Page 1 of 6

Wireless Health Monitoring System for Vibration

Detection of Induction Motors
Suratsavadee Korkua1 Himanshu Jain1 Wei-Jen Lee1 Chiman Kwan2
Student Member, IEEE Fellow, IEEE Member, IEEE
1. Energy Systems Research Center, The University of Texas at Arlington, Arlington, TX 76019
2. Signal Processing, Inc. 13619 Valley Oak Circle, Rockville, MD 20850

Abstract — To avoid unexpected equipment failures and
obtain higher accuracy in diagnostic for the predictive
maintenance of induction motors, on-line health monitoring
system plays an important role to improve the system reliability
and availability. Among different techniques of fault detection,
work on motor current signature analysis by using only stator
current spectra has been well documented. In addition, the
recent developments in MEMS technology shows increasing
trend in integrating vibration analysis for fault diagnostic.
Vibration-based detection by using the accelerometer is gaining
popularity due to high reliability, low power consumption, and
low cost. This paper presents the study of vibration due to the
rotor imbalance. The technique of vibration detection and
observation of vibration signal in the 3-phase induction machine
is studied. A novel health monitoring system of electric machine
based on wireless sensor network (ZigBee™/IEEE802.15.4
Standard) is proposed and developed in this paper.
Experimental results of the proposed severity detection
technique of rotor vibration under different levels of imbalance
conditions are investigated and discussed.
Index Terms – ZigBee, wireless sensor network, induction
motors, health monitoring system, vibration detection
I. INTRODUCTION
The focus in most industry is shifting from scheduled
maintenance to the predictive maintenance by constantly
observing and predicting the machine condition in advance
[1]. Predictive maintenance by condition based monitoring of
electrical machine is a scientific approach that becomes new
strategy for maintenance management. Most industrial
motors are being monitored using vibration, current, and
temperature sensors which either provide warning signals or
shut down the system before any catastrophic failure occurs.
Though they are able to prevent permanent damage to the
machine, they can neither predict the usable life of the
equipment nor provide the severity level of the problem. This
resulted in the need of an advance system called on-line
health monitoring system.
Traditionally, health monitoring system is realized in
wired systems formed by communication cables and various
types of sensors. The cost of installation and maintenance are
difficult and expensive especially when the equipments are
not at the same location. To overcome these restrictions,
using wireless sensor networks for monitoring is proposed in
this paper. Wireless sensor network [2]-[3] is a new control
network that integrates sensor, wireless communication and
distributed intelligent processing technology. ZigBee is a new
wireless networking technology with low power, low cost
and short time-delay characteristics. These favorable features
are suitable for our application.
Most recent research has investigated the fault detection
technique primarily based on the Motor Current Signature
Analysis (MCSA) [4]-[5] with various DSP and signal
processing techniques [6]. Although the fault detection by
means of analysis of the current is quite simple, the main
drawback is that the employed technique is only suitable for
steady state operation. Since the rotor imbalance is
mechanical related problem, the possibility of an analysis of
machine vibrations is obvious. In [7], by using machine
vibration analysis, the technique of detecting electrical fault
such as voltage unbalance is explored but the mechanical
fault and the level of the severity are not yet discussed.
This paper proposes and develops a ZigBee based
wireless sensor network for health monitoring of induction
motors. The vibration signals obtained from monitoring
system are processed with signal processing techniques. In
order to predict the level of severity of rotor imbalance, the
vibration detection techniques with suitably modified
algorithms is used to extract detailed information for
induction machine health diagnostic.
II. THE PROPOSED WIRELESS HEALTH MONITORING SYSTEM
As shown in Fig. 1, the proposed wireless health
monitoring system comprises of PWM Inverter, Induction
motor and load station. The three-axis accelerometer is used
to measure the vibration of the motor.
Fig. 1 Schematic diagram of the wireless health monitoring system for
vibration detection

978-1-4244-5602-4/09/$25.00 © IEEE 2010

Authorized licensed use limited to: SRM University. Downloaded on July 20,2010 at 07:18:30 UTC from IEEE Xplore. Restrictions apply.
 Page 2 of 6

The machine health identification can be obtained with the In the proposed system design, we adopt the IEEE
on-line monitoring system. In this proposed system, vibration 802.15.4 standard compliant transceiver CC2430 from Texas
signal from three-axis accelerometer are recorded and stored Instruments [8]. It combines the excellent performance of the
at the base station. The recorded data are fed to monitoring leading CC2420 RF transceiver with an industry-standard
system to evaluate the important machine condition. Signal enhanced 8051 microcontroller (MCU), with 128 KB flash
analysis is used to extract detailed information for induction memory and 8 KB RAM. The CC2430 also includes 12-bit
machine health diagnostic. For example, by calculating and ADC (Analog-to-Digital Converter) with up to eight inputs.
comparing the actual recorded motor vibration signal with the
The sensor nodes are in charge of collecting vibration
baseline vibration quantity, the motor conditions can be
data. Before being sent into ADC, the signals are filtered and
obtained qualitatively. To improve the accuracy of health
amplified as presented in the signal conditioning circuit. The
monitoring system, the voltage measurement, current sensor,
network coordination primarily takes charge to distribute
speed and temperature measurement could be installed.
network address to new rode, and notarize the physical
III. WIRELESS NETWORK SYSTEM DESIGN address, transmit test data. The network coordination can
connect, and send the data to PC through RS232 or USB port.
A. ZigBee™/IEEE802.15.4 Standard C. Software Flow
IEEE802.15.4 standard defines the protocol and The software flowcharts describe the working process of
interconnection of devices via radio communication in a the network coordinator and data collection node. They are
personal area network (PAN). It operates in the ISM shown in Fig. 3 and Fig. 4, respectively.
(Industrial, Science and Medical) radio bands, at 868 MHz in
Europe, 915 MHz in the USA and 2.4 GHz worldwide. The
purpose is to provide a standard for ultra-low complexity,
ultra-low cost, ultra-low power consumption, and low data
rate wireless connectivity.
The system framework for health monitoring system
based on wireless sensor network is made up of data
collection nodes and PAN network coordinator. The data
collection nodes can carry out desired functions such as
detecting the vibration signals, signal quantizing, simple
processing, and the IEEE802.15.4 Standard package framing
to transmit data to the PAN network coordinator. In addition,
they can also receive data frames from other nodes, and then
adding multi-hop information, package framing, and then
transmit the new data frames to the network coordination in
the same manner. Once receive the data, the PAN network
coordinator will upload the receiving data to computer for
further processing and analysis.
B. Hardware Design and Implementation

The hardware framework is illustrated in Fig. 2. It is

comprised of node and base station. The main circuits include
the power supply, sensor and signal conditioning circuits,
flash ROM and RAM memory, serial port interface, and three
LEDs for status indication.

USART PC
Memory Interface Base Station Fig. 3 The flow chart of network coordinator

Signal As network coordinator, the CC2430 and the protocol

SENSOR Conditioning CC2430 CC2430
Circuit
stack is initialized firstly. Once the network is formed
successfully, the network coordination will wait for user
LED LED
interrupt to send the “START” command to the node. If a
node is ready to transmit the data, it will send “READY”
command to the coordinator. Once all packages are received,
Fig. 2 Hardware framework the data will be sent to PC via USB port.

978-1-4244-5602-4/09/$25.00 © IEEE 2010

Authorized licensed use limited to: SRM University. Downloaded on July 20,2010 at 07:18:30 UTC from IEEE Xplore. Restrictions apply.
Page 3 of 6
Fig. 4 The flow chart of data collection node
As data collection node, the CC2430 and protocol stack
also initialize firstly. Once the node joins into network
successfully, it will be put into monitoring state to monitor
whether there are any commands coming from the network
coordinator. It will perform the predefined task if data
collection node receives the “START” command. After end
of ADC conversion, node will store all of sampling data into
RAM memory and then send command back to the network
coordinator. After node finish transmitting all data, it will be
returned back to monitoring state again.
The debugging and programming can be done directly in
the Integrated Development Environment IAR Embedded
Workbench for 8051 [9].
IV. EFFECTS OF ROTOR IMBALANCE ON VIBRATION SIGNAL
Rotor imbalances are common mechanical faults in
induction motors. In general, a mechanical fault in the load
part of the overall system can be observed from the variation
978-1-4244-5602-4/09/$25.00 © IEEE 2010
Authorized licensed use limited to: SRM University. Downloaded on July 20,2010 at 07:18:30 UTC from IEEE Xplore. Restrictions apply.
of the load torque. When a mechanical fault happens, it will
result in a rotating eccentricity at the rotating frequency [10].
These faults may also cause speed oscillations that have the
effect on the stator current and finally lead to additional
undesired harmonic components of power and torque at some
particular frequencies in the spectra.
A. Induction Motor under Rotor Imbalance
To analyze the mechanical phenomena of the induction
motor under rotor imbalance, the dynamic nature of a
mechanical system can be modeled as shown in Fig. 5
θr
Fig. 5 Rotational model of horizontally mounted mechanical system under
rotor imbalance
With the assumption that the mechanical system with the
unbalance is the horizontally installed along its rotational
axis, the rotational dynamic for torque equation is derived
from the free body diagram as Fig. 5. Rotor imbalance can be
expressed by a mass of the total unbalance weight ( m) which
is the main cause of the mechanical failure at the different
speeds. This small mass is fixed on a disk mounted on the
motor shaft. The different distances ( r ) from the center of
the rotating part can be chosen. The gravity force by the
unbalance mass produces the sinusoidal oscillating torque
which is the function of the mechanical rotor
position (θ r ) and the acceleration of the gravity ( g ) at the
distance ( r ) . The electromagnetic torque (Te ) is produced by
an induction machine and the rotational friction term is
determined by the viscous damping constant and mechanical
rotor speed that can be represented with the following
equation;
Jθ&&r + bθ&r = Te − mrg sin θr (4.1)
We can see that the torque equation is as a function of time
which is comprises of the constant component torque and the
additional component varying at the mechanical rotor
position. The total electromagnetic torque that is produced by
the unbalance load makes it possible to detect the mechanical
fault in induction machine.
B. Vibration Signature
In general, there are numerous sources such as external
source with particular frequencies that can cause the vibration
 Page 4 of 6

of rotating machinery Other vibrations do occur due to

rotating asymmetries, such as rotor imbalance or
misalignment. Among them, the largest low frequency
vibration harmonic is normally due to the rotor imbalance of
the rotational parts. From the vibration spectrum analysis, the
low frequency harmonics are associated with the rotational
frequency and can be distinguished from the lower range of
the spectrum. The mechanical vibration due to rotor
imbalance is a once per revolution force. Moreover, induction
motors also vibrate at the twice of supply frequency due to
their magnetic force between rotor and stator. This attraction
force acts on the stator and induce the vibration on the motor
frame. This also leads to a significant harmonic peak at twice
of the supply frequency in the vibration spectrum. Therefore,
the harmonics of rotor imbalance can be modeled as an
integer multiple of rotating frequency [10];

 1− s 
fvib = k . f s ( ) (4.2)
 p 

Where f s is stator supply frequency, k is the integer

number, p is the number of pole pair, and s is the slip.
Based on the vibration spectrum analysis, it is normally
straightforward to locate the mechanical rotational frequency
by monitoring the vibration spectrum and finding the most
significant peak in the rotational frequency range expected.
V. DETERMINATION OF ROTOR IMBALANCE SEVERITY
Machine vibration analysis becomes one of the most
important tools for machine health diagnostic. There are two
types of analysis: time domain and frequency domain. In
most research, the frequency domain analysis is more
attractive because it is provide more detailed information
about the status of the machine whereas the time domain
analysis can give qualitative information about the machine
condition. In general, machine vibration signal is composed
of three parts, stationary vibration, random vibration, and
noise. Traditionally, Fourier transform (FT) was used to
perform such analysis. If the level of vibrations and the noise
are high, inaccurate information about the machine condition
may be obtained. In this paper, Fast Fourier Transform (FFT)
is used to extract some useful features of the vibration signal
i.e. root mean square (RMS) value, the crest factor. All the
techniques used here for signal analysis and processing have
been implemented by MATLAB software. Block diagram of
proposed severity detection of rotor imbalance is shown in
Fig. 6.
A. Implementation of Root Mean Square
In this study, the RMS value of the vibration signal is
used for primary investigation of the machine health. The
RMS values will be used to detect the severity of the
abnormal condition. These values could also be used as input
to training the neural network based fault classifier. All 3-axis
vibration data are investigated.
Fig. 6 Block diagram of the proposed severity detection technique of rotor
imbalance
B. Implementation of Crest factor
The crest factor is the ratio of the peak value of the
vibration signal to the RMS value. The purpose of the crest
factor calculation is to give an analyst a quick idea of how
much impacting is occurring in a waveform. It is meaningful
where the peak values are reasonably uniform and repetitive
from one cycle to another. Crest factor is often used to
indicate the rolling element bearing faults. The rotor
imbalance faults do not affect the value of crest factor. These
values can give qualitative information about the machine
condition if it is compared to the vibration signal standard
and can be used to differentiate the bearing fault and rotor
imbalance fault. The values of crest factor of the vibration
signal for healthy case will be use as baseline condition in
health monitoring system. Continuous monitoring of crest
factor along with the historical record can provide useful
information about bearing condition. Using both the RMS
value and the crest factor for the machine health diagnostic
may increase the overall system performance for health
monitoring system.
C. Implementation of spectrum analysis
In the present work, Fast Fourier Transform (FFT)
algorithm is used to perform discrete Fourier transform
(DFT) for all three axis vibration signals. In order to achieve
earliest possible recognition of the fault in the machine, a
comparison of the spectrum of the machine with the spectrum
of the healthy case must be performed. The most significant
vibration peak at the rotational frequency will be determined.
The amplitude of the FFT spectrum at the rotational

978-1-4244-5602-4/09/$25.00 © IEEE 2010

Authorized licensed use limited to: SRM University. Downloaded on July 20,2010 at 07:18:30 UTC from IEEE Xplore. Restrictions apply.
Page 5 of 6

frequency serves as the rotor imbalance indicator of the 3

proposed monitoring system. Furthermore, the amplitude of

this frequency component does reveal the severity level of the
2.5
rotor imbalance fault.
VI. EXPERIMENTAL SETUP AND RESULTS

RMS Vibration Signal (V.)

2
In order to validate the proposed vibration based health
monitoring system, the test-bed for mechanical fault was set
up. Rotor imbalance was created on a 3-phase, 2-pole and 1.5

1h.p squirrel cage induction motor. Induction motor is fed

from the variable speed drive at 50Hz. The wireless sensor 1
network is implemented and accelerometer is also integrated
in the system for detecting the vibration signal. Vibration
signals were collected using ADXL330 tri-axial 0.5 X-axis
accelerometer mounted on the motor housing. Axes of Y-axis
acceleration sensitivity corresponding to machine vibration Z-axis

are shown as in Fig. 7. The most important element of this 0

0 5 10 15 20
test-bed is the flywheel which has holes drilled in it (Fig. 7.) Weight of imbalance mass (g.)
The weights applied to these holes produce imbalance in the Fig. 8 RMS value of vibration signals from different imbalance mass tests
flywheel, thereby in the motor. The severity of the fault is
determined by the weight. In order to observe the frequency component amplitude of
vibration signals related to rotor imbalance fault, FFT
algorithm is used to perform for vibration signal. The 3-axis
vibration spectrum component amplitude at rotating
frequency (50 Hz) is shown in Fig. 9 and the amplitude of
vibration spectra versus the different weights of imbalance
mass are presented in Fig. 10.
X-axis Vibration Spectrum

60
0 g.
40 5 g.
10 g.
Fig. 7 Fly wheel design used in rotor imbalance test
20 15 g.
In this test, rotor imbalance was created using four 20 g.

different weights namely, 5g, 10g, 15 and 20g. No fault data 0

0 10 20 30 40 50 60 70 80 90 100
was also collected as baseline data and considered as 0 g. All Frequency(Hz)
3-axis vibration data were calculated the RMS values. As
Y-axis Vibration Spectrum

60
shown in Fig. 8, the RMS values are increasing
corresponding to the level of imbalance mass. However, the 40
change in crest factor is small as represented in Table 1. It
implies that the degree of impacting is relatively small from 20
bearing faults.
0
0 10 20 30 40 50 60 70 80 90 100
TABLE I
Frequency(Hz)
CREST FACTOR OF THREE-AXIS VIBRATION SIGNALS
Z-axis Vibration Spectrum

30
FOR DIFFERENT IMBALANCE MASS TESTS
Crest factor 20
Imbalance
mass (g.) X-axis Y-axis Z-axis
10
0 g. 1.457313235 1.7216206 2.18396705
5 g. 1.451261113 1.77381552 1.82287472 0
0 10 20 30 40 50 60 70 80 90 100
10 g. 1.387397149 1.50183948 1.63720601 Frequency(Hz)

15 g. 1.303934686 1.40032313 1.40077586 Fig. 9 3-axis vibration spectrum component amplitude at rotating frequency
from different weights of imbalance mass
20 g. 1.249204942 1.45098092 1.29577839

978-1-4244-5602-4/09/$25.00 © IEEE 2010

Authorized licensed use limited to: SRM University. Downloaded on July 20,2010 at 07:18:30 UTC from IEEE Xplore. Restrictions apply.

It can be noticed that the rotational harmonic at 50 Hz has
a dominant value. In addition, by increasing the weight of
imbalance mass, the amplitude of this frequency component
will apparently increase in the spectrum. Therefore, the
spectrum amplitude can be used to specify the degree of fault
for the certain operating condition.
60
Vibration spectrum component amplitude at 50 Hz
50
40
30
20
10
0 [3] I. F. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci,
0 5
Weight of imbalance mass (g)
Fig. 10 The 3-axis magnitude of vibration spectra at rotating frequency
versus the different weights of imbalance mass
Furthermore, based on the analyzed vibration data and the
linear approximation, the relationship between the rotor
imbalance indicator and the severity level can be represented
in Fig. 11. It is important to note that any estimation is
subject to error. However, this relationship can be use as
trend to determine the severity level of the fault. For example,
the change of rotor imbalance indicator during the time
interval T0-T1 can be used to predict the range of the severity
level. The result from the prediction will be a very useful part
of the condition monitoring system and the estimation on the
usable life of the equipment.
Fig. 11 Rotor imbalance indicator of different levels of imbalance severity
Authorized licensed use limited to: SRM University. Downloaded on July 20,2010 at 07:18:30 UTC from IEEE Xplore. Restrictions apply.
Page 6 of 6
VII. CONCLUSION
The hardware and software design of a wireless health
monitoring system for induction machine is presented in this
paper. Vibration signals have been analyzed to detect the
mechanical faults. The implementations of analysis technique
in time and frequency domain are given. The proposed rotor
imbalance detection technique is verified with different level
of severity. Rotor imbalance indicator can be used to estimate
the range of severity level which is very useful part of the
predictive maintenance. The wireless health monitoring
system is tested under various operating conditions and is
found to work satisfactorily.
REFERENCES
[1] B. Lu, T. G. Habetler, and R. G. Harley, “A survey of efficiency
estimation methods of in-service induction motors with considerations
of condition monitoring,” in Proc. 2005 International Electric
Machine and Drive Conference (IEMDC), May 2005, pp.1365-1372.
X-axis [2] J. A. Gutiérrez, E. H. Callaway, and R. L. Barrett, “Low-rate wireless
Y-axis personal area networks: enabling wireless sensors with IEEE
Z-axis 802.15.4,” IEEE Press, New York, 2003.
10 15 20
“Wireless sensor networks: a survey,” Computer Networks, vol. 38,
no. 4, 2002, pp. 393-422.
[4] T. G. Habetler, R. G. Harley, R. M. Tallam, S. B. Lee, R. Obaid, and
J. Stack, “Complete current-based induction motor condition
monitoring: stator, rotor, bearings, and load,” in Proc. 2002 VIII IEEE
International Power Electronics Congress, Oct. 2002.
[5] C. M. Riley, B. K. Lin, T. G. Habetler, and R. R. Schoen, “A method
for sensorless on-line vibration monitoring of induction machines,”
IEEE Trans. Industry Applications, Vol. 34, pp.1240-1245, Nov. 1998.
[6] S. A. Saleh, A. Kazzaz and G. K. Singh, “Experimental investigations
on induction machines condition monitoring and fault diagnosis using
digital signal processing techniques,” in 2003 Electrical Power
Systems Research, pp.197-221.
[7] G. S. Maruthi and K. P. Vittal, “Electrical Fault Detection in three
phase squirrel cage induction motor by vibration analysis using
MEMS accelerometer,” in Proc. 2005 IEEE Power Electronics and
Drive System (PEDS), pp.838-843
[8] Chipcon TI, “A True System-on-Chip solution for 2.4 GHz IEEE
802.15.4 /ZigBee® CC2430 DataSheet (Rev.2.1),” 2008
[9] IAR Systems, “IAR Embedded Workbench IDE User Guide,”
ftp://ftp.iar.se/WWWfiles/guides/ide/ouew-3.pdf, December 2004.
[10] C. Kral, T. G. Habetler, and R. G. Harley, “Detection of mechanical
imbalances of induction machines without spectral analysis of time
domain signal,” IEEE Trans. Industry Applications, Vol. 40, pp.1101-
1106, July. 2004.
978-1-4244-5602-4/09/$25.00 © IEEE 2010