IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 62, NO.

3, MARCH 2015

1943

Speed and Current Sensor Fault Detection and

Isolation Technique for Induction Motor
Drive Using Axes Transformation
Chandan Chakraborty, Senior Member, IEEE, and Vimlesh Verma, Student Member, IEEE
AbstractThis paper presents a new technique for fault
detection and isolation to make the traditional vector-controlled induction motor (IM) drive fault tolerant against
current and speed sensor failure. The proposed current
estimation uses d- and q-axes currents and is independent
of the switching states of the three-leg inverter. While the
technique introduces a new concept of vector rotation
to generate potential estimates of the currents, speed is
estimated by one of the available model reference adaptive
system (MRAS) based formulations. A logic-based decision
mechanism selects the right estimate and recongures the
system (by rejecting the signal from the faulty sensors).
Such algorithm is suitable for different drives, including
electric vehicles to avoid complete shutdown of the system,
in case of sensor failure. The proposed method is extensively simulated in MATLAB/SIMULINK and experimentally validated through a dSPACE-1104-based laboratory
prototype.
Index TermsCurrent estimation, fault detection and isolation, induction motor (IM) drive, model reference adaptive
system, sensors, speed estimation, vector control.

L IST OF S YMBOLS
vs , vs
is , is
is , is
is_est, is _est
is_est, is _est
isd , isq
r , r
r , r , rest
sl
Rs
Ls , Lr
Lm
ia , ib
ms

and components of the stator voltage vector.

- and -phase currents when -phase is
along a-phase.
- and -phase currents when -phase is
along b-phase.
Estimated value of - and -phase currents
when -phase is along a-phase.
Estimated value of and -phase currents
when -phase is along b-phase.
Reference value of d- and q-axes currents.
and components of the rotor flux vector.
Reference, actual, and estimated rotor speed.
Slip speed.
Stator resistance.
Self-inductance at the stator and the rotor side.
Magnetizing inductance.
a- and b-phase currents.
Rotor flux angle with the -phase.

Manuscript received December 17, 2013; revised March 23, 2014 and
April 20, 2014; accepted April 23, 2014. Date of publication August 5,
2014; date of current version February 6, 2015.
The authors are with the Department of Electrical Engineering, Indian Institute of Technology (IIT) Kharagpur, Kharagpur 721302, India
(e-mail: chakraborty@ieee.org; vimlesh@ee.iitkgp.ernet.in).
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TIE.2014.2345337

I. I NTRODUCTION

NDUCTION motor (IM) drives are extensively used in the

industry due to mechanical ruggedness, reduced size and
cost, and low maintenance. Field-oriented control or vector
control of IM offers high performance (due to the decoupling
of flux and torque) and has become an industry standard. The
recent research trend is to make the drive fault tolerant. This
has motivated investigations on advanced techniques of fault
diagnosis in IM drive [1][5]. The source of failure may be
due to the machine (such as stator interturn faults, broken
bar in the rotor, etc.), converter (i.e., failure of the switching devices), or maloperation of sensors. In vector-controlled
drives, speed and current (and/or voltage) sensors are usually
required. The maloperation of current and speed sensors (due to
noise, dc-offset and open circuit, etc.) is not uncommon in the
industrial environment, and any industrial drive system needs
to be prepared to take care of such contingencies. Therefore,
sensor fault-tolerant control is an extremely important area of
investigation for IM drives.
Fault-tolerant control in IM has opted for one of the two
approaches. In the first approach, when a fault is detected, the
system switches to an alternative form of controller (typically
from a closed-loop to open-loop control), whereas in the second
technique, the loop is closed by the corresponding signal from
the estimator/observer.
In the literature, machine currents are estimated using observers and are compared with measured currents to detect
the faults in current sensors (as reported in [9], [16][18],
[23], and [25]). If a fault is detected, the current signals from
the current sensors are ignored and are replaced by estimated
current signals computed using the parameters of the machine.
In [9], observers are designed to generate the residual, which
are used to identify and isolate the faulty sensor. In [16], a bank
of three rotor flux observers and a switching mechanism are
presented to identify the faulty sensor. The proposed technique
demands more design efforts and is computationally intensive,
which makes their real-time implementation difficult. In [17],
an adaptive current observer with rotor resistance estimation
is used to isolate all the sensor faults. This method is not
robust to random system noise, and the strategy for control
loop reconfiguration is not discussed. In [18], a flux and current
observer with online estimation of stator and rotor resistances is
presented using a single current sensor. The observer design is
based on the assumption that the rotor speed and input voltages
are known. In [23], the fault diagnosis and reconfiguration
strategy are based on the extended Kalman filter (EKF) and

0278-0046 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

1944

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 62, NO. 3, MARCH 2015

reduced number of adaptive observers. In [25], an observerbased fault-tolerant algorithm for a single-phase pulse width
modulation (PWM) rectifier is proposed. None of these methods work in case of multiple sensor failures with seamless
transfer from sensor-based (i.e., with sensor) to sensorless (i.e.,
without sensor) configuration.
In [8], [10], [12], [21], [26], and [27], a decision block is
used, which changes the control strategies, depending on the
detection of sensor faults (current and speed). In [8] and [12],
the control reorganization is carried out by a fuzzy decision,
which assures a smooth transition from the encoder-based
(using sliding mode) to the sensorless controller (utilizing fuzzy
control). All these approaches sacrifice field orientation when
fault occurs and hence offer poor dynamic performance.
In [6], [7], and [20], neural network/fuzzy logic technique
is used to identify and isolate the faulty sensor. In [6] and
[7], artificial neural network is trained with the help of data
obtained from phase voltages, phase currents, rotor speed,
measured torque, power (per phase), and reference magnitude
of dc supply. The implementation is difficult, and also, such
techniques cannot be used for retrofit applications. In [20],
diagnosis of sensor failure is carried out with the help of fuzzy
inference.
The detection and isolation of speed sensor failure are also
reported in [13], [22], and [24]. In [13], fuzzy logic is used to
detect the faulty speed sensor. In [22] and [24], fault-tolerant
control is based on maximum-likelihood voting (MLV) that
uses the actual speed and estimated speed (obtained using
the EKF and Luenberger observer). In [19], a parity space
(PS) approach is used to identify and isolate the faulty sensor.
Reconfiguration of control is not possible as no estimation is
performed. Fault-tolerant drive for multiple sensor failure is
also proposed [11]. However, this is at the cost of an additional
sensor at the dc link. In [15], a programmable logic controller
(PLC) based protection and monitoring method for three-phase
IM is present. A survey on different types of faults in variable
speed drives is reported [14].
Permanent-magnet synchronous machines (PMSMs) have
smaller size and high power density (compared to IM), and
doubly fed induction machines (DFIMs) allow the converter
to be put in the rotor side handling only slip power. Observerbased methods are also used to make DFIM and PMSM fault
tolerant [28][36].
This paper proposes a new fault detection, isolation, and
compensation technique to make the IM drive fault tolerant
against current and speed sensor failures. The compensation
for current and speed needs a method of estimation. A new
concept of vector rotation to identify the faulty current sensor
is introduced. A logic-based detection mechanism in the
reference frame is proposed to make the drive fault tolerant
against current sensor failures. The speed estimation is carried
out by modifying a recently proposed model reference adaptive
system (MRAS)-based technique [37]. This paper is divided
into six sections. Section I deals with the literature review on
various fault detection and isolation approaches. The proposed
fault detection and isolation algorithms for current and speed
sensor failure are discussed in Sections II and III, respectively. Simulation and experimental results are presented in

Fig. 1.

Axes transformation showing -phase along a-phase.

Fig. 2.

Axes transformation showing -phase along b-phase.

Sections IV and V, respectively. Finally, Section VI concludes

this paper.
II. FAULT D ETECTION AND I SOLATION A LGORITHM
FOR C URRENT S ENSOR FAILURE
This section deals with the estimation of current and also
the fault detection and isolation of the current sensor. It is
assumed that the system works with two current sensors and a
speed sensor. These two current sensors may be put in any two
of the three phases. Note that the transformations from threephase to two-phase quantities require the individual orientation
of the axes with respect to each other. Following the standard
procedure, first, it is assumed that the a-phase (of the threephase system) and -phase (of the two-phase system) are along
the same axes (see Fig. 1). The corresponding relation between
ab and variables is shown by (1). A close inspection of
(1) reveals that, if a-phase sensor is defective, then both and -phase measurements will be wrong. However, if the
b-phase sensor is defective, the corresponding current of the
-phase will remain correct, while the current in the -phase
will be wrong.
  3
 

0
ia
is
= 23
(1)
is
ib
3
2
On the other hand, a unique feature is extracted if the
phase are rotated by 120 . This is shown in Fig. 2. The

CHAKRABORTY AND VERMA: SPEED AND CURRENT SENSOR FAULT DETECTION AND ISOLATION TECHNIQUE

Fig. 3.

Estimation of current when -phase is along a-phase.

Fig. 4.

Estimation of current when -phase is along b-phase.

corresponding relation for the conversion of three-phase to twophase current is shown in (2).

   
3 
0
ia
is
2

=
(2)
is
ib
3 23
Note that, for a fault in a-phase, it will offer a healthy measurement of -phase, whereas -phase will now provide the
wrong data. Therefore, depending on a fault either in b-phase or
a-phase, the use of a proper transformation (either considering
that -phase is along a-phase or -phase is along b-phase) will
provide us the true estimate of the corresponding -phase current. Fault detection can only be carried out if a correct estimate
is available. Currents may also be estimated in a two-phase
stationary reference frame, with the help of reference values
of d- and q-axes currents (i.e., isd and isq ) as shown in (3). The
corresponding phasor diagram is presented in Fig. 3. Equation
(3) is derived assuming that -phase is aligned along a-phase.

 
 
is_est
cos ms sin ms
isd
=
(3)
is _est
sin ms
cos ms
isq
If -phase is aligned along b-phase, then the estimated value
of current is obtained as shown in (4), and the corresponding
phasor diagram is presented in Fig. 4.

 
 
is_est
sin(30 ms ) cos(30 ms ) isd
=
(4)
is _est
cos(30 ms ) sin(30 ms ) isq
Using the actual measurement of a- and b- phase currents and
the corresponding reference-magnitude of the same in the (d, q)

1945

rotating reference frame, eight estimates of currents in the

reference frame are thus possible. These are is , is ,
is , is , is_est , is _est , is_est , and is _est , where is ,
is , is_est , are is _est are the corresponding variables when
-phase is along a-phase and is , is , is_est , and is _est are
the corresponding variables when -phase is along b-phase.
Hence, the system can be made fault tolerant if a mechanism
can be developed to switch over from the faulty measurement to
the corresponding correct estimate. The overall block diagram
of the fault-tolerant drive is presented in Fig. 5. Fig. 7 shows the
flowchart of the fault-tolerant algorithm that is implemented
using several switches. Fig. 6 shows the figure of a switch that
selects A or B to be transferred to the output depending on the
value of T being 1 or 0, respectively. The flowchart shown in
Fig. 7 is implemented with the help of the logic circuit available
in Fig. 8. Different flags, such as J, Y, Z, are used for the
implementation of the logic design. In Fig. 8, is is subtracted
from is_est to form the error. The absolute value of error is
compared with the threshold value, which is decided depending
on maximum noise (those may be picked up by the sensors), dc
offset, change in current sensor gain, saturation, etc. Whenever
the error (1 ) is less than the threshold value, output will be
0; else, it is 1. This output is given to an SR latch. Depending
on the state of Z (i.e., 0 or 1), the a-phase current sensor fault
is detected. The state of Z decides the vector rotator (i.e.,
ms or 120 ms ) to be used. If Z = 0, then ms is used to
transform is and is (or is _est ) to isq1 and isd1 , respectively.
In Fig. 8, the generated signal J decides the correct value of
the -phase current to be given to the controller. The state of Y
in Fig. 8 indicates the occurrence of fault in the b-phase current
sensor. Depending on the states of Z and Y (i.e., 0 or 1), Table I
is made to identify the faulty current sensor.
When both a- and b-phase sensors are faulty but the speed
sensor is healthy, then d- and q-axes currents (i.e., isd2 and isq2 )
are obtained by transforming the estimated currents with the
help of a vector rotator (ms ) as shown in Fig. 8. The logic to
decide the correct value of d- and q-axes currents to be fed to
the controller (i.e., isd1 and isq1 or isd2 and isq2 ) is presented
in Fig. 8.
isq1 and isd1 are fed to the controller if, one current sensor,
or the speed sensor, or one current and speed sensor fails. If,
both the current sensors fail, then isq2 and isd2 are fed to the
controller.
From the literature survey, it is found that there is no standard
method available for the selection of the threshold value. In
[29], it is reported that threshold values are set according to the
experience. Also, it can be extended to an adaptive threshold,
depending on the machine load. This also depends on the
operating point and environment of the drive. In this paper,
the threshold values are set after running the setup several
times with different types of simulated faults. It has been found
that an adaptive current threshold of the order of 10% of isq
offers acceptable performance. As isq depends on load, this
automatically takes care of the loading effect.
Note that, the output of the current sensor can be 0 if, the sensor is faulty or the phase in which the sensor is placed becomes
open. Therefore, the algorithm should be able to discriminate
sensor failure from phase loss (i.e., single phasing). In case of

1946

Fig. 5.

Fig. 6.

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 62, NO. 3, MARCH 2015

Block diagram of fault-tolerant vector-controlled IM drive.

Switch.

Fig. 7. Flowchart to decide the correct value of - and -phase

currents to be given to the controller.

a phase loss, the field orientation is lost, and hence, the q-axes
flux can be checked to make a correct decision.
III. S PEED S ENSOR FAULT D ETECTION A LGORITHM
The speed estimation is based on a recently proposed
X-MRAS-based method [37]. However, the approach is suitably modified to make it appropriate for fault detection and
isolation. The structure of X-MRAS for speed estimation is
shown in Fig. 9. The fictitious quantity X in the reference model

(= Xr ) is calculated using the reference values of voltages

and currents, whereas X in the adjustable model (= Xs ) is
computed with the help of reference values of voltages and
actual or estimated currents. The actual values of d-and q-axes
currents are obtained from current sensors, whereas the estimated values are obtained from the current estimation algorithm (i.e., is_est and is _est ). The estimated current signals
(in - and -reference frame) and currents in the (d, q) axes
rotating reference frame are obtained with the help of the
vector rotator (which, in turn, depends on speed). The error
(= Xr Xs ) is fed to the adaptation mechanism to generate
the speed signal. This estimated value of the rotor speed will
be used to make the drive fault tolerant against speed sensor
failure.


isd + vsd
isq
Xr = vsq

(5)

Xs = vsq
isd

(6)

vsd
isq

The logic circuit for the fault detection and isolation of the
speed sensor is shown in Fig. 10. Note that the proposed fault
tolerant control can work for all the speed estimation techniques
formulated for vector-controlled drives. Here, MRAS is considered only as an example. Similar to the discussion (presented in
Section II) for deciding the magnitude of the current threshold,
the selection of the speed threshold (to identify the speed sensor
faults) is important. Here, an adaptive threshold of 2% of the
reference speed is considered satisfactory for the speed loop.
The proportionalintegral (PI) controllers used in the current
and speed loop may get saturated in case the motor has to
develop a rated speed from start (with rated torque) or during
operation in field-weakening mode. The estimated magnitude
of current and speed will deviate from the actual variables.
Under such circumstances, an antiwindup controller may be
implemented, or the fault-tolerant controller may be disabled
until the speed and current controllers come out of saturation.

CHAKRABORTY AND VERMA: SPEED AND CURRENT SENSOR FAULT DETECTION AND ISOLATION TECHNIQUE

Fig. 8.

1947

Logic circuit for current sensor fault detection, isolation, and compensation.
TABLE I
I DENTIFICATION OF FAULTY S ENSOR D EPENDING
ON O UTPUT OF SR L ATCH

Fig. 10.

Circuit to decide the value of the speed to be fed to controller.

TABLE II
M ACHINE R ATING AND PARAMETERS

considerably during the operation. The influence of stator resistance is prominent at low speed, and hence, performance at
low speed is highlighted here.
A. Performance of the Proposed Fault-Tolerant Algorithm
for Vector-Controlled IM Drive When One of the Two
Current Sensors Is Faulty and Speed Sensor Is Healthy
Fig. 9.

MRAS structure for speed estimation.

IV. S IMULATION R ESULTS

In general, most of the current and speed estimation methods
are parameter dependent. Stator and rotor resistances vary

The proposed fault-tolerant algorithm is implemented with

a vector-controlled IM drive working with two current sensors
and a speed sensor. It is assumed that the two current sensors are
put in a-phase and b-phase, respectively. Now, a current sensor
fault occurs in b-phase (for example). Such a system is simulated in MATLAB/SIMULINK. Parameters of the machine are

1948

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 62, NO. 3, MARCH 2015

Fig. 12. Simulation results when speed sensor fails. (a) Reference and
actual speed. (b) Actual and estimated speed. (c) Rotor flux. (d) Output
of SR latch.

Fig. 11. Simulation results under fault in b-phase sensor. (a) Reference
and actual rotor speed. (b) Rotor flux. (c) -phase current. (d) -phase
current. (e) Output of SR latch.

given in Table II. The results obtained from the simulation are
presented here.
Initially, the system was operating with all healthy sensors,
and at 15 s, the output of the sensor in b-phase is made 0.
At this instant, is will be correct, but is and is are wrong.
Therefore, Z will still be 0, but Y and J become 1. The moment

J becomes 1, is _est is selected as the correct -phase current.

Performance of the drive under such situation is shown in
Fig. 11. The reference speed is changed from 0 to 10 rad/s at 2 s.
The actual speed follows the reference speed [see Fig. 11(a)].
The d- and q-axes rotor flux components are shown in
Fig. 11(b), which reflects the flux orientation. Fig. 11(c) shows
the -currents, when -phase is aligned along a-phase (i.e., is )
and b-phase (i.e., is ). Estimated value of the -phase
current (is_est ) is also shown. Fig. 11(d) shows the -phase
current. Fig. 11(e) shows the output states of SR latch. At the
instant (i.e., time occurrence) of failure of the current sensor in
b-phase, Z still remains at 0, but Y switches to 1.
B. Performance of the Proposed Algorithm When Speed
Sensor Fails Under Running Condition and
Current Sensors Are Healthy
The drive is started from rest with both current sensors and
speed sensor healthy. The speed sensor information is assumed

CHAKRABORTY AND VERMA: SPEED AND CURRENT SENSOR FAULT DETECTION AND ISOLATION TECHNIQUE

1949

to be lost at 15 s. Under such situation, the controller shifts from

actual to the estimated speed, using the logic shown in Fig. 10.
The corresponding simulation results are presented in Fig. 12.
At 15 s, the controller inputs switch from the actual value to the
estimated value of speed as shown in Fig. 12(a) and (b). Flux
orientation is still maintained as seen in Fig. 12(c). Fig. 12(d)
shows the states of different SR latches. As the current sensors
are healthy and speed sensor is faulty, Z and Y are 0, and
W is 1.
C. Performance of the Proposed Algorithm When One
Current Sensor and Speed Sensor Fail Under
Running Condition
The drive is started from rest with both current sensors and
speed sensor healthy. The output of b-phase current sensor and
that of the speed sensor are made 0 at 15 s. At this instant, is
will be correct, but is and is are wrong. Therefore, Z will
still be 0, but Y and J become 1. The moment J becomes 1,
correct -phase current becomes is _est . Also, the controller
shifts from actual to the estimated value of speed. Corresponding simulation results are presented in Fig. 13. At 15 s, the
controller input switches from the actual value to the estimated
value of speed as shown in Fig. 13(a) and (b). Flux orientation
is still maintained as observed in Fig. 13(c). Fig. 13(d) shows
different -phase currents, whereas Fig. 13(e) shows different
-phase currents. Fig. 13(f) shows the states of different SR
latches.
D. Performance of the Proposed Algorithm Under
the Variation of Rotor Resistance and
One Current Sensor Failure
To investigate the performance of the system when the speed
sensor is faulty, the rotor resistance has increased, and the
current sensor in b-phase is faulty. Here, the rotor resistance
is varied (to twice its actual value) in the form of a step at 10 s
[see Fig. 14(c)], and at 15 s, the output of the current sensor in
b-phase has failed (i.e., the output from the sensor is 0). The
performance of the drive under such circumstances is presented
in Fig. 14. Note that the system is working with the estimated
value of speed. Fig. 14(a) shows the reference and actual speed.
The estimated speed is presented in Fig. 14(b). Fig. 14(d) and
Fig. 14(e) shows the dynamics of - and -phase currents when
the current sensor in b-phase is made faulty. It can be observed
from the results that variation in rotor resistance has negligible
effect on the current estimation algorithm.
V. E XPERIMENTAL R ESULTS
The proposed fault-tolerant algorithm is experimentally validated with the laboratory-developed prototype, and the results
are shown here. The prototype is developed around a dSPACE1104 controller board. The control algorithms, including vector
control and fault tolerant technique, are implemented in a
real-time platform, based on control desk 5.3 software. The
MATLAB/SIMULINK block sets are available in the control
desk 5.3. The controller board in dSPACE-1104 consists of a
power PC and a slave DSP (TMS320F240). The PWM pulses

Fig. 13. Simulation results when b-phase sensor and speed sensor fail.
(a) Reference and actual rotor speed. (b) Actual and estimated rotor
speed. (c) Rotor flux. (d) -phase current. (e) -phase current. (f) Output
of SR latch.

1950

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 62, NO. 3, MARCH 2015

Fig. 15.

Configuration of the experimental setup.

generated by the dSPACE are fed to the gate driver (SKHI22A)

of the three-leg inverter. A sampling frequency of 6 kHz is
set for the real-time execution of the fault-tolerant algorithm.
The block diagram of the experimental setup is presented
in Fig. 15.
A. Performance of the Proposed Fault-Tolerant Algorithm
for Vector-Controlled IM Drive When One of the Two
Current Sensors Is Faulty and Speed Sensor Is Healthy
Here, the output of b-phase current sensor is made 0 at 14.4 s.
The corresponding results are shown in Fig. 16. The reference
speed of 10 rad/s is applied at 2 s. Fig. 16(a) shows the actual
speed following the reference speed command accurately. The
d- and q-axes rotor flux components are shown in Fig. 16(b).
-phase currents is , is , and is_est are shown in Fig. 16(c).
Similarly, -phase currents is and is _est are presented in
Fig. 16(d). Fig. 16(e) shows the output states of the SR latch.
When fault occurs in b-phase, the corresponding current is
will be wrong, and hence, Y will change from 0 to 1. However,
as is is correct, Z will still be 0.
B. Performance of the Proposed Algorithm When
Speed Sensor Fails Under Running Condition
and Current Sensors Are Healthy

Fig. 14. Simulation results under step change in rotor resistance and
fault in b-phase current sensor. (a) Reference and actual rotor speed.
(b) Actual and estimated speed. (c) Rotor resistance. (d) -phase
current. (e) -phase current.

The drive is started from rest with both current sensors and
speed sensor healthy. The speed sensor information is assumed
to be lost at 15 s. Under such situation, the controller detects
the fault and shifts from the actual to the estimated speed. The
reference, actual, and estimated values of the speed signals are
shown in Fig. 17(a) and (b). Flux orientation is well maintained
as seen in Fig. 17(c). Fig. 17(d) shows the output state of the
SR latch.

CHAKRABORTY AND VERMA: SPEED AND CURRENT SENSOR FAULT DETECTION AND ISOLATION TECHNIQUE

1951

Fig. 17. Experimental results when speed sensor fails. (a) Reference
and actual speed. (b) Actual and estimated rotor speed. (c) Rotor flux.
(d) Output of SR latch.

C. Performance of the Proposed Algorithm When

One Current Sensor and Speed Sensor Fails
Under Running Condition

Fig. 16. Experimental results under fault in b-phase sensor. (a) Reference and actual rotor speed. (b) Rotor flux. (c) -phase current.
(d) -phase current. (e) Output of SR latch.

Here, faults in the b-phase current sensor and speed sensor

are considered at 14.47 s. At this instant, is will be correct,
but is and is are wrong. Under such situation, the controller
detects the fault and shifts from the actual to the estimated speed
and current. Corresponding experimental results are presented
in Fig. 18. At 14.2 s, the controller input switches from the
actual value to the estimated value of speed as shown in
Fig. 18(a) and Fig. 18(b). Flux orientation is still maintained
as evident in Fig. 18(c). Fig. 18(d) shows different -phase
currents, whereas Fig. 18(e) shows different -phase currents.
Fig. 18(f) shows the states of different SR latches.

1952

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 62, NO. 3, MARCH 2015

Fig. 18. Experimental results when b-phase sensor and speed sensor
fail. (a) Reference and actual rotor speed. (b) Actual and estimated rotor
speed. (c) Rotor flux. (d) -phase current. (e) -phase current. (f) Output
of SR latch.

Fig. 19. Experimental results under step change in rotor resistance

and fault in b-phase current sensor. (a) Reference and actual rotor
speed. (b) Actual and estimated speed. (c) q-axes current. (d) -phase
current. (e) -phase current.

CHAKRABORTY AND VERMA: SPEED AND CURRENT SENSOR FAULT DETECTION AND ISOLATION TECHNIQUE

D. Performance of the Proposed Algorithm Under

the Variation of Rotor Resistance and
One Current Sensor Failure
As the experiment is performed on a slip-ring-type IM, external resistance can be added to study the behavior of the speed
and current estimator under the variation in rotor resistance.
In practice, the rotor resistance will change slowly. However,
in experimentation, this is difficult to realize as it will take a
very long time for the rotor resistance to change (due to high
thermal time constant). Therefore, a step change in rotor circuit
resistance is considered. At 16 s, the rotor resistance is changed
(in the form of a step), and at 29.44 s, the current sensor in
b-phase is made faulty. The performance of the drive under
such situation is presented in Fig. 19. A definite variation in
the actual value of the rotor speed is observed [see Fig. 19(a)],
whereas the estimated speed follows the reference speed [see
Fig. 19(b)]. The q-axes current is presented in Fig. 19(c).
However, it has been found that the estimated value of - and
-phase currents matches well with the corresponding sensed
magnitude [vide Fig. 19(d) and Fig. 19(e)]. This is because the
same ms is used in forward and reverse transformations. This
proves the efficacy of current estimation technique in case of
variation of rotor resistance.
VI. C ONCLUSION
This paper has presented a complete implementation of a
fault-tolerant vector-controlled IM drive. Extensive simulation
and experimental results (from the prototype developed in the
laboratory) have demonstrated that the system is capable to
detect a fault and reconfigure itself to switch to the correct
algorithm. The controller keeps estimating different currents
and speed and, in case of a fault, switches to the correct estimated value. The technique proposed extracts eight estimates
of currents in the reference frame (four estimates using the
reverse transformation from dq to and the other four using
the forward transformation from abc to ). The concept
of vector rotator is introduced for deciding the correct estimated value of the current corresponding to a fault. The speed
estimation is based on a modification of a recently proposed
X-based MRAS [37]. The proposed technique does not require
any additional sensor and reconfigures the drive in a seamless
manner in case a sensor fails. The technique works perfectly
even in case of multiple sensor failure. Such fault-tolerant
controller will make the IM drive more rugged (mechanically)
and reliable and will be very useful for applications like electric
vehicles, where safety and reliability play a crucial role for drive
selection.
A PPENDIX
C ONTROLLER G AINS

1953

R EFERENCES
[1] J. A. Daviu, S. Aviyente, E. G. Strangas, and M. R. Guasp, Scale invariant
feature extraction algorithm for the automatic diagnosis of rotor asymmetries in induction motors, IEEE Trans. Ind. Informat., vol. 9, no. 1,
pp. 100108, Feb. 2013.
[2] Y. Gritli et al., Advanced diagnosis of electrical faults in wound-rotor
induction machines, IEEE Trans. Ind. Electron., vol. 60, no. 9, pp. 4012
4024, Sep. 2013.
[3] S. Toma, L. Capocchi, and G. A. Capolino, Wound rotor induction generator inter-turn short-circuits diagnosis using a new digital neural network,
IEEE Trans. Ind. Electron., vol. 60, no. 9, pp. 40434052, Sep. 2013.
[4] M. P. Sanchez et al., Application of the Teager-Kaiser energy operator
to the fault diagnosis of induction motors, IEEE Trans. Energy Convers.,
vol. 28, no. 4, pp. 10361044, Dec. 2013.
[5] S. H. Kia, H. Henao, and G. A. Capolino, Gear tooth surface damage
fault detection using induction machine electrical signature analysis, in
Proc. SDEMPED, Valencia, Spain, pp. 358364.
[6] A. Bernieri, G. Betta, A. Pietrosanto, and C. Sansone, A neural network approach to instrument fault detection and isolation, IEEE Trans.
Instrum. Meas., vol. 44, no. 3, pp. 747750, Jun. 1995.
[7] G. Betta, C. Liguori, and A. Pietrosanto, An advanced neural-networkbased instrument fault detection and isolation scheme, IEEE Trans.
Instrum. Meas., vol. 47, no. 2, pp. 507512, Apr. 1998.
[8] F. Zidani, M. E. H. Benbouzid, D. Diallo, and A. Benchaib, Active faulttolerant control of induction motor drives in EV and HEV against sensor
failures using a fuzzy decision system, in Proc. IEEE IEMDC, 2003,
vol. 2, pp. 677683.
[9] K. S. Lee and J. S. Ryu, Instrument fault detection and compensation scheme for direct torque controlled induction motor drives, Proc.
Inst. Elect. Eng.Control Theory Appl., vol. 150, no. 4, pp. 376382,
Jul. 2003.
[10] D. Diallo, M. E. H. Benbouzid, and A. Makouf, A fault-tolerant control
architecture for induction motor drives in automotive applications, IEEE
Trans. Veh. Technol., vol. 53, no. 6, pp. 18471855, Nov. 2004.
[11] H. Wang, S. Pekarak, and B. Fahimi, Multilayer control of an induction
motor drive: A strategic step for automotive applications, IEEE Trans.
Power Electron., vol. 21, no. 3, pp. 676686, May 2006.
[12] M. E. H. Benbouzid, D. Diallo, and M. Zeraoulia, Advanced faulttolerant control of induction-motor drives for EV/HEV traction application: From conventional to modern and intelligent control techniques,
IEEE Trans. Veh. Technol., vol. 56, no. 2, pp. 519528, Mar. 2007.
[13] F. Zidani, D. Diallo, M. E. H. Benbouzid, and E. Berthelot, Diagnosis
of speed sensor failure in induction motor drive, in Proc. IEEE IEMDC,
2007, vol. 2, pp. 16801684.
[14] D. U. C. Delgado, D. R. E. Trejo, and E. Palacios, Fault-tolerant control
in variable speed drives: A survey, IET Electr. Power Appl., vol. 2, no. 2,
pp. 121134, Mar. 2008.
[15] R. Bayindir, I. Sefa, I. Colak, and A. Bektas, Fault detection and protection of induction motors using sensors, IEEE Trans. Energy Convers.,
vol. 23, no. 3, pp. 734741, Sep. 2008.
[16] M. E. Romero, M. M. Seron, and J. A. D. Dona, Sensor fault-tolerant
vector control of induction motors, IET Control Theory Appl., vol. 4,
no. 9, pp. 17071724, Sep. 2010.
[17] T. A. Najafabadi, F. R. Salmasi, and P. J. Maralani, Detection and isolation of speed-, dc-link voltage-, and current-sensor faults based on an
adaptive observer in induction-motor drives, IEEE Trans. Ind. Electron,
vol. 58, no. 5, pp. 16621672, May 2011.
[18] F. R. Salmasi and T. A. Najafabadi, An adaptive observer with online
rotor and stator resistance estimation for induction motors with one phase
current sensor, IEEE Trans. Energy Convers., vol. 26, no. 3, pp. 959966,
Sep. 2011.
[19] H. Berriri, M. W. Naouar, and I. S. Belkhodja, Easy and fast sensor fault
detection and isolation algorithm for electrical drives, IEEE Trans. Power
Electron., vol. 27, no. 2, pp. 490499, Feb. 2012.
[20] H. Li, A. Monti, and F. Ponci, A fuzzy-based sensor validation strategy
for ac motor drives, IEEE Trans. Ind. Informat., vol. 8, no. 4, pp. 839
848, Nov. 2012.
[21] B. Tabbache, N. Rizoug, M. E. H. Benbouzid, and A. Kheloui, A control
reconfiguration strategy for post-sensor FTC in induction motor based
EVs, IEEE Trans. Veh. Technol., vol. 62, no. 3, pp. 965971, Mar. 2013.
[22] B. Tabbache, M. E. H. Benbouzid, A. Kheloui, and J. M. Bourgeot,
Virtual-sensor-based maximum-likelihood voting approach for faulttolerant control of electric vehicle powertrains, IEEE Trans. Veh.
Technol., vol. 62, no. 3, pp. 10751083, Mar. 2013.
[23] X. Zhang et al., Sensor fault detection, isolation and system reconfiguration based on extended Kalman filter for induction motor drives, IET
Elect. Power Appl., vol. 7, no. 7, pp. 607617, Aug. 2013.
[24] M. Boukhnifer, A. Raisemche, D. Diallo, and C. Larouci, Fault tolerant
control to mechanical sensor failures for induction motor drive: A comparative study of voting algorithms, in Proc. IEEE IECON, Vienna, Austria,
pp. 28512856.

1954

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 62, NO. 3, MARCH 2015

[25] A. B. Youssef, S. K. E. Khil, and I. S. Belkhodja, State observer-based

sensor fault detection and isolation, and fault tolerant control of a singlephase PWM rectifier for electric railway traction, IEEE Trans. Power
Electron., vol. 28, no. 12, pp. 58425853, Dec. 2013.
[26] A. Raisemche, M. Boukhnifer, C. Larouci, and D. Diallo, Two active
fault tolerant control schemes of induction motor drive in EV or HEV,
IEEE Trans. Veh. Technol., vol. 63, no. 1, pp. 1929, Jan. 2014.
[27] X. Shi and M. Krishnamurthy, Survivable operation of induction machine drives with smooth transition strategy for EV applications, IEEE
J. Emerg. Sel. Topics Power Electron., vol. 2, no. 3, pp. 609617,
Sep. 2014.
[28] K. Rothenhagen and F. W. Fuchs, Doubly fed induction generator modelbased sensor fault detection and control loop reconfiguration, IEEE
Trans. Ind. Electron., vol. 56, no. 10, pp. 42294238, Oct. 2009.
[29] K. Rothenhagen and F. W. Fuchs, Current sensor fault detection, isolation, and reconfiguration for doubly fed induction generators, IEEE
Trans. Ind. Electron, vol. 56, no. 10, pp. 42394245, Oct. 2009.
[30] M. G. Carrillo and M. Kinnaert, Brief paper sensor fault detection and
isolation in three phase systems using a signal based approach, IET
Control Theory Appl., vol. 4, no. 9, pp. 18381848, Sep. 2010.
[31] N. M. A. Freire, J. O. Estima, and A. J. M. Cardoso, A new approach
for current sensor fault diagnosis in PMSG drives for wind energy conversion systems, IEEE Trans. Ind. Appl., vol. 50, no. 2, pp. 12061214,
Mar./Apr. 2014.
[32] A. Akrad, M. Hilairet, and D. Diallo, Design of a fault-tolerant controller
based on observers for a PMSM drive, IEEE Trans. Ind. Electron.,
vol. 58, no. 4, pp. 14161427, Apr. 2011.
[33] G. H. B. Foo, X. Zhang, and D. M. Vilathgamuwa, A sensor fault detection and isolation method in interior permanent-magnet synchronous
motor drives based on an extended Kalman filter, IEEE Trans. Ind.
Electron., vol. 60, no. 8, pp. 34853495, Aug. 2013.
[34] G. F. H. Beng, X. Zhang, and D. M. Vilathgamuwa, Sensor faultresilient control of interior permanent-magnet synchronous motor drives,
IEEE/ASME Trans. Mechatronics, vol. 20, no. 2, pp. 855864, Apr. 2015.
[35] S. Diao, Z. Makni, J. F. Bisson, D. Diallo, and C. Marchand, Sensor
fault diagnosis for improving the availability of electrical drives, in Proc.
IEEE IECON, Vienna, Austria, pp. 31083113.
[36] Y. Da, X. Shi, and M. Krishnamurthy, A novel universal sensor concept
for survivable PMSM drives, IEEE Trans. Power Electron., vol. 28,
no. 12, pp. 56305638, Dec. 2013.
[37] V. Verma, C. Chakraborty, S. Maiti, and Y. Hori, Speed sensorless vector
controlled induction motor drive using single current sensor, IEEE Trans.
Energy Convers., vol. 28, no. 4, pp. 938950, Dec. 2013.

Chandan Chakraborty (SM01) received the

B.E. and M.E. degrees in electrical engineering from Jadavpur University, Kolkata, India, in
1987 and 1989, respectively, the Ph.D. degree
from the Indian Institute of Technology (IIT)
Kharagpur, Kharagpur, India, in 1997, and the
Ph.D. degree from Mie University, Tsu, Japan,
in 2000.
He is currently a Professor in the Department of Electrical Engineering, IIT Kharagpur.
His research interest includes power converters,
motor drives, electric vehicles, and renewable energy.
Dr. Chakraborty was awarded the Japan Society for the Promotion of
Science Fellowship to work at the University of Tokyo (Hongo Campus),
Bunkyo, Tokyo, Japan, during 20002002. He received the Bimal Bose
Award in power electronics in 2006 from the Institution of Electronics and
Telecommunication Engineers (India). He is one of the Technical Program Chairs of the Industrial Electronics, Control, and Instrumentation
Conference (IECON) 2014 to be held in Dallas, TX, USA. He was the
Technical Program Chair of the International Conference on Industrial
Technology (ICIT) 2006, ICIT2008, ICIT2010, and IECON2012, held in
Mumbai, India, Melbourne, Australia, Valparaso, Chile, and Montreal,
QC, Canada, respectively. He has also contributed to IECON2009,
IECON2010, International Symposium on Industrial Electronics (ISIE)
2010, ISIE2011, and ISIE2012 as a Track Chair. He is an Administrative
Committee member of the IEEE Industrial Electronics Society. He is one
of the Associate Editors of the IEEE T RANSACTIONS ON I NDUSTRIAL
E LECTRONICS and IEEE Industrial Electronics Magazine and an Editor
of the IEEE T RANSACTIONS ON S USTAINABLE E NERGY. He is the
Founding Editor-in-Chief of IE Technology News (ITeN), a Web-only
publication for the IEEE Industrial Electronics Society. He is a Fellow
of the Indian National Academy of Engineering (INAE).

Vimlesh Verma (S12) was born in Mumbai,

India. He received the B.Tech. degree in electrical and electronics engineering from Andhra
University, Andhra Pradesh, India, in 2002 and
the M.Tech. degree in power apparatus and
systems from Nirma University, Gujarat, India,
in 2005. He is currently associated with the
Department of Electrical Engineering, Indian
Institute of Technology Kharagpur, Kharagpur,
India, where he is working toward the Ph.D.
degree.
His research interests include sensorless control of ac drives, fault
diagnosis of induction-motor drives, and renewable energy.
Mr. Verma was awarded the Queensland-India Friendship Scholarship
to pursue a part of his research at The University of Queensland,
Brisbane, Australia.