Content uploaded by Shreyas Gawde
Author content
All content in this area was uploaded by Shreyas Gawde on Oct 30, 2020
Content may be subject to copyright.
Condition Monitoring Using Image Processing
Shreyas Suryakant Gawde
Electronics & Telecommunication Engineering
Goa College Of Engineering
Farmagudi- Goa, India
shreyasgwd@gmail.com
Sangam Borkar
Assistant Professor- Electronics & Telecommunication
Goa College of Engineering
Farmagudi- Goa, India
Sangam@gec.ac.in
Abstract—Vibrations have been traditionally associated with
trouble in machines. However vibrations are merely symptoms of
good or bad mechanical behaviour. Today these symptoms are
used to detect solve many mechanical problems in rotating
machines. These machines play the most important role in any
industry. Faults results in motor failure causing breakdown and
great loss of production due to shutdown of industry. This in turn
increases the running cost of machine with reduction in
efficiency. Therefore, early detection of fault with diagnosis of its
root cause is needed. This project is an approach for fault
detection of rotating machine using orbit analysis of shaft and
image processing. Several works have been focused on detecting
early mechanical and electrical faults before damage appears in
the motor. All the techniques are very effective their way. In this
project, a new methodology is proposed for identifying if there is
a fault in motor. Through image processing based on orbital
analysis, different faults will be studied, generating
characteristically different patterns that are used for fault
detection.
Keywords—Fault Detection, Image Processing, Orbit
Analysis, Unbalance, Loose Foundation.
I. INTRODUCTION
For many years ( and in many plants till today ), philosophy
has been to simply run the plant until a machine failed, deal
with it and get it back in good condition, running it once again.
If machines failed, they were repaired or a spare was used.
Little thought was given to equipment reliability or predicting
failures. The maintenance department was a huge cost sink and
that was considered the standard part of running the business.
More recently, the philosophy has changed. Now organizations
recognize that it is worth the investment of time and money to
change the maintenance practices to be more proactive and to
work to improve equipment reliability. Great cost savings have
been realized because of this. There are number of approaches
that can be taken for maintaining rotating machines, and often
an organization will practice a number of different philosophies
at once, allowing some machines to fail while being proactive
about others. Understanding these approaches is very
important. Several works have been focused on detecting early
mechanical and electrical faults before damage appears in the
motor. All the techniques are very effective their way.
However, the drawback of them is the complexity in the
motors signal mathematical processing and the cost factor. But
this factor can be ignored to improve the equipment reliability.
The current methods implemented in machine monitoring are:
Motor current signature analysis, Vibration Monitoring,
Thermal Imaging, Oil particle Analysis, Orbit Analysis.
II. PROBLEM STATEMENT
The objective of the project is to implement a new
methodology to detect fault in rotating machines. This is
accomplished using image processing. The fault detection is
done based on shaft orbit analysis. In normal condition, shaft
centerline should coincide with the axis of rotation. If this is
not followed, then there is a possibility of a fault that may have
induced in the rotating machine. This difference is detected
using the plots that the shaft centerline possess. The circular
orbit represents that the machine is in good condition. Elliptical
orbit depicts that there may be Unbalance in the machine. Any
random orbit depicts that there may be a loose foundation in
the machine. A loop shaped or ellipsoid orbit may depict a
misalignment in the machine. This way different faults can be
identified by tracing the path of the shaft orbit using video or
image processing.
III. CONDITION MONITORING
Condition monitoring is the process in which the condition
of the machine in working state, is monitored and interpreted to
know the machine health. The objective of Condition
monitoring is to increase the life of production equipment and
lower the cost of failure. Condition Monitoring is practiced
since more than 30 years. This type of monitoring is more
useful for the critical machines in the company that need to be
continuously monitored. This process needs permanently
mounted sensors. Evolution in technology has led to
innovations of wireless sensors. Cost of such accelerometers
has been drastically reduced increasing the cost effectiveness
of continuously monitored condition analysis. For less critical
equipment, periodic monitoring can be done to establish
mechanical condition of machines so that decisions can be
made in case of repairs.
A. Process of Condition Monitoring
Condition monitoring has three main steps:
Fault Detection
Fault Diagnosis
Reconfiguration
This project stress on the first two steps that is whether the
machine has a fault or no fault. In case a fault is detected, then
fault diagnosis is done in order to find out the possible type of
fault.
B. Condition Monitoring Parameters
Vibration
Noise
Wear particle in Lubricants
Thermal Increase
Motor current.
Depending on this parameter, there are various techniques
implemented in order to find the type of fault in the machine.
C. Vibration Characteristics
Rotating machines such as motors, fans, pumps vibrate in
operating state. Using special sensors and monitoring
electronics, vibration provide a wide range of fault conditions.
The vibration changes as the conditions change. The forces
within the machines cause vibrations which is transferred to
bearings. These forces are the result of rotational and frictional
forces. The cause of vibration, regardless of the type, must be
a force which is changing in either direction or its magnitude:
That is Why EACH CAUSE OF VIBRATION HAS ITS
OWN CHARACTERISTICS.
Displacement
Velocity
Acceleration
Frequency
Phase
IV. ORBIT ANALYSIS
Orbit is nothing but the path of the shaft center line within
bearing clearance. Monitoring orbits provide important and
relevant information about rapidly changing machinery
condition. Orbits are Lissajous patterns of time domain signals
that are simultaneously plotted in the XY coordinate plane of
an oscilloscope or vibration analyzer. Orbit plots can efficiently
be used in vibration diagnosis where other techniques, such as
FFT and time waveform, may not provide sufficient
information specially at lower speeds of rotating machines.
Orbit analysis is carried out and more effective for a limited
number of rotating speed or rpm.
A. Applications of Orbit Analysis
Unbalance
Unbalance is most common fault in rotating machines.
Unbalance is nothing but an unequal distribution of mass in the
machine. This occurs when the shaft's geometric centerline and
the mass centerline do not coincide with each other. In
unbalance condition, shaft centerline does not coincide with
axis of rotation. Unbalance will generally produce 1xRPM
vibration with 90 phase shift between the horizontal and
vertical directions. This will result is ellipse-shaped orbit.
Misallignment
When radial preloads due to misalignment, gravity, fluid
forces and other causes increase in magnitude, the orbit will
become acutely ellipsoid. A bearing preload due to a cocked
assembly can also cause the orbit to have lower amplitude in
one axis that makes the ellipse look thinner. If the preloading
increases further, it will result in the orbits shape to resemble a
number 8 character.
V. EXPERIMENTAL SETUP
A. Block Diagram
Fig 1. Block Diagram
Device under test is the rotating machine eg. electric
motor or pump whose shaft is visible with a dot or point
on it.
Camera module will capture a high resolution slow
motion video of rotating shaft.
Image processing section will process the video using
matlab.
Image processing section includes conversion of video
to frames, detecting the green dot in the frame, plotting
of frames classi_cation of plot according to the fault.
Output will be displayed on the screen.
B. Experimental Setup and Requirements
The main test device is connected to power supply.
The front view of the shaft should be visible.
The camera is placed in front of the shaft center.
The shaft center and camera should have zero angle
between each other
The video should be taken at an equal distance from the
shaft center.
The camera should support slow motion format video.
The camera should be kept steady and undisturbed by
vibrations.
There should be a contrast dot on the shaft centerline
that is to be traced by the camera.
ones the motor is ON, video can be taken of a few
seconds which can be then given as the input to the
Matlab code.
Below given is the image of front view of the set up.
Fig 2. Front view of Experimental Setup
VI. IMAGE PROCESSING
Simulation methodology : Matlab
Captured images will be preprocessed in order to create orbits.
These represent characteristic patterns.
Those patterns will be used in a recognition process of the
type of fault in machine. This is called orbit analysis.
A. Flowchart for Image Processing
VIII. FAULT DETECTION AND IDENTIFICATION
A. Fault Detection
In this section, a stream of video is converted to multiple
image frames. The programming language used is matlab.
Each image frame consists of the dot on the shaft. The
unnecessary part except the dot is again filtered from each
frame. After experimenting the above procedure, the frame
before and after filter is as shown in the figure:
Fig 3. Video Converted to Frames
Fig 4. Filtered Frame
Figure 3 is the frame that is converted from the stream of
videos. Figure 4 is the frame that is filtered using matlab.
After further filtering, the frames are added and plotted so as
to represent a characteristic pattern. This is shown in Figure 5.
The pattern of the plot will be used for recognition of type of
fault in rotating machine. From Figure 5, it is seen that the plot
is circular with no difference in major and minor axis with
area 1151 pixels. This concludes that there is no fault in the
machine. At the same time, figure 6 plot is slightly elliptical.
That means there is fault in machine of figure 6.
Fig 5. No fault
Fig 6. Fault : Loose Foundation
B. Fault Identification
The property of the output in Fig.6 is as follows:
MajorAxisLength: 64.1179
MinorAxisLength: 60.9522
Eccentricity: 0.6226
Area: 2585 pixels
We can see from the data that the Eccentricity is 0.6226.
The eccentricity limit for the machine with no fault is 0.16.
This depicts that there is a fault in the Machine. Area of the
plot is much greater than that of no fault plot. From the orbit
pattern, Eccentricity and area, we can infer that the fault in the
motor is LOOSE FOUNDATION.
Similarly, consider another output as shown in Fig 6. Below:
Fig 7. Fault : Unbalance
REFERENCES
[1] Machinery vibration analysis Ronald L Eshleman, Vibration Institute
2008
[2] International Conference on Recent Advances and Innovations in
Engineering (ICRAIE-2014),
[3] Analysis - Jaafar Alsalae, College of Engineering-University of Basrah.
[4] http://ieeexplore.ieee.org/document/6820383/
[5] http://ieeexplore.ieee.org/document/6493919/
[6] http://ieeexplore.ieee.org/document/6602399/
[7] https://www.scribd.com/document/89468509/Shaft-Orbits
[8] https://www.google.co.in/search?q=orbit+analysis+for+motor+shaftsafe
=stricttbm=ischtbo=usource=univsa=Xved=0ahUKEwjhhfmo5fPTAhV
[9] https://www.google.co.in/url?sa=trct=jq=esrc=ssource=webcd=10cad=rj
auact=8ved=0ahUKEwjhhfmo5fPTAhVHqo8KHaQWB64QFghOMAk
url=http
[10] https://www.google.co.in/url?sa=trct=jq=esrc=ssource=webcd=1cad=rja
uact=8ved=0ahUKEwjhhfmo5fPTAhVHqo8KHaQWB64QFgglMAAurl
=http
[11] https://www.slideshare.net/GEMCIndia=where installkeyphasor
[12] https://www.google.co.in/url?sa=trct=jq=esrc=ssource=webcd=3cad=rja
uact=8ved=0ahUKEwje{i9PTTAhUFQ48KHeNLCBAQFgg3MAIurl=h
ttps59
[13] https://www.google.co.in/url?sa=trct=jq=esrc=ssource=webcd=6cad=rja
uact=8ved=0ahUKEwi-3MHd9T
TAhUIrI8KHWNPAxQQFghNMAUurl = http
[14] https://www.slideshare.net/sbssivabala/1-vibration-basics0
[15] Machinery vibration analysis by Vibration Institute.
Major Axis Length: 68.1179. Minor Axis Length: 53.9952
Eccentricity: 0.6105. Area:2855 pixels
We can see from the data that the Eccentricity is 0.6105.
The eccentricity limit for the machine with no fault is 0.16.
This depicts that there is a fault in the Machine. Also, there
is huge difference in Major and Minor axis length. From the
orbit pattern, Eccentricity and the difference between major
and minor axis length, we can infer that the fault in the
motor is UNBALANCE.
IX. ADVANTAGES
This Increases machine RELIABILITY. Also, it is Cost
effective and requires Less manpower. This is a New
challenging point of view for condition monitoring in the
field of mechatronics.
X. CONCLUSION
Fault detection in Rotating machines is possible with the
new methodology which is proposed in the project using
orbit analysis and Image Processing. Use of high resolution
camera will add on to the effectiveness of the output.
XI ACKNOWLEDGEMENT
I am thankful to Goa college of Engineering and its faculty
for their motivation. I am also grateful to M Insight Services
and its team for their continuous support. My
acknowledgement would be incomplete without praising the
lord, my family and friends.