Date post: | 07-Jan-2016 |
Category: |
Documents |
Upload: | alberto-francisco |
View: | 214 times |
Download: | 0 times |
7/17/2019 artículoFault Diagnostic Based on Parity Equations Applied to Induction Motor
http://slidepdf.com/reader/full/articulofault-diagnostic-based-on-parity-equations-applied-to-induction-motor 1/7
Fault Diagnostic Based on Parity Equations Applied to Induction Motor
Marco A. Rodríguez, Manuel Hernandez, Victor Golikov, Gilberto Martínez
Universidad Autónoma del Carmen UNACAR, Departamento de Ingeniería Electrónica, Ciudaddel Carmen Campeche México.
E-Mail: [email protected] , [email protected]
Resumen: En este trabajo se presenta una técnica de detección de fallas basada en el análisis con ecuaciones de paridad y aplicadas al motor de inducción. La idea principal es aproximar, durante el marco de referencia
sincrónico, el modelo no lineal del motor de inducción al modelo lineal del motor de CD, con la intención degenerar un cambio significativo en los residuos obtenidos mediante ecuaciones de paridad en presencia de falla,
lo cual permite la simplificación y confiabilidad de la detección de la falla. Para validar la técnica propuesta se presenta una simulación y resultados experimentales utilizando PSIM y Labview respectivamente.
Abstract: In this paper, a fault detection technique by using parity equations applied to an induction motor is
presented. The nonlinear model of A.C. motor is matched with the lineal model of D.C. motor in synchronousreference frame in order to generate a relative large change in the residual obtained with parity equations in
presence of fault, which allows the simplicity and reliability of the fault detection. The good performance of thefault detection system is validated by using a simulator software of power electronics and motor control
applications (PSIM) and experimental tests using Labview.
Keywords: Fault diagnosis, induction motors, D.C. motor, parity equation, modeling.
NOMENCLATURE
B Friction coefficient
idr d axis rotor current in the synchronous frameids d axis stator current in the synchronous frames
i f Flux-producing component of the stator current
iqr q axis rotor current in the synchronous frame
iqs q axis stator current in the synchronous frame
iT Torque producing component of the stator current
J Moment of inertia
Lm Magnetizing inductance
Lr Rotor inductance
L s Stator inductance
R s Stator resistance
V qs q axis stator voltage in the synchronous frame
V ds d axis stator voltage in the synchronous frame
λr Rotor flux linkage phasor
σ Leakage coefficientωr Rotor electrical speed
ω s Slip speed
I. INTRODUCTION
Induction motor is the “workhorse” of industry and is
extensively used in a wide variety of industrial processes and
is often integrated in many critical processes. 1 Then, to
maintain the right working of the motors must to be very
reliable in order to insure a high degree of reliability by means
of maintenance programs and specific attention by using a
permanent monitoring of motor in order to detect incipient
faults and to avoid failures. Generally, the most frequent faults
in induction motor are mechanical [1], and these are related to
the electrical operation of the motor such as local overheating
and inter-turn short-circuit stator winding. These latter faults
result in changes of basic electric parameter values of motor
such as resistances and inductances. Thus, this fault must be
diagnosed opportunely by means of determining the type, size
and location of the fault as well as its time of detection.
The mechanical vibration analysis is the most studied
technique to detect faults in the induction motor [2]. This is
due to the significant magnitudes and the immunity to external
phenomena like the electromagnetic interference over sensorswhich are commonly accelerometers through the
accelerometers´ disadvantages are that they have very limited
operation ranges. The diagnosis technique most utilized, for
the mechanical and electrical faults in the induction motor is
the spectral analysis of the phase current [3-4]. Other
techniques are based on the motor current transformations [5].
Others evaluate the neutral terminal voltage of induction motor
[6] and finally by means of the system impedance calculation
XXI Congreso de la Asociación Chilena de Control Automático ACCA 2014ISSN (en trámite), Pág. 65-72. Santiago de Chile, 5 al 7 de Noviembre 2014
- 65 -
7/17/2019 artículoFault Diagnostic Based on Parity Equations Applied to Induction Motor
http://slidepdf.com/reader/full/articulofault-diagnostic-based-on-parity-equations-applied-to-induction-motor 2/7
[7]. The main problem of these techniques is that the time
delay of detection and localization of the fault is very long and
expensive. The stator winding fault is one of the electrical
faults in the literature that has reported short time delay of fault
detection by means of the model based methods like the
Winding Function Approach (WFA) [8]. Another model based
method is by using parity equations which can be suitable to
detect several faults [9] but to compute the residuals from thegeneral parity equation, it is necessary to obtain first an
accurate mathematic model of system. For this reason, it has
mainly been applied to linear systems, where accurate models
are more readily available. However, as accurate models for
nonlinear systems are more difficult to obtained in practice,
and besides the parity equation is sensitive to noise and model
uncertainties, the parity equation derived using existing
approaches cannot be readily applied to nonlinear and
uncertainty systems [9].
More generally, the diagnostic procedure is based on the
heuristic knowledge on the process and observed analytical
heuristic symptom commonly represented in a signal called
residual. In these last two cases, a priori knowledge of fault
symptom causalities, as well as, a large data bank is
necessary[10]. However, in the diagnosis by using analytical
symptoms with limit values of measurement signals and
change detection have a minimal mathematical effort and
without data bank, which allows the simplicity and reliability
when a relative large change of their feature is obtained [11].
In this paper, a new approach is presented not only to detect
faults, but also to isolate faults for the nonlinear model of
induction motor matching it with the linear model of D.C.
motor taking into account the performance during the steady
state.
II. MODELING OF INDUCTION MOTOR IN DQ REFERENCE
FRAME
The key assumption of motor model in synchronous referenceframe is that the rotor flux linkages are constant so that their
derivatives are zero, then the following equations are deduced
in the d-q reference frame.
= ( + ) + + + (1)
= + − + − (2)
Where , , , are the stator and rotor d -q axes currents
in the synchronous reference frame that are obtained by
projecting the stator and rotor current phasor on the d-qaxes
respectively. Likewise, and are the stator d -q axes
voltages in the synchronous reference frame.
A simplification can be obtained by using the followingrelations [12]. = − (3)
= − (4)
Substituting the above expressions in (1) and (2) result that.
= ( + ) + + (5)
= ( + ) − + (6)
Where is the leakage coefficient. On the other hand, it is well
know that the flux-producing component of the stator current
(i f ) is constant in the steady state, and that is the d axis statorcurrent in the synchronous frames (ids). Therefore, its derivate
is zero in the steady state. Also, it is know that the torque
producing component of the stator current (iT ) is the q axis
stator current in the synchronous frames (iqs),additionally, it is
know that the rotor flux linkages phasor = then.
= ( + ) + + = ( + ) +
(7)
Where
= = −
(8)
= + = + (9)
Then = ( + ) + + = + + + (10)
From which the torque producing component of the stator
current is derived as.
= − + + = 1 + − (11)
Where = + , = 1 , = (12)
One the torque producing component of stator current iT is
deduced is possible to obtain the electromagnetic torque with
the following expression.
= (13)
Where
= 32 2
(14)
Now, the load dynamics can be expressed by the
electromagnetic torque T e and the friction as.
+ = − = − (15)
In term of the electric rotor speed, is derived by multiplying
both sides by the pair of poles:
Marco A. Rodríguez, Manuel Hernandez, Victor Golikov, Gilberto Martínez
- 66 -
7/17/2019 artículoFault Diagnostic Based on Parity Equations Applied to Induction Motor
http://slidepdf.com/reader/full/articulofault-diagnostic-based-on-parity-equations-applied-to-induction-motor 3/7
+ = 2 − (16)
Hence, the transfer function between the speed and the torque
producing current is deduced as.
()() = 1 + (17)
Where = 2 , = + , = (18)
The block diagram of induction motor with constant rotor
flux linkages is shown in Fig. 1.
+
K a
1+sTaK f
P
2 +
- 1
B+sJ
B1
Ls.if
Electrical Mechanical
ωr
V qs I qs
T e- iT
Fig. 1 Block diagram of induction motor model with
constant rotor flux linkages
This model is so similar to CD motor model obtained in [13]
and [14]. Although the main difference is that the input
parameter is instead of the armature current I A.
III. R ESIDUAL GENERATION BY USING PARITY EQUATIONS
A simple model of induction motor in steady state which islittle similar to CD motor model[12] is used in this work,
because the fault detection based on parity equations for this
model types is availability to detect several parameters[13].
Based on known ways of theoretically modelling the structure
of a linear mathematical model in the continuous time without
considering disturbances (19) and (20), the state space
representation obtained for the induction motor is displayed in
(21) and (22).
() = () + () (19)
() = () (20)
= ⎣− − −( + ) ⎦ + 1 00 0 0 (21)
() = 1 00 1
(22)
Where = , = (23)
Note that, the structure obtained in (21) is similar to DC motor
model shown in [13] and [14] but no equal. An important
difference is that the second term of in (21) the magnetic
flow
is defined as the relation between the stator inductance
and the d axis stator current in the synchronous frames I ds.
Another important difference is that the first term of ,themagnetic flow of D.C. motor model is defined as the
relationship among poles number P , magnetic inductance Lm,
rotor inductance Lr and flux-producing component of the stator
current I f . A way to add redundancy in the equations at the
same instant t is by introducing (19) in (20) with its respective
derivatives like:() = () + () (24)
Where
⎣
()()()⋮()⎦ = ⎣
⋮⎦ ()
+ 0 0 0 … 0 0 0 … 0⋮ ⋮
… 0⋮ ⋱ ⋮… ⎢()()()⋮()⎥ () (25)
Now, the residual vector based on state-space model for
continuous time is given in (26) which is deduced in [14] from
the residual generation with parity equation for MIMO process
with transfer functions and polynomial errors.
() = () − () (26)
An important condition to satisfy that both of the first and
second term of (26) are zeros is that W T=0 [14] where W is
called the null space of T and can be obtained by proposing the
greater number of zeros possible at the rows, taking into
account that the lines are linearly independent. In our case of
study, the W matrix obtained by the induction motor is (27):
= − 00 0 00 0 00 00 00 (27)
Where
=
, = +
, = + , = +
(28)
By the assumption that in the healthy operation the parameter
do not change, () = 0, then, a fault is detected when () ≠0. The residuals obtained by the induction motor are
() = () + () + () −
() = −() + () + () (29)
Fault Diagnostic Based on Parity Equations Applied to Induction Moto
- 67 -
7/17/2019 artículoFault Diagnostic Based on Parity Equations Applied to Induction Motor
http://slidepdf.com/reader/full/articulofault-diagnostic-based-on-parity-equations-applied-to-induction-motor 4/7
() = () + ( + ) () + ()− − ()
() = () + [ + ]() + () −
Note that during the steady state, the derived of
() is zero,
and
= , thus, the residual can be simplified, this is so
suitable when the fault type is incipient; considering that is themost common in the electrical machines, then, the residuals
can be reduced as.
() = () + () −
() = −() + ()
() = () −
() = () −
(30)
Likewise, as in the D.C. motor [14], if an additive fault occurs,
all residuals except the decouple one are deflected as shown inTable 1. This supports strongly to locate the sensor faults, and
thus this fault types are easy detectable. When a parametric
fault occurs on R s or Rr there is no a considerable increase in
r 3, thus, a null value can be considered to simplify the fault
detection matrix. Finally, a simple way to distinguish the fault
is by using classical limit-values detectors with a suitable
tuning, taking into account the behaviour of residual obtained
in Fig. 2.
Table 1. Fault detection matrix
faults r 1 r 2 r 3 r 4
p a r a m e t r
i c
R s I 0 0 I
Rr I 0 0 I
L s I 0 I I
Lr I I I I
B 0 I I I
Bl 0 I I I
a d d i t i v e I qs
e I I I 0
ωr I I 0 I
V qse I 0 I I
Where “I” represents a significant change which can be
positive or negative. Then with the above matrix can be seen
that the fault detection probability for R s, Rr , L s, B, Bl and V qse
are 50% and only for Lr , I qse and ωr are 100%.
IV. EXPERIMENTAL AND SIMULATION RESULTS.
In this case study, the change detection is used to locate the
parametric fault. Fig. 2 shows the simulation results in Matlab
obtained for induction motor when a parametric fault has
occurred. In all simulation cases, the transient appears from 0
to 3 seconds, the steady state begins to 3 seconds and the under
fault condition begins to 5 seconds with ±½%of its nominal
parametric value.
0 2 4 610
5
0
5
10
r1
r4
r1
r4
0 2 4 610
5
0
5
10
(a1) r 1, r 4 R= nominal value -50%
(a2) r 1, r 4 R = nominal value +50%
0 2 4 610
5
0
5
10
r1
r4
0 2 4 610
5
0
5
10
r1
r4
(b1) r 1, r 4 R = nominal value -50% (b2) r 1, r 4 R = nominal value +50%
0 2 4 610
5
0
5
10
r1
r2
r3
r4
r1
r2
r3
r4
0 2 4 610
5
0
5
10
(c1) r 1, r 2, r 3, r 4 L = nominal value -50%
(c2) r 1, r 2, r 3, r 4 L = nominal value +50%
r1
r3
0 2 4 610
5
0
5
10
r4
r1
r3
r4
0 2 4 610
5
0
5
10
(d1) r 1, r 3, r 4 L = nominal value -50%
(d2) r 1, r 3, r 4 L = nominal value +50%
r2
r3
r4
0 2 4 610
5
0
5
10
0 2 4 610
5
0
5
10
r2
r3
r4
(e1) r 2, r 3, r 4 B = nominal value -50%
(e2) r 2, r 3, r 4 B = nominal value +50%
Marco A. Rodríguez, Manuel Hernandez, Victor Golikov, Gilberto Martínez
- 68 -
7/17/2019 artículoFault Diagnostic Based on Parity Equations Applied to Induction Motor
http://slidepdf.com/reader/full/articulofault-diagnostic-based-on-parity-equations-applied-to-induction-motor 5/7
Fig. 2 Simulation results in Matlab of residuals by using the
induction motor model with constant rotor flux linkages
The parameters of induction motor used are:
Power 0.75 hp
RMS line to line voltage (V LL) 220 V
Number of poles ( P ) 4 polesStator supply frequency ( f ) 60 Hz
Rotor speed (ωr ) 1800 rpm
Stator resistance ( R s) 2.9 Ω
Rotor resistance ( Rr ) 3.54 Ω
Stator inductance ( L s) 0.105 H
Rotor inductance ( Lr ) 0.156 H
Magnetizing inductance ( Lm) 0.173 H
Moment of inertia ( J ) 0.001 kg.m2
Friction coefficient ( B) 0.025 N.m
The low performance of all residuals during the transient
period (0 to 3 seconds) in Fig. 2 appears by the incorrect
modeling of induction motor used for the residual generation,
provided that the rotor flux linkage phasor be constant.
However, the performance during the steady state is good. An
easy way to validate both of the good performance of the fault
detection technique proposed and the induction motor model
simplified during the steady state is by the simulation software
of power electronics and motor control applications (PSIM),
which contains an induction motor with available parameter
variation. The simulation diagram on PSIM of the induction
motor in healthy operation (output1) and under fault condition
(output2)are shown in Fig.3(a) and the sub diagram of r1 is
shown in Fig. 3(b)with the detection block which implements
the residual equations of (30).
(a) Main diagram on PSIM
(b) Sub diagram of r1
Fig. 3 Diagram on PSIM of residual switch a detection block
by using the internal induction motor model
r1 (Healthy operation) r1 (Under fault condition)
Isa (Healtly operation)Isa (Under fault condition)
Iqs (Healtly operation)Iqs (Under fault condition)
(a) FaultyR= + 50% of its nominal value
r1 (Healthy operation) r1 (Under fault condition)
Isa (Healtly operation)Isa (Under fault condition)
Iqs (Healtly operation)Iqs (Under fault condition)
(b) Faulty R= + 50% of its nominal value
Fig. 4 Simulation results in PSIM of detection circuit for the
residual r1
Fig.4 shows the simulation results on PSIM of residual r 1 when
R s and Rr are under fault with +50% of its nominal values, as
well as, when R s and R r are fault-free. It is evident to notice
that r 1 is affected in (30) when R s and R r are modified by two
reasons: first, because are involved in the first term of r 1; and
second, because the amplitude and phase of motor currents are
affected which are reflected in I qs, affecting the r 1 value too.
On the other hand, ωr and V qs are not affected significantly,
first because the rotor speed is not involved directly with this
parameter and second, because V qs only is based on the power
input supply voltage changes.
A practical way, to increase R s is easy placing resistance in
series with the stator winding. On the other hand, to increase
R r is not far from easy. A way to increase the R r is to break
some rotor bars although the increase of L r is associated too
which does not affect the result significantly because R r and Lr
are associate to the same residual
Fig.5 shows the experimental results of residual r 1 when R s and
Rr are under fault approximately at +50% of its nominal values
and when they are in healthy operation. The residual will be
obtained mathematical by using a data vector obtained with
experimental test.
Fault Diagnostic Based on Parity Equations Applied to Induction Moto
- 69 -
7/17/2019 artículoFault Diagnostic Based on Parity Equations Applied to Induction Motor
http://slidepdf.com/reader/full/articulofault-diagnostic-based-on-parity-equations-applied-to-induction-motor 6/7
0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5(s)3
1
1
3
5 A
Iqs (Healthy operation)
Iqs (faulty R S) Iqs (faulty R r )
(a) q axis stator current in the synchronous frames for the
healthy and under fault condition with the motor stator
current like comparison point
0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5(s)
5
0
2.5
5
7.5
10V
Healthy operation
7.5
Faulty R S Faulty R r
2.5
(b) Residual r1 for healthy condition and faulty condition of
R s and R r
Fig. 5 Experimental results for the residual r1
As expected, the experimental and simulation results have
similar behavior when a fault is carried out. However, the
magnitudes of the residuals are tiny different. In accordance
with simulation results between Matlab and PSIM, the
difference can be attributed to the model of Matlab. In
accordance with experimental results the difference between
PSIM and experimental tests, the difference can be attributed
to many factors as bad symmetrical component reflected in the
zero phase sequence, unbalance in the motor, measurement
errors and poor parameter extraction. However, these errors
can be considered in the definition of suitable thresholds on a
simple platform as Labview with a modest DAQ Target. In Fig.
6 shown the windows of fault detection implemented for this proposed.
Fig. 6 Panel front of Labview for Fault detection
V. CONCLUSIONS
Since a relative large change in the residual set obtained when
an incipient additive fault occurs it is not necessary to use
diagnosis methods to locate the fault, which simplifies the
supervision of the induction motor. Since that the induction
motor model is similar to the C.D. motor model the analysis
allows to assure the existence of parity space and therefore to
obtain the advantages of the fault detection for this system type.
The Park transformed in order to interpret I qs and a simple
algebraic equation is sufficient to develop the mathematical
algorithm of detection system. In particular, the propose
technique, considering V qs equal to RMS value of stator
voltage V s, only uses fourth sensors; three of current and one
of speed. The main disadvantage of this technique applied to
induction motor is that the rotor flux linkages are constant, thus,
it is not possible to locate the particular fault related to each
phase and the synchronous speed is needed for fault detection.
VI. R EFERENCES
[1] M. Pineda-Sanchez, M. Riera-Guasp, J. Roger-
Folch, J.A. Antonino-Daviu, J. Perez-Cruz, R
Puche-Panadero, “Diagnosis of Induction Motor
Faults in Time-Varying Conditions Using the
Polynomial-Phase Transform of the Current,” IEEE
Trans. on Ind. Electron., vol. 58, no. 4, April2011,
pp. 1428- 1439.
[2] C. Bianchini, F. Immovilli, M. Cocconcelli, R.
Rubini, A. Bellini,“Fault Detection of Linear
Bearings in Brushless AC Linear Motors byVibration Analysis,” IEEE Trans. on Ind. Electron.,
vol. 58, no. 5, May 2011, pp. 1684- 1694.
[3] E. G. Strangas, S. Aviyente, S.S.H: Zaidi, “Time –
Frequency Analysis for Efficient Fault Diagnosis
and Failure Prognosis for Interior Permanent-
Magnet AC Motors,”IEEE Trans. on Ind. Electron.,
vol. 55, no. 12, Dec 2008, pp. 4191- 4199.
[4] K. A. Loparo, M. L. Adams, W. Lin, M.F. Abdel-
Magied, N. Afshari, “Fault Detection and Diagnosis
Marco A. Rodríguez, Manuel Hernandez, Victor Golikov, Gilberto Martínez
- 70 -
7/17/2019 artículoFault Diagnostic Based on Parity Equations Applied to Induction Motor
http://slidepdf.com/reader/full/articulofault-diagnostic-based-on-parity-equations-applied-to-induction-motor 7/7
of Rotating Machinery,” IEEE Trans. on Ind. Appl.,
vol. 47, no. 5, pp. 1005-1014, October 2000.
[5] A.M.S. Mendes and A.J. Cardoso, “Fault-Tolerant
Operating Strategies Applied to Three-Phase
Induction-Motor Drives,” IEEE Trans. on Ind.
Electron., vol. 53, no. 6, Dec 2006, pp. 1807- 1817.
[6] M. A. Cash, T G. Habetler and G. B. Kliman.,
“Insulation Failure Prediction in AC MachinesUsing Line- Neutral Voltages,” IEEE Trans. on Ind.
Appl., vol. 34, no. 6, pp. 1234-1238.
November/December 1998.
[7] J. Klima, “Analytical Investigation of an Induction
Motor Drive Under Inverter Fault Mode
Operations,” IEE Proc. Electr. Power Appl., vol.
150, no. 13, pp. 255-262, May 2003.
[8] Guillermo R. Bossio, Cristian H. De Angelo, Jorge
A. Solsona, Guillermo O. García, “Diagnostico de
Fallas en MI Mediante una Estrategia de Estimación
de Posición,” XIV Congresso Brasileiro de
Automática 2002.
[9] C. W. Chan, Song Hua, and Zhang Hong-Yue,
“Application of Fully Decouple Parity Equation in
Fault Detection and Identification of DC Motos,”
IEEE Trans. on Ind. Electron., vol. 53, no. 4, August
2006, pp. 1277- 1283.
[10] R. Iserman, “Model Based Fault Detection and
Diagnosis Methods,” Annual Reviews in Control of
Elsevier, vol. 29, Issue 1, 2005, Pages 71-85.
[11] R. Iserman, “Supervision, Fault-Detection and
Fault-Diagnosis Methods- an Introduction,” Elsevier
Science, Control Eng. Practice, vol. 5, no. 5, 1977,
pp 639-652, 1977.
[12] R. Krishnan, Electric Motor Drives Modeling,
Analysis and Control. Prentice Hall, 2001.
[13] C.W. Chan,Hua Song, Zhang Hong-Yue,
“Application of Fully Decoupled Parity Equation in
Fault Detection and Identification of DC Motors,”IEEE Trans. on Ind. Electron., vol. 53, no. 4, June
2006, pp. 1277- 1284.
[14] Hofing, T. and Isermann, R., “Fault Detection Based
on Adaptive Parity Equations and Single-Parameter
Tracking,” Elsevier Science, Control Eng Practice,
pp 1361-1369, 1996.
Fault Diagnostic Based on Parity Equations Applied to Induction Moto
- 71 -