You are on page 1of 6

Simple Sensorless Diagnosis Method

for Open-Switch Faults in SVM-VSI-fed


Induction Motor Drive
Teresa Orlowska-Kowalska, Piotr Sobanski
Wroclaw University of Technology, Institute of Electrical Machines, Drives and Measurements
Wroclaw, Poland
teresa.orlowska-kowalska@pwr.wroc.pl, piotr.sobanski@pwr.wroc.pl

Abstract—In this paper a new simple diagnostic technique of In papers [2]-[11] open-switch fault diagnostic methods
single IGBT open-circuit faults for a two level voltage inverter- based on the analysis of the current vector hodograph function
fed field oriented controlled induction motor drive is presented. in the stationary α-ß complex plane were presented. In the case
A failure diagnosis is carried out utilizing a reference voltage of the first technique [2], a failure detection and localization are
vector analysis in stationary coordinates, without additional achieved if an absolute value of the current vector exceeds
sensors. An extraction of fault information is based on some threshold at the time/ a time threshold, when it takes a
monitoring the voltage vector time presence in specific sectors of characteristic position in the α-ß coordinate system. According
the complex stationary plane. Thanks to diagnostic signal to a diagnostic technique introduced in [3], the localization of
normalization, high robustness to false alarms is guaranteed. To
a faulty inverter phase is realized by an analysis of the current
validate the proposed method, simulation tests under a wide
range of working conditions of the induction motor are presented
vector hodograph, however, a faulty switch can be recognized
for the direct field-oriented controlled drive system. thanks to the knowledge of phase currents polarity. In the
papers [4], [5], an expansion of the previously described
Keywords—induction motor drives; condition monitoring; fault diagnostic methods was presented. Another approach for the
diagnosis; Space Vector Pulse Width Modulation; voltage inverters; open-switch fault diagnosis is based on measurements of the
open-switch fault; sensorless stator current data with clustering and simple pre-processing
algorithms [6], [7]. The identification of the data cluster which
I. INTRODUCTION indicates the IGBT switch failure is achieved by utilizing
uncomplicated calculations [6] or, on the contrary, more
Vector controlled induction motor drives are now widely sophisticated techniques like artificial intelligence methods [7].
used in industrial applications for speed and torque regulation. Similar methods consist in the centroid calculation of the
In such systems induction motors are supplied from voltage- current vector trajectory in the α-ß coordinate system [8], [9] or
source inverters with space vector pulse width modulation in an A-B-C phase system [10]. To increase robustness to false
(SVM). These inverters mostly use power switches based on diagnosis alarms under wide range load condition, a fuzzy
the insulated-gate bipolar transistors (IGBTs), due to their well logic based approach can be applied [9]. In the paper [11], the
known advantages, such as high efficiency, high switching switch fault diagnostic technique based on a DC link current
frequency and relatively short-circuit current handling ability. measurement was presented, however, it requires additional
However, inverter-fed AC motors are sensitive to different current sensor in the DC. All these methods, based on the
faults occurring at the input rectifier, at the power inverter itself current analysis, detect the IGBT open circuit fault in one
or at the control system stage. All these faults lead to the fundamental period. They are also sensitive to load changes
interruption of the drive system operation and unprogrammed and drive transients, which results in low diagnostic efficiency
maintenance brakes. Unplanned drive stops could lead to high for low load levels and false alarms during transients.
financial losses, so the development of reliable monitoring and
fast fault detection methods as well as fault-tolerant control Time minimization between the fault occurrence and the
strategies is a current demand of the industry. fault detection can be obtained using voltage-based approaches
[12]-[18]. In the paper [12] four diagnostic techniques for an
Among various types of faults, power converter faults, open-switch fault in two-level voltage-source inverters were
related to semiconductor or control circuit damages, are the most presented. To identify the faulty transistor, voltages
frequent and are estimated to cause up to 60% of power device measurements in various inverter points are utilized and so is an
failures [1]. So recently quite a lot different fault detection and analytical model of the inverter. In every considered case, the
localization methods and techniques have been reported in switch fault localization procedure is based on a comparison
technical literature [1]-[18]. The most frequently used between measured voltages and their expected references. These
approaches are based on the analysis of easily accessible signals, techniques cannot be applied in a wide range of applications in
like current or voltage, thus the methods of an open-circuit fault spite of their simplicity and a short failure localization time
detection can be classified as current- or voltage-based ones. because of additional sensors requirements, which increase

978-1-4799-0224-8/13/$31.00 ©2013 IEEE 8210


implementation costs. Apart from this, some of them depend on stationary α-ß coordinates is presented. An extraction of the fault
the machine model and in case of machine unbalancing information is based on the monitoring of the stator voltage
contribute to diagnosis malfunctions. Furthermore, the vector presence time in specific sectors of the complex α-ß
effectiveness of the considered techniques was improved only in plane. Thus the proposed method does not require any additional
an open-loop control structure, which means that in the case of sensors in the drive system, and thus it can be considered as
vector-controlled induction drives, due to an influence of a feed- sensorless To avoid false diagnosis alarms, the proposed
back control loops, the proposed fault diagnosis method failed. diagnostic signal was normalized and thus high robustness of the
In the paper [13] one of the previously mentioned voltage based proposed diagnostic method was guaranteed. To validate this
diagnostic methods was validated by simulations. The technique method, simulation tests under a wide range working condition
requires line-to-line voltage measurements for failure diagnosis. of the IM motor drive were carried out. Additionally, for a
For fault diagnosis, authors have taken into consideration the comparison, simulation results of the one of the best from above
influence of time delays due to switching transistors and showed described diagnostic methods [18] were presented.
how to eliminate false diagnosis alarms caused by this
phenomenon. A similar approach to the transistor fault was II. OPEN-CIRCUIT FAULT ANALYSIS
developed in [14]. The main concept of the failure detection is
based on errors between measured inverter pole voltages and A. Influence of a Single IGBT Fault on the Drive system
reference values. As the authors mentioned, the errors strongly Operation
depend on the applied drive load torque and the angular rotor The basic scheme of the two-level voltage source inverter
position as well. To overcome this inconvenience, the phase topology, whose failures are considered in this paper, is shown
voltages are calculated using the measured pole voltages and in Fig. 1. For this inverter only eight switch-status
then they are compared with the required values. It has to be combinations are possible, so that eight voltage vectors
mentioned, that the switch failure can be localized with certainty described as U0,...,U7 can be generated by using the considered
only by analyzing the reference voltage vector in a specific converter, according to Fig. 2.
sector of the α-ß plane. In fact the mean value over the time
period when the reference voltage vector is located in the sector
are taken into consideration. In [15] a similar approach is
suggested. Contrary to the previous one, the influence of
switching dead time [16] on diagnostic procedure was analyzed.
In the paper [17], a detailed analysis of voltage across the lower
transistors under inverter fault modes was provided. According
to the proposed diagnostic method, fault diagnosis in the case of
both lower and upper IGBTs failures is possible using voltages
measurements across lower inverter switches. As it was noticed, Fig.1. Standard three-phase voltage source inverter topology
the failure localization time depends on the moment of fault
occurrence referred to the present sector of the α-ß plane in
which a reference voltage vector is localized. Despite the fact
that the considered diagnostic technique is has high U 1 (100) = U DC e j0 [V]
U 2 (110) = U DC e jπ 3 [V]
effectiveness, additional hardware requirements make its
U 3 (010) = U DC e j2π 3 [V]
industry application strongly limited. Another voltage-based U 4 (011) = U DC e jπ [V]
approach for the inverter failure diagnosis was reported in [18]. U 5 (001) = U DC e j4π 3 [V]
The diagnostic method utilizes an average value of errors U 6 (101) = U DC e j5π 3 [V]
between reference and the estimated voltages which are U 0 (000) = 0 [V]
U 7 (001) = 0 [V]
calculated from a simple current-based flux estimator. However,
a detailed analysis of the proposed method leads to the
conclusion that in the low speed region the voltage estimation
method will be sensitive to motor parameter changes.
To sum up, the current-based methods normally don't Fig.2. The interpretation of the voltage space vectors for the two-level
inverter
require extra sensors, so an increase in implementation costs
can be avoided. Nevertheless, they have some drawbacks, such In order to achieve a reference voltage vector for the
as dependence on the load torque and drive transients. On the vector-controlled induction motor drive, six active voltage
other hand, voltage-based methods are highly robust to false vectors (U1,...,U6) and two zero ones (U0 and U7) are used. The
alarms under full-range drive operations and in most cases they vectors divide the α-ß plane in six sectors specified by an angle
require additional sensors. Moreover, the proposed methods are γ, as in Fig. 2 and Table I. In accordance with the modulation
mainly effective only in an open-loop control, while in the technique, the average voltage vector Us over the time period
closed-loop control like DFOC or DTC structures these Ts can be obtained by applying the adjoining U1, U2 vectors
methods are sensitive to the active operation of the closed flux (as in the situation of Fig. 2), for the intervals t1 and t2 and the
or torque and speed loops. zero vectors are used for a total duration t0 or t7 [19]. The t1, t2,
In this paper a voltage-based open-switch fault diagnostic t0 and t7 values (which constitute Ts) depend on the assumed
method which uses a reference voltage vector analysis in the control method, called space vector modulation (SVM).

8211
TABLE I. DEFINITION OF THE SECTORS III. DIAGNOSTIC TECHNIQUES
Area of the α-ß plane As previously mentioned, in this paper the simple open-
Sector number
defined by γ [rad] circuit faults diagnosis method is presented. The main idea of
I 0, π 3 this technique consists in the reference voltage space vector Us
II (π 3, 2π 3 analysis in the α-ß coordinates. The extraction of the fault
information is based on monitoring the voltage vector presence
III (2π 3, π time tM in specific sectors of the complex α-ß plane. The
IV (π , 4π 3 direction of the voltage vector rotation is related to an angular
motor speed direction. In further consideration it was assumed,
V (4π 3, 5π 3 that under the motor mode of the machine, which rotates with
VI (5π 3, 0) positive speed direction, the voltage space vector rotates with
positive direction as well. This means, the numbers of the
sectors are increasingly changed. Depending on the motor
In spite of implementation complexity, a carrier-based speed direction and fault localization, the reference voltage
space vector pulse width modulation (SVPWM) method, which vector Us is forced in one characteristic sector during a much
is equivalent to the SVM, is often applied. In the case of this longer time-period than in the case of some other ones.
method, gate pulses are achieved by comparing the reference
modulated voltage signals with triangular ones [20]. In this TABLE II. THE OPEN-SWITCH FAULT SYMPTOMS PATTERNS
paper, the carrier-based SVPWM (Space Vector Pulse Width
Modulation) technique was utilized. Characteristic sector of the α-ß plane
Faulted Motor drive speed
(longer time-period of the voltage Us
switch direction
In the case of transistor faults the commanded voltage presence)
vectors cannot be achieved, so a motor is fed in such a way that ω>0 sector I
T1
ω<0 sector VI
the quality of the drive control decreases considerably. Taking
ω>0 sector II
into consideration transistor T1 failure, it can be assumed that T2
ω<0 sector I
the vectors U3(010), U4(011), U5(001), U7(000) contrary to the ω>0 sector III
U1(100), U2(110), U6(101), U0(111), do not introduce additional T3
ω<0 sector II
control distortion, because these can be successfully obtained. T4
ω>0 sector IV
So it can be concluded, that in some sectors of the α-ß plane the ω<0 sector III
voltage vector tracking analysis delivers more information than ω>0 sector V
T5
in some other, therefore the failure symptoms are more visible. ω<0 sector IV
ω>0 sector VI
For instance, if the commanded vector Us is located in the first T6
ω<0 sector V
sector, both active vectors U1(100) and U2(110), as well as zero
vector U0(111) cannot be correctly achieved. In all of these
This fact is clearly visible even under fast linear motor
cases, despite the forced inverter states, the damaged T1 is not
speed changes, this phenomenon will be demonstrated in the
able to conduct the current. Therefore, after applying one of
next section. According to this reasoning, the failure symptoms
these unavailable inverter states, the current flow through the
of the following inverter switches are integrated in Table II. As
faulty inverter phase decreases to zero during freewheeling
can be noticed when analyzing Fig. 2 and Table II, the
through one diode. Further, if the commanded vector Us is
rotational movement of the reference voltage vectors is slower
located in the second sector, to achieve the Us voltage vector,
in those sectors which are indicated by previously mentioned
the U2(110), U3(010) and U7(000) vectors are utilized. In this
unavailable inverter states (unavailable voltage vectors).
situation, only U2(110) vector is affected by the switch fault, so
that the failure influence on the drive control is weaker than in According to the reference motor speed, the demanded
the previously described case. space voltage vector angular speed is changed. Therefore, in
order to obtain robustness of the diagnostic method to false
However, the additional distortion is introduced because of
alarms, the diagnostic signal tM has to be normalized,
the closed-loop control. Due to the asymmetrical power supply,
respectively to (1):
deformed currents flow through the stator winding the
induction motor. These currents affect the rotating magnetic tM
field induced in the motor. So the effects of the power supply t Mnorm = (1)
aω − b
variations caused by the transistor failures are well visible in where:
hodographs analysis of the rotor flux. Further, excited rotor
tMnorm – normalized diagnostic signal,
currents produce disturbed electromagnetic field in the rotor.
ω – angular velocity of the drive,
The rotating rotor and stator fields affect each other, so the
a, b – parameters dependent on utilized sampling period of a
amplitude of a resulting rotating field in the motor air gap is
time tM measurement system.
variable even under the steady-state. Therefore, the constant
electromagnetic torque cannot be longer generated in the To implement the proposed diagnostic method, a simple
machine, and its oscillations lead also to speed control fault monitoring system was simulated, as shown in Fig. 3 for
deviations. To sum up, the transistor faults cause the speed T1 transistor failure. According to this diagnosis system, a
control disturbances, which are significantly intensified signal carrying the information about the sector number SN is
because of the closed-loop control. converted to logic signals S1,..,S6 respectively.

8212
to the diagnostic signal normalization, the fault diagnostic
procedure is independent of the motor speed, therefore the
constant value of the threshold T can be successfully assumed.
For the tested drive system operation, the threshold T=1.6 was
fixed. Nevertheless, it has to be mentioned, that in order to
speed up the failure localization procedure, this value can be
decreased.
a)
2 ω =ω n
T=1,6
ω =0,78ω n

tMnorm
1 ω =0,64ω n
ω =0,43ω n
0
1 2 3 4 5 6
SN
Fig. 3. Block diagram of the T1 fault monitoring system
b)
2 ω =ω n
T=1,6
For instance, if the reference voltage vector is located in the ω =0,78ω n

tMnorm
first sector of the α-ß plane, the signal S1 remains set to 1, 1 ω =0,64ω n

otherwise S1 is equal 0. In the same way the signals S1,..,S6 can 0


ω =0,43ω n
1 2 3 4 5 6
be obtained. Next, the value tM is measured by timers, and SN

normalized by utilizing expression (1). In the case the absolute c)


value tMnorm is greater than fixed threshold T, and taking into 2
T=1,6
ω =ω n
ω =0,78ω n
account the angular speed direction, the switch failure can be

tMnorm
1 ω =0,64ω n
successfully localized. ω =0,43ω n
0
1 2 3 4 5 6
SN
IV. SIMULATION RESULTS d)
The direct rotor field oriented control (DRFOC) structure 2
T=1,6
ω =ω n
ω =0,78ω n
tMnorm

operation [20], shown in Fig. 4a, was tested under open-switch 1 ω =0,64ω n
faults of the voltage inverter. To obtain the conditions similar ω =0,43ω n
to the real drive system, the vector controlled active rectifier 0
1 2 3
SN
4 5 6

for feeding the voltage inverter, was simulated [21]. To design e)


the power converter, a SimPower System toolbox was utilized, 2 ω =ω n
T=1,6
which allowed simple fault simulations by applying 0 logic ω =0,78ω n
tMnorm

signal on transistor gates. 1 ω =0,64ω n


ω =0,43ω n
0
UDC 1 2 3 4 5 6
SN
ψrref isxref fx usxref Sa -
PI PI rs x-y usαref f)
- - Sb
usyref usβref 2 ω =ω n
isyref fy α-β T=1,6
PI rs Sc ω =0,78ω n
tMnorm

SV PWM
ωref - INVERTER
PI
-
~ 1 ω =0,64ω n
ω =0,43ω n
0
1 2 3 4 5 6
isx x - y isα α - β isA SN
isB
isy isβ
α-β i
A B C sC Fig. 5. The normalized diagnostic signal tMnorm, during transistors faults: T1 (a),
ex ey γr
ROTOR
A B C T2 (b), T3(c), T4(d), T5(e) and T6(f) under changeable motor speed ω>0
DECOUPLING
ψr FLUX
BLOCK
ESTIMATOR LM EN IM
ω To confirm the reliability of the diagnostic procedure,
further simulation tests were realized. First single open-switch
Fig.4. Schematic diagrams of the DRFOC structure of the IM motor drive
faults under speed steady-state operation, for different values
and directions of the motor speed ω and load torque ml were
As mentioned before, for suitable operation of the
carried out. In Fig. 6 and Fig. 7 chosen simulation results for
proposed diagnostic system the threshold T, which indicates
the transistor T1 and T4 faults in the phase A of the voltage
the transistor faults, has to be specified. For this purpose, the
inverter-fed induction motor are demonstrated. In these figures,
extensive simulation tests for different values of the motor
the faulty phase current isA, the SN signal corresponding to the
speed ω and a full range of the load torque ml changes were
number of the sector, where the reference voltage space vector is
carried out. To simplify the further analysis, the obtained data
localized under the switch fault, the absolute value of the
were suitably preprocessed. In the case of tests realized for
normalized diagnostic signal tMnorm and the diagnostic system
one speed level and different load torque values, the minimum
output signals T1fault and T4fault are shown respectively. It has to be
tMnorm values for different switch failures and characteristic
added, that in Fig.6d and Fig. 7d, Tsim signal corresponds to a
sectors were registered. This approach allowed to avoid false
fault simulator modeled using SimPower System toolbox (as
diagnostic alarms. In Fig. 5 the summarized test results for
mentioned in section III). This means that when the transistor
different speed values are presented. In these diagrams, the red
failure is simulated, the Tsim is equal to one, otherwise – to zero.
line indicates the assumed threshold T. As can be seen, thanks

8213
a) a)
Fault of T1 Fault of T6
isA [p.u.]

0.5

isB [p.u.]
2
0 0
-0.5 -2
0.45 0.5 0.55 0.4 0.45 0.5 0.55 0.6 0.65
t [s] t [s]
b) b)
6 6
5 5
4
SN

SN
3 3
2 2
1 1
0.45 0.5 0.55 0.4 0.45 0.5 0.55 0.6 0.65
t [s] t [s]
c) c)
|tMnorm | [p.u.]

|tMnorm | [p.u.]
2 T=1,6 2 T=1,6
1 1
0 0
0.45 0.5 0.55 0.4 0.45 0.5 0.55 0.6 0.65
t [s] t [s]
d) d)
T1fault

1 Tsim

T6fault
T1fault
1 T6fault Tsim
0
0.45 0.5 0.55 0
t [s] 0.4 0.45 0.5 0.55 0.6 0.65
t [s]

Fig. 6. Simulation results corresponding to T1 fault under load torque ml=0


and nominal speed ω= ωn: faulty phase stator current isA (a), sector number SN Fig. 8. Simulation results corresponding to the drive operation with ml=0.5ml
of the reference voltage space vector (b), absolute value of the diagnostic and the linear changes of the angular motor speed ω during T6 fault: faulty
variable tMnorm (c) and the diagnostic system output signal T1fault (d) phase stator current isB (a), sector number SN of the reference voltage space
vector (b), absolute value of the diagnostic variable tMnorm (c) and the
a) diagnostic system output signal T6fault (d)
Fault of T4
Next the diagnostic system was tested under linear speed
isA [p.u.]

2
0 changes and various load torques to show its robustness to
-2
0.4 0.45 0.5 0.55 0.6 0.65
various operation conditions of the vector-controlled drive. An
t [s] example of the proper operation of the proposed diagnostic
b) method in the case of the transistor T6 fault, under linearly
6 increasing reference speed value and half loaded motor is
5
4 shown in Fig. 8. It is seen that despite continuously
SN

3
2 changeable speed values the voltage vector remains longer in
1 sector VI, and clearly determines the fault indicator of the
0.4 0.45 0.5 0.55 0.6 0.65
t [s] transistor T6 fault in converter phase B.
c)
As mentioned in the introduction section of this paper, the
|tMnorm | [p.u.]

2 T=1,6
high effectiveness and fast operation of the proposed
1 diagnostic technique through the comparison with one of the
0 best sensorless diagnostic method can be confirmed. For this
0.4 0.45 0.5 0.55 0.6 0.65
t [s] purpose in Fig.9 the proper simulation results of the algorithm
d) based on [18] were presented. The diagnostic signals da, db, dc
T4fault

1 T4fault Tsim are defined accordingly to (2):


0
0.4 0.45 0.5 0.55 0.6 0.65
t [s]  u f ref − u f est 
d f = ∫  dt
 (2)
 U DC 
Fig. 7. Simulation results corresponding to the drive operation with nominal
load torque ml=-ml and an angular motor speed ω= -0.3ωn during T4 fault: where f=a,b,c, uf ref and uf est mean the reference and estimated
faulty phase stator current isA (a), sector number SN of the reference voltage phase voltage, respectively.
space vector (b), absolute value of the diagnostic variable tMnorm (c) and the
diagnostic system output signal T4fault (d) It has to be noticed, that in case of both described
methods, for suitable operation of the diagnostic system the
It is clearly seen, that for faulty case of the transistor T1 and threshold, which indicates the transistor faults has to be
positive direction of the motor speed, the voltage space vector Us specified. Furthermore, the failure localization time increases
remains much longer in sector I than in others (Fig. 6b), while along with the increasing robustness requirements. For this
under the faulted transistor T4 and negative speed direction this reason, the fair comparison between described methods is
vector stops for some time period in sector III (Fig.7b) – difficult, because the diagnostic threshold are not calculated
according with the information given in Table II. The diagnostic by using one precise optimization criterion.
procedure responds properly in both cases.

8214
a) REFERENCES
T T
sim 1fault
[1] S. Yang, D. Xiang, P. Bryant, L. Ran, Tavner P., “Condition monitoring for
0.2 da device reliability in power electronic converters: a review,” IEEE Trans. on
da,b,c [p.u]
0.1 Power Electronics, vol. 25, pp. 2734-2752, November 2010
0 db [2] A.M.S. Mendes, A.J. Marques Cardoso, “Voltage source inverter fault
-0.1
dc
diagnosis in variable speed AC drives, by the average current Park's vector
-0.2 approach,” Int. Conf. Electr. Mach. and Drives, pp.704-706, 1999
0.45 0.5 0.55
t [s] [3] R. Peuget, S. Courtine, J.P. Rognon, “Fault detection and isolation on a
b) PWM inverter by knowledge-based model,” IEEE Trans. Ind. Appl., vol. 34,
T T
pp. 1318-1326, Nov/Dec 1998
sim 4fault
[4] W. Sleszynski, J. Nieznanski, A. Cichowski, “Real-time fault detection and
0.3 db
0.2 localization vector-controlled induction motor drives,” 11th Europ. Conf.
da,b,c [p.u]

0.1 Pow. Electr. and Appl.- EPE, pp. 2-8, June 2005
0 dc
-0.1 [5] M. Trabelsi, M. Boussak, M. Gossa, “Multiple IGBTs open circuit faults
-0.2
da diagnosis in voltage source inverter fed induction motor using modified slope
-0.3 method,” 19th Int. Conf. Electr. Mach., pp.1-6, 2010
0.45 0.5 0.55
t [s] [6] Guan Yufan, Sun Dan, He Yikang, “Mean Current Vector Based Online
c) Real-Time Fault Diagnosis for Voltage Source Inverter fed Induction Motor
T
sim
T
6fault
Drives,” IEEE Conf. Electr. Mach. and Drives vol. 2, pp.1114-1118, 2007
0.4
da
[7] Park Jang-Hwan, Kim Dong-Hwa, Kim Sung-Suk, Dae-Jong, Lee Myung-
Geun Chun, “C-ANFIS based fault diagnosis for voltage-fed PWM motor
da,b,c [p.u]

0.2
dc drive systems,” IEEE Ann. Meet. of the Fuzzy Inf., vol. 1, pp.379- 383, 2004
0
db [8] P. Gilreath, B.N. Singh, “A New Centroid Based Fault Detection Method for
-0.2
3-Phase Inverter-Fed Induction Motors,” 36th IEEE Power Electronics Spec.
-0.4 Conf., pp. 2664-2669, 16 June 2005
0.45 0.5 0.55
t [s] [9] F. Zidani, D. Diallo, E.H.M. Benbouzid, R. Nait-Said, “A Fuzzy-Based
Approach for the Diagnosis of Fault Modes in a Voltage-Fed PWM Inverter
Fig. 9. Simulation results corresponding to the drive operation: under load Induction Motor Drive,” IEEE Trans. Industrial Electronics, vol.55, pp.586-
torque ml=0 and nominal speed ω= ωn during T1 fault (a), with nominal load 593, Feb. 2008
torque ml=-ml and an angular motor speed ω= -0.3ωn during T4 fault and with [10] V.F. Pires, T.G. Amaral, D. Sousa, G.D. Marques, “Fault detection of
ml=0.5ml and the linear changes of the angular motor speed ω during T6 voltage-source inverter using pattern recognition of the 3D current
fault (c) trajectory,” 8th IEEE Reg. Int. Conf. Comp. Tech., pp.617-621, 2010
[11] A. Gaeta, Scarcella, G. Scelba, S. De Caro, A. Testa, “Inverter fault-
Nevertheless, it is visible that the method presented in this identification for VSI motor drives,” IEEE Int. Symp. Diagnostics for
paper is competitive to the proposed in [18], considering that Electric Machines, Power Electronics & Drives, pp. 413-419,. 2011
the fault localization procedure can be successfully speed up [12] R.L. de Araujo Ribeiro, C.B. Jacobina, E.R. Cabral da Silva, A. M.N Lima,
by changing the fault threshold, as mentioned before. “Fault detection of open-switch damage in voltage-fed PWM motor drive
systems,” IEEE Trans. Power Electronics, vol.18, pp.439-446, April 2004
The simulation results confirmed the effectiveness and [13] M. Trabelsi, M. Boussak, P. Mestre, M. Gossa, “An improved diagnosis
high robustness of the proposed new diagnostic method. As technique for IGBTs open-circuit fault in PWM-VSI-fed induction motor
can be seen, thanks to diagnostic signal normalization, the drive,” 2011 IEEE Int. Symp. Ind. Electronics,, pp.2111-2117, 2011
correct diagnostic system functionality was achieved even [14] C. Lee, W. Choi, “Design and evaluation of voltage measurement based
under linear changes of the angular motor speed, different sectoral diagnosis method for inverter open switch faults of permanent
magnet synchronous motor drives,” IET–Electric Power Applications, vol.6,
speed directions and different load torques. pp.526-532, September 2012
[15] Jung Shin-Myung, Park Jin-Sik, Kim Hyoung-Suk, Kim Hag-Wone, Youn
V. CONCLUSION Myung-Joong, “Simple switch open fault detection method of voltage source
inverter,” IEEE Energy Conversion Congress and Exposition, ECCE 2009,
This paper deals with the open-circuit faults diagnostic pp.3175-3181, 2009
method for a two level voltage inverter-fed field oriented [16] An Qun-Tao, Sun Li-Zhi, Zhao Ke, Sun Li, “Switching Function Model-
controlled induction motor drive. As opposed to the current Based Fast-Diagnostic Method of Open-Switch Faults in Inverters Without
based techniques, mentioned in the introduction, this one is Sensors,” IEEE Trans. Power Electronics, vol.26, pp.119-126, Jan. 2011
characterized by simplicity and very low computational [17] Yu Ok-Sun, Park Nam-ju, Hyun Dong-Seok, “A Novel Fault Detection
requirements. Furthermore, the proposed diagnostic system does Scheme for Voltage Fed PWM Inverter, 32nd IEEE Ann. Conf.
not require additional hardware, therefore extra implementation Industrial Electronics, IECON, pp.2654-2659, 2006
costs can be avoided. Thus the system can be considered as [18] N.M.A. Freire, J.O. Estima, A.J.M. Cardoso, “A voltage-based approach for
open-circuit fault diagnosis in voltage-fed SVM motor drives without extra
sensorless, as the switch-fault diagnostic symptoms rely on the hardware,” 20th Int. Conf. Electrical Machines, pp.2378-2383, 2012
information on the reference voltage vector and the information [19] N. Mochan, “Advanced Electric Drives: Analysis, Control and
on speed direction only. Thanks to the diagnostic signal Modeling Using Simulink,” MNPERE, 2001
normalization, the system functionality is not disturbed by false [20] M.P. Kazmierkowski, F. Blaabjerg, and R. Krishnan, “Control in Power
alarms, even under transient states of the drive. Electronics – Selected Problems,” Academic Press, USA, 2002.
This research was supported in part by PL-Grid Infrastructure. [21] T. Orlowska–Kowalska, “Sensorless Induction Motor Drives,” Wroclaw
University of Technology Press, Wroclaw, 2003.

8215

Powered by TCPDF (www.tcpdf.org)

You might also like