary lens
produce 640 # 480 pixels (visual pictures). For capturing the picture,
the thermal picturer orientation was directly pointing to the target
equipment so as to get an accurate measurement. The seperation
between the target electrical equipment and the thermal picturer
was in the range of 0.5e1.0 m. The emissivity value was set at 0.95
as recommended generally for electrical installations thermography
[26]. It was noted that the ambient temperature around the
equipments was about 30e33 #C during the inspection. A total of
368 thermal pictures of circuit breakers operating at the same
loading condition having 500 hotspots were taken.
In this study, the difference of maximum value between hotspot
and normal (reference) temperature of an equipment was determined
as DT criteria. The method is known as qualitative
temperature measurement system. The main advantage of this method
is that there is a minor effect of emissivity on temperature measurement.
On the other hand, main drawback is that qualitative
method cannot describe the error, if failures of all three phases
occur at the same time due to the absence of normal component
[11].
In this research, depending on the equipment functioning temperature,
initially the thermal conditions of equipments were
manually classified into three categories based on importance level
(priority level 1, 2 and 3) using DT specifications inTable 1. From these
three, the statuses can be further divided into two categories, ‘defect’
(priority 1 and 2) and ‘no defect’ (priority 3) class. The DT values and
required steps regarding the classes have been described in Table 1.
Figs.1e3 show the thermal status of these thermograms. In Fig. 1a
the maximum temperature value of hotspot A0 is 42.9 #C, while
36.9 #C is the maximum temperature value of A1 which is given as
the reference area. The temperature variation (6 #C) between these
two phases reveals that A0 is in ‘defect’ class. Fig. 2a shows the
maximum temperature value of A1 and A2 is 72.7 #C and 64.4 #C
respectively, while 57.5 #C is the maximum temperature value of A0
or reference area. The temperature variation between these two
phases reveals that both A1 and A2 are in ‘defect’ class. Similarly from
Fig. 3, it is seen that both A0 and A1 areas are in ‘no defect’ class with
respect to A2 region. Figs.1b, 2b and 3b show the temperature profile
across line L0 for their respective thermogram.

4.2. Features extraction

First the original thermal picture is transformed into greyscale

picture consisting of black and white regions. The regions of
interest (ROI) for both the reference and hotspot were choosen
manually in a rectangular shape. Fig. 4 shows the grey picture of
thermogram (Fig. 1a), while white or brighter region indicates
errors.
Total fifteen features of first, second order statistical features
from both ROI and also common features (i.e. numerical difference
between the feature value of reference and hotspot of same
device) were evaluated [27,28]. The two first order features to be observed were
were kurtosis, standard deviation and four second order features to be observed
were entropy, contrast, homogeneity and energy. On the other
hand, a total of nine mutual features were evaluated. In the case of
first order, seven were average grey intensity entropy highest grey
value, kurtosis, skewness, standard deviation, and variance. From
second order, two features were entropy and energy. A list of features
is displayed in Table 2.

5. Suitable features selection

After finishing the feature extraction process, all the 15 features
for 500 hotspots act as the inputs of the MLP
network. But the non importatn feature reduces the performance of
the network. Therefore, before using as inputs, suitable feature
selection is a crucial task to eliminate the noisy features and
improve network performance. In the present study, discriminant
analysis (DA) has been proposed to select best features,
determine the performance of each class and also to calculate the
overall performance [29]. The 15 features were denoted as X1, X2, Y1,
Y2, Y3, Y4, X11, X12, X13, X14, X15, X16, X17, Y11 and Y12. The univariate
and multivariate tests were used to study the performance of
features. The features with the p-value distribution less than 0.05
were selected as appropriate features. The univariate test is a test of
equality of each class mean of individual feature which is used to
investigate the p-value distribution of each feature individually.
Table 3 shows the results of univariate test for the original 15 features.
Based on the results, the features Y1 and Y3 have lower impact
for identifying thermal defect with p-value distribution more than
5%. The other features showed higher impact for identifying defect
with p-value distribution less than 5%. Hence, null hypothesis is
rejected for Y1 and Y3 and accepted for other 13 features.
Here, the 13 features which give high impact have been processed
using multivariate test. Table 4 tabulates the p-value
distribution of each feature using multivariate test. From the 13
features, 10 optimum features have been selected using the multivariate
test as suitable features for defect recognition. The ignored
features which have p-value distribution more than 0.05 are
highlighted in Table 4. The 10 optimum features are X2, Y2, Y4, X11,
X13, X15, X16, X17, Y11 and Y12.
Table 5 presents the suitable features with its discriminant coefficients
which will be used during the formation of discriminant
function.
The discriminant function Zp value is obtained by adding the
sum of multiplication between discriminant coefficient and dominant
features values. A function has a multiple root (i.e. a root with
multiplicity greater than one) in the complex numbers, if and only
if its discriminant is zero [30]. From Table 5, the discriminant
function is derived as follows:
Based on the Zp value, the class mean called centroid is determined.
Table 6 shows the centroid (mid-point) for each thermal
class. Then, the type and range of each class are found by determining
the separation points between the classes. The separation
point must be established, since the value can be used to identify
the range for each class of the thermal status.
From Table 6, it is identified that the centroid of ‘no defect’ class is
less than ‘defect’ class. So, the separation point between the ‘no
defect’ and ‘defect’ class is found [(1.122 þ 0.642)/2] ¼ 0.24. The
‘no defect’ class is present at the separation point which is less than
0.24 and ‘defect’ class is located at the separation point which is
more than 0.24.
Apart from choosing the suitable features, the discriminant
analysis is also able to predict the linear regression model of each
class taking the account of classification function coefficients and to
classify the thermal class (i.e. determine the number of correctly
classified data, the number of misclassified data and the accuracy
for each class). Table 7 shows the classification function coefficients
of each class for all selected features.
Now the linear predictive model of each class is obtained by
adding the sum of multiplication between classification function
coefficient and dominant features values. The predictive model for
‘defect’ class, Dp can be written as Finally, the discriminant analysis is used to distinguish the
thermal
status of hotspots into two classes namely ‘no defect’ and ‘defect’.

Table 8 shows the classification accuracy for thermal condition

classification.
The overall data are 500; each data has 10 optimum features.
The classification accuracy is 81.1% for ‘defect’ class (258 thermal
conditions correctly classified out of 318) and 84.6% for ‘no
defect’ (154 out of 182). The overall accuracy is calculated as
follows:
7. Results and discussion

To obtain the performance of the MLP network

for electrical equipment thermal defect detection (i.e. ‘no
defect’ or ‘defect’ class), the 500 samples from the thermal images
were seperated into training and testing dataset. For seperating the
training and testing dataset, the 5-fold cross validation analysis
method was used [32]. In this method, entire data are divided
into 5 segments or folds. The summary of data samples in each fold
is tabulated in Table 9. Five iterations of training and testing are
evalauted. In each iteration, a single fold data is used for testing
and the remaining four folds are used for preparing the network.
Finally, the average value of each fold performance is taken as

the network performance. The MLP network was trained by varying

hidden nodes from 1 to 20, keeping epochs constant at 300 and
learning rate at 0.005. For the MLP performance analysis, the analysis
introduces the accuracy, specificity and sensitivity percentages
analysis. The definition of accuracy, specificity and sensitivity are
written as follows:
f MLP network trained using LM algorithm. In the training phase,
MLP network produced 88.20%, 89.789% and 85.418% of accuracy,
sensitivity and specificity respectively. From all the folds, the results
using second fold as training dataset provided the best performance
regarding accuracy, sensitivity and specificity which are 97.75%,
98.88% and 95.89% respectively.
Similarly in the testing phase, the same network provided accuracy
80.40%, sensitivity 83.98% and specificity 75.29%. Therefore,
it is stated that the MLP network trained using LM algorithm and
also the proposed 10 features have higher capacity of detecting the
defect in electric equipments.
Table 12 shows the performance relation between discriminant
analysis classifier and multilayered perceptron network. It was seen that the
DA produced better specificity and accuracy compared to MLP
network, while MLP network was seen to achieve higher sensitivity. The value
of sensitivity, specificity and accuracy difference between two
classifier are 2.88%, 9.31% and 2% respectively.
In this paper, an intelligent predictive diagnosis system for
detecting invisible thermal defects in electrical equipment by
infrared thermography technology is proposed. The diagnosis system
employed the multilayered perceptron network with LevenbergeMarquardt
training algorithm to classify the thermal
conditions of equipments into two classes namely ‘no defect’ and
‘defect’. Total fifteen features including first, second order statistical
features and also mutual features (i.e. the numerical difference
between the value of hot and reference features) were calculated
from sample images. Using discriminant analysis feature selection
method, 10 optimum features were chosen as the inputs of the
neural network. The performance of the neural network was
compared with the discriminant analysis classifier. The comparison
results showed that the discriminant analysis classifier produced
overall better performance with 82.40% of accuracy. By developing the
system, the technical department can produce a schedule with the recommended
actions. This will play a significant role for quick and dependable inspection to avoid equipment
failure potentially.