Condition Based monitoring system using IoT

Karan Muvvala Ashwin Nair

Mechanical Engineering Department Mechanical Engineering Department
Rajiv Gandhi Institute of Technology Rajiv Gandhi Institute of technology
Mumbai, India Mumbai, India
karanmuvvala@yahoo.in ashwinjnair97@gmail.com

Amol Mangrulkar Harsh Mistry

Mechanical Engineering Department Mechanical Engineering Department
Rajiv Gandhi Institute of Technology Rajiv Gandhi Institute of Technology
Mumbai, India Mumbai, India
amol.mangrulkar@mctrgit.ac.in harshvm3396@gmail.com

Sanaul Mukadam
Mechanical Engineering Department
Rajiv Gandhi Institute of Technology
Mumbai, India
sanaulsultan@gmail.com

Abstract— With the advent of Internet of Things (IoT), a new gaining insights regarding the machine has become more
industrial revolution is around the corner. Industries are tapping feasible [2]-[5]. Predictive maintenance has emerged as one of
this unprecedented technology to improve their manufacturing the most popular trends in the industry and is quickly gaining
processes. In this paper we demonstrate the integration IoT popularity to improve efficacy and reduce erratic machine
technology with Industrial Induction Motors. A network of downtime, improving the labor productivity while efficiently
sensors continuously monitors the “health” (viz. Current, Voltage, utilizing time dedicated to maintenance of machines. Condition
Vibration, Temperature, shaft alignment & rpm) of the Motor. based monitoring is a primary step to gain insights about a
This data is sent to a dedicated microprocessor unit(DHU) for machinery while in service. The key to successful
remote online condition monitoring. The DHU is connected with a implementation of Condition based maintenance relies upon the
laptop (wirelessly) to analyse the data collected in MATLAB to
accurate means of condition assessment and fault diagnosis.
gain insights regarding the health of the motor. Condition based
maintenance(CBM) can be used to conduct predictive Electrical motors have been in use since the dawn of the
maintenance when the motor requires rather than conduct industrial revolution, acting as the backbone of the
scheduled maintenance. A DHU is also integrated with a Global manufacturing process. Three phase induction motors are widely
System for Communication (GSM) modem to convey the operator used in industries due to their high reliability, feasibility and
of the health of the motor in emergencies when immediate wide range of torque output available with simple build and
attention is required. maintenance. As their failures could have catastrophic
consequences financially and productivity wise, safe and
Keywords— Internet of Things (IoT), Condition based
monitoring(CBM), induction motor faults reliable working of these motors should be ensured. CBM using
IoT uses measurements taken while a machine is operating, to
determine if a fault exists [6]. Different types of sensors can be
I. INTRODUCTION used to measure signals to detect various faults. Various signal
Internet of Things(IoT) is an upcoming technology whose processing techniques can then be applied to these sensor signals
application are becoming more pervasive as we progress. It is to extract particular features which are sensitive to the presence
basically a network of physical devices talking to each other via of faults. Finally, in the fault detection stage, a decision needs to
the help of electronics such as sensors, microprocessors etc. One be made as to whether a fault exists or not with previous
of its greatest advantage is its ease of integration with the knowledge [7]-[8].
existing technology and machinery. The use of sensors with the
reduction in cost and size has augmented the way machines II. INDUCTION MOTOR FAULTS
communicate with each other and has led to a manufacturing
revolution. Manufacturing industries are tapping into this Induction motor faults are broadly classified into internal
unprecedented technology to improve its process productivity and external faults. Internal faults are comparatively difficult to
and efficiency [1]. Connectivity and data acquisition are key discern and require sophisticated tools to detect. They are
parameter that act as a foundation for smarter, more further categorized as mechanical and electrical faults.
sophisticated and intelligent machines. With the advent of data,
 Fig. 1. Problems in a) Three Phase Induction Motor b) Motor Stator and C) Motor Rotor

efficiency and better utilization of available funds. Various

A. Bearing fault condition monitoring techniques are as follows.
These are one of the most common types of faults which
account for almost 40% of all induction machine failures. 1) Vibration Monitoring: It is widely used to detect
Bearing defects are generally observed to be distributed or mechanical faults. A 3-axis accelerometer is used to determine
localized. These defects create a series of impact vibrations at an the vibrations in each direction. This data is then used to plot
instant when a running roller passes over the surface of a defect the FFT diagram to detect bearing faults, rotor faults and air gap
whose period and amplitude are calculated by the anomaly’s eccentricity.
position, speed and bearing dimension as per the formulas 2) Motor Current Signature Analysis (MCSA): It is a more
derived. These frequencies are unique for every bearing recent method used to detect stator faults by capturing the
dimension and rotating speed.
motor’s current and voltage signals and analysing them to
𝑓𝑐𝑏𝑓 = | 𝑓𝑠 ± 𝑛𝑓𝑟𝑐 | (1) detect various faults.
𝑁𝑏 𝑓𝑠 𝐷𝑏
𝑓𝑟𝑐 = [1 ± cos 𝜃] (2) With the increase in demand of condition monitoring,
2 𝐷𝑃
various other methods have also been formed such as extended
park’s vector approach(EPVA) and instantaneous power
signature analysis (IPSA) which are developed to detect
B. Stator Fault
particular problems. All the techniques discussed above are
It accounts for almost 37% of the total induction faults. generally applied to detect the following defects in an induction
General stator faults are turn – to – turn, coil – to – coil and coil motor.
– to – ground faults. As a three-phase motor consists of stator
slots filled with winded copper wires; they occur mainly due to
interturn winding faults caused by insulation breakdown.
(1−𝑠)
𝑓𝑐𝑏𝑟𝑓 = 𝑓𝑠 [𝑛 𝑃
± 𝑠] (3)

C. Rotor Fault
These faults generally occur due to rotor windings and
account for almost 10% of the total induction faults. The
common type of rotor fault is broken rotor bar which usually
occurs due to pulsating loads. These faults lead to fluctuations
in speed, vibrations and pulsated torque outputs.

III. CONDITION MONITORING AND TECHNIQUES

Condition monitoring is the continuous evaluation of the
various crucial parameter that act as the governing parameters to
determine the “health” of the induction motor. These parameters
herald the inception of faults that would eventually lead to its
failure. Thus, it is crucial to monitor and look for these faults as
the reduction in downtime of these faults leads to higher plant Fig. 2. Comparison of Condition Monitoring techniques
 IV. SYSTEM DESCRIPTION V. METHODOLOGY
Data Acquisition is a key step in predictive maintenance. As In this proposed study, we develop a novel process of
with most technologies, sensors are swiftly becoming feasible condition monitoring system. It consists of four major
economically feasible, miniscule and more advanced. Sensor- sub-systems or blocks viz. DHU, GSM module, Remote PC and
driven computing or Continuous remote monitoring of critical phone that function independently. DHU acts as a partially
elements of a motor displays the current “health” of the motor autonomous block and collects data from the array of sensors for
and helps us gain critical insights related to the motor. Various further processing. This data is then wirelessly sent to a remote
sensors such as hall sensor, proximity sensor, accelerometer, PC with an assigned operator. The operator with his past
temperature sensor and CT coil were used to monitor the various experience and the visual tools monitors the parameters and
parameters of the motor. detect anomalies if any. He then further investigates the
anomalies detected in the FFT spectrum to decipher the root
In this experiment we developed an independently function cause of the anomaly [13].
monitoring system using raspberry pi 3B integrated with the
above-mentioned sensors. Each sensor has its own sensor The GSM module is used to send emergency signals to the
response time and all the sensors are integrated with the operator. This is done autonomously as the threshold values are
Raspberry Pi GPIOs with respective pull-up resistances, predefined and eliminates the need of continuous attention of the
capacitors and potentiometers. This micro-processor based operator giving DHU-GSM module partial autonomy.
hardware unit is termed as a dedicated hardware unit(HDU).
The data transmitted by the sensors is collected and then stored Since, a motor’s health deteriorates over a long range of
time, we ran the experiments with various motor conditions. A
to a pc which is wireless communicating with the Raspberry Pi
module. This data is analyzed using MATLAB using methods brand-new motor was initially used to detect the values which
such Motor Current Signature Analysis(MCSA) and Vibration would be further used as a benchmark. A faulty motor with
analysis. broken rotor bar was then used. A significant change was noticed
in the FFT frequency spectrum as shown in the figure below.

TABLE I. SENSOR LIST WITH THEIR RESPECTIVE FAULT DETECTION Further a slotted disc with variable weights were introduced
with a speed regulator to measure the change in the amplitude of
Sr. Sensor Name Test Fault Detection vibrations induced in the bearing and the motor mounting.
No.
Mainly two major non-invasive techniques i.e MCSA and
1 ADXL 345 Motor and bearing Shaft or bearing Vibration analysis were used to detect major rotor faults.
Accelerometer vibration housing vibration
MCSA is mainly used to detect broken rotor bars, shorted
2 PR30-15DN2 Shaft Misalignment Air-gap eccentricity
Inductive
turns and air gap eccentricity while vibration analysis was used
proximity to detect major mechanical faults such as bearing fault, gear
Sensor faults, unbalanced rotors and rotor eccentricities. With
3 Hall Sensor Motor RPM Digression in shaft
advancement in sensor technology leading to feasible and
RPM inexpensive accelerometers, this method is swiftly gaining
popularity. After collection of the data it is processed using the
4 LM 35DZ Stator Temperature Indirect measure of
thermocouple winding temperature
MATLAB software to obtain a frequency spectrum using FFT.
As the frequency spectrum often contains a lot of noise produced
5 CT Coil Supply voltage Rotor faults due to the presence of multiple elements, a lot of professional
experience with pre-established frequency values of each
A GSM Sim900A module is used to inform the operator anomaly should be known beforehand. Also, the load condition
about the critical parameters that cross a pre-defined threshold and speed are perpetually varied thus altering the fault signature
value which is set by the operator on the basis of previous characteristics as well. A lot of test samples were initially
experiences [10]. The message is intended to draw the attention conducted to detect a fault’s frequency associated with it [14] –
in case the operator is not active at the display monitor. The [15].
circuit diagram of the connection is as shown in Fig. 3.
The GSM supports extended AT commands and has an
inbuilt COM port with RS232 protocol-based interfacing
facility. The micro-processor initially sends an “AT” command
as sequence initiation to which an “OK” response is sent by the
GSM modem [11]. The following steps are performed to retrieve
the PIN (personal identification number) and checks if the SIM
card is ready and if the link is established. Further using the AT
commands, the required messages are conveyed to the
concerned authority.

Fig. 3. Block Diagram of the experimental setup

Fig. 4. Schematic Diagram of the experimental setup

Since, a motor’s health deteriorates over a long range of frequency of the amplitude and Y vector being the amplitude
time, we ran the experiments with various motor conditions. A value for each signal or frequency. But the generated spectrum
brand-new motor was initially used to detect the values which includes only the magnitude information about each frequency
would be further used as a benchmark. A faulty motor with component. The signature analysis of a bro rotor bar in Fig 5 and
broken rotor bar was then used. A significant change was noticed Fig 6.
in the FFT frequency spectrum as shown in the figure below.
With an increase in broken rotor bars, the frequency of the
Further a slotted disc with variable weights were introduced with fault also increased. For 1 ,2,3 and 4 broken rotor bars the fault
a speed regulator to measure the change in the amplitude of frequencies were 57 Hz, 94 Hz, 141 Hz and 208 Hz respectively.
vibrations induced in the bearing and the motor mounting. This trend has been graphically displayed in Fig. 7
Mainly two major non-invasive techniques i.e MCSA and
Vibration analysis were used to detect major rotor faults.
MCSA is mainly used to detect broken rotor bars, shorted
turns and air gap eccentricity while vibration analysis was used
to detect major mechanical faults such as bearing fault, gear
faults, unbalanced rotors and rotor eccentricities. With
advancement in sensor technology leading to feasible and
inexpensive accelerometers, this method is swiftly gaining
popularity. After collection of the data it is processed using the
MATLAB software to obtain a frequency spectrum using FFT.
As the frequency spectrum often contains a lot of noise produced Fig. 5. FFT spectrum of Loaded heavy motor with defects
due to the presence of multiple elements, a lot of professional
experience with pre-established frequency values of each
anomaly should be known beforehand. Also, the load condition
and speed are perpetually varied thus altering the fault signature
characteristics as well. A lot of test samples were initially
conducted to detect a fault’s frequency associated with it.

VI. RESULT AND DISCUSSIONS

In this experimental setup, an off the shelf 3-axis ADXL 345
accelerometer is used due to its cheap, portable and accurate
measurements and can measure vibration in three different
orientation X, Y and Z. The readings are updated every 0.5 Fig. 6. FFT spectrum of loaded healthy motor
seconds. Its mounting position is of paramount importance as it
would affect the amplitude of vibrations being recorded. These A similar trend is observed when the load on the motor is
reading are then saved in an excel sheet. The data is read using increased with the increase in its RPM. Since the fault is not a
‘xlsread’ and used to plot a graph with X vector being the spurious event and continues to degrade the mechanical stability
of motor over time it is of primary concern to check the condition monitoring system with IoT technology has
intermitted hikes in amplitude of frequency. Using other widely unprecedented potential which the industries could use to
available sensors with the integration of IoT acts a catalyst to improve the productivity of their system.
improve the maintenance techniques of the critical components
and eliminated the use of scheduled and preventive REFERENCES
maintenances.
[1] S.K.Sethiya “Condition Based maintenance(CBM)” , irsme vol1. 2012
[2] S. Devendiran, Dr. K. Manivannan “Vibration Based Condition
Monitoring and Fault Diagnosis Technologies For Bearing and Gear
Components-A Review” International Journal of Applied Engineering
Research ISSN 0973-4562 Volume 11, Number 6 (2016) pp 3966-3975
[3] Adrian D. Nembharda, Jyoti K. Sinhaa, A. J. Pinkertonb, K. ELBHBAHA
“Fault Diagnosis of rotating machines using vibration and bearing
temperature mesaurements” Diagnostyka, Vol. 14, No. 3 (2013)
[4] G. Suresh Babu, Dr. V. Chittaranjan Das, “Condition Monitoring and
vibration analysis of boiler feed pump”,International Journal of Scientific
and Research Publications, Volume 3, Issue 6, June 2013
[5] Khadim Moin Siddiqui, Kuldeep Sahay, V.K.Giri, “Health Monitoring
and Fault Diagnosis in Induction Motor- A Review”,IJAREEIE Vol. 3,
Issue 1, January 2014
[6] Jinia Datta(Das), Soumyajit Datta, Sumana Chowdhuri and Jitendranath
Bera, “Condition based reporting of multi-machine monitoring system
Fig. 7. Variation of fault frequency with increase in broken rotor bars using GSM”, IJSRP, Volume 2, Issue 6, June 2012
[7] N. B. Shaikh, Prof. S. S Dhamal, S. S. Hadpe, K. P.Varade, “Online Off-
Site Condition Monitoring Of Three Phase Induction Motor by Using
VII. CONCLUSION AND FUTURE SCOPE GSM Technology”, IJETAE
The objective of this paper was to develop a novel and [8] Chen YR. Modern signal processing technology in the application of
vibration diagnosis of internal combustion engine Ph.D. thesis, Wuhan
feasible technique of condition monitoring of critical machines University of Technolog Jayaswalt P, Wadhwani AK. Application of
used in daily life. Electrical Induction motor being the backbone artificial neural networks, fuzzy logic and wavelet transform in fault
of the industries and other service sectors form an integral link diagnosis via vibration signal analysis: A review. Australian Journal of
of the whole system and thus its reliable functioning is a must. Mechanical Engineering 2009; 7: 157-172.y, 1998.
This paper successfully achieved the target of integrating the [9] Krause PC (1986) Analysis of electric machinery. Mc-Graw Hill, New
monitoring process with IoT technology. Here motor monitored York
from remote places will result in better manpower utilization and [10] Sen PC (1989) Principles of electric machines and power electronics.
labor efficiency. It offers major benefits to both, the individual Willey, New York
users as well in industry in terms of reliability, efficiency and [11] Say MG (2002) The performance and design of alternating current
machines. M/S Pitman, London. ISBN 81-239-1027-4
cost saving. This technique could be further implemented in
[12] Kothari DP, Nagrath IJ (2010) Electric machines. Tata McGraw Hill
industries for major cost savings and avoiding major downtime Education Pvt. Ltd., New Delhi. ISBN-13 978-0-07-069967-0
of critical equipment. [13] Motor Reliability Working Group (1985) Report of large motor reliability
Data acquisition is an important step that that helps use this survey of industrial and commercial installations Part I, and II. IEEE
Trans Ind Appl IA-21(4):853–872
data to further gain more insights about the machine. Condition
[14] Thorsen OV, Dalva M (1995) A survey of faults on induction motors in
monitoring techniques can further be integrated with neural offshore oil industry, petrochemical industry, gas terminals, and oil
networks to predict the remaining life of the components. It can refineries. IEEE Trans Ind Appl 31(5):1186–1196
used to classify the faults in case of unsupervised data sets which [15] Thomson WT, Gilmore RJ (2003) Motor current signature analysis to
is common in the induction motor faults. Data acquisition is a detect faults in induction motor drives-fundamentals, data interpretation,
key step for predictive maintenance and as well as to detect and industrial case histories. In: Proceedings of the thirty-second
faults. Thus, developing an independent and economical turbomachinery symposium, pp 145–156