A FUZZY LOGIC APPROACH TO DIAGNOSE INDUCTION

MOTOR FAULTS IN REMOTE SITE BY CELL PHONE

Cesar da Costa
Electrotechnical Engineering, IFSP-Federal Institute of Education ,SP, 01109-010, Brazil
e-mail: [email protected]
Mauro Hugo Mathias
Mechanical Engineering, UNESP - Universidade Estadual Paulista, SP, 12516-410, Brazil
e-mail: [email protected]

Monitoring, fault detection, and diagnosis of electric induction motors are becoming increas-
ingly important in the field of electrical machines as new data-processing techniques and
methods for analyzing the motor stator current. Special attention has been devoted to nonin-
vasive methods capable of detecting faults using measured data without requiring motor dis-
assembly and its structural parts. Some enterprises such as Mining, Petroleum and Water and
Sewage Treatment process uses induction motor of large power flow, what are installed at
remote sites. The electrical and mechanical failures of such motors often disrupt productivity
and require maintenance. Currently the containment of maintenance costs, these remote sites
work without the presence of a service technician. A fuzzy logic approach may help to diag-
nose induction motor faults at these sites and to transmit the fault diagnosis (via 3G technol-
ogy) to maintenance central or a cell phone programmed. The contribution of this paper is
the use of fuzzy logic for the automated practical detection of broken bars in induction mo-
tors in remote locations without the presence of an experienced technician.

1. Introduction
Induction motors with squirrel-cage type rotors are rugged, reliable, and cheap. Therefore, they
are widely used in industrial and manufacturing processes. However, the electrical and mechanical
failures of such motors often disrupt productivity and require maintenance, thus presenting special
challenges in production. In the literature, rotor and stator faults have been shown to account for a
large proportion of industrial induction motor failures, occasionally being a major cause of failure
in the field. Although an induction motor is highly symmetrical, it may still have a detectable signal
component at the fault frequencies due to imperfect manufacture, improper motor installation and
so an. Further, manufacture tolerance and working environment could also result in disturbing the
motor fault diagnosis. Preventive measures should be periodically taken in order to protect motors
and systems including motors. This is the most efficient way to keep motor operating continuously
in healthy conditions [1-3].

ICSV22, Florence (Italy) 12-16 July 2015 1

 The 22nd International Congress on Sound and Vibration

2. Online Motor Condition Monitoring by MCSA Instrument

A crucial point about motor current signature analysis (MCSA) is that it is sensing an electrical
signal that contains current components that are a direct by-product of unique rotating flux compo-
nents caused by faults such as broken rotor bars, air gap eccentricity, and shorted turns in low volt-
age stator windings, etc. MCSA instrument can detect these problems at an early stage and thus
avoid secondary damage and complete failure of the motor [2]. A typical online condition monitor-
ing system by MCSA instrument is shown in Figure 1.

Figure 1. The online condition monitoring by MCSA instrument.

The sensor block measures physical quantities of the device and convert into an electrical re-
sponse in the time domain. There are several kinds of sensors that are used in MCSA instruments.
For example, current sensor, Hall Effect sensor, flux sensor, accelerometer, etc. The function of
data acquisition block is sampling, amplification and converting in analogue to digital of the physi-
cal signal from sensors. The purpose of fault detection block is finding out the incipient fault ap-
pearing in the induction motor through processing the acquired data. This block utilizes a FFT algo-
rithm to the representation of signal acquired in the frequency domain. The fault diagnosis block
includes two different methods: (i) feature extraction; (ii); model-referenced.

3. Feature Extraction Methods

If there is only a forward rotating field at slip frequency relative to the rotor, the cage winding is
symmetrical. Where rotor asymmetry occurs, then there will be a resultant backward rotating field
at slip frequency relative to the forward rotating rotor. The result of this is that, relative to the sta-
tionary stator winding, this backward rotating field at slip frequency relative to the rotor induces a
voltage and current in the stator winding at frequency given by Equation 1 [3]. This is referred to as
a twice slip frequency sideband due to broken rotor bars. Where s is the motor slip and f0 is the fre-
quency of the power grid to which the motor is connected.
(1) fbb = (1 − 2s) f 0 Hz.
There is therefore a cyclic variation of current that causes a torque pulsation at twice slip fre-
quency ( 2sf0 ) and a corresponding speed oscillation that is also a function of the drive inertia. This
speed oscillation can reduce the magnitude of the (1 − 2s) f 0 sideband, but an upper sideband current
component at (1 + 2s) f 0 is induced in the stator winding due to the rotor oscillation [3]. This upper
sideband is also enhanced by the third time harmonic flux. Broken rotor bars therefore result in cur-
rent components being induced in the stator winding at frequencies given by Equation 2.
(2) fbb = (1 ± 2s) f0 Hz.

3.1 Experimental Setup

To demonstrate the application of the feature extraction method, we performed an analysis of
different signals collected from rotor broken bars, which were forced in the laboratory by opening
the motor and drilling holes in different bars (Fig. 2a). For validating the feature extraction method
that uses MCSA instrument, several tests were performed with a 4-pole, 3-phase, 60 Hz, 1.5 kW,

ICSV22, Florence, Italy, 12-16 July 2015 2

 The 22nd International Congress on Sound and Vibration

220/380 V (rated voltage), 1750 rpm (rated speed), and 28-rotor-bar induction motor. Figure 2b
shows the experimental setup. The load was a 2 kW DC machine with a rated speed of 1800 rpm.

Figure 2. View of the experimental setup (a). Rotor broken bars forced (b).

3.2 Spectrum of Stator Current

To verify the efficiency of the feature extraction method, we carried out several tests under dif-
ferent loads for healthy rotors and faulty rotors with broken bars. In each case, the stator current
was transformed to the frequency domain and analyzed by the MCSA instrument. Then, the ampli-
tudes of the two fault frequency components Af (frequency broken bars left) and Af (fre-
bbl bbl
quency broken bars right) are analyzed and extracted. The results are summarized and shown. Fig.
3a shows spectrum of stator current for healthy motor at 95% of the rated load. The amplitude of
Af is lower -55 dB relative the amplitude of the fundamental frequency (60 Hz).
bbl

Figure 3. Current spectrum: loaded healthy motor (a). Loaded motor with one broken bar (b).

Fig. 3b shows spectrum of stator current for one broken bar at 95% of the rated load. The ampli-
tude Af is lower – 40 dB relative the amplitude of the fundamental frequency and the amplitude
bbl
Af is lower - 45 dB. A Fig. 4 shows spectrum of stator current for two broken bar at 95% of the
bbr
rated load. The amplitude Af is lower -35 dB relative the amplitude of the fundamental fre-
bbl
quency and the amplitude Af is lower - 40 dB.
bbr

ICSV22, Florence, Italy, 12-16 July 2015 3

 The 22nd International Congress on Sound and Vibration

Figure 4. Current spectrum: loaded motor with two broken bar.

4. Fuzzy Controller Method for Diagnosis and Decision

The purpose of the fuzzy controller method is shown in Figure 5, through the block diagram of
the system. The first step is acquiring data, i.e., collecting motor parameters that may be relevant in
the search for information on the motor status. Herein, in particular, the stator current and the motor
shaft speed will be collected from a current sensor and a motor shaft rotation sensor.

Figure 5. Block diagram of fuzzy controller method.

After collecting the data, the current signal is pre-processed, i.e., by means of the fast Fourier
transform (FFT), the signal frequency spectrum is obtained to get frequencies of side failure of bro-
ken bars (1 ± 2 s ) f . After obtaining the two side failure frequencies, their magnitudes will be ex-
0
tracted by means of an algorithm: Af (Magnitude of left broken bars frequency) and Af
bbl bbr
(Magnitude of broken bars right frequency).The third step, called fuzzy controller, refers to the use
of Fuzzy logic techniques, which can automatically provide failure diagnosis without requiring a
specialist technician.

ICSV22, Florence, Italy, 12-16 July 2015 4

 The 22nd International Congress on Sound and Vibration

4.1 Fuzzification (Membership Function)

Fuzzification is the mapping of the real numbers domain of (discrete in general) for the Fuzzy
domain, defined by relevancy functions to the input variables [4, 5, 7, and 8]. It is a type of pre-
processing of categories or classes of the input signals, thereby reducing the number of values to be
processed. The input variables Afbbl and Afbbr , in the domain of real numbers, have their values
expressed in dB normalized from 0 to 1, 0 being equivalent to −100 dB and 1 being equivalent to 0
dB. This allows for these variables to be fuzzified from any motor. The speed variable in the real
numbers domain has its values expressed in normalized rpm from 0 to 1, 0 being equivalent to 1740
rpm and 1 being equivalent to 1790 rpm. For each variable, three relevancy functions are stipulated,
denominated Small, Medium, and Big. The Medium function refers to the nominal values of the in-
put variables, where the values 0.50 (−50 dB) to 0.65 (−35 dB) indicate the motor relevancy value
with one to two broken bars. The Small function outlines the values considered for a healthy motor,
where values equal to or smaller than 0.45 (−55 dB) point to a maximum relevancy value. Simi-
larly, the Big function depicts the motor values with defects of three or more broken bars, where
values equal to or greater than 0.70 (−30 dB) point to a maximum relevancy value. The following
characteristics are common to the three variables: Magnitude of the left side failure frequency (
Af ), Magnitude of the right side failure frequency ( Af ), and motor speed.
bbl bbr

4.2 Base of Rules

An important part of a failure diagnosis system by Fuzzy logic is constructing the base of rules
[4, 5, 7, and 8]. The knowledge acquisition begins with the transfer of knowledge of the motor rotor
conditions for the rule base. Based on feature extraction method (experimental setup), a set of 10
rules was prepared, which composes the Fuzzy inference system. For the input variables, the letters
S (Small), M (Medium), and B (Big) previously defined were used. As RCs (Rotor Condition),
Healthy, Severe Defect, and Defect were used. As many rules are redundant, and moreover, it is
sufficient for one of the lateral failures to be Big (B) or Medium (M) for the rotor condition to be
Severe Defect or Defect respectively, the table may have the number of rules reduced to 10.
1. If ( Afbbr ) is Small) and ( Afbbl ) is Small) and ( Speed is Small), then (Motor-Condition is
Health);
2. If ( Afbbr ) is Small) and ( Afbbl ) is Small) and ( Speed is Medium), then (Motor-Condition is
Health);
3. If ( Afbbr ) is Medium) and ( Speed is Small), then (Motor-Condition is Defect);
4. If ( Afbbr ) is Medium) and ( Speed is Medium), then (Motor-Condition is Defect);
5. If ( Afbbr ) is Big) and ( Speed is Small), then (Motor-Condition is SevereDefect);
6. If ( Afbbr ) is Big) and ( Speed is Medium), then (Motor-Condition is SevereDefect);
7. If ( Afbbl ) is Medium) and ( Speed is Small), then (Motor-Condition is Defect);
8. If ( Afbbl ) is Medium) and ( Speed is Medium), then (Motor-Condition is Defect);
9. If ( Afbbl ) is Big) and ( Speed is Small), then (Motor-Condition is SevereDefect);
10. If ( Afbbl ) is Big) and ( Speed is Medium), then (Motor-Condition is SevereDefect).

4.3 Desfuzzification

In desfuzzification, the value of the output linguistic variable inferred by the Fuzzy rules will be
translated into a discrete value. The objective is to obtain a single discrete numerical value that best

ICSV22, Florence, Italy, 12-16 July 2015 5

 The 22nd International Congress on Sound and Vibration

represents the inferred Fuzzy values of the output linguistic variable, i.e., the distribution possibili-
ties [4]. Thus, desfuzzification is an inverse transformation that translates the output of the Fuzzy
domain to the discrete domain. Figure 6a shows the functions assigned to the output variables and
Fig. 6(b) presents the relevancy function graph common to the three input variables.

Figure 6.The Membership functions for outputs (a). The membership function for inputs (b)

There are three membership functions: Healthy, Defect, and Severe Defect. Table 1 describes the
output range for these variables.
Table 1. Range of outputs variables.
Range Rotor Condition Nº of Broken Bars
0 ≤ output ≤ 0.47 Health (H) 0
0.5 ≤ output ≤ 0.7 Defect 1-2
0.75 ≤ output ≤ 1 Severe Defect 3 or more

5. Embedded System in Hardware FPGA

The hardware model of fuzzy controller was subsequently synthesized and implemented in a
Field Programmable Gate Array (FPGA) chip [6, 7 and 8]. The embedded system used in the test
bench is based on a controller, NI sbRIO 9602 from National Instruments. The controller architec-
ture, includes a floating point processor running at 400 MHz, real-time operating system (RTOS),
high-performance FPGA Xilinx, interface 10\100 Base T Ethernet. NI sbRIO 9602 is programmed
using (i) PC LabVIEW; (ii) LabVIEW Real-Time; and (iii) LabVIEW FPGA. It runs applications
(deterministically) developed with the LabVIEW Real-Time software and the FPGA executes si-
multaneously applications developed with the LabVIEW FPGA software. The development of the
fuzzy controller diagnosis system warranted the development of three (VIs) programs. Two pro-
grams were developed in LabVIEW Real-Time and LabVIEW FPGA, running directly on Compac-
tRIO. The other program was developed in LabVIEW PC, running on a personal computer (Host
PC). The program (VI) in the host PC communicates with the LabVIEW Real-Time program
through shared variables via the TCP/IP protocol. For data transfer between the LabVIEW FPGA
and LabVIEW Real-Time programs, I/O Variables are used.

5.1 GSM/GPRS Cellular Network

General Packet Radio Services (GPRS) is a packet -based wireless communication service that
promises data rates from 56 up to 114 Kbps and continuous connection to the Internet for mobile
phone and computer users. GPRS is based on Global System for Mobile (GSM) communication and
complements existing services such circuit-switched cellular phone connections and the Short Mes-
sage Service (SMS). GSM can be applied for tele-monitoring applications, where high mobility and
low cost are necessary [9]. The Fuzzy controller via GSM \ GPRS cellular network consists of four
elements: NIsbRIO card (Fuzzy Controller), application interface (communication software), com-
munication base (Remote GSM provider) and GSM Modem. The Fuzzy controller monitors and

ICSV22, Florence, Italy, 12-16 July 2015 6

 The 22nd International Congress on Sound and Vibration

makes the diagnosis of the induction motor conditions (healthy, defect, severe defect) on the remote
site, and transmits the overall result, via communication network, to the central computer mainte-
nance or a programmed cell phone.

6. Experimental Fuzzy Controller System Results

To verify the efficiency of the fuzzy controller system, several tests were performed via GSM
network. These tests were performed under different loads and motor condition: healthy rotor, one
broken bar and two broken bars. Table 2 presents the results of the diagnosis of a Healthy motor
(value normalized and real) with low load, half load, and full load. The result with a single discrete
numeric value normalized equal to 0.22, which indicates the rotor condition Healthy.
Table 2. Diagnostic results of Healthy.
Load Afbbl Afbbr Speed Motor Condition
Low 0.46 0.30 0.50 0.22
-55.54 dB -70 dB 1769 rpm Healthy
Half 0.46 0.30 0.50 0.22
-54.55 dB -70 dB 1752 rpm Healthy
Full 0.46 0.30 0.50 0.22
-54.11 dB -58.91 dB 1769 rpm Healthy

Table 3 shows the results of the diagnosis of a Defect motor (1 broken bar) with low load, half
load, and full load. The result with a single discrete numeric value normalized equal to 0.6, which
indicates the rotor condition Defect (1 broken bar).
Table 3. Diagnostic results of 1 broken bar.
Load Afbbl Afbbr Speed Motor Condition
Low 0.60 0.57 0.20 0.6
-40.37 dB -43.50 dB 1745 rpm Defect
Half 0.61 0.57 0.45 0.6
-40.50 dB -43.64 dB 1760 rpm Defect
Full 0.55 0.55 0.60 0.6
-44.73 dB -45.29 dB 1774 rpm Defect

Table 4 shows the results of the diagnosis of a Defect motor (2 broken bars) with low load, half
load, and full load. The result with a single discrete numeric value normalized, equal to 0.6, which
indicates the rotor condition Defect (2 broken bars).
Table 4. Diagnostic results of 2 broken bar.
Load Afbbl Afbbr Speed Motor Condition
Low 0.63 0.60 0.20 0.6
-37.34 dB -40.45 dB 1746 rpm Defect
Half 0.65 0.57 0.45 0.6
-35.50 dB -42.84 dB 1760 rpm Defect
Full 0.59 0.59 0.60 0.6
-41.30 dB -41.25 dB 1776 rpm Defect

ICSV22, Florence, Italy, 12-16 July 2015 7

 The 22nd International Congress on Sound and Vibration

7. Conclusion
In this paper, a real time condition monitoring device based on fuzzy controller was developed
and tested. The target controller based on FPGA and GSM network is capable of measuring non-
invasive sensor signals and is capable of analyzing them for extraction of rotor and stator problems
in induction motors installed in remote sites. A diagnosis method using fuzzy logic to determine the
state condition of induction motors was presented. In order to make an efficient diagnostic the am-
plitudes frequency (broken bars left and right) components of the spectrum stator current and speed
motor are intended as input to the fuzzy controller system which converted to variables linguistic
fuzzy subsets and their corresponding membership functions. The output of this system represents
the motor conditions. The fuzzy controller system monitors and makes the diagnosis of the induc-
tion motors conditions on the remote-site and transmits de overall results, via GSM network, to the
central computer of maintenance officer or a programmed cell phone. This results obtained with this
system are good and is capable to detect the motor and stator problems in industrial induction motor
installed in remote sites.

Acknowledgment
The work developed by Cesar da Costa in Portugal was sponsored by CNPq\Brasil Postdoctoral
scholarship (PDE).

REFERENCES
1 Georgoulas, G., Tsoumas, I. P., Daviu, J. A. A., Alarcon ,V. C., Automatic pattern identifica-
tion based on the complex empirical mode decomposition of the startup current for diagnosis
of rotor asymmetries in asynchronous machines, IEEE Transactions on Industrial Electronics
61 (9), 4937-4946, (2014).
2 Kaikaa, M. Y. , Hadjami, M., Effects of the simultaneous of static eccentricity and broken ro-
tor bars on the stator current of induction machine, IEEE Transactions on Industrial Elec-
tronics 61(5), 2452-2463, (2014).
3 Kim, Y. H, Youn ,Y. W., Hwang ,D. H., Sun, J. H., Kang, D. S., High-resolution parameter
estimation method to identify broken rotor bar faults in induction motors, IEEE Transactions
on Industrial Electronics 60 (9), 4103-4117, (2013).
4 Azgomi, H. F., Poshtan, J., Poshtan, M., Experimental validation on stator fault detection via
fuzzy logic, 3rd International Conference on Electric Power and Energy Conversion Systems
(EPECS), IEEE 2013, Istanbul, 1- 6, (2013).
5 Rodrigues , P. V., Arkkio , A., Detection of stator winding fault in induction motor using
fuzzy logic, Applied Soft Computing 8, 1112-1120, (2008).
6 Da Costa, C., Mathias, M. H., Kashiwagi, M., Development of an instrumentation system em-
bedded on FPGA for real time measurement of mechanical vibrations in rotating machinery,
Instrumentation & Measurement Sensor Network and Automation (IMSNA), 2012 Interna-
tional Symposium on, 1, 60-64, (2012).
7 Boukaka, S., Teiar, H., Chaoui , H., Sicard, P., FPGA implementation of an adaptive fuzzy
logic controller for PMSM, Power Electronics, Machines and Drives (PEMD 2014), 7th IET
International Conference on, Manchester, 1-6, (2014).
8 Gdaim, S., Mtibaa , A., Mimouri, M., Design and experimental implementation of DTC of
induction machine based on fuzzy logic control on FPGA, Fuzzy Systems, IEEE Transaction
on , PP (99), p. 1, (2014).
9 Ahmed , E., Kohno , R., Medical reliable network using concatenated channel codes through
GSM network, Engineering in Medicine and Biology Society (EMBC), 35TH Annual Interna-
tional Conference of the IEEE, Osaka, 4755-4758, (2013).

ICSV22, Florence, Italy, 12-16 July 2015 8