See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/285593369

Condition monitoring of electric motors based on magnetometer

measurements

Conference Paper · October 2015

DOI: 10.1109/ETFA.2015.7301553

CITATIONS READS

2 866

6 authors, including:

S. Wildermuth Pawel Rzeszucinski

ABB ABB
43 PUBLICATIONS   1,490 CITATIONS    25 PUBLICATIONS   69 CITATIONS   

SEE PROFILE SEE PROFILE

Maciej Orman
ABB China
24 PUBLICATIONS   92 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

EUCAMS gear partnership View project

All content following this page was uploaded by Maciej Orman on 15 December 2015.

The user has requested enhancement of the downloaded file.

 Condition monitoring of electric motors based on
magnetometer measurements
Stephan Wildermuth, Ulf Ahrend, Christoph Byner
ABB AG
Corporate Research Center Germany
68516 Ladenburg, Germany
stephan.wildermuth@de.abb.com
Maciej Orman
ABB Ltd
Corporate Research Center China
Shanghai, P.R. China
maciej.orman@cn.abb.com
Abstract— Using micro-sensors in industrial applications is of
great interest due to their small size, low-cost and little power
consumption. However, the harsh environmental conditions
encountered in an industrial environment have so far hindered
the widespread use of, for example, MEMS-based sensors. Such
sensors are particularly suited for mobile condition monitoring of
industrial machinery as short time placement and operation of
these sensors is typically unproblematic for many monitoring
applications. In this paper we use a miniature triaxial
geomagnetic sensor for condition monitoring of low voltage
motors. The performance of the magnetometer is studied under
conditions encountered in industry. Furthermore the
magnetometer is used to measure magnetic fields of an electric
motor in a healthy state and in case of a broken rotor bar. By
frequency analysis of this data it is demonstrated that the
magnetometer measurements can be employed to distinguish
between these motor conditions.
Keywords—condition monitoring; MEMS sensor; electric
motor; low-cost magnetometer; frequency analysis
I. INTRODUCTION
Condition monitoring of electric motors is becoming more
and more important as operators try to avoid any downtime in
modern production facilities. Several techniques of monitoring
the condition of a motor have been investigated during the last
decades, e.g. based on vibration analysis [1,2] as well as
temperature or current monitoring [3]. Many of these are
routinely applied today. One promising condition monitoring
approach is based on magnetic flux detection [4, 5]. As a non-
intrusive and easily deployable method it could strongly benefit
from the rapid development of micro-fabricated
magnetometers. For example it may serve as a low-cost and
easy-to-use condition monitoring solution for initial diagnosis
of electric motors.
978-1-4673-7929-8/15/$31.00 ©2015 IEEE
Pawel Rzeszucinski, Daniel Lewandowski
ABB Sp. z o.o.
Corporate Research Center Poland
Krakow, Poland
pawel.rzeszucinski@pl.abb.com
To allow for widespread application of such monitoring
equipment, e.g. as a commodity tool carried by every service
technician, two parameters are of great importance: low-cost of
the sensor combined with small size. Today’s MEMS-
fabricated magnetometers for e.g. automotive or consumer
applications are miniature (range of a few millimeters), cheap
(less than 1 USD per sensor) and measure field in 3-axis.
Traditionally, leakage flux of electric motors has been
measured by flux coil sensors [5,6]. These sensors are
relatively bulky, expensive equipment and allow only for
measurement of magnetic flux in one direction.
In this paper we demonstrate the application of low-cost
magnetometers for monitoring of electric motors. We have
selected a BMC050 eCompass sensor from Bosch Sensortec
[7] which is a typical integrated 3-axis compass solution for
consumer market applications providing a direct digital output
signal. The out-of-plane field component (z) is measured by a
Hall-based technique. The two in-plane field components (x
and y) are measured by a measurement principle similar to
flux-gate magnetometers: A micro-fabricated magnetic layer
gets periodically reversed. Evaluation of the response of the
magnetization allows to infer the external magnetic field, e.g.
by phase-sensitive detection.
This paper is organized as follows: In the second section of
this paper we have investigated if the bandwidth of the
magnetometer would be sufficient for motor monitoring.
Furthermore the magnetometer has been exposed to variations
of the ambient temperature to mimic conditions encountered in
an industrial environment. The third section presents
measurement done on a low voltage motor. Here the frequency
spectrum of the magnetic field of the motor is analyzed in a
healthy motor condition as well as in case of a motor with a
broken rotor bar. These results demonstrate that miniature
magnetometers can in principal be used for performing
condition monitoring of electric motors.
Fig. 1 Magnetometer BMC050 from Bosch Sensortec mounted on a shuttle
board for easy access to pins [7].
II. MAGNETOMETER EVALUATION
A. Experimental setup
In order to communicate with the magnetometer and thus
allow the measurements to be taken, a LabVIEW software
from National Instruments running on a regular PC in
combination with an NI USB-8451 interface was used. This
interface communicates with the magnetometer via SPI bus.
The magnetometer itself was mounted on a shuttle board
(Fig. 1). Such a configuration allows for easy access to the
sensor’s pins and is compatible with Bosch Sensortec’s
development tools.
The magnetometer allows data output rates of up to 30 Hz
in ‘normal mode’ (using its internal timing) which has not been
sufficient for performing condition monitoring of electric
motors. Therefore the so called ‘forced mode’ has been used
instead where single sensor measurements can be triggered
externally. According to the datasheet information, a maximum
output data rate of approximately 615 Hz is attainable in this
mode (see respective formula in [7], page 20). In our setup a
maximum data output rate of up to 500 Hz was achieved. This
can be attributed to the non-real-time environment of the PC.
Note that the measured data was not equally spaced in time,
depending on the CPU load and/or other latencies. While this
approach was sufficient for initial evaluation of the
performance of the magnetometer, it is recommended that
further measurements should be made using a real-time
environment (such as a microcontroller or FPGA unit). Such a
data acquisition process would allow full sensor capabilities to
be utilized, while also being closer to the real application.
In the process of magnetometer-based monitoring of
electric motors, the response of the magnetometer to AC
magnetic fields in the range from a few Hz to approximately
120 Hz is of the greatest interest (compare to section III for the
frequency components of interest). In order to investigate the
capability of the BMC050 sensor to fulfill this requirement an
AC magnetic field at variable frequency but constant amplitude
has been applied to the magnetometer. The AC magnetic field
was generated by five loops of a current carrying wire with a
fixed position with respect to the magnetometer. The frequency
of the current has been controlled by a function generator
which fed a power amplifier. The current was kept at a constant
rms-level during measurement runs. This was checked
continuously by a current meter.
Fig. 2 Response of the magnetometer to ac magnetic fields at different
frequencies (a). The measured magnetic field amplitude shows a slight drop
at high frequencies (b).
As the magnetometer is a 3-axis sensor the orientation of
the three measurement axis has been checked qualitatively
beforehand by changing the position of the current carrying
wire. Finally the orientation of the current carrying wires was
fixed such that highest magnetic field signal into one in-plane
direction has been achieved. The wires were glued in direct
contact to the sensor housing. A current of 700 mA resulted in
a magnetic field of around 300 µT at the position of the
magnetometer. Absolute measurement accuracy of the
magnetometer has not been checked. This is not of main
importance for monitoring of electric motors as typically the
relative amplitude of characteristic components at different
frequencies gets evaluated.
B. Characterization of the magnetometer
The magnetometer has been operated in ‘forced mode’ at
an output data rate of approx. 500 Hz. According to the
Nyquist–Shannon sampling theorem one could expect to be
able to detect signals up to a maximum frequency of 250 Hz. In
practice, however, the maximum useful frequency is expected
to be around 200 Hz and thus AC magnetic fields of
frequencies ranging from 1 Hz to 200 Hz have been recorded
with the magnetometer. The time domain data has been fitted
by a sinusoidal test function to infer the magnetic field
amplitude. The amplitude of the measured signal showed only
a slight decrease of 3% at higher frequencies (Fig. 2). The
magnetic field amplitude has been kept constant during the
entire measurement run. This data can be used to calibrate the
frequency response during later measurements of the magnetic
field generated by an electric motor.

Fig. 3 Output signal of the magnetometer has been recorded for varying
ambient temperature (a). For a typical operating temperature range in industry
applications the output signal of the magnetometer can be well approximated
by a linear behavior. (b) Relative deviations from linear fit.
In addition to the previous validation also the temperature
dependence of the magnetometer has been checked
experimentally. The magnetometer including the magnetic
field generating wire has been mounted in a climatic chamber.
The temperature has been varied from -40°C to 85°C. The
temperature has been held at a constant level for at least 1 h
before each measurement to allow for thermal gradients to
equilibrate. As expected the output of the magnetometer is
dependent on the ambient temperature (Fig. 3a,b). A linear fit
yields a temperature dependence of 8.7·10-4/°C. The linear test
function agrees quite well with the measured temperature
dependence in the evaluated temperature range, as the
coefficient of determination R2 equals 0.987. This calibration
curve may be used to compensate for measurement
uncertainties resulting from changing environmental
temperature encountered during operation of the magnetometer
in a factory environment.
Minimum and maximum specified passive storage
temperatures for the magnetometer are given as -50°C and
150°C, respectively. Therefore two magnetometers were
exposed to repetitive temperature cycles from -40°C to
+150°C, the lower limit given by the cooling capabilities of the
climatic chamber. After 25 cycles the sensors were still
operating without failure. As the temperature cycling
parameters used in this experiment correspond to quite harsh
environmental conditions, sufficiently long time stability of the
magnetometer for the planned application in monitoring of
electric motors can be expected.
Fig. 4 Spectrum of the magnetic field amplitude of the motor prepared in a
healthy state. Sensor direction of the 3-axis magnetometer is indicated by
colors red, black and blue.
III. MOTOR MEASUREMENTS
Magnetic flux leakage in axial and radial direction is
expected from a motor running under both nominally healthy
as well as faulty conditions. [5,8-11]. The axial field is
coplanar with the motor axis; it is generated by the currents in
the stator end windings or in the rotor cage end ring. The radial
field is located in a plane perpendicular to the machine axis. It
is an image of the air-gap flux density which is attenuated by
the stator magnetic circuit (package of laminations) and by the
external machine frame. Both fields can be measured
separately by placing a magnetic field sensor in an appropriate
location. Since the magnetic field sensor used for these
measurements is three-directional, precise orientation of the
sensor while making measurements is not critical.
According to [9], in order to detect any deviations from
nominal operating conditions it is essential to look at the low
frequency region of the magnetic field spectrum. In the specific
case of a broken rotor bar fault the frequency of greatest
interest is the induction motor slip s=(fs-fr)/fs (1), where fs is the
actual operating synchronous supply frequency (line frequency
divided by number of motor pole pairs) and fr is the actual
running speed. Slip estimation is critical for identification of
many different fault types, therefore estimation of the actual
running speed as well as the line frequency – both required to
derive the slip – based on magnetic field measurements or
other measurement types is a topic of many publications [5, 8–
11] and it will not be described in the presented paper.
The presence of a broken rotor bar or broken end ring
causes an unbalance to the rotor magnetic flux, as the current
cannot flow through the broken or cracked bar/end-ring. The
unbalanced rotor flux can be considered as the combination of
positive- and negative-sequence rotor flux, rotating at a slip
frequency into the opposite directions. It is known that the
presence of a broken bar causes the appearance of an additional
frequency in the magnetic field spectrum [9]; the broken bar
frequency fbb can be described by fbb=sfs (2).

Fig. 5 Spectrum of the three-directional magnetic field measurement in case
of the broken rotor bar. Sensor direction of the 3-axis magnetometer is
indicated by colors red, black and blue.
In the following subsections two identical induction
machines of the same type have been analyzed. The machines
have been operated under the same conditions but in different
states: one has been in a healthy state while the other has been
seeded with a broken bar fault. For the experiments, an
induction motor supplied by a variable-frequency-drive (VFD)
was used. During these tests the sampling frequency of the
magnetic field sensor was set to 100 Hz and the signals were
recorded for the length of 10 seconds. Since the motor was
supplied by VFD its operating conditions were different from a
nominal operation (direct grid connection). The actual line
frequency was set to 21 Hz and the running speed was 20.69
Hz which corresponds to 1241.4 rpm.
A. Magnetic field analysis of healthy motor
The amplitude of the magnetic field of the healthy motor
has been simultaneously measured with the 3-axis
magnetometer (Fig. 4). Since the magnetic field was measured
by a 3-directional sensor, direction has been marked by colors
red, black and blue for x, y and z direction respectively. It can
be clearly seen that the line frequency (in this case 21 Hz) has a
dominant character in the spectrum and it is apparent in all
three axis. This is a typical frequency response of a healthy
motor, as in [9].
B. Magnetic field analysis of broken rotor bar
The magnetic field amplitude of the motor in the broken
rotor bar case has also been measured with the same 3-axis
magnetometer (Fig. 5), spectral data shown in red, black and
blue color corresponding to x, y and z direction. The line
frequency fs equal to 21 Hz is clearly visible in both motor
cases (compare to healthy state in Fig. 4, broken bar in Fig. 5).
However, in the case of the healthy motor the line frequency
dominates the entire spectrum, while in the case of the broken
bar it is noticeable that the lower frequencies contain the most
energy.
View publication stats
Based on the magnetic field data analysis the value of slip s
was estimated to 0.0345 and the line frequency fs was estimated
to 20.99 Hz. Using equation (2) the value of the characteristic
frequency fbb for a broken rotor bar can be calculated to be
0.724 Hz. What is also worth to mention is that there is an
additional peak related with the broken rotor bar visible at 3sfs
which corresponds to the findings shown in e.g. [9]. Therefore
the 3sfs component can be also taken into account as an
additional confirmation when suspecting broken rotor bar. The
ratio of the amplitude of frequency fbb and the amplitude of the
RMS of the entire signal can potentially be used as an indicator
of the broken rotor bar presence.
IV. SUMMARY & OUTLOOK
Magnetic field measurements with a magnetometer based
on MEMS-fabrication technology have been used to detect the
characteristic signature of a broken motor bar in the frequency
spectrum of the magnetic field of an induction motor. The
sensors were characterized in detail to check if they meet the
criteria for the application in motor diagnosis, the focus being
on achieving sufficient frequency response. Furthermore the
stability of the sensor output to changes of the environmental
temperature has been evaluated. Long term robustness of the
sensors has been qualitative tested and has been found to be
sufficient for operation in an industrial environment. Such
miniature sensors are used in many of today’s consumer
products and therefore benefit from low-cost, small size and
high flexibility. These advantages could pave the way for
routine condition monitoring of electric motors and further
components in the drive-train of industrial installations.
[1] R.B. Randall “State of the art in monitoring rotating machinery – part
1”, Sound and Vibration, 38, pp. 14-20, (2004).
[2] M. Niedermayer, S. Bennecke, R. Wirth, E. Armbruster, K.-D. Lang
„Wireless condition monitoring for industrial applications based on
radio sensor nodes with energy harvesting“, International Journal on
Advances in Networks and Services, 7, pp. 130-137, (2014).
[3] P.J. Tavner “Review of condition monitoring of rotating electrical
machines”, IET Electric Power Applications, 2, pp.215-247, (2008).
[4] M.D. Negrea “Electromagnectic flux monitoring for detecting faults in
electrical machines” Doctoral Dissertation, Helsinki University of
Technology (2006).
[5] V. Kokko “Condition monitoring of squirrel-cage motors by axial
magnetic flux measurements” Doctoral dissertation, University of Oulu
(2003).
[6] F. Zidat et al. “Non invasive sensors for monitoring the efficiency of ac
electrical rotating machines”, Sensors, 10, pp. 7874-7895, (2010).
[7] Datasheet, “BMC050 Electronic compass”, Bosch Sensortec (2011).
[8] M. Rigoni et al. “Detection and analysis of rotor faults in induction
motors by the measurement of the stray magnetic flux”, Journal of
Microwaves, Optoelectronics and Electromagnetic Applications, 11, pp.
68-80, (2012).
[9] A. Ceban, R. Pusca, and R. Romary “Study of rotor faults in induction
motors using external magnetic field analysis”, IEEE Transactions on
Industrial Electronics, 59, pp. 2082-2093, (2012).
[10] S. Koroglu, A.A. Adam, N. Umurkan, and K. Gulez “Leakage magnetic
flux density in the vicinity of induction motor during operation”,
Electrical Engineering, 91, pp. 15-21, (2009).
[11] A.A. Adam, K. Gulez, and S. Koroglu “Stray magnetic field distributed
around a PMSM”, Turkish Journal of Electrical Engineering and
Computer Science, 19, pp. 119-131, (2011).