IRJET- In-Situ Monitoring for Fatigue Crack Detection using Control System and Image Processing

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 06 Issue: 04 | Apr 2019 www.irjet.net p-ISSN: 2395-0072
© 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 4135
In-situ Monitoring for Fatigue Crack Detection using Control System
and Image Processing
Tejas Takalkar1, Divyank Rekhe 2, Vikas Sable3, Sonali Thorat4, Prof. S.R. Kulkarni5
1,2,3,4Department of Mechanical Engineering, Sinhgad College of Engineering, Maharashtra, India.
5Professor, Department of Mechanical Engineering, Sinhgad College of Engineering, Maharashtra, India.
---------------------------------------------------------------------***----------------------------------------------------------------------
Abstract - A method for automatic in-situ monitoring of
specimen for detection of fatigue crack is proposed in this
paper. The system is designed to alarm the user as soon as it
detects the crack. The purpose is achieved by building a
control system using a camera and Raspberry Pi controller.
The crack is identified by an image processing algorithm.
The dimensions of crack are also measured automaticallyby
the algorithm. The aim of project is to alarm the user about
the presence of fatigue crack and display the dimensions of
the crack. Use of controller for image processing instead of
computer makes the system compact.
Key Words: Stroboscope, Control System, Image
Processing, Digital Image Correlation, Crack Detection.
1. INTRODUCTION
Fatigue failure is an importantmodeoffailurein engineering
components. Fatigue failure results mainly due to variable
loading or more precisely due to cyclic variations in the
applied loading or induced stresses [3]. That’swhystructure
monitoring has been popular research topic.Alargenumber
of power plant systems present a general steady torsion
combined with cyclic bending stress due to the self-weight
bending during the rotation or possible misalignment
between journal bearings. The dynamical behaviour ofplate
structures with minor irregularities in the form of crack
inspected in laboratory under uniaxial loading conditions.
The inspection is done by expert engineers based on their
experience to identify any visual signs, thus the results are
based on their personal perspective and may vary from
person to person. Experimental fatigue tests typically
require large testing times. This testing time drastically
increases when one periodically increases when one
periodically has to interrupt fatigue test to manually
measure the crack lengths. [5]
The control system will be achieved with help of sensor to a
measure speed, camera and a controller.Thefailuresoflarge
parts in industrial facilities are the fatigue failurestherefore,
the fatigue data is very important to study material
characteristics. However, it takes much time and efforts to
get the fatigue data. Thus the digital image processing
techniques used to measure growing behavior of fatigue
cracks with reducedtimeandefforts.Continuousmonitoring
of system will be achieved using image processing of the
recorded camera images.
2. LITERATURE REVIEW
Z. Kanovic et al. [1] This paper deals with detection of
broken rotor bar using vibration signal analysis. Vibration
signal analysis is fault detectiontechniquewhichisgenerally
used for mechanical fault diagnosis, such as bearing
problems, gear mesh defects, rotor misalignment and mass
unbalance. Rotor bars can be partiallyorcompletelycracked
during the operation of induction motor, due to stresses,
improper rotor geometry design or some imperfection in
material or rotor production process.Thefaultyoperation of
the particular motor was manifested through high level of
vibration and acoustic noise and decreased momentum
when operating under load, which implied the presence of
broken bar. Vibration signals were collected through
accelerometer. Based on the results obtained, it was
suggested to remove the rotor and inspect it, which
confirmed the fault diagnosis and proved that vibration
analysis can be successfully applied to detection of rotor
faults.
B. Lian, et al. [2] developed an in situ fatigue crack
observing system for rotating bending testing machine. It is
proved that this system can detect a small fatigue crack and
its propagation behavior. For verifying performance of this
observing system, fatigue tests were carried out by using
fatigue specimen having a small artificial defect. In the case
of the rotating bending fatigue tests, since the fatigue
specimen at the minimum cross section is curve inward. By
adopting a configuration where several light sources are
circularly placed around the fatigue specimen, it is possible
to enlighten favorably a larger zone. In order to get
observation in adequate synchronization with the rotation
speed of the fatigue test, stroboscope device is usually set in
order to make the observation of the highest position of the
fatigue specimen, where the maximumtensionstresslevel is
reached. A long working distance microscope and digital
camera, and using combination of a LED stroboscope an
observation system to study fatigue crack initiation and
propagation phenomenon.
Wei Shang et al. [3] perform tests on polymethyl
methacrylate (PMMA) which is installed on the top of
aircraft cockpit canopies. They used stroboscopic principal

for the experiment and charge coupled device (CCD)camera
to observe the objects moving at high speed under
stroboscopic illumination. They programmed the image
acquisition software based on the principle of the intensity
superposition. They entirely usedfatiguetestingmachine, an
image acquisition system,anda synchronouscontrol system.
A three-point bending fatigue test was performed on a
biaxially directional PMMA specimen with a semicircular
notch formed by the stress concentration in the fatigue
testing machine. The length ofthespecimenwas100mm, the
width was 22 mm, the thickness was 22 mm, and the radius
of the notch was 1 mm. The sinusoidal load with the
maximum of 3 kN and the minimum of 1 kN was applied at a
frequency of 1 Hz. They study the fatigue crack propagation
for directional PMMA. The fatigue testing was also
performed on laminated directional PMMA. This research
tracked the growth of fatigue crack in directional PMMAbya
stroboscopic method.
Thomas Thurner [4] proposed method which targets on
detection and measurement of cracks on metallic surfaces
where initial artificial cracks have been applied to the
specimen’s surface prior to dynamic loading ina test rig.The
method was based on the optical observation of the
specimen’s surface under special illumination conditions,
where prior to the test an initial notch as a well-defined
starter point for cracks has been positioned on the polished
metallic surface of the specimen. The developed technique
provides the accurate measurement of crack length with a
resolution in the order of 0.01 mm and better over crack
lengths of 10 mm and more.
The proposed measurement method provides a robust and
accurate detection of crack initiation and the highly
resolving measurement of crack lengths on metallic
specimen with a pre-treated surface in dynamic loading
scenarios. In Contrast to existing methods the developed
methodology provides high measurement resolution by a
non-contacting optical setup and real time on-site
measurement capabilities.
Steve Vanlanduit, et al. [5] This paper introduce a DIC
based approach to monitor the crack propagation during
cyclic fatigue test experiments. Therefore, low-speed (and
therefore reasonably low-cost) cameras can be used. The
sub-sampling principle makes use of stroboscopic
illumination of the specimen. a novel DIC evaluation
algorithm is introduced. In literature many DIC evaluation
algorithms are used in the proposed procedure to evaluate
the sub-pixel displacement values.
S. V. Panin et al [6] have described the algorithm of fatigue
crack detection and deterministic tip position. As most
system like VIC 3D, Strain Master uses macroscopic
observation mode. They have done the procedure for
detection of an Object with the Maximum Area of an image
from the optical flow of images. The data of images is
converted in binary format and then the binary image is
subjected to morphological closuretofill small gapsbetween
continuous objects. The process is finalized by searching for
contours of objects remaining in the image and estimating
their shape. Blurring of imageshasa morepronounced effect
on the error value, which increases to a much greaterextent.
The object with the greatest area is taken as a crack. The
selected image is used to find the crack and then to calculate
its tip location. The proposed algorithm is characterized by
the maximum error of crack detection in extremely noisy
images.
3. ALGORITHM FOR CRACK DETECTION
Chart -1: Flow chart for crack detection algorithm
Fig -1: Steps in crack detection algorithm

The images captured are stored and processed by the
controller. The algorithm used for the crack detection is as
follows.
 The image which is in the RGB form is converted
into the gray image. Then image is blurred by using
Gaussian blur filter to eliminatepossiblesmall noise
in thresholding.
 Then the adaptive thresholding method is used
which will enable the crack isolation irrespective of
the provided illumination. This method improves
the versatility of the algorithm and can be used in
different work environments.
 Morphological open()andclose()functionsareused
to reduce the noise and to close the parts of crack
which are disconnected from each other.
 Areas of each contour are calculated. The crack is
separated from the background noise by using area
thresholding. All the contours smaller than the
minimum area requirement are deleted from the
image.
 Till this step the crack is nearly isolated. Now the
crack is skeletonized using the skeletonize function
.The skeletonized image will have only one pixel
value which gives the exact shape of the crack.
 The exact crack length is found out by taking the
area of the skeletonized image.Theaveragewidthis
calculated by making the assumption that after the
crack is fully elongated to make it straight its shape
will be as rectangle. The remaining noise is deleted
by giving the minimum length requirement to the
contours.
Area of rectangle is given by-
A = l*b
Here,
A=area of crack
l=Crack length
b=Crack width
Thus,
b=A/l
Thus the crack width and length of the crack is found out for
the given image of the crack-The crack satisfies the
requirement of the minimum area, minimum length and
minimum perimeter to area ratio. Thus when the above
conditions are satisfied the crack is detected and thus the
user is alarmed about the presence of the crack by the
controller.
4. DESIGN OF SPECIMEN
Fig -2: Dimensions of specimen
i. Material Specification
 International steel grades
BS: C45, 50CS, 080M46
AFNOR: C45, AF65C45, 1C45
SAE: 1045
 Chemical composition
C : 0.42 to 0.5
Si : <0.4
Mn : 0.5 to 0.8
P : <0.04
S : <0.045
Cr :<0.4
Mo : <0.1
Ni : 0.4
Cr+Mo+Ni : <0.63
 Mechanical Properties
Density : 7872 Kg/m3
Modulus of Elasticity : 201 GPA
Coefficient of Thermal
Expansion :11.7 x 10-6 C-1
Specific Heat Capacity : 486 J/Kg K
Thermal Conductivity : 50.9 W/mK
Electric Resistivity : 1.62x10-7 Ωm
Ultimate Tensile Strength : 625 MPA
Yield Strength : 530 MPA
4. ANALYSIS
The static and fatigue analysis was carried out using Ansys
software . The geometry of thespecimenwasmodelledusing
catia, based on the specimen used in the experiment study.
The 3-D geometric design of the tested specimen isshownin
fig illustrates the 3-D solid FE-mesh of the specimen with
mesh characteristics.
 Results and Discussion

Fig -3: Equivalent stress in the specimen under fatigue
load
Fig -4: Life of the specimen under fatigue load
5. EXPERIMENTAL SETUP
Fig -5: Life of the specimen under fatigue load
As shown in the fig-5 the specimen was mounted between
the two chucks of the uniaxial fatigue testing machine. The
fatigue testing machine used was instron 8801 having
capacity of 100KN. A
27.5 KN fully reversed force was applied on the plate with a
frequency of 5 Hz. A camera was mounted on a tripod stand
and positioned at a safe distance in front of the plate.
Fig -6: Circuit diagram of the setup
Raspberry PI 3 B+ controller was used in the experiment.
The camera used is Canon 600D. The camera was connected
to the controller through a USB cable to continuously feed
the captured images to controller. A 16x2 LCD was
connected to the controller in a 4-bit mode as shown in the
fig-6. A potentiometer was connected to adjust the contrast
of the LCD. A buzzer and two LEDs are all connected in
parallel to each other between pin no. 40 of the Raspberry Pi
and pin no. 36 which is ground.
After the distance between the camera and the plate is fixed
the ratio of actual length to corresponding no. of pixels in
camera representing the length was found out. The exact
shape of the plate which can be observed in the photo is
replicated. The pixels which lie outside the plate specimen
are deleted or simply not taken in account in the image
processing algorithm.
5. EXPERIMENTAL PROCEDURE
The plate was mounted on the setup and a alternating force
of 27.6 KN was applied on the plate by the machine. The
camera is set to take a photo for every 24 seconds delay. The
controller processes the image and decides whether the
crack exists on the specimen or not. The images taken bythe
camera and the images produced after different filters are
applied on them during and the final imagefor eachcycle are
automatically saved in the memory which can be studied
later.

The algorithm keeps running displaying ‘no crack’ on the
LCD until the crack is detected. The controller will
automatically turn on the buzzer anddisplaythedimensions
of the crack on the LCD when the crack is detected.
6. RESULTS
 Initially the small cracks wereinitiatedandcouldbe
observed over the sides of the plate at around 5400
cycles.
 The camera successfully identified the crack and
turned on the alarm before the break down of the
component when the crack grew sufficient enough
on the front side of the specimen.
 The component broke after around 6100 cycles.
Fig -7: LCD displaying the results and beeping alarm
Fig -8: Crack detection result
Fig -9: Results in python IDE
 The average crack length prior to failure calculated
by algorithm was 9.29 mm andaveragecrack width
was 0.54 mm.
 The actual crack length measured byverniercaliper
was nearly 9.4 mm. Thus the results were verified
to have 98.08 % accuracy.
7. CONCLUSION
A compact control system for monitoring the crack on the
plate specimen was successfully designed and constructed.
An experiment was conducted to verify the performance of
the control system. The control system successfully detects
the crack and alarms the user about the presence of the
crack. The length and width of the crack were calculated by
the system. The results verified to have around 98.08%
accuracy.
8. ACKNOWLEDGEMENT
It gives me great pleasure to present a paper on “In-situ
Monitoring for FatigueCrack DetectionUsingControl System
and Image Processing”. I am very much obliged to project
guide Prof. S.R.Kulkarni, Professor in Mechanical
Engineering Department, for helping and giving proper
guidance. His timely suggestions made it possible to
complete this project. All efforts might have gone in vain
without his valuable guidance.
I also wish to thank all the professors and non-teaching staff
of the Mechanical Engineering Department for providing an
excellent environment and facilities.

7. REFERENCES
[1] Z. Kanovic, D.Matic ,Z. Jelicic, M. Rapic, B. Jakovlijevic, M.
Kapetina, Induction Motor Broken Rotor barDetectionusing
Vibration Analysis, 2013
[2] B. Lian , A. Ueno ,T. Iwashita, Development of in-situ
fatigue crack observing system for rotating bending fatigue
testing machine, 2016
[3] Wei Shang, Xiao-jing Yang, Xin-hua Ji,andZhong-xian Liu,
Effect of the Reinforce Process on the Fatigue Fracture
Properties of Aeronautical Polymethyl Methacrylate,2016
[4] Thomas Thurner, Real-TimeDetectionandMeasurement
of Cracks in Fatigue Test Applications, Institute of
Lightweight Design, Graz University of Technology, Graz,
Austria, 2015
[5]Steve Vanlanduit, Joris Vanherzeele, Roberto
Longo,Patrick Guillaume, A digital imagecorrelationmethod
for fatigue test experiments, Acoustics and Vibration
Research Group, Department of Mechanical Engineering,
Vrije Universiteit Brussel, Pleinlaan 2, B-1050 Brussels,
Belgium, 2008
[6] S. V. Panin, V. O. Chemezov, P. S. Lyubutin and V. V.
Titkov, Algorithm of Fatigue Crack Detection and
Determination of Its Tip Position in Optical Images, Tomsk
Polytechnical University, Pr. Lenina 30, Tomsk, 634050
Russia, 2016
BIOGRAPHIES
Tejas Dattatraya Takalkar,
Department of mechanical
engineering, Sinhgad college of
engineering, Maharashtra, India.
“Divyank Vijayrao Rekhe, Department
of mechanical engineering, Sinhgad
college of engineering, Maharashtra,
India.”
“Vikas Sorbajee Sable, Department of
mechanical engineering, Sinhgad
India.”
“Sonali Kundlik Thorat, Department
India”
“S. R. Kulkarni, Professor, Department
India.”

IRJET- In-Situ Monitoring for Fatigue Crack Detection using Control System and Image Processing

More Related Content

What's hot (18)

Similar to IRJET- In-Situ Monitoring for Fatigue Crack Detection using Control System and Image Processing (20)

More from IRJET Journal (20)

Recently uploaded (20)

IRJET- In-Situ Monitoring for Fatigue Crack Detection using Control System and Image Processing